text
stringlengths 10
379k
|
|---|
Melanocytes are specialized pigment-producing cells residing in the skin and eyes of mammals. Melanin is synthesized and stored in melanosomes, membrane-bound subcellular organelles. These share some characteristics with lysosomes, but also possess several melanosome-specific components, including matrix proteins and melanin-synthesizing enzymes. Melanosome biogenesis can be divided into four morphologically distinct stages. Stage I premelanosomes are nonpigmented vacuoles derived from endosomes, which then acquire internal striations (stage II). Melanin formation results in pigment being deposited onto these striations, giving rise to stage III, and eventually fully melanized electron-dense stage IV melanosomes. The highly dendritic architecture of melanocytes in the skin allows transfer of mature pigment granules to a large number of keratinocytes, resulting in normal pigmentation (; ; ).
The study of mouse mutants displaying alterations in coat color has identified a large number of gene products that affect pigmentation (). Many of these are involved in melanosome function. They include the enzymes required for melanin synthesis, primarily tyrosinase, as well as tyrosinase-related protein (Tyrp/TRP) 1 and Dct (TRP2; ), and molecules with structural roles in melanosome formation, such as Pmel17 (). Others are involved in the regulation of intracellular protein trafficking and organelle biogenesis. Most mouse models for Hermansky-Pudlak syndrome (HPS), a heterogeneous group of disorders associated with albinism in humans, fall into this category. They include subunits of the biogenesis of lysosome-related organelles complexes (BLOCs), of adaptor protein 3 (AP-3), or of the homotypic vacuolar protein sorting complex (; ; ). Others again are required for correct intracellular distribution of melanosomes, for example, Rab27a, melanophilin, and myosinVa ().
Analysis of the coat color mutant “chocolate” () identified the small GTPase Rab38 as another gene involved in the regulation of pigmentation (). The Rab family of proteins consists of >60 members in mammalian cells (). Some are ubiquitously expressed, whereas others are highly tissue specific, reflecting more specialized roles. Each Rab protein shows a characteristic subcellular distribution and may thus represent an important determinant of organelle identity (; ; ). Rabs serve as key regulators of vesicular trafficking between subcellular compartments. They have been implicated in the formation of vesicles at the donor membrane and in the tethering, docking, and fusion of vesicles or organelles with their target membranes (; ). They are also involved in a variety of other processes, for example, the regulation of early endosome motility by Rab5 (), the capture of melanosomes at the cell periphery by Rab27a (), or the remodeling of the cytoskeleton by Rab8 (). The regulatory capacity of Rab proteins is based on their ability to act as GTP-dependent molecular switches, with activation coupled to reversible association with intracellular membranes. In this way, Rabs can control the recruitment of downstream effectors to specific subcellular compartments in an activation-dependent way ().
Rab38 is predominantly expressed in melanocytes and retinal pigment epithelial cells and is localized to pigmented melanosomes. In the mouse, a recessive Gly19 to Val point mutation was identified in Rab38. The resulting coat color phenotype resembles that of the mouse, which carries a mutation in Tyrp1, and reduced levels of Tyrp1 were reported in melanosomes of cultured melanocytes, suggesting an involvement of Rab38 in Tyrp1 transport (). The present study characterizes Rab38 in the mutant and defines a role for Rab38 in melanosome biogenesis. We demonstrate that Rab38 and the closely related Rab32 are important, functionally redundant regulators of melanosomal protein trafficking and melanocyte pigmentation.
Primary skin melanocytes isolated from homozygous mice showed strikingly reduced levels of pigmentation compared with cells from + littermate controls when observed during the initial 2–3 wks in culture. After prolonged culture of 4 or 5 wks, however, melanocytes appeared similar to + controls (). Ultrastructural analysis of these cells showed no major differences between and + in the extent of melanization or melanosome size (). Immortal melanocyte cell lines derived from mice (melan-cht) are also well pigmented ().
Because the phenotype results from a single amino acid change in Rab38, the presence of substantial numbers of mature pigment granules in melanocytes could be due to the mutant protein (Rab38) retaining some functional activity. Rabs are peripheral membrane proteins that rely on geranylgeranylation for association with cellular membranes and function (). We therefore analyzed the expression and subcellular localization of Rab38 in control BL/6-derived melan-Ink4a melanocytes (BL/6) and in melan-cht melanocytes () by subcellular fractionation (). In BL/6 cells, the majority of Rab38 was pelleted by centrifugation at 100,000 , demonstrating association with cellular membranes, and partitioned into the detergent phase upon extraction with Triton X-114, indicating lipid modification. In cells, the total levels of Rab38 were substantially reduced. More important, the protein was exclusively detected in the cytosolic fraction or in the aqueous phase. A control Rab, Rab27a, was pelleted and detergent extracted equally in both cell lines. Recombinant EGFP-tagged Rab38 expressed in BL/6 melanocytes showed a characteristic punctate pattern. It localized primarily to perinuclear membranes but also to the cell periphery, where it was frequently observed overlapping with or in close proximity to pigmented melanosomes (; and see , , and ). EGFP-Rab38 was diffusely distributed throughout the cytoplasm, with a concentration in the nucleus that is characteristic of soluble EGFP-tagged Rab mutants. These results suggest that mutant Rab38 is unstable, lacks posttranslational lipid modification, and does not associate with cellular membranes, rendering it functionally inactive.
A lower rate of pigment synthesis in melanocytes may implicate Rab38 in fine-tuning the kinetics of this pathway. However, it is also possible that Rab38 regulates a more fundamental step in melanosome biogenesis, but that alternative mechanisms can at least partially compensate for the loss of Rab38 function in . Rab38 is closely related to Rab32, another member of the Rab family (; ; ). The mouse Rab38 and Rab32 proteins are 67% identical. This is more comparable to the degree of sequence identity between Rab isoforms such as Rab27a and -27b (72%), Rab3a and -3b (78%), or Rab5a and -5b (82%), which are able to functionally compensate for each other, than to that between more distantly related Rabs, which are typically only between 20 and 30% identical. To determine whether there could potentially be redundancy between Rab38 and Rab32, we examined the expression and subcellular localization of Rab32 in melanocytes. cells expressed similar levels of Rab32 to control BL/6 cells (). In melanocytes cotransfected with EGFP-tagged Rab38 and mRFP-tagged Rab32, almost complete colocalization was observed, with both Rabs being targeted to the same vesicular structures at the cell periphery, as well as within the cell body ().
In addition to melanocytes, Rab38 was expressed in bone marrow mast cells and basophil-derived RBL cells and at lower levels in lung and lung alveolar type II–derived MLE-12 cells (). It was not detected in liver, brain, spleen, or kidney, or in bone marrow dendritic cells, RAW (macrophage), AR42J (exocrine pancreas), MIN6 and INS-1 (insulinoma), or AtT20 (anterior pituitary) cells. Low levels of Rab32 were observed in liver, lung, and kidney, which may indicate its presence in a subpopulation of cells within these tissues. Rab32 was also seen in spleen, bone marrow mast cells, bone marrow dendritic cells, and macrophage-derived RAW cells. Other cell types tested were negative (), suggesting a highly tissue-specific expression pattern for both Rab38 and Rab32.
As shown in , melanocytes possess substantial levels of pigment. Recombinant EGFP-tagged Rab38 was overexpressed in these cells (:EGFP-Rab38) using lentiviral vectors. This resulted in a small but reproducible increase in melanin content relative to noninfected (not depicted) or EGFP-expressing cells (:EGFP; ), indicating that Rab38 activity is indeed able to stimulate pigment biosynthesis.
We then used Rab32-specific siRNA oligonucleotides to investigate the potential contribution of Rab32 to pigment biogenesis. Transfection with four different siRNAs (including 32–1 and 32–2, and others not depicted) resulted in a substantial reduction in Rab32 protein levels, whereas control oligonucleotides showed no effect (). Rab32 siRNA–transfected cells, but not cells treated with control siRNA, gradually lost pigmentation. Pigment was measured in :EGFP cells 10 d after the first of two rounds of transfection. A 70 or 87% decrease in melanin content, respectively, was observed in cells treated with oligos 32-1 and 32-2 (). In contrast, in :EGFP-Rab38 cells, where Rab38 function was restored by stably expressing exogenous Rab38, a knock down of Rab32 had no effect on melanin levels (). This suggests a role for Rab32 in regulating pigment biogenesis and demonstrates that Rab38 and Rab32 can functionally compensate for each other, whereas the loss of both Rab proteins leads to a dramatic reduction in melanocyte pigmentation.
Pigment generation requires the formation of an immature melanosome and its subsequent maturation. To analyze further how Rab38 and Rab32 may contribute to this complex process, the subcellular localization of these Rab proteins was examined in more detail. EGFP-tagged Rab38 (and EGFP-Rab32; and not depicted) displays a characteristic distribution in melanocytes, where it is present at the cell periphery and on a population of vesicular structures in the perinuclear region (). A similar pattern was observed for endogenous Rab38 (). Labeling of EGFP-Rab38–expressing cells with antibodies to tyrosinase or Tyrp1 showed extensive colocalization with both melanosome markers (). At the cell periphery, EGFP-Rab38 partially colocalized with mature pigmented melanosomes. More frequently, however, Rab38-positive vesicles contained tyrosinase or Tyrp1 but were devoid of pigment, especially in the perinuclear region (). The distribution of Rab38 overlapped with that of Rab27a, another melanosome marker, but with Rab38 more prominent on perinuclear membranes and Rab27a more concentrated on peripheral, pigmented melanosomes ().
To determine if nonpigmented Rab38-positive structures could represent newly synthesized, immature melanosomes, we used MNT-1 human melanoma cells, a particularly good model for analyzing early-stage organelles because of their relative abundance (). Again, Rab38 colocalized extensively with Tyrp1, both in the cell periphery and in the perinuclear region (). MNT-1 cells were also double labeled for Rab38 and the melanosomal matrix protein Pmel17 (). Antibody HMB45 is specific for a maturation-dependent cleavage form of Pmel17 and selectively labels stage II (striated but nonpigmented) melanosomes (). Little colocalization was observed between Rab38 and Pmel17 by immunofluorescence (IF) microscopy. Immunogold EM on cells stably expressing EGFP-tagged Rab38 did not detect Rab38 on stage II melanosomes either (). Instead, in addition to mature pigmented (stage IV) melanosomes, EGFP-Rab38 was present on small cytoplasmic vesicles and tubules, frequently seen near the Golgi apparatus or in close proximity to melanosomes (). Many of these vesicular structures also contained Tyrp1. Similarly, EGFP-Rab38 was observed on Tyrp1-positive (not depicted) and on tyrosinase-positive cytoplasmic vesicles in BL/6-derived mouse melanocytes (). We detected little colocalization between Rab38 or Rab32 and markers for the Golgi, the TGN, or the early endosome (unpublished data). Collectively, its subcellular localization suggests a possible role for Rab38 in regulating a vesicular transport step involved in the delivery of both Tyrp1 and tyrosinase from the TGN to the maturing melanosome.
Melanocytes lacking functional Rab38 and Rab32, i.e., cells treated with Rab32-specific siRNAs (Rab38/Rab32-deficient cells), are characterized by a drastic reduction in the number of pigmented melanosomes. The abundance and distribution of Pmel17-positive vesicles, however, resembled that in control cells, as assessed by IF microscopy of HMB45-labeled cells (). Ultrastructural analysis confirmed that nonpigmented stage II melanosomes were still being formed ().
In control cells, Tyrp1 and tyrosinase localized to vesicles distributed throughout the cytoplasm and along the cell periphery, with an additional pool in the Golgi region. In contrast, in Rab38/Rab32-deficient cells, Tyrp1 and tyrosinase were almost exclusively restricted to the perinuclear region (). Little Tyrp1, and virtually no tyrosinase, was detected on peripheral structures in these cells. This change in subcellular distribution suggests that the loss of pigmentation induced by Rab32 knockdown is the result of a disruption in the intracellular trafficking of Tyrp1 and, more important, tyrosinase.
The perinuclear pool of tyrosinase in control cells as well as tyrosinase in Rab38/Rab32-deficient cells showed a high degree of colocalization with TGN38, a marker for the TGN (). Post-Golgi sorting of tyrosinase is thought to involve trafficking through an endosomal compartment (). In Rab38/Rab32-deficient cells, some additional tyrosinase-positive structures were frequently observed in close proximity to the TGN38-labeled compartment. However, these did not appear to possess either AP-3 or the transferrin receptor (unpublished data); thus, it is not clear if they could represent such endosomes.
To distinguish between newly synthesized protein in transit through the Golgi stack and protein retained in this compartment because of a block in transport, protein synthesis was inhibited with cycloheximide for 3 h. In control cells, this resulted in depletion of the perinuclear pool of tyrosinase, whereas protein in peripheral melanosomes was not affected, and no substantial changes in overall levels of tyrosinase were observed. In contrast, the majority of Rab38/Rab32-deficient cells (∼70% of depigmented cells compared with <10% of control cells) showed drastically reduced tyrosinase labeling after incubation with cycloheximide (), indicating degradation of the protein after its exit from the TGN. Any residual tyrosinase seen in these cells remained largely colocalized with TGN38, with no obvious accumulation in a post-TGN compartment. This argues for a critical Rab38/Rab32-dependent step in the trafficking of tyrosinase from the TGN to the melanosome.
cells were transfected with Rab32 siRNA to induce loss of pigmentation and were subsequently infected with virus introducing either EGFP or EGFP-Rab38. Expression of EGFP alone did not affect levels of pigmentation or the characteristic perinuclear localization of tyrosinase. In contrast, in cells expressing EGFP-Rab38, trafficking of tyrosinase to cytoplasmic and peripheral vesicles was restored, concomitant with the cells recovering normal levels of pigmentation (). Approximately 59% of EGFP-expressing cells in this experiment were depigmented and showed restriction of tyrosinase to the TGN region, whereas only 10% of EGFP-Rab38–positive cells had not recovered both pigmentation and a normal subcellular distribution of tyrosinase (). This further supports the conclusion that Rab38 can rescue melanosome biogenesis in Rab38/Rab32-deficient cells and that it restores pigment biosynthesis by mediating tyrosinase transport from a perinuclear compartment to the melanosome.
We show here for the first time that Rab38 and Rab32 act in a functionally redundant way in regulating skin melanocyte pigmentation. These Rab proteins control an important post-Golgi step in the trafficking of key melanogenic enzymes and are therefore critical for melanosome maturation.
The identification of the mutation as Rab38 implicated Rab38 in the regulation of pigment biogenesis (), but the effect of this mutation on Rab38 function remained unclear. Although not an invariant residue within the Rab family, glycine 19 is conserved in >50% of Rabs and is situated within the generally highly conserved GTP binding pocket. The corresponding Rab5 mutant, Rab5, showed increased activity (). The biochemical properties of Rab38, however, are not compatible with a functionally active Rab, and we therefore regard the mouse as equivalent to a Rab38-null mutant.
In comparison with other mouse pigmentation mutants (), the phenotype is very mild. The similarity in coat color, and reduced Tyrp1 levels in melanosomes, led to the suggestion that the mouse may be a phenocopy of the mouse, with the phenotype arising from a defect in Tyrp1 trafficking (). We now demonstrate that pigmentation in melanocytes is dependent on Rab32. The dramatic loss of pigment in the absence of both Rab38 and Rab32 is consistent with a critical role not only in the trafficking of Tyrp1 but also of tyrosinase, the key enzyme in melanin synthesis, which requires the melanosomal environment for catalytic activity (; ). On the basis of these observations, we would predict a Rab38/Rab32 double-knockout mouse to show severe hypopigmentation. Compensation by Rab32 may also explain why mutations in Rab38 have not been identified in human patients with oculocutaneous albinism (). However, the presence of a detectable coat color phenotype in the mouse may indicate subtle functional differences between the two Rab proteins, possibly accounting for the differential expression patterns observed in some specialized cell types.
Functional redundancy between mammalian Rab38 and Rab32 is consistent with the presence of only a single homologue in other species, such as Rab-RP1 in () or Glo-1 in (). Rab-RP1 is mutated in , a eye color phenotype exhibiting defects in pigment granule synthesis (), and Glo-1 mutants lack lysosome-like gut granules (). These observations suggest an evolutionarily conserved role for the Rab38/Rab32-related subgroup of Rab proteins in the biogenesis of specialized lysosome-related organelles (LROs) such as gut granules in , eye pigment granules in , and mammalian melanosomes.
Deficiencies in the biogenesis of LROs are also characteristic of HPS and the mouse models for HPS, with melanocytes and platelets most severely affected. The corresponding proteins (AP-3, Vps33, and the BLOC subunits), however, are expressed ubiquitously, suggesting more general roles in the biogenesis of lysosomes (; ). In contrast, Rab38 and Rab32 (; ; ) were restricted to cell types characterized by the presence of LROs, a morphologically and functionally diverse group of organelles that includes melanosomes, platelet-dense granules, mast cell granules, lamellar bodies of lung epithelial cells, lytic granules of cytotoxic T-lymphocytes, and MHC class II compartments of antigen-presenting cells (; ). This further supports a highly specialized role for Rab38 and Rab32 in LRO function.
Recent studies investigating the intracellular trafficking of melanosomal integral membrane proteins like Pmel17, tyrosinase, and Tyrp1 have revealed much about the formation of melanosomes and the complex sorting pathways involved (; ,; ). Premelanosomes appear to be largely derived from endosomal precursors, which progress to stage II organelles through the Pmel17-dependent formation of lumenal striations. These provide the fibrillar matrix for the subsequent deposition of melanin, initiated by the delivery of melanogenic enzymes, i.e., tyrosinase and tyrosinase-related proteins. The differential enrichment for Pmel17 in early-stage organelles and tyrosinase and Tyrp1 in later-stage pigmented structures provides evidence for the existence of distinct sorting pathways to premelanosomes and to more mature organelles (). Our results suggest that Rab38 and Rab32 are not required for the formation of early-stage melanosomes. Trafficking of tyrosinase and Tyrp1 to these organelles, on the other hand, is dependent on Rab38 or Rab32.
The subcellular localization of these Rab proteins argues for their recruitment to post-TGN transport vesicles and a role in regulating the subsequent delivery of such vesicles to maturing melanosomes. Given the loss of tyrosinase in the absence of Rab38 and Rab32, this may represent a tissue-specific trafficking route critical for diverting proteins destined for LROs away from the degradative pathway to lysosomes. The TGN localization of tyrosinase in Rab38/Rab32-deficient cells is consistent with the recruitment of these Rabs to carrier vesicles derived directly from the TGN. Melanosomal targeting of tyrosinase is thought to involve transit through an early endosomal compartment (), which may implicate Rab38/Rab32 in TGN to endosome trafficking. Alternatively, tyrosinase may traverse the endosome normally in the absence of Rab38 and Rab32, but may require their subsequent recruitment to endosome-derived transport vesicles. The latter scenario is supported by the continued presence of Rab38 on or in close proximity to mature melanosomes. A detailed analysis of the different subpopulations making up the endosomal system should prove informative with regard to pinpointing the precise site of action of Rab38 and Rab32.
Morphological studies of melanosomes in a range of HPS mutants documented disruption of organelle maturation at distinct stages in different mutants (; ; ). The molecular mechanisms and sites of action of most of the corresponding proteins, in particular the BLOC components, are still unknown. However, aberrant localization of tyrosinase and Tyrp1 is seen in many forms of HPS (; ). A more detailed comparison of melanosomal protein trafficking in the different mutants should shed light on the potential interplay between Rab38/Rab32 and other HPS proteins during melanosome biogenesis.
(Rab38/Rab32 homologue) and or (AP-3) double mutants show considerably reduced eye pigmentation compared with the corresponding single mutants, indicating that may function in an AP-3–independent pathway (). The association of Rab38 with Tyrp1 trafficking () and the mistargeting of tyrosinase but not Tyrp1 observed in melanocytes lacking AP-3 () raised the possibility that tyrosinase may traffic via an AP-3–dependent pathway, whereas Tyrp1 followed a separate Rab38-dependent route. Our results, however, demonstrate a role for Rab38/Rab32 in tyrosinase trafficking as well. The lack of colocalization between Rab38/Rab32 and AP-3 (unpublished data) is consistent with the regulation of distinct steps. The plasticity of the sorting pathways involved, though (), could certainly result in additive effects upon removal of two components that normally act sequentially within the same pathway. Melanosome targeting of tyrosinase was more severely impaired than that of Tyrp1 upon loss of Rab38/Rab32, possibly indicating that the latter can access alternative pathways more efficiently. For a better understanding of Rab38/Rab32 function and the trafficking pathways regulated by this subfamily of Rab proteins in melanocytes and other cell types, the identification of interacting partners for Rab38 and Rab32 will be of great interest.
C57BL/6J-Rab38 (+/) mice were obtained from The Jackson Laboratory and were maintained and propagated under UK project licenses 70/5071 and 70/6210 at the Central Biomedical Services of Imperial College London. Primary mouse melanocytes were derived as described previously () and were maintained in RPMI-1640 supplemented with 5% FCS, 200 nM phorbol 12-myristate 13-acetate, 200 pM cholera toxin, 100 U/ml penicillin G, and 100 U/ml streptomycin at 37°C with 10% CO. The immortal melanocyte cell line melan-Ink4a was derived from C57BL/6J mice homozygous for an exon 2 deletion (). Eight independent melan-cht lines (melan-cht-1–8) were generated by crossing C57BL/6J-Rab38/Rab38 mice with –null mice (to circumvent cell senescence). Melanocyte cultures were prepared from individual F2 neonatal mice homozygous for both mutations (provided by L. Lamoreux, Texas A & M University, College Station, TX) as described previously (). Cells were maintained in RPMI-1640 supplemented with 10% FCS, 200 nM phorbol 12-myristate 13-acetate, and 200 pM cholera toxin at 37°C with 10% CO. MNT-1 human melanoma cells were maintained in DME supplemented with 10% AIM-V medium, 20% FCS, nonessential amino acids, and sodium pyruvate.
Mouse tissues were collected from C57BL/6J mice perfused with PBS and were homogenized in 3 volumes of 50 mM Hepes, pH 7.2, 10 mM NaCl, 1 mM dithiothreitol, and protease inhibitor cocktail (Roche). Homogenates were centrifuged at 1,000 for 10 min at 4°C to sediment unbroken cells and nuclei. Cultured cells were pelleted and disrupted by sonication in the aforementioned buffer. Protein concentrations of postnuclear supernatants or whole cell lysates were determined using the Bio-Rad protein assay (Bio-Rad Laboratories). For subcellular fractionation, cells were disrupted as above. Membranes were pelleted by centrifugation at 100,000 for 1 h at 4°C. Alternatively, Triton X-114 (Calbiochem) was added to 1%, samples were incubated on ice for 10 min, followed by 5 min at 37°C, and centrifuged at 16,000 for 3 min at room temperature. Aqueous and detergent phases were recovered, and the detergent phase was reextracted once with buffer. All fractions were adjusted to the same volume, and equal volumes were analyzed by SDS-PAGE and immunoblotting.
To generate polyclonal antibodies to Rab38 and Rab32, the C-terminal hypervariable regions of rat Rab38 (residues 162–211) or mouse Rab32 (residues 175–223) were fused to GST and purified on glutathione–Sepharose beads as previously described (). Sera from rabbits immunized with these fusion proteins were preabsorbed against GST and affinity purified using the same antigens immobilized on AminoLink Coupling Gel (Pierce Chemical Co.) as described previously (). Anti-Rab27a monoclonal antibody 4B12 was described previously (). Rabbit polyclonal anti-tyrosinase PEP7 (IF, 1:150) was a generous gift from V. Hearing (National Institutes of Health, Bethesda, MD). Other antibodies were as follows: rabbit anti-tyrosinase (EM; provided by A. Theos and M. Marks, University of Pennsylvania, Philadelphia, PA; ), mouse anti-Tyrp1 TA99 (IF, 1:200 [ID Labs]; EM [American Type Culture Collection]), mouse anti-Pmel17 HMB45 (IF, 1:50 [DakoCytomation]; EM [Labvision]), sheep anti-TGN38 (1:100; Serotec), rabbit anti-EGFP (EM; Invitrogen), mouse anti-EGFP (immunoblot; 1:1,000; Roche), mouse anti–α tubulin and mouse anti–γ tubulin (1:1,000; Sigma-Aldrich).
Samples were fractionated on 13% SDS-PAGE gels and transferred to polyvinylidene difluoride membrane (Millipore). Membranes were blocked in PBS/0.1% Tween-20 (PBS/T) with 4% nonfat dried milk, incubated with primary antibody in PBS/T, washed four times in blocking solution, and incubated with horseradish peroxidase–conjugated secondary antibody (anti–rabbit or anti–mouse; 1:5,000; DakoCytomation) followed by washing in PBS/T. Bound antibody was detected using the ECL Plus Western blotting detection system (GE Healthcare).
pEGFP-Rab27a, -Rab38, and -Rab32 were generated by subcloning cDNA for rat Rab27a, rat Rab38, or human Rab32 into pEGFP (CLONTECH Laboratories, Inc.), using standard cloning techniques. pEGFP-Rab38 was generated from pEGFP-Rab38 using the QuikChange site-directed mutagenesis system (Stratagene) to introduce a single nucleotide substitution in codon 19 (GGT to GTT). For mRFP-Rab38, Rab38 was amplified from pEGFP-Rab38 by PCR using polymerase (Stratagene) and was subcloned into pRFP-C (monomeric red fluorescent protein; a gift from R. Tsien, University of California, San Diego, La Jolla, CA).
For the introduction of plasmid constructs, cells were transfected with Fugene 6 (Roche) according to the manufacturer's recommendations. Per 16-mm well, 0.3 μg DNA and 1 μl transfection reagent were used. Cells were analyzed after 48 h. For siRNA oligonucleotides, mouse melanocytes were transfected with Oligofectamine (Invitrogen). Per 16-mm well, 0.625 μl of reagent was used, and the final concentration of siRNA oligos was 100 nM. For most experiments shown, cells were subjected to two rounds of transfection on days 0 and 5, were passaged as required, and were analyzed on day 10 or 11. siRNAs were purchased from Dharmacon. Oligo 32-1 corresponded to nucleotides 261–280 and oligo 32-2 to nucleotides 602–621 of mouse Rab32. As control, a pool of four nontargeting control siRNAs (Dharmacon) was used.
The lentivirus vector pHR′SIN-cPPT-SEW (a gift from A. Thrasher, Institute of Child Health, London, UK) expresses the EGFP gene under the control of the spleen focus forming virus U3 promoter/enhancer and carries a modified Woodchuck posttranscriptional regulatory element to improve virus titre, and the central polypurine tract sequence of HIV-1 to aid nuclear entry (). EGFP-Rab38 was amplified from pEGFP-Rab38 by PCR using polymerase and was subcloned into this plasmid using standard techniques. Lentivirus particles were produced and titred as previously described (). Cells were plated in RPMI-1640 with 5% FCS (5 × 10 cells per well) and were infected with virus at an approximate MOI of 50 for MNT-1 cells and MOI of 100 for mouse melanocytes.
Cells were disrupted by sonication in 50 mM Tris-HCl, pH 7.4, 2 mM EDTA, 150 mM NaCl, 1 mM dithiothreitol, and protease inhibitors. Pigment was pelleted at 20,000 for 15 min at 4°C, rinsed once in ethanol/ether (1:1), and dissolved in 2 M NaOH/20% dimethylsulfoxide at 60°C. Melanin content was measured as optical density at 492 nm.
Cells were grown on glass coverslips (coated with polylysine for MNT-1 cells) and transfected with plasmid constructs where indicated. For inhibition of protein synthesis, cycloheximide was added to the growth medium to 100 μg/ml, and cells were incubated for a further 3 h. Cells on coverslips were rinsed in PBS, fixed with 3% PFA in PBS for 30 min, rinsed again, and permeabilized with 0.05% saponin in PBS. Coverslips were blocked for 30 min, incubated with primary antibodies for 1.5 h, washed four times and incubated with Alexa 488– or Alexa 568–conjugated anti–mouse, anti–rabbit or anti–sheep secondary antibodies (1:200; Invitrogen), and washed again. All incubations and wash steps were in PBS/BSA/0.01% saponin. Coverslips were mounted in Mowiol (Calbiochem) mounting medium and viewed on a microscope (DM-IRBE; Leica) with a PL Fluotar 40× 1.0 oil objective. Images were acquired using a confocal system (TCS NT; Leica) and processed with Photoshop (Adobe).
For conventional EM, cells grown on coverslips were fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer for 24 h. After several washes with 0.1 M cacodylate buffer, the cells were postfixed with 2% OsO, dehydrated in ethanol, and embedded in Epon while on the coverslips. Ultrathin sections were prepared and counterstained with uranyl acetate and lead citrate before observation. For immunogold labeling, cells were fixed with 2% PFA or with a mixture of 2% PFA and 0.2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. Cells were processed for ultracryomicrotomy as described previously (). Ultrathin sections were prepared with an Ultracut FCS ultracryomicrotome (Leica), and single- or double-immunogold labeled with antibodies and protein A coupled to 5 or 10-nm gold, as indicated in the figure legends. Sections were observed and photographed under an electron microscope (Philips CM120; FEI Company). Digital acquisitions were made with a numeric camera (Keen View; Soft Imaging System). |
Skeletal muscle differentiation begins in the embryonic somites. Soon after their separation from the presomitic mesoderm, somites become partitioned into the dermomyotome and sclerotome (; ; ). Sclerotome cells form the cartilages of the vertebral bodies and ribs. The dermomyotome gives rise to the differentiated skeletal muscle of the myotome and the connective tissues of the dermatome. The dorsomedial region of the dermomyotome is the site of early expression of the skeletal muscle–specific transcription factors MyoD and Myf5 (; ; ) and is a source of cells for the myotome (; ; ).
Skeletal muscle differentiation in the somites is promoted by members of the Wnt family released from the neural tube and overlying ectoderm and by Sonic Hedgehog produced in the notochord (; ; ; , ; ; ). Myogenesis is also regulated by Noggin and Wnt5b that are synthesized within the segmental plate and somites (; ; ; ; ; ; ; ). Noggin promotes myogenesis by inhibiting bone morphogenetic proteins (BMPs) diffusing from the lateral plate mesoderm (; ; ; ; ; ; ; ).
Although inductive molecules are required for the up-regulation of MyoD and Myf5 in the somite and the onset of skeletal muscle differentiation (), both transcription factors are weakly expressed in the presomitic mesoderm (; ; ; ). Cells expressing MyoD mRNA are also present in the epiblast of the chick embryo (; ; ). The epiblast gives rise to all tissues of the embryo (; ; ) and is a source for embryonic stem cell lines (). When MyoD-positive (MyoD) cells are isolated from the epiblast and placed in culture, nearly all differentiate into skeletal muscle (). This population recruits pluripotent epiblast cells to the skeletal muscle lineage in vitro by releasing an inhibitor of the BMP signaling pathway (). In this study, we examined the role that MyoD-expressing epiblast cells play in regulating myogenesis in vivo.
Given that MyoD-expressing epiblast cells produce an inhibitor of the BMP signaling pathway in vitro, and that Noggin is important for muscle differentiation in vivo, we hypothesized that MyoD cells would be incorporated into the somites and produce Noggin. To test this hypothesis, we examined the sites of incorporation of MyoD epiblast cells in the developing chick embryo and determined whether they expressed Noggin. MyoD cells were tracked in the embryo by tagging them with the G8 mAb. G8 recognizes a surface antigen specifically expressed in cells that express MyoD mRNA in the epiblast and fetal organs (, ; ).
Most cells labeled with the G8 mAb in the epiblast () were later found in the somites (). G8-positive (G8) cells were concentrated in the dorsomedial and ventrolateral regions of the dermomyotome and myotome (), and some expressed sarcomeric myosin, which is a marker for differentiation ().
The majority of cells that had been prelabled with G8 in the stage 2 embryo expressed Noggin mRNA and protein in the somites, and most cells expressing Noggin mRNA were labeled with G8 (). Labeling for Noggin protein was more extensive than the distribution of G8 cells (), most likely a reflection of diffusion. A few G8/Noggin-positive (Noggin) cells were also found in the mesenchyme of the head, neural tube, and eyes (). The neural tube contained cells that expressed Noggin, but not G8 (). This pattern of Noggin expression is similar to that reported by . Our double-labeling experiments suggest that cells that express MyoD mRNA in the epiblast become a primary source of Noggin in the somite.
To determine if MyoD epiblast cells are critical for skeletal myogenesis, they were ablated in the embryo by labeling them with the G8 mAb, followed by lysis with complement. Approximately 70 cells (∼0.4% of the total epiblast cells) were ablated in the posterior epiblast of the stage 2 embryo (), which is where MyoD/G8-expressing cells are located (). Embryos treated with complement () or G8 alone (not depicted) had only a few dead cells throughout the entire epiblast. The specificity of G8/complement treatment was demonstrated by incubating embryos with the E12 mAb and complement. E12 labels a subpopulation of epiblast cells that expresses the neurogenic transcription factor NeuroM, but not MyoD (). Cells lysed with E12 and complement were located in the central/anterior and posterior regions of the epiblast ().
Morphogenesis appeared to progress normally for the first 2 d after ablation of MyoD epiblast cells with G8 and complement (unpublished data). Similar numbers of somites formed in treated and control embryos. However, differences were observed as development progressed. Whereas the ventral body wall was closed in 5-d control embryos, it remained open in G8/complement-treated embryos (). Embryos died between the fifth and seventh day after elimination of MyoD cells in the epiblast, and all had herniations of the heart and abdominal organs through the ventral body wall. Treated embryos also exhibited malformations of both eyes, or more commonly, the right eye only (), and twisting of the neural tube at the cervical or sacral level (not depicted). In some embryos, abnormalities were observed in the facial prominences (). None of the control embryos had ventral body wall, eye, or facial defects (). The malformations resulting from ablation of MyoD cells in the epiblast are consistent with their locations in older embryos, as determined by cell-tracking experiments ().
Histological analyses of embryos treated with G8 and complement revealed that the somites partitioned into the dermomyotome, myotome, and sclerotome. However, abnormalities were observed in the morphology of the dermomyotome and myotome, ranging from an enlargement of the dorsomedial and ventrolateral regions (not depicted) to a thickening along the entire length () compared with controls (). The boundary between the dermomyotome and myotome in treated embryos was less discrete than that of control embryos. These morphological differences were paralleled by an increase in cell number in the dermomyotome and myotome-like structure of G8/complement-treated embryos (). The dermatomes of treated embryos were also expanded, whereas the sclerotomes of control embryos contained more cells than those of treated embryos (; ). The notochord () and cartilage rudiments of the limbs (not depicted) appeared similar in treated and control embryos. The neural tube was properly positioned in ablated embryos (), except for twisting at cranial or sacral levels (not depicted). In contrast, the somites of embryos treated with the E12 mAb and complement were similar to other control embryos, except that the neural tube was kinked along its length (). These experiments demonstrate that ablating separate subpopulations of cells in the epiblast results in distinctly different malformations.
The effect of ablating G8/MyoD cells in the epiblast on skeletal muscle differentiation was examined by staining for sarcomeric myosin. Whereas control embryos contained an abundance of muscle in the dorsomedial and ventrolateral myotomes and limbs, G8/complement-treated embryos had reduced myosin staining in the dorsomedial myotome, and severely diminished or no detectable myosin in the ventrolateral myotome (; ). Differentiated skeletal muscle was also decreased in the limbs (). In all G8/complement-treated embryos, differentiation was affected more severely on the right side of the embryo than the left (; ).
The effect of ablating G8/MyoD epiblast cells on muscle differentiation was accompanied by a decrease in MyoD and Myf5 mRNAs in the 5-d embryo (). At this time, few cells in control embryos expressed the marker for myogenic precursor cells Pax-3 () in the dermomyotome; however, there was an abundance of Pax3-positive cells in the dermomyotome of treated embryos (). In contrast, in the 2.5-d embryo, more Pax-3–positive cells were present in the somites of control embryos than those treated with G8 and complement (G and H).
Ablation of MyoD epiblast cells also resulted in a decreased expression of Noggin in the somites compared with control embryos (). The reduction in Noggin was more pronounced on the right side of the embryo than on the left (not depicted). In contrast, Noggin expression in the neural tube appeared increased in treated embryos compared with controls ().
Overall, these results complement those of published studies. First, an inverse relationship was found between expression of MyoD/Myf5 and Pax-3 in the somite (; ; ). Second, inhibition of MyoD and Myf5 expression in Pax-3-positive cells occurs in response to BMP signaling (). Third, Pax-3–positive cells are present in Noggin-null mice ().
To test whether exogenous Noggin could compensate for the loss of MyoD epiblast cells, Noggin-soaked beads were implanted into embryos 2 d after ablating G8/MyoD cells in the epiblast. 4 d later, the aforementioned gross malformations of the ventral body wall, neural tube, and facial prominences were not observed (). The eyes had normal pigmentation after Noggin supplementation, although in some embryos the right eye remained smaller than the left ().
The amount of myosin staining in the ventrolateral and dorsolateral myotomes of G8/complement-treated embryos supplemented with Noggin appeared similar to, or increased, compared with that of buffer-treated embryos implanted with control beads (). Enhanced myogenesis in the myotomes may reflect the accumulation of Pax-3– positive cells before the addition of Noggin. Exogenous Noggin also promoted muscle differentiation in the limbs of ablated embryos, but not to the level of that observed in control embryos (). These results demonstrate that Noggin can, to a significant extent, replace the ablated MyoD epiblast cells to promote both skeletal myogenesis and normal morphogenesis.
The purpose of this study was to determine whether cells that express MyoD mRNA in the epiblast play a role in regulating skeletal myogenesis in the somites. Initially, these cells were found in the posterior epiblast of the two-layered embryo (; ). A few hours later, MyoD cells were present within and lateral to the primitive streak (; ), in a pattern similar to that revealed by fate map analyses of prospective paraxial mesoderm cells (; ; ; ; ). As expected, most MyoD epiblast cells were incorporated into the somites.
A few MyoD cells were observed in areas of the epiblast fated for nonsomitic tissues, and later, in the sclerotome, neural tube, and fetal organs lacking skeletal muscle (, ; ). Cells expressing MyoD and Myf5 outside of skeletal muscle remain undifferentiated (, ; ). This raises the possibility that ectopically placed myogenic precursors, possibly originating in the epiblast, could be a potential source of rhabdomyosarcomas. These malignancies are characterized by the expression of myogenic genes and often arise in structures lacking skeletal muscle ().
MyoD-expressing cells of the epiblast become a major source of Noggin within the somites. Elimination of this population in the epiblast resulted in a reduction of Noggin and skeletal muscle in the myotomes and limbs in older embryos. These results are consistent with previous studies demonstrating that Noggin regulates myogenesis in the somites by inhibiting BMPs (; ; ; ; ; ; ; ; ). Although some epiblast cells initiated the expression of MyoD and the G8 antigen after stage 2 (see Materials and methods), they were unable to completely compensate for those that were ablated earlier in development.
Ventrolateral hypaxial muscles were more severely affected by ablating MyoD epiblast cells than the dorsomedial epaxial muscles. This is opposite to what was observed in Noggin-deficient mice (); however, in G8/complement-treated chick embryos, there was an alternative source of Noggin in the neural tube in close proximity to the epaxial myotome. The paucity of muscle in the ventral myotome of G8/complement-treated embryos was the most likely cause of herniation of organs through the body wall. Malformations of the eye and facial prominences arising from ablation of MyoD epiblast cells may be secondary to a disturbance in the differentiation of facial and extraocular muscles, although it is possible that Noggin produced by MyoD epiblast cells affects nonmyogenic cells.
The malformations that arise as a result of ablation of MyoD cells in the epiblast resemble those that are present in humans with Axenfeld-Rieger syndrome, an autosomal dominant haploinsufficiency of the Pitx2 gene (; ; ). Mutations in Pitx2 affect the development of left-sided structures (; ; ; ; ), whereas elimination of MyoD cells in the chick epiblast affected myogenesis, eye development, and Noggin expression more severely on the right side of the embryo. Although Pitx2 is involved in left–right asymmetry, it is expressed symmetrically in myogenic cells of the eye, limb, and myotome of the mouse (; , ). Pitx2-null mice exhibit dysgenesis of the extraocular muscles and thinning of the abdominal wall (). Given the similarities between the malformations in the chick embryo arising from ablation of MyoD epiblast cells and those that result from perturbation of Pitx2, it is possible that inhibition of BMP signaling by Noggin is involved in Pitx2 expression in the somite during myogenesis. This notion is supported by the fact that BMP2 induces, and Noggin inhibits, the expression of Snail/snr which represses Pitx2 expression (; ; ; ).
One possible explanation for the asymmetric effects of ablating G8/MyoD cells in the stage 2 embryo is that cells initiating expression of MyoD in the epiblast after treatment are preferentially incorporated into the left side of the embryo. It also is possible that MyoD epiblast cells are involved in signaling before their incorporation into the somites. This notion is consistent with our finding more MyoD cells on the right side of Hensen's node (), which is a structure located at the rostral end of the primitive streak that is a rich source of signaling molecules regulating laterality in the embryo (; ). The importance of the node in myogenesis was illustrated in the mouse inverted viscerus mutant embryo, in which defects in the flow of molecules across the node reversed the asymmetric expression of α-skeletal actin and myosin light chain 3F in the myotome (). Cell-tracking experiments that follow the pathways of migration of stages 1–3 MyoD epiblast cells may shed light on the mechanism whereby muscle differentiation is asymmetrically perturbed after their ablation.
We have identified skeletal muscle stem cells in the epiblast based on their expression of MyoD and demonstrated a critical role for these cells in regulating myogenesis in vitro and in vivo (; ; ; this study). postulated that the early chick embryo contains founder cells for the myogenic lineage based on their observation that skeletal muscle emerges in chick blastoderm cultures. In this context, founder cells were defined as those that give rise to all cells of a given lineage. In contrast, MyoD epiblast cells of the stage 2 embryo do not appear to be the sole source of myogenic precursors in the somite. In our cell-tracking experiments, the G8 mAb that had been applied in the stage 2 embryo was not detected in most cells of the myotome. It is possible that asymmetric divisions occurring in the dermomyotome or degradation of the antibody may have reduced the signal in myotome cells. However, cells expressing Pax-3 did emerge in the dermomyotome after ablating MyoD cells in the epiblast, and their ability to differentiate was revealed when ablated embryos were supplemented with exogenous Noggin. Therefore, under these experimental conditions, most Pax-3 precursors do not appear to be the direct descendants of MyoD epiblast cells.
Muscle founder/pioneer cells also have been defined in the avian embryo as those that are the first to differentiate and enter the myotome (). In invertebrate embryos, founder/pioneer cells enter the myotome and serve as a scaffold for myoblast fusion (; ; ). Although G8/MyoD epiblast cells may be involved in both of these processes, their influence clearly extends beyond that of playing a structural role in the myotome. Collectively, the data indicate that the primary function of MyoD epiblast cells within the somites is to promote the differentiation of myogenic precursors by releasing Noggin.
White Leghorn chick embryos were obtained from BE Eggs and staged according to the method of . Stage 2 embryos were removed from the shell on the yolk and placed in a tissue culture dish. 100 μl of G8 mAb diluted 1:40 in PBS was applied to the embryo for 45 min. After rinsing in PBS, embryos were incubated for 30 min in 100 μl rhodamine-conjugated goat anti–mouse FAb′2 fragments (Jackson ImmunoResearch Laboratories) diluted 1:400 and rinsed. Embryos on the yolk were poured into an empty host shell, covered, and incubated at 37°C for 2–4 d. Embryos were fixed in 4% formaldehyde overnight, embedded in paraffin, sectioned transversely at 10 μm, and applied to gelatin-coated slides. Sections were labeled with an antibody to Noggin (R&D Systems) or processed for expression of Noggin mRNA as described in the following paragraphs. Sections were mounted in Gelmount (Biomeda) and observed with an epifluorescence microscope (Eclipse E800; Nikon) using 4×/0.2 NA and 60×/1.4 NA oil objectives. Photomicrographs were produced with the video camera (Evolution QE; Media Cybernetics) and Image-Pro Plus image analysis software (Phase 3 Imaging Systems). Tracking of G8 epiblast cells was conducted in three embryos.
To test for the presence of residual unbound G8 mAb after the initial labeling period, we first determined the number of G8 cells directly after labeling stage 2 embryos. An average of 76 cells in three embryos was labeled with G8. When embryos were labeled with G8 and an Alexa Fluor 488–conjugated secondary antibody, incubated for 3 h at 37°C, and labeled with a rhodamine-conjugated secondary antibody, there was an average of 77 cells labeled with both Alexa Fluor 488 and rhodamine, and no cells were labeled with either fluorochrome alone. A third group of embryos was labeled with G8 and Alexa Fluor 488 secondary antibody, incubated for 3 h, and then exposed to more G8 mAb, followed by the rhodamine secondary antibody. In this case, there was an average of 80 cells with both Alexa Fluor 488 and rhodamine, and 17 cells were labeled with rhodamine alone. These results demonstrate that additional cells are expressing the G8 antigen after the initial labeling period; however, there is insufficient residual G8 mAb present to label those cells. Therefore, unbound G8 mAb does appear to be washed out of the embryo during our labeling procedure, thereby demonstrating the feasibility of tracking MyoD cells from the stage 2 epiblast into the mesoderm.
Stage 2 embryos were removed from the shell on the yolk, labeled with the G8 mAb, and rinsed as described in the previous section. 100 μl of baby rabbit complement (Cedar Lane, Inc.) diluted 1:40 in Hanks' buffered saline containing 0.1% BSA was applied to the embryo for 30 min at room temperature. Control embryos received Hanks' buffer with BSA, G8 mAb, or complement alone. An additional control involved incubating embryos in the E12 mAb that labels a subpopulation of cells expressing NeuroM mRNA in the epiblast (), followed by the addition of complement. The presence of lysed cells was determined directly after treatment by incubating embryos in 0.2% trypan blue in PBS for 15 min at 37°C and counting the number of blue cells. After treatment, embryos not exposed to trypan blue were poured into an empty shell and incubated for 2–7 d. Embryos were analyzed at the gross level and in 10-μm serial sections after embedding in paraffin. The number of embryos in each treatment group is listed in . Some sections were stained with hematoxylin 2 and eosin-Y (Richard Allan Scientific). Other sections were stained with antibodies or analyzed for mRNA expression by in situ hybridization, as described in In situ hybridization.
MyoD cells of the stage 2 epiblast were lysed with G8 and complement and placed in a shell as described in the previous section. Control embryos were incubated in Hanks' buffer only. Embryos were incubated at 37°C to reach stages 11–14. Embryos on the yolk were removed from the shell and placed in a 100-mm tissue culture dish in preparation for addition of Noggin-soaked beads.
1 μl Affigel blue agarose beads (BioRad) was soaked in either 10 μl of a 100-ng human recombinant Noggin (PeproTech) per milliliter of PBS solution or PBS alone (). A PBS- or Noggin-soaked 70-μm bead was inserted on the right side of the embryo, lateral to the fifth and sixth rostral somites, 11th and 12th somites, and between the most caudal somite and the rostral end of the presomitic mesoderm. Embryos were placed in 60-ml capacity glass bowls, covered, and cultured at 37°C for 2–4 d. Embryos were analyzed at the gross level and in transverse, 10-μm serial sections. The number of embryos in each treatment group is listed in .
Paraffin sections were applied to Teflon-printed, three-well glass slides (Electron Microscopy Sciences) coated with 0.2% gelatin. The in situ hybridization procedure is described in detail in . Messenger RNAs for MyoD, Myf5, and Noggin were detected with DNA dendrimers conjugated with Cy3 and the following antisense oligonucleotide sequences: chicken MyoD, 5′-TTCTCAAGAGCAAATACTCACCATTTGGTGA TTCCGTGTAGTA-3′ (L34006; ); chicken Noggin, 5′-TCTCGTTAAGATCCTTCTCCTTGGGGTCAAA-3′ (NM_204123; ); and chicken Myf5, 5′-ATATAGTGGATGGCAGAGCTGAGG ATTTCG-3′ (S53719; ). Fluorescent dendrimers were obtained from Genisphere, Inc. Nuclei were stained with Hoechst dye. Double labeling with dendrimers and antibodies was performed as previously described (, ,; ).
Paraffin sections were labeled with the MF20 mAb to sarcomeric myosin heavy chain () diluted 1:60, a goat anti–mouse polyclonal antiserum to Noggin (R & D Systems) diluted 1:200, or a mAb to Pax-3 () diluted 1:150. Primary antibodies were labeled with rhodamine-conjugated goat anti–mouse FAb′2 fragments (Jackson ImmunoResearch Laboratories) or fluorescein-conjugated donkey anti–goat IgG (CHEMICON International, Inc.), as previously described (). The MF20 and Pax-3 mAbs were obtained from the Developmental Studies Hybridoma Bank.
The number of nuclei present in the dermatome, dermomyotome, myotome, and sclerotome of 20–24 sections through the wing level of two embryos treated with the G8 mAb and complement and two embryos incubated with Hanks' buffer was determined in sections stained with Hoechst dye or hematoxylin and eosin using the Image-Pro Plus image analysis software. The accuracy of cell counting via software analysis was validated by comparing cell numbers to those obtained by manually counting cells. The number of myosin-positive cells was determined in 11–17 sections from the wing level of two treated and two control embryos by manually counting MF20-labeled cells via microscopy. Statistically significant differences in populations were determined using the test. |
African sleeping sickness threatens >60 million people in sub-Saharan Africa, resulting in >70,000 deaths per year, and additional impact is felt through its effects on livestock. The disease is always fatal if untreated, and existing drugs are unacceptably toxic and often require hospitalization (). With outdated existing drugs, new targets for drug development are urgently needed. This search is being aided by a better understanding of the biology of the causative agent, spp.
The life cycle of is divided between a mammalian host and the tsetse fly (). In mammals, the parasite population consists of proliferative slender cells and nonproliferating stumpy forms, which are arrested in G1/G0 (). The transition between these forms occurs in response to a parasite-derived signal in the bloodstream (), resulting in an accumulation of stumpy forms. This optimizes parasite transmissibility because stumpy forms are preadapted for uptake into the tsetse fly (). In particular, they have up-regulated certain metabolic activities in preparation for life in the fly and are more resistant to proteolytic attack and pH fluctuations (; ). In the tsetse, stumpy forms differentiate to procyclic forms, a transition that can be efficiently reproduced in culture by cis aconitate and a reduction of temperature (). Once stimulated to differentiate, stumpy forms embark on a precisely programmed developmental pathway involving changes in cell morphology, metabolic activity, surface antigen expression, and gene expression. Importantly, the generation of stumpy forms in the bloodstream represents an irreversible commitment to differentiate; stumpy forms not taken up in a fly blood meal ultimately degenerate in the bloodstream ().
In higher eukaryotes, tyrosine phosphorylation is a well-characterized mechanism for regulating cell growth and differentiation, as well as many other aspects of cell life (; ). However, less is known about tyrosine phosphorylation events in lower eukaryotes and prokaryotes. For example, in bacteria, protein tyrosine phosphorylation is a rare occurrence and yet tyrosine phosphatases are essential for infection and survival of pathogenic species like , , or (; ; ). Kinetoplastid parasites such as spp. and spp. occupy an interesting evolutionary niche, being unicellular organisms and among the most diverged representatives of the eukaryotic world. Although intracellular signaling events have not yet been described in detail for these organisms, it is likely that tyrosine phosphorylation will also play a role in cellular processes as in higher eukaryotes. Supporting this, there is evidence that several proteins are phosphorylated on tyrosine residues in kinetoplastids (; ) presumably through the activity of dual-specificity protein kinases, as kinetoplastid genomes do not encode any recognizable tyrosine-specific kinases (). Tyrosine phosphatase activity also shows marked differences among different life cycle stages in both and ().
From the precedent in higher eukaryotes, it is likely that phosphotyrosine phosphatases will be also relevant in the control of cell growth and development in kinetoplastids. Supporting this idea, it was recently reported that the heterologous expression of the human PTP1B gene in , together with the inhibition of tyrosine kinases, promoted partial differentiation from promastigote to amastigote forms (). Here, we demonstrate that the activity of a protein tyrosine phosphatase, PTP1, exhibits a pivotal function in parasite differentiation. Biochemical characterization of the enzyme demonstrates that it is a tyrosine-specific phosphatase whose enzymatic activity is regulated by pH and changes in its oxidation state. Importantly, when PTP1 activity is inhibited by RNAi or biochemically, differentiation to procyclic forms occurs spontaneously in the absence of any exogenous trigger. This response is restricted to a subset of bloodstream cells, which we propose are those already committed to differentiation. Supporting this, tyrosine phosphatase inhibition in a homogeneous population of stumpy forms triggers synchronous, efficient, and complete differentiation to proliferative procyclic forms. These data reveal that a tyrosine phosphatase activity is a key molecular regulator of the initiation of trypanosome differentiation, providing a potential pharmacological target to restrict parasite transmissibility and virulence.
To identify molecules implicated in the differentiation competence of bloodstream stumpy forms, we searched the genome database for molecules that define G1/G0 arrest in other organisms. This revealed a 595-bp fragment with limited sequence similarity to the protein tyrosine phosphatase PTPROt (). PTPROt was first identified in mammalian lymphoid organs and is up-regulated in quiescent B cells. The intact gene was then isolated by PCR from cDNA, and the complete gene sequence was determined. This was subsequently verified upon completion of the genome project. This gene, which we have named PTP1 ( phosphotyrosine phosphatase 1), is positioned on chromosome 10 (Tb10.70.0070).
Previous evidence that protein phosphatase activities were differentially regulated during the trypanosome life cycle () prompted us to examine the developmental mRNA expression profile of PTP1 by Northern blotting. Total RNA was prepared from monomorphic bloodstream slender forms, bloodstream stumpy forms, and in vitro–cultured procyclic forms. This revealed that the PTP1 mRNA was expressed in all stages examined, although up-regulated in stumpy forms (1.5–3-fold), which was somewhat variable between samples (). To relate mRNA expression to that of the protein, an anti-peptide antibody was raised against the sequence N-AMKQKRFGMVQRLEQ-C from the amino acid sequence at position 265–279 in PTP1. When reacted against lysates derived from isogenic monomorphic slender and stumpy forms of EATRO 2340 and procyclic forms of Lister 427, approximately equal expression of PTP1 protein was detected. PTP1 expression was also analyzed during synchronous differentiation of bloodstream stumpy forms to procyclic forms. This revealed no transient changes in the protein expression levels of PTP1 during the events of differentiation (). Finally, analysis of the subcellular localization of PTP1 by differential detergent extraction revealed that PTP1 associated predominantly with the cytoskeletal fraction in bloodstream forms (). Although no association with any discrete structure (e.g., the flagellum) could be detected (Fig. S1, available at ), a cytoskeletal association for PTP1 matches the distribution of several tyrosine phosphatases characterized in other organisms (; ). This compartmentalization is believed to contribute to their substrate specificity.
The 1.8-kb PTP1 mRNA contains an ORF of 921 nucleotides encoding a protein of 306 amino acids with a predicted molecular mass of 34 kD. A predicted PEST sequence region is located between residues 135 and 152 (, underlined). Analysis of the amino acid sequence of PTP1 revealed an N-terminal region (43 residues) with no apparent homology to any other protein and a C-terminal region (263 residues) that is homologous to the human protein tyrosine phosphatases PTP1B and PTPROt (24 and 23% identity over the catalytic domain of each molecule, respectively). An extensive search of different trypanosomatid databases, identified orthologues of PTP1 in (PTP1, with 74.1% identity), (PTP1, with 62.0% identity), and (PTP1, with 61.3% identity). The sequence alignment of these orthologues together with human PTP1B is shown in . When sequence similarity is considered, these values rise to 71–80% between PTP1 and trypanosomal PTP1s, compared with 38% to human PTP1B. In contrast, has a syntenic gene encoding a predicted tyrosine phophatase less closely related to PTP1 (42% similarity; LmjF36.2180).
Previous sequence analysis of human PTPs have led to the identification of 10 conserved motifs, some of which are important in substrate binding and catalysis (). The trypanosomal PTP1 subfamily contains all the landmark motifs present in classical tyrosine-specific phosphatases (). These include the phospho-Tyr binding motif (); the WPD loop (M8), which contains the catalytic aspartic acid (the general acid in catalysis); the catalytic P-loop or PTP signature motif (V/I)HCSAGXGR (T/S) (M9); and the Q-loop (M10), which is part of the active site in classic PTPs. Motifs 3–7 (M3–M7) are also present in trypanosomal PTP1s with a high percentage of conservation, consistent with their role as structural motifs located in the core of the PTP catalytic domain. In total, 9 of the 10 well-conserved PTP motifs identified in the mammalian enzymes are present in all the examined trypanosome PTP1s, with only the less-conserved motif 2 missing. Instead of motif 2, a trypanosome-specific motif, T1, replaces this structural motif.
Other distinct motifs were also identified in the trypanosomal PTP1s, generating a total of four trypanosome-specific motifs in the catalytic region (). These are as follows: T1 at position 55–63, “LANEXTIYP”; T2 at position 165–167, “EVD”; T3 at position 239–244, “LIGAYA”; and T4 at position 290–295, “RLGV (D/S) (I/V).” In addition, two other motifs have been identified in the precatalytic (Pc) region: PcT1 “R (M/L)QREFXQLQ” at position 20–29 and PcT2 “ENPRXI (D/N)FTTSL” at position 32–43. These precatalytic motifs are well conserved in all members of the clade (, , and ) and less conserved in . A BLAST search failed to identify any of these motifs in other proteins, indicating that they are unique to the trypanosomal PTP1 family. We hypothesize that these trypanosome-specific motifs may be relevant to regulation of PTP1 or molecular recognition of other cellular targets.
Having identified PTP1 as a putative tyrosine phosphatase, it was important to characterize its enzymatic activity profile and substrate specificity. For this, recombinant protein was produced for the wild-type protein and for two PTP1 mutants, one for the putative catalytic cysteine, Cys 229 (in the P-loop) and one for the catalytic Asp 199 (in the WPD loop). The catalytic mutants of PTP1 were generated by mutating Cys 229 to serine (C229S) and Asp 199 to alanine (D199A). The wild-type enzyme and mutant enzymes were expressed as His-tagged fusions and purified to >95% homogeneity. After removal of the His tag by enterokinase digestion, multiangle light scattering analysis of the enzyme showed that PTP1 behaves as a monomeric protein in solution, with a determined mass of 35,140 D.
The phosphatase activities of the purified recombinant wild-type, D199A, and C229S enzymes were assayed using various concentrations of -nitrophenylphosphate (pNPP), a widely used substrate of tyrosine phosphatases. The pH profile analysis of pNPP dephosphorylation demonstrated that the wild-type enzyme has the highest specific activity at pH 6.0 (), in agreement with the fact that PTPs in general have optimal enzymatic activities at low pH values. The mutant enzyme D199A showed low activity (5–8% of the wild type), and the C229S mutant enzyme was totally inactive under any assay condition and substrate concentration tested (unpublished data). This confirms our predictions from the sequence analysis and demonstrates that both residues, Cys 229 and Asp 199, are essential in the mechanism of catalysis of PTP1, matching their orthologues in mammalian PTPs. Steady-state kinetic analysis of the wild-type enzyme yielded a V value of 0.12 mM min and a k/K of 3.57 mM × s for the dephosphorylation of pNPP. Addition of 2 mM of DTT had a marked effect on the catalytic rate, increasing the V by threefold and the k/K by 14-fold ( and Table S1, available at ), consistent with previous observations suggesting that the catalytic cysteine needs to be in a reduced state for efficient nucleophilic attack of the substrate (). To investigate whether redox events (; ; ) could influence the enzymatic activity of PTP1, we performed a time course activity assay using increasing amounts of the oxidizing agent, hydrogen peroxide (HO). shows that after 10 min of incubation, the addition of 0.25 mM HO completely inactivated PTP1 and 0.025 and 0.1 mM HO reduced the enzyme activity by 37 and 58%, respectively. Importantly, the enzymatic activity of the inactive PTP1 toward pNPP was completely restored after 5 min of incubation with 10 mM DTT. These experiments indicate that PTP1 is sensitive to reversible redox regulation, consistent with several characterized mammalian PTPs ().
To test the specific activity of PTP1, a full kinetic characterization of the wild-type enzyme was undertaken using a range of phosphorylated substrates. These included tyrosine-, serine-, and threonine-phosphorylated peptides; phospho–amino acids; nucleotides; phospholipids; and inorganic phosphorylated compounds ( and Table S1). Most important, this analysis showed that PTP1 favored tyrosine-phosphorylated substrates in the dephosphorylation assays while showing 10–40-fold less activity against the Ser/Thr-phosphorylated substrates. In contrast, PTP1 exhibited little or no activity toward the phospholipids, nucleotides, or inorganic phosphocompounds tested, suggesting that the purified enzyme does not exhibit either lipid phosphatase or phosphoesterase activity. Analysis of the saturation kinetics of PTP1 was also performed using the Tyr-phosphorylated EGF receptor peptide (pEGFR), the Insulin receptor peptide (pInsulin), phospho-Tyr (pTyr), and phospho-Ser (pSer). The results obtained showed that PTP1 dephosphorylates the pEGFR peptide with the highest V, 1.5–2 times higher than for pInsulin peptide and pTyr, and six times higher than for pSer. As activity toward nucleotides, phospholipids, and inorganic phospocompounds was considerably lower or nonmeasurable, no further kinetic analysis was considered relevant for these compounds.
In a final series of biochemical experiments, the inhibitor profile of PTP1 was investigated. The phosphatase activity of PTP1 was assayed using pNPP or pEGFR peptide in the presence of different inhibitors. The PTP-specific inhibitor sodium orthovanadate impaired the catalytic activity of PTP1 in a concentration-dependent manner (). Dephosphorylation of both pNPP and of the pEGFR peptide was reduced by 90% by 1 mM sodium orthovanadate, and the activity was completely abolished by 10 mM of this PTP inhibitor. In contrast, the Ser/Thr phosphatase inhibitor sodium fluoride had no measurable effect on the activity of PTP1 toward either pNPP or pEGFR, even at concentrations of 50 mM. Likewise, no effect was observed with the PP1- and PP2A-specific inhibitor okadaic acid or with the alkaline phosphatase inhibitor tetramisole. Combined, these experiments confirmed our prediction, based on sequence analysis, that PTP1 exhibits all the biochemical characteristics typical of a tyrosine-specific phosphatase.
To initiate a functional analysis in vivo, PTP1 was targeted in cultured bloodstream forms by RNAi. The PTP1 coding region was inserted into the vector pZJM () between two opposing T7 RNA polymerase promoters and transfected in to the “single marker” bloodstream line, which expresses T7 RNA polymerase (). This results in PTP1 double-stranded RNA expression, thereby invoking RNAi. We successfully generated several viable cell lines, although in only one case was effective ablation of PTP1 mRNA and loss of protein observed, and this generated a subtle growth phenotype (Fig. S2, available at ). During phenotypic characterization of this cell line, we observed that a small proportion (2–12%) of cells spontaneously differentiated to procyclic forms despite maintenance of the cells in bloodstream form culture conditions (HMI-9 culture medium at 37°C). This phenotype manifested itself as the presence of cells in the population that stained strongly with antibodies specific for both EP and GPEET () procyclin (). Significantly, this did not represent loss of procyclin gene expression control alone (as can be observed at a frequency of up to 0.1% in wild-type bloodstream populations); analysis of the procyclin-stained cells indicated that they also exhibited characteristic procyclic-form morphology, with their mitochondrial genome (kinetoplast) being positioned away from the cell posterior (; ). They also expressed the procyclic stage–specific cytoskeletal antigen CAP5.5 (; unpublished data). Unfortunately, the differentiation phenotype was rapidly unstable and thus difficult to maintain in culture, with the proportion of differentiated cells decreasing in successive passages, even in the absence of induction.
The instability of the phenotype observed with PTP1 RNAi prompted a search for alternative approaches to target the activity of this enzyme. Recently, a series of highly specific sulphonamido-benzabromarone allosteric inhibitors of PTP1B have been reported (), of which 3-(3,5-Dibromo-4-hydroxy-benzoyl)-2-ethyl-benzofuran-6-sulfonicacid-(4-(thiazol-2-ylsulfamyl)-phenyl)-amide (BZ3) exhibits good cell permeability. This compound prevents closure of the PTP1B WPD loop, thereby preventing dephosphorylation of substrates. To investigate whether this would provide a useful reagent for functional analysis of PTP1, we tested the inhibition profile and specificity of BZ3 against PTP1 and other PTPs. We found that 10 μM of BZ3 reduced the activity of PTP1 at least 50%, consistent with its activity against human PTP1B (IC = 8 μM; ), whereas a dual-specificity phosphatase and the human low molecular weight phosphatase HCPTPB, which lacks the WPD loop, showed no inhibition at this concentration ().
Having verified the specificity of BZ3 for PTP1, we exposed monomorphic bloodstream forms in culture to a titration of BZ3 ranging from 50 to 150 μM, which was in the same range as the concentration effective against mammalian PTP1B in CHO cells in culture (250 μM; ). Strikingly, this also resulted in spontaneous differentiation of a subset of the parasites in response to the inhibitor, with 150 μM BZ3 stimulating procyclin expression in the bloodstream population at a frequency of 9.4% (). Moreover, the phenotype of the cells was consistent with that observed in the RNAi line; i.e., the procyclin expressers in the population also exhibited kinetoplast repositioning, indicative of morphological differentiation (unpublished data). In contrast, exposure of monomorphic bloodstream forms to a non–cell-permeable PTP inhibitor, sodium orthovanadate (at 1 mM), did not result in detectable differentiation, nor did exposure to 1 mM NaF, a Ser/Thr phophatase inhibitor (). Thus, PTP1 inhibition promotes spontaneous differentiation to procyclic forms in the absence of exogenous trigger in a subset of bloodstream forms in culture.
To verify that PTP1 was a target of BZ3 in vivo and that this was linked to the differentiation phenotype observed, transgenic bloodstream cells lines were generated which ectopically expressed wild type or a D199A mutant of PTP1, which binds but not release substrates (), acting as a dominant-negative mutation (). When exposed to 150 μM BZ3, cells expressing the D199A mutant reproducibly exhibited significantly enhanced differentiation when compared with the same cell line in the absence of tetracycline induction (18.1% in the induced population versus 11.6% in the uninduced population; Tukey post hoc comparison, P = 0.049). In contrast, no enhanced differentiation was observed in cells that ectopically expressed the wild-type PTP1 (9% in the induced population versus 10.3% in the uninduced population; P = 0.86). This demonstrates that expression of the D199A mutant of PTP1 increases differentiation in response to BZ3. This provides further evidence that PTP1 is a target of BZ3 in vivo and that this is linked to the observed differentiation phenotype.
We wished to investigate why only a subset of monomorphic bloodstream trypanosomes underwent spontaneous differentiation to procyclic forms when PTP1 was inhibited. Monomorphic bloodstream forms are so named because they have lost the capacity to generate morphologically stumpy forms through prolonged laboratory passage. However, they do retain the capacity to differentiate to procyclic forms when stimulated with cis aconitate, although this is asynchronous in the population and of variable efficiency. We previously proposed that this asynchrony arises from the requirement for individual cells to undergo commitment to differentiation (; ). These committed cells remain slender in morphology but are functionally equivalent to stumpy forms in terms of their ability to differentiate, a condition we have termed stumpy* (; ). We hypothesized that the small proportion of monomorphic cells that differentiated in response to BZ3 or PTP1 RNAi were these stumpy* forms, which was supported by the increased differentiation of these RNAi lines when incubated with a cAMP analogue reported to promote stumpy formation (Fig. S2 C). To further evaluate this, we investigated the response to BZ3 of homogenous populations of stumpy forms, which are uniformly and irreversibly committed to differentiation to procyclic forms. Our prediction was that these cells would show a highly efficient differentiation to procyclic forms in response to BZ3 when compared with cultured monomorphic lines.
Initially, we assayed BZ3-induced differentiation in pleomorphic slender cells derived from a rodent infection. This resulted in an ∼7% differentiation after 24 h, consistent with the response of cultured monomorphic cells (Fig. S3, available at ). Thereafter, trypanosome populations highly enriched for stumpy forms were harvested from a mouse infection, and these cells were exposed either to 150 μM BZ3, 6 mM cis aconitate, or 0.3% vol/vol DMSO (used to solubilize BZ3). In all cases, the cells were then maintained in bloodstream-form culture conditions (37°C in HMI-9) over a period of 24 h, conditions that are compatible with both efficient differentiation of cis aconitate–stimulated cells () and maintenance of the viability of undifferentiated stumpy forms. Samples were harvested at various time points and analyzed for the expression of EP procyclin by immunofluorescence and for morphological differentiation by quantitative analysis of the kinetoplast–posterior dimension in each cell population. and demonstrate the remarkable result of this analysis: stumpy populations differentiated to procyclic forms synchronously and efficiently when exposed to BZ3, with the kinetics of this being equivalent to cells exposed to cis aconitate (). Thus, by 6 h in 150 μM BZ3, >70% of cells were EP procyclin positive (), and by 24 h the kinetoplast–posterior dimension had increased from ∼1 μm (equivalent to that seen in bloodstream forms) to ∼5 μm (equivalent to that in procyclic forms; ; ). In contrast, cells exposed to DMSO alone remained as bloodstream stumpy forms and showed no evidence of differentiation to procyclic forms (). Moreover, by 24 h, differentiated cells in the 150 μM BZ3–treated population were undergoing cell division as differentiated procyclic forms and had induced expression of the procyclic stage–specific cytoskeletal protein CAP5.5 (; ; ). We conclude that inhibition of PTP1 stimulates committed bloodstream-form trypanosomes to differentiate to procyclic forms in the absence of any previously described trigger for this process.
In this study, we report the identification, biochemical characterization, and functional analysis of a novel protein tyrosine phosphatase, PTP1, implicated in life cycle differentiation control in PTP1 was originally identified in a search for molecular markers of the nondividing bloodstream form of , by analogy to other molecules associated with G0 arrest in eukaryotic cells. Bioinformatics analysis of the PTP1 sequence and further searches in other kinetoplastid genomes revealed the existence of orthologue protein tyrosine phosphatases in the other trypanosomatids , , and . All of them contain the well-defined motifs conserved in classical PTPs, necessary for catalysis and substrate binding. In addition, distinct motifs are conserved in the trypanosomal subfamily of phosphatases, both in the catalytic region and in the N-terminal region, upstream of the catalytic domain. We suggest that these spp.–specific motifs could be important in functional regulation of PTP1.
When expressed as a recombinant protein, PTP1 preferentially dephosphorylated phosphotyrosine substrates and was inhibited in vitro by sodium orthovanadate, a known tyrosine-specific phosphatase inhibitor. Moreover, the enzyme was regulated by reversible oxidation and mutation of the predicted essential catalytic residues C229 and D199 resulted in proteins with either severely impaired enzymatic activity (D199A) or total inactivity (C229S). Together, these results confirm the assignment of PTP1 as a phosphotyrosine-specific phosphatase. This contrasts with previously reported phosphatases in , for which our sequence analysis failed to find any of the conserved PTP-specific motifs or any homology to the classical PTPs (). Thus, PTP1 is to our knowledge the first report of a cloned gene encoding a bona fide nonreceptor tyrosine-specific phosphatase in these organisms.
The biological function of PTP1 was investigated by two experimental approaches: gene-specific RNAi and using a cell-permeable PTP1B-specific inhibitor in vivo. The first approach implicated PTP1 in life cycle regulation, with spontaneous differentiation being observed in a subset of cultured bloodstream-form cells. However, this phenotype was unstable, and the proportion of differentiated cells was reduced to wild-type levels with continued culture of the parasites whether cells were induced or not. This is not surprising, as transfection procedures and continued passage at 37°C in bloodstream-specific culture media would quickly select against populations that generate procyclic forms. Moreover, other functions of PTP1 in proliferative bloodstream forms cannot be excluded. Therefore, we also analyzed the effect of a recently characterized cell-permeable allosteric inhibitor of PTP1B, a benzabromarone derivative (). This inhibitor prevents the closure of the WPD loop of PTPs, therefore keeping the enzyme in a catalytically inactive form. We demonstrated that this inhibitor was effective against PTP1, but not against the human low molecular weight PTP (LMWPTP) or a dual-specificity phosphatase. Analysis of the phosphatase complement of the complete genome identified only two putative proteins with an intact WPD loop motif and predicted tyrosine phosphatase activity. One of these is PTP1, and we have confirmed its BZ3 sensitivity here. The second enzyme, which we term PTP2 (Tb11.01.5450), is not likely to contribute to the observed differentiation phenotype, as RNAi against this enzyme does not elicit differentiation of bloodstream forms (unpublished data). Although we cannot exclude other potential uncharacterized targets in trypanosomes, the compatible phenotypes resulting from genetic and pharmacological inhibition of PTP1 implicate this enzyme as being responsible for the differentiation phenotype. Moreover, specificity of the response was further supported by the enhanced differentiation observed in BZ3-treated cells expressing the dominant-negative D199A mutant of PTP1.
Only a small subset of bloodstream-form cells in asynchronous in vitro cultures were stimulated to differentiate to procyclic forms by BZ3. However, the most striking observation was the effect of this inhibitor on uniform populations of stumpy forms. Here, treatment with the PTP1B inhibitor resulted in efficient differentiation to procyclic forms. We observed that stage-regulated protein expression (EP procyclin; CAP5.5), kinetoplast repositioning, and cell cycle progression all occurred with equivalent efficiency and on the same time scale as cells treated with cis aconitate, an established trigger for trypanosome differentiation. We interpret the differentiation of the small proportion of cells in bloodstream cultures as the response of the subset of cells in this population, which have entered division arrest or committed to other early events in stumpy formation, before morphological transformation. We previously referred to these cells as stumpy* forms ().
In , we present a model for the role of PTP1 in the control of bloodstream to procyclic differentiation. In this model, bloodstream parasites commit to stumpy formation and become competent (stumpy*) to differentiate to procyclic forms. However, they are held arrested in this state in the bloodstream by the action of PTP1. Then, upon uptake by the tsetse fly, this inhibition is removed and the parasites progress unhindered to procyclic forms, progressing through a highly programmed developmental pathway in which cellular events occur on a predetermined pathway and time scale. This model has two important implications. First, it supports the concept that proliferative slender cells are not competent to differentiate to procyclic forms unless they commit to the early events of stumpy formation, in particular, cell cycle arrest. Also, it suggests that the default pathway for stumpy forms is to differentiate directly to procyclic forms and that this differentiation is inhibited in the bloodstream by the action of PTP1. In other words, differentiation primarily results from release of the “brake,” rather than application of the “accelerator.” This scenario is compatible with the role of PTP1 in slug development, where removal of a PTP1 precipitates accelerated differentiation via cell aggregation ().
This model does not predict what is the natural signal that triggers the inactivation of PTP1, although one possibility is the different conditions to which the trypanosome is exposed upon entering the fly. For example, pH fluctuations have been observed in the tsetse digestive tract from pH 9.0 to pH 10.2 (), at which the activity of PTP1 is considerably lower. Potentially combined with other specific signals, for example, changes in redox conditions upon tsetse uptake, this would then inactivate PTP1 and license the trypanosome for cell cycle reentry and differentiation into a proliferative procyclic form.
The proposed role of PTP1 in preventing inappropriate differentiation of stumpy forms in the mammalian bloodstream is essential for transmissibility of the parasite. If not strictly controlled, the progression to procyclic forms would result in rapid death of the stumpy population in the bloodstream because of either the activation of complement by the alternative pathway (killing cells that have lost the variant surface glycoprotein) or the generation of antibodies to invariant procyclic surface antigens. This makes PTP1 a key component of the regulation of the trypanosome life cycle and hence a potentially important pharmacological target for controlling trypanosome transmission, for example, in epidemic foci involving intensive human–human or human–livestock transmission. If coupled with factors to promote stumpy formation, targeting PTP1 may also help limit parasite virulence and promote trypanosome clearance. There is intense interest in the development of PTP1B inhibitors in the pharmaceutical industry because of the importance of these enzymes in diabetes and obesity (; ), and we are currently investigating the opportunity for piggyback strategies to target kinetoplastid PTPs. Significantly, the presence of trypanosome-specific motifs in this enzyme offers the potential for developing inhibitors with specificity for the parasite enzyme with respect to their mammalian counterparts.
Phosphosubstrates and other chemicals were purchased from Sigma-Aldrich. The Threonine (KRpTIRR) and Serine (RRApSVA) phosphopeptides were obtained from Upstate. The PTP1B inhibitor BZ3 was purchased from Calbiochem.
Parasite lines used were EATRO 2340 GUP2965 (monomorphic slender forms) or, for stumpy generation, the isogenic pleomorphic line GUP2962 (). For RNAi analyses, Lister 427 single-marker cells were used (a gift from G. Cross, Rockefeller University, New York, NY; ), whereas established procyclic forms were Lister 427.
For BZ3 inhibition assays, cells in HMI-9 medium () at 37°C were exposed to DMSO, PTP1B inhibitor BZ3, or 6 mM cis aconitate. Samples were assayed by immunofluorescence at room temperature using antibody to EP procyclin (diluted 1:500 in PBS; Cedar Lane Laboratories), antibody to GPEET (diluted 1:200; a gift from I. Roditi, University of Bern, Bern, Switzerland), or the procyclic form cytoskeletal protein CAP5.5 (undiluted hybridoma supernatant; a gift from K. Gull, University of Oxford, Oxford, England; ). Secondary antibodies used were goat anti–mouse (EP procyclin; CAP5.5) or –rabbit (GPEET procyclin) conjugated to either FITC (1:50; Sigma-Aldrich) or Alexa 488 (1:200; Invitrogen). Slides were stained with 1 μg/ml DAPI and mounted in MOWIOL (Harlow Chemical Co.). Morphometric measurements of kinetoplast repositioning used Scion Image 1.62. Images were captured using a Cohu charged-coupled device camera attached to an Axioscope 2 (Carl Zeiss MicroImaging, Inc.) using either Plan Neofluar 63× (1.25 NA) or Plan Neofluar 100× (1.30 NA) phase-contrast objectives. Images were processed using Photoshop CS (Adobe).
A 595-bp fragment related to PTPROt () formed part of the sheared strain TREU 927/4 genomic DNA clone 18E19 (available from GenBank/EMBL/DDBJ under accession no. ). Oligonucleotides were designed to the 5′ (PTP A; 5′-CATCAACCTCGTACGACG-3′) and 3′ (PTP B; 5′-CAAGCCATACAATAATTG-3′) ends of the gene fragment, and were these used in independent amplifications with primers specific for the spliced leader sequence or polyA tail to amplify the gene in two overlapping halves from cDNA. The sequence of the full gene was verified after completion of the trypanosome genome sequence.
RNA preparation and Northern blotting was performed as described by using a digoxygenin-labeled PTP1-specific riboprobe (Roche). Stumpy cell fractionation used the Qproteome kit (QIAGEN) with each fraction being probed with antibodies to α-tubulin (cytoskeletal fraction; a gift from K. Gull), cytosolic PGK (cytosol; a gift from P. Michels, Université Catholique de Louvain, Brussels, Belgium), and BiP (membrane and organelles; a gift from J. Bangs, University of Wisconsin, Madison, WI) as controls. For detection of PTP1, an anti-peptide (NH-AMKQKRFGMVQRLEQ-COOH) antibody was raised in rabbits and affinity purified against the immunogen (Eurogentec). Preimmune serum detected no trypanosome protein. Western blotting was performed according to .
PTP1 transcript ablation in bloodstream forms was achieved by cloning the PTP1 coding sequence into pZJM (), which was then linearized by NotI digestion and transfected as previously described. Transformants were selected with 2.5 μg/ml phleomycin. Ectopic overexpression of PTP1 and D199A mutants used the trypanosome expression vector pHD451.
Dephosphorylation of phosphosubstrates was detected by measuring the release of inorganic phosphate with the malachite green detection system according to the manufacturer's protocol (Protein tyrosine phosphatase assay kit; Sigma-Aldrich). Reaction mixtures (50 μl) contained 5–10 μg of purified PTP1 with concentrations from 0.01 to 0.2 mM of phosphopeptides in 50 mM Hepes and 150 mM NaCl, pH 7.0. Reactions were incubated at 37°C for 15–30 min and quenched by adding of 50 μl of malachite green reagent. After 15 min of further incubation at room temperature, the absorbance of the samples were measured at 620 nm in a microplate reader (Opsys MR; Dynex Technologies), with released inorganic phosphate being determined using a phosphate standard curve. The specific activity is defined as pmoles of inorganic phosphate released in a minute per milligram of protein. Kinetic constants K and V were calculated using the Lineweaver-Burke plot of the reciprocal initial velocity versus the reciprocal concentration of substrates. See the supplemental text (available at ) for information about protein expression constructs and recombinant protein purification.
Enzyme activity was measured by monitoring PTP1 (5–20 μg) catalyzed hydrolysis of pNPP to -nitrophenol (). A final concentration of 20 mM pNPP was present in the assay. The pNPP assay buffer was 50 mM Tris, 50 mM bisTris, and 100 mM Na acetate, pH 5–7.5. Each reaction (400 μl) was performed in triplicate, being incubated at 37°C for 15 min and quenched by adding 500 μl of 1 M NaOH. The concentration of released -nitrophenol, determined at 405 nm, was converted to millimolar units using a millimolar extinction coefficient of 18.0 mM cm.
The assays (120 μl) contained 60 μg of PTP1 and 0–0.25 mM HO in 50 mM Hepes, pH 7.0, and 150 mM NaCl. Reactions, initiated by addition of HO, were incubated at room temperature. After 15 min, 10 mM DTT was added. 20-μl samples removed at 0, 5, 10, 15, 20, 30, and 40 min were assayed for residual activity using pNPP as substrate.
Table S1 shows kinetic constants for the hydrolysis of phosphosubstrates by PTP1. Fig. S1 shows immunofluorescence localization of ectopically expressed PTP1. Fig. S2 shows growth kinetics, RNA, and protein expression of the PTP1 RNAi line and the response of the PTP RNAi line to pCPTcAMP. Fig. S3 shows BZ3 response of pleomorphic slender cells. Online supplemental material is available at . |
The maintenance of epithelial integrity is closely integrated with the regulation of cell proliferation in a variety of biological contexts, including normal development, tissue regeneration, and tumor progression. During mammalian development, there is close linkage between regulation of the cell cycle and the ability of neural crest progenitors to delaminate from the neurepithelium and initiate migratory behavior (). In addition, epithelial wounding produces a local stimulation of proliferation as a result of the disruption of cell contacts (; ). Most importantly, recent studies have revealed that a number of neoplastic tumor suppressor mutations result simultaneously in the disruption of epithelial polarity, tissue integrity, and normal controls on proliferation. For example, loss of the tumor suppressor gene results in highly disorganized cell masses that display uncontrolled proliferation (; ). The underlying basis for the observed tight linkage between epithelial organization and cell proliferation remains unclear, but current models include cell contact–mediated mechanisms for growth arrest, compartmentalized distribution of growth factors, their receptors, and/or intracellular transducers, and the existence of components that have dual but separable roles in epithelial integrity and cell signaling (for example, β-catenin; ). These studies highlight the importance of cellular architecture, particularly the cytoskeleton and its ability to organize the cell membrane through linkage with transmembrane proteins, to regulate both epithelial integrity and proliferation.
The neurofibromatosis 2 tumor suppressor protein Merlin and its close relatives Ezrin/Radixin/Moesin (ERM; ; ) function as membrane-cytoskeletal linkers and regulators of multiple signaling pathways (; ; ). Merlin and ERMs share ∼45% sequence identity and a similar domain organization with an N-terminal 4.1 ERM domain, a putative coiled-coil spacer, and a C-terminal domain that in ERMs binds to filamentous actin (). Merlin has a clear role in regulating proliferation (; ), whereas Moesin and its paralogues Ezrin and Radixin are thought to maintain epithelial integrity by organizing the apical cytoskeleton ().
A central question in the study of these proteins has been how their interaction with binding partners is regulated. For both Merlin and ERMs, there is abundant evidence for an intramolecular interaction between the 4.1 ERM domain and the C-terminal domain (; ; ; ; ; ). In ERM proteins, this interaction produces a closed, inactive form of the protein that does not interact with either transmembrane binding partners or filamentous actin (; ). For Merlin, studies in mammalian cells suggest that the closed form is active in inhibiting proliferation (; ; ; ), whereas studies in suggest that, as with ERMs, the open form of Merlin retains all essential genetic functions (). Whether this apparent distinction between flies and mammals represents a true functional difference or reflects methodological differences remains to be resolved.
Phosphorylation of a conserved threonine (Thr) in the actin-binding domain of ERM proteins has been demonstrated to be important for their activation by relieving the head to tail interaction (; ; ; ; ). The precise kinase responsible for this event is unclear, although its activity seems to be positively regulated by Rho activation in mammalian cells. In , the Ste20 family kinase Slik is necessary for the phosphorylation of Moesin, although, again, it is not clear whether Slik phosphorylates Moesin directly or via intermediate kinases (). In mammalian cells, Merlin activity is regulated by a phosphorylation event at serine 518 that blocks head to tail interactions (). However, unlike ERMs, it appears that the phosphorylated form of Merlin is inactive in that it does not suppress growth (). In contrast, hypophosphorylated Merlin is enriched under conditions of serum starvation or confluency, suggesting that this form is growth suppressive (; ; ). Serine 518 is thought to be phosphorylated by the p21-activated kinase (PAK) downstream of Rac activity (Kissil et al., 2002; ), although the possibility of other mechanisms regulating Merlin phosphorylation cannot be excluded. In addition, evidence to date has failed to demonstrate phosphorylation of the equivalent Thr residue to the one phosphorylated in ERMs, although this residue is conserved in both mammalian and fly Merlin.
Many questions remain about the regulation of Merlin activity, particularly in the context of developing tissues undergoing normal proliferation. To better understand how Merlin is regulated, we have investigated the mechanism by which Merlin phosphorylation and, thus, its activity are controlled in . In particular, we have examined the possibility that Merlin and Moesin are regulated by the same molecular mechanism. In this study, we show that Slik kinase, which positively regulates Moesin function, also regulates Merlin but in the opposite direction. In addition, our observations suggest a competitive interaction between Moesin and Merlin for Slik activity. These results provide in vivo evidence of a kinase-based regulation of Merlin and suggest that Merlin and Moesin are coordinately regulated in developing tissues.
Previous studies in and mammalian cells have demonstrated that Merlin displays complex subcellular localizations, being found both at the apical plasma membrane and in punctate cytoplasmic structures that are associated with endocytic compartments (; ; ; ). Deletion mutagenesis indicates that the C-terminal domain is important in regulating Merlin's subcellular localization and its activity in rescue assays (). This domain is similar in structure to the C-terminal domain of ERM proteins, and, although it does not bind actin, the Thr residue that is phosphorylated in ERMs is conserved in both fly and human Merlin (). Collectively, these observations raise the possibility that the phosphorylation state and, therefore, Merlin subcellular localization and function are modulated similarly to Moesin. A previous study has shown that the phosphorylation of Moesin is regulated by the Ste20 family kinase Slik and that like Moesin and Merlin, Slik is localized in the apical region of epithelial cells (). Based on these observations, we investigated possible functional interactions between Slik and Merlin.
cells in heterozygous
(wild type) imaginal epithelia. Induction of a homozygous mutant clone by mitotic recombination simultaneously produces a homozygous wild-type (
) sister clone, thus allowing side by side comparisons between cells containing two, one, or no functional copies of the gene.
cells lack this marker. Optical sections taken below the apical surface of the epithelium () show a clear inverse correlation between gene dosage and Merlin staining.
clones and decreased in homozygous wild-type sister clones relative to the surrounding heterozygous
cells.
clones might reflect altered subcellular localization.
cells. Such redistribution might reflect an altered phosphorylation state for Merlin, as has been observed for the ERM proteins (). For this experiment, we fixed tissues using a TCA treatment that has previously been shown to preserve the phosphorylation state in mammalian ERM proteins (). Our initial experiments indicated that this protocol considerably enhanced detection of the phosphorylated form of Moesin and confirmed the previous report that Moesin phosphorylation is dependent on Slik activity (; ). In these preparations, phospho-Moesin staining was decreased both in apical () and basolateral () optical sections.
epithelial cells ().
clones was decreased at the apical surface of the epithelium, where much of the protein is normally found (). Similar results were also observed using standard PFA fixation in optical cross sections through clones ().
cells is associated with punctate structures (). Thus, the loss of Slik function results in a redistribution of Merlin from a close association with the apical membrane to the basolateral domain of the cell. Similar effects are also observed in clones induced in the follicle cell epithelium that surrounds the developing oocyte ().
To further examine the effects of Slik activity on Merlin subcellular localization, we performed coexpression experiments in cultured S2 cells. Previous studies (; ) have shown that upon induction, Merlin initially localizes to the membrane of S2 cells and then, within 3 h, traffics into punctate cytoplasmic structures that are associated with endocytic vesicles (). Perturbation of the C-terminal domain of Merlin alters its localization and trafficking pattern (). To determine whether Slik affects the subcellular localization and movement of Merlin, a pulse-chase assay was performed in S2 cells using a heat shock–inducible GFP-tagged Merlin expression construct (). Control experiments in which cells were induced to express a pulse of Mer exhibited a similar pattern of Merlin localization to that reported previously (; ). In contrast, the coexpression of Slik with Mer results in a shift in the temporal pattern of Merlin localization. In this case, a substantial proportion of cells displayed Merlin that associated with the plasma membrane even 6 h after induction (). Thus, Slik activity prevents the normal trafficking of Merlin off the plasma membrane and into cytoplasmic punctate structures. Coexpression of a kinase-inactive version of Slik has no effect on Merlin localization or trafficking (, compare G with E). Together with the loss of function clonal analysis, these results indicate that Slik kinase activity controls the localization and trafficking of Merlin.
clones and S2 cells could reflect changes in its phosphorylation state. Therefore, we used immunoprecipitation and immunoblot analysis to examine Merlin phosphorylation under varying levels of Slik activity. Previous studies in mammalian cells have shown that Merlin exists in several isoforms, representing at least two and, under certain conditions, three phosphorylated states (, ). Merlin produces a similar pattern on immunoblots (), where at least three forms can be visualized. Treatment with λ phosphatase converted the slower migrating bands to the most rapidly migrating form (), indicating that the slower migrating forms represent differentially phosphorylated forms of the protein.
When () was expressed in wing imaginal discs under the apterous GAL4 driver, the ratio of phosphorylated to nonphosphorylated Merlin increased compared with wild-type imaginal discs (6.3 ± 1.6 vs. 4.3 ± 0.7; = 6; P = 0.009). In contrast, Merlin isolated from wing discs that overexpressed kinase-inactive Slik showed a phosphorylation pattern that was indistinguishable from wild type (ratio of phosphorylated to nonphosphorylated = 4.6 ± 1.0; = 4; P = 0.44; ). This indicates that kinase activity of Slik is required for the observed effect on the phosphorylation of Merlin protein.
To better characterize Slik effects on Merlin phosphorylation, we next examined these proteins when expressed in cultured S2 cells. A similar pattern of Merlin isoforms is observed on immunoblots when Merlin is expressed in S2 cells, as was seen in wing imaginal discs (unpublished data). Increased phosphorylation of Merlin in the presence of Slik kinase is also observed in S2 cells, albeit with a more subtle effect.
As the Thr residue near the C terminus of Moesin (Thr) is also conserved in Merlin (Thr; ), we wondered whether Slik activity might control the phosphorylation of this site in Merlin. To address this question, we used site-directed mutagenesis to construct phosphomimetic (Mer) and nonphosphorylatable (Mer) versions of the Merlin protein and examined their effect on Merlin phosphorylation in the presence of Slik kinase in S2 cells. Expressed Mer displays a prominent hyperphosphorylated band, whereas this band is much less prominent in expressed Mer (). These results indicate that sites in addition to Thr are phosphorylated in Merlin and suggest that the phosphorylation state of Thr may regulate the phosphorylation of these sites by other kinases. The addition of Slik kinase does not appear to alter the phosphorylation pattern of either mutant (), which is consistent with the notion that Slik acts on Merlin via phosphorylation of the Thr residue.
If Slik's effects on Merlin localization are mediated by phosphorylation, phosphomimetic Merlin mutations should affect subcellular localization in a similar manner to the cotransfection of wild-type Merlin with Slik. To examine this, Mer and Mer were tested in the aforementioned S2 cell trafficking assay. As we observed for wild-type Merlin in the presence of Slik kinase (), Mer alone trafficked very slowly off the plasma membrane (). However, Mer internalized from the plasma membrane to the cytoplasm with even faster kinetics than Mer or Mer coexpressed with kinase-inactive Slik (, compare I with E and G). These results indicate that one effect of phosphorylation is to regulate Merlin trafficking and subcellular localization. They also suggest that phosphorylated Merlin remains closely associated with the plasma membrane, whereas hypophosphorylated Merlin rapidly traffics off of the membrane, possibly in association with transmembrane proteins.
To ask whether Slik interacts directly with Moesin and Merlin, we used an in vitro GST pull-down assay (). The results indicate that bacterially expressed Merlin and Moesin both bind to Slik in vitro. In addition, we attempted to determine whether purified Slik can phosphorylate either Moesin or Merlin in vitro. However, as previously shown for Moesin (), we were unable to detect direct phosphorylation of Merlin or Moesin by Slik kinase (unpublished data). Whether this indicates that Slik acts in vivo via intermediary kinases or requires unidentified cofactors not present in our experiments is unknown, but the observation that Slik interacts directly with both Moesin and Merlin is consistent with the idea that they serve as substrates for Slik's kinase activity.
As a further test of functional interaction between Merlin and Slik, we examined genetic interactions between and mutations. Specifically, we asked whether reducing slik function genetically modifies the phenotype of an activated transgene (; ) that confers growth suppression. The expression of in wild- type wings causes a reduction in size by a mean of 15% from wild type (P = 0.01; ). Using this sensitized genetic background, we asked whether manipulating gene dose affects the activity of endogenously expressed wild-type Merlin. The reduction of dose ( is completely recessive) by one half in this genetic background reduced wing size by a mean of 18% (P = 0.002; ). Thus, reduction in Slik function enhances the phenotype from expressing an activated form of Merlin, suggesting that Slik antagonizes Merlin function. This phenotypic interaction is most likely mediated through Slik's effects on endogenously expressed wild-type Merlin acting synergistically with the coexpressed Mer, which lacks the Thr residue.
The evidence presented thus far supports a model whereby Slik controls Merlin subcellular localization and function by regulating its phosphorylation state. In this model, Slik directly affects Merlin function. Alternatively, it is possible that Slik alters Merlin activity indirectly through its previously documented effects on Moesin function (). To address this question, we asked whether the expression of a phosphomimetic Moesin mutation, Moe, which has been shown to be active even in the absence of function (), could rescue the effects of the loss of on Merlin subcellular localization (). For this experiment, we used the mosaic analysis with a repressible cell marker (MARCM) technique to express Moe specifically in somatic mosaic clones. This technique allows the overexpression of one protein (Moe) while removing the expression of another protein (Slik) in the same set of cells.
cells, indicating that the effect of Slik on Merlin is not mediated through its effects on Moesin activation.
If Merlin and Moesin are substrates for Slik-dependent phosphorylation, one might predict that Moesin and Merlin act competitively for Slik activity. To address this, we examined the effect of Slik overexpression on Merlin localization in the apical domain in the presence or absence of Moesin protein. The expression of Slik alone under the control of the GAL4 driver in the posterior compartment resulted in no discernable effect on Merlin (). To address the potential role of Moesin, we simultaneously reduced Moesin function using a transgene that produces double-stranded RNA for Moesin (). Reduction of Moesin expression using this RNAi transgene alone results in a subtle increase in Merlin protein staining in the apical domain (). However, the expression of wild-type Slik in combination with a reduction in Moesin produced a marked increase in Merlin protein staining in the apical domain (), indicating a shift toward increased apical localization. In contrast, coexpression of kinase-dead (kd) Slik (Slik) and the Moesin RNAi transgene had no apparent effect on Merlin (), indicating again that Slik kinase activity is necessary for these effects.
We also addressed the relationship between Merlin and Moesin using the aforementioned S2 cell trafficking assay. Coexpression of Merlin and Moesin does not alter the subcellular trafficking of Merlin (). However, the coexpression of Moesin blocks the effect of Slik on Merlin trafficking (, compare E with K), which is consistent with the hypothesis that Moesin and Merlin act as competitive substrates.
A previous study has shown that the Slik kinase positively regulates Moesin activity via phosphorylation near the C terminus, thereby inhibiting activation of the Rho small GTPase and promoting epithelial integrity (). Overexpression of Slik in imaginal tissues results in the hyperphosphorylation of Merlin, suggesting that in addition to Moesin, Slik regulates the phosphorylation state of Merlin. Interestingly, in mammalian cells, Merlin phosphorylation is affected by PAK, which, like Slik, is a member of the Ste20 family of kinases (). Current models of Merlin function predict that hyperphosphorylated Merlin is inactive (), which is consistent with our observation that functions antagonistically to in genetic interaction tests. In accord with this notion, was originally identified in a misexpression screen by its ability to cause overproliferation when expressed ectopically in imaginal epithelia (). Collectively, the data presented here leads us to predict that activity of the Slik kinase coordinately regulates both epithelial morphology and, at the same time, cell proliferation (for summary see ). To our knowledge, this is the first demonstration of a single mechanism with the potential to regulate both processes simultaneously in developing tissues.
We speculate that the observed coordinate regulation of Merlin and Moesin may be important in the developing imaginal discs during larval and pupal development. In larval stages, most imaginal epithelia proliferate rapidly and at the same time maintain a highly structured epithelial monolayer (). At this stage, Slik activity could allow high rates of proliferation and simultaneously promote epithelial integrity that is necessary to prevent the unregulated growth or invasive cell behavior. At the end of larval life and at the onset of metamorphosis, the cell cycle slows dramatically, and, at the same time, the imaginal discs radically change shape during a morphogenetic process termed eversion. Previous studies have shown that these shape changes require rearrangements of local contacts between cells (; ; ), suggesting that epithelial integrity must be modulated. We predict that at this stage, Slik function may be decreased to coordinate these changes in the imaginal epithelium. Further studies to examine Slik expression and the regulation of its function will be of interest in this regard.
This study also provides the first genetic evidence that Moesin and Merlin functionally interact through competition for Slik kinase activity, although previous studies have shown physical interactions between these proteins (; ; ). It is interesting to note that in mammalian Schwann RT4 cell lines, expressing constitutively phosphorylated Merlin not only impairs the ability of Merlin to suppress proliferation and motility but also induces a novel ERM-like phenotype (). attribute this phenotype to the conversion of Merlin to an ERM-like molecule. However, if Merlin and Moesin are also coordinately regulated in mammalian cells, an alternative possibility is that overexpression of a phosphomimetic Merlin could affect the phosphorylation state of endogenous ERM proteins, thereby increasing their level of activity.
We found that the loss of function results in a dramatic shift in Merlin localization from the apical plasma membrane to punctate cytoplasmic structures. We have previously shown that Merlin traffics from the plasma membrane with endocytic vesicles in cultured cells (), raising the possibility that in the absence of Slik, activated Merlin is more stably associated with endocytic compartments than in normal cells. If this is so, inactive Merlin may reside at the plasma membrane and, in response to activation, traffics internally, presumably in association with transmembrane proteins. If this model is correct, it suggests that Merlin may function in tumor suppression by facilitating removal from the plasma membrane of receptors that promote cell proliferation. This model fits well with our recent observation that several receptors, including Notch and the EGF receptor, accumulate to abnormal levels on the surface of cells that are mutant for and the functionally redundant related tumor suppressor ().
Several important questions remain regarding the regulation of Moesin and Merlin that we have described in this study. It remains unclear whether Slik itself can directly phosphorylate either protein or whether there are one or more kinases operating downstream of Slik. Additionally, the dual functions described here may provide novel insights into the role of the mammalian orthologues of Slik, such as PAK, in the malignant transformation of epithelial cells. Equally important will be to elucidate how Slik activity is itself controlled. Given its ability to simultaneously regulate epithelial integrity and proliferation in developing epithelial tissues, Slik may function as a central integrator of the multitude of signals that converge to regulate growth and morphology during development.
The UAS- and kinase-inactive transgenes are described in .
; FRT42D,
/ (). For overexpression studies, UAS-Myc-Mer (), UAS-Myc-Moe, UAS-Myc-Moe, and UAS-Myc-Moe () were expressed by crossing to GAL4 flies (). A Moesin RNAi transgene () was crossed to flies. All other stocks were obtained from the Bloomington Stock Center.
to allow the simultaneous expression of wild-type Slik ubiquitously and a heat shock–driven pulse (30 min at 37°C) of expression of wild-type hsGFP-tagged Merlin (hsMer). hsMer retains wild-type function (). pCasperHS Myc Mer and pCasperHS Myc Mer were made by mutating Thr 616 to alanine or aspartic acid using complementary oligonucleotides and the QuikChange method (Stratagene; constructed by R. Kulikauskas, Duke University, Durham, NC). Mutations were confirmed by sequencing. Cells were collected, fixed in 2% PFA for 20 min at room temperature, and Merlin GFP patterns were analyzed at 1, 3, and 6 h after heat shock. At least three independent replicates were scored for each experiment. For each combination and time point analyzed, a minimum of 150 transfected cells were counted. Myc-tagged constructs were detected using monoclonal anti-Myc at 1:4,000 (9B10; Cell Signaling). Slik was detected using a polyclonal antibody (). Myc and Slik were then visualized using cyanine dye CY3, FITC secondary antibodies (Jackson ImmunoResearch Laboratories), and cells mounted in ProLong (Invitrogen). Cells were analyzed using a confocal microscope (LSM510; Carl Zeiss MicroImaging, Inc.) and a plan-Apo 63× NA 1.4 lens.
To characterize the phosphorylation patterns of Merlin protein, late third instar wing imaginal discs were dissected in serum-free media (Invitrogen) and homogenized in lysis buffer (20 mM Hepes, pH 7.0, 50 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 10 mM DTT, 1.0% Triton X-100, Complete Protease Inhibitor [Roche], 50 mM NaF, 30 mM Na pyrophosphate, and 100 μm Na orthovanadate). Merlin protein complexes were subsequently immunoprecipitated using guinea pig anti-Merlin linked to Sepharose protein A beads () and separated on 8% (118:1) polyacrylamide gels (). For phosphatase treatment after immunoprecipitation, the protein A beads were precipitated, and one half was then treated with 400 U λ phosphatase (New England Biolabs, Inc.) at 30°C for 45 min followed by Western blot analysis.
Wandering third instar larvae were dissected in serum-free media and fixed in either 4% PFA or ice-cold 10% TCA () for 20 min. For Western analysis (W) and immunolocalization (I), antibodies used were as follows: guinea pig anti-Slik at 1:40,000 (W) and 1:10,000 (I; provided by S. Cohen and D. Hipfner, European Molecular Biology Laboratory, Heidelberg, Germany), rabbit anti-Moesin D44 at 1:40,000 (W) and 1:20,000 (I; provided by D. Kiehart, Duke University, Durham, NC), rabbit antiphospho-Moesin at 1:10,000 (I; obtained from D. Ready, Purdue University, West Lafayette, IN), guinea pig anti-Merlin at 1:10,000, rhodamine phalloidin at 1:1,000 (Invitrogen), mouse anti-coracle at 1:500, and mouse anti–β-tubulin at 1:5,000 (W; E7; developed by M. Klymkowsky and obtained from the Developmental Studies Hybridoma Bank, The University of Iowa, Iowa City, IO). Appropriate secondary fluorescent antibodies (FITC and cyanine dyes CY3 and CY5) were obtained from Jackson ImmunoResearch Laboratories and were used at 1:1,000. Western blots were visualized and quantified using an infrared imaging system (Odyssey; LI-COR). Immunostained tissues were mounted in ProLong (Invitrogen) and analyzed using either an LSM410 or LSM510 confocal microscope (Carl Zeiss MicroImaging, Inc.) with a plan-Apo 63× NA 1.4 lens. Figures were compiled in Photoshop 7.0.1 (Adobe).
GST, GST-Merlin, and GST-Moesin fusion proteins were grown in BL21 cells overnight at 37°C. Cultures were diluted 1:100, grown to an OD of 1, and GST constructs were induced by adding 1 mM IPTG and grown at 18°C for 3 h. Lysates were sonicated and batch incubated with glutathione–Sepharose 4B for 3 h at 4°C and washed in columns with an excess of 10 bed volumes of 1× PBS. [S]methionine-labeled probe protein (Slik) was prepared using the T7 TNT Quick Coupled Transcription/Translation System (Promega) according to the manufacturer's instructions. Proteins were incubated at 4°C for 4 h and boiled in SDS sample buffer, and proteins were separated on a 10% SDS-PAGE gel, transferred to nitrocellulose, and exposed to film.
Larvae of the genotype or were heat shocked at 36 ± 12 h after egg laying for 1 h at 37°C, 1 h at 25°C, and 1 h at 37°C. Wing imaginal discs were dissected from wandering third instar larval stages and fixed in 4% PFA. GFP was visualized directly. Moesin was detected with rabbit anti-Moesin D44 at 1:20,000 (provided by D. Kiehart), and Merlin was detected with guinea pig anti-Merlin at 1:10,000. Moesin and Merlin were then visualized using cyanine dye CY3 and FITC secondary antibodies, respectively (Jackson ImmunoResearch Laboratories), and cells were mounted in ProLong (Invitrogen). Cells were analyzed using a confocal microscope (LSM510; Carl Zeiss MicroImaging, Inc.) with a plan-Apo 63× NA 1.4 lens.
Crosses with flies of the appropriate genotypes were raised at 25°C, and wings were analyzed as described previously (). Images were collected on a camera (AxioCam HRm; Carl Zeiss MicroImaging, Inc.) mounted on a microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) using a Fluar 5× NA 0.25 lens. Area measurements of each wing were obtained from images using the free draw tool in ImageJ software (National Institutes of Health). Statistics were calculated using Excel (Microsoft), and figures were compiled in Photoshop 7.0.1 (Adobe). |
Actin filaments rapidly polymerize in some regions within cells while rapidly depolymerizing in others. In some cases, such as at the tips and bases of filopodia, polymerization and depolymerization are spatially separated (). In others, such as lamellipodia, they are close together (). Simultaneous polymerization and depolymerization of a protein polymer in the same solution is a nonequilibrium behavior that requires an energy source. The most likely source of this energy is ATP hydrolysis by actin during polymerization, but how this energy is used in cells to promote actin dynamics is poorly understood ().
propels itself through the host cell's cytoplasm by assembling an actin filament–based comet tail whose characteristic morphology depends on spatially separating actin assembly reactions from disassembly reactions. Comet tail assembly is restricted to the interface between the bacterial surface and the comet tail, where local activation of Arp2/3 rapidly nucleates a dendritic network of actin filaments (; ; ; ). The fast rate of propulsion indicates that mammalian cytoplasm contains a high concentration of available actin monomer that drives fast actin polymerization. Everywhere else in the tail, actin depolymerizes rapidly. The rate of subunit loss is proportional to the local concentration of actin filaments, leading to the exponential decay of filament density with a time constant of ∼25 s, suggesting that disassembly is a first-order reaction that proceeds through a single rate-limiting step (). However, the pool of actin monomer presents a challenge to the depolymerization reaction. How can filaments in the comet tail depolymerize rapidly in an environment that favors fast assembly? Providing a mechanistic answer to this question requires identification of the cellular factors that perform the reaction.
Members of the cofilin family of actin-binding proteins have been identified as key factors accelerating actin disassembly in all eukaryotic cells. Cofilin is necessary for the rapid turnover of actin arrays in cells (), including the disassembly of actin comet tails (). Whether cofilin alone is sufficient for disassembling actin arrays in cells is not known. Cofilin alone seems unlikely to explain the disassembly behavior of actin comet tails. Cofilin can sever actin filaments, creating new ends that grow in the presence of G actin (), but filament growth is not detectable in comet tails. Furthermore, the experimentally triggered activation of cofilin in cells results in a burst of net actin assembly, not disassembly (). Therefore, the question of what, in addition to cofilin, is required to depolymerize the actin comet tail in the presence of high concentrations of polymerizable actin remains open. In this study, we use the actin comet tail as a model substrate to identify cellular factors capable of rapidly disassembling it.
To simplify the analysis of actin comet tail disassembly, we experimentally separated it from the assembly reaction. High speed supernatants of detergent extracts from HeLa cells readily assemble comet tails, but they have no detectable disassembly activity (unpublished data). To assay comet tail disassembly, comet tails were first assembled on adhered to coverslips in perfusion chambers by flowing in HeLa cell extract mixed with rhodamine-actin to mark the comet tails. Disassembly was then induced by replacing this solution with solutions containing actin-depolymerizing factors. The reaction was followed by time-lapse imaging, and disassembly was quantified by measuring the decrease of comet tail fluorescence over time.
Comet tails disassemble very slowly when the assembly mixture is replaced with buffer alone or buffer containing latrunculin to sequester free G actin ().
is >15 min).
of ∼30 s in 500 μg/ml of total protein), implying that thymus contains high concentrations of depolymerization factors (). By Western blotting, undiluted thymus extract contains ∼20 μM cofilin (unpublished data). Pure –expressed cofilin was also able to depolymerize actin comet tails under these conditions, but the rates of disassembly appeared to be slower compared with those detected with dilute thymus cytosol. To quantitatively compare thymus extract to pure cofilin, we normalized activity to cofilin concentration ().
of disassembly, and we plot the data as an apparent disassembly rate (apparent k) versus concentration, where apparent k = ln2/
. From this analysis, thymus extract is 5–10 times more potent than pure cofilin for disassembling actin comet tails, implying that it contains factors in addition to cofilin that accelerate actin disassembly.
To identify these factors, thymus cytosol was separated on an anion exchange column (DE52), and the depolymerizing activity of the flow through fraction was compared with that of the bound material (). Although the flow through fraction was sufficient to depolymerize comet tails, it was less potent than the starting material. The DE52-bound fraction contained no detectable depolymerizing activity; however, it potentiated the activity of the flow through. By Western blotting, all of the endogenous cofilin was present in the DE52 flow through (unpublished data). Normalizing the flow through fraction to cofilin concentration revealed that thymus DE52 flow through is only twice as potent as pure –expressed cofilin alone. To test whether pure cofilin could substitute for DE52 flow through, we titrated pure cofilin down to a concentration that is insufficient to disassemble comet tails on its own. Potent depolymerizing activity could be regenerated by combining this subthreshold amount of pure cofilin with the DE52-bound cofilin-free fraction (). These results suggest that besides cofilin, thymus cytosol contains at least one additional factor involved in actin depolymerization that can be separated from cofilin by fractionation on DE52.
The ability of the DE52-bound material to induce a subthreshold amount of cofilin to disassemble actin comet tails provided a simple visual assay to fractionate for the potentiating factors. However, substantial amounts of activity were lost upon further fractionation of the DE52-bound cofilin-potentiating fraction. Therefore, we screened several columns and conditions to test the possibility that the DE52-bound material in fact contains multiple factors that must act together to facilitate cofilin-mediated depolymerization. Separation of the DE52-bound fraction on cation exchange resulted in the loss of all activity (). However, combining limiting cofilin with the S flow resulted in the appearance of a new peak of activity bound to the S column. Therefore, the cofilin-potentiating activity in thymus cytosol consists of at least two factors. The S flow through fraction was designated fraction X, and the S-bound material was designated fraction Y.
Factor X was purified from fraction X through a series of chromatographic separations. The activity of factor X was monitored using the comet tail depolymerization assay in the presence of limiting cofilin and fraction Y. With this assay, factor X fractionated as a single biochemical activity (), resulting in the isolation of a single polypeptide () that was identified by mass spectrometry as Aip1 (55.4% amino acid and 54.9% mass coverage; Fig. S1, available at ).
A limiting amount of cofilin was used in combination with Aip1 to isolate factor Y through a series of chromatographic separations. With this assay, factor Y fractionated as a single biochemical activity (), resulting in the isolation of two closely migrating polypeptides on SDS-PAGE () that were both identified by mass spectrometry as coronin-1A (43.4% amino acid and 43.2% mass coverage; Fig. S2, available at ).
To simplify the assay during the purification procedure, depolymerization activity was assessed by determining the amount of Aip1, coronin, or cofilin required to depolymerize all of the comet tails to completion in a set amount of time. Although this assay facilitated purification, it is presumably nonlinear and, thus, is not sufficiently accurate to determine the amount of each factor required to disassemble comet tails. To obtain a better measurement of the effective concentrations of each component, we varied the concentration of one of the three factors while holding the other two constant and measured the apparent rate of actin comet tail disassembly by quantifying the total fluorescence within the comet tails as a function of time using time-lapse fluorescence microscopy.
displays the apparent rate of comet tail disassembly as a function of cofilin concentration using cofilin alone or in the presence of the other depolymerizing factors. The coronin and Aip1 concentrations were set at saturating levels for this experiment (1 and 0.2 μM, respectively). Cofilin alone was sufficient to disassemble the comet tails with an IC50 of ∼3–4 μM. The addition of coronin alone or Aip1 alone to cofilin had no substantial effect on the amount of cofilin required to disassemble the comet tails. However, the addition of both coronin and Aip1 accelerated cofilin-mediated depolymerization and lowered the IC50 of cofilin 10 times to ∼0.3 μM. To determine the IC50s of Aip1 and coronin, we used a constant limiting amount of cofilin (1 μM) and a constant saturating amount of either Aip1 (0.2 μM) or coronin (2 μM) while varying the concentration of the other potentiating factor. By this method, we estimate an IC50 for Aip1 of 30 nM () and 150 nM for coronin (). These potent effects argue for biochemical specificity of the reaction.
All of the aforementioned experiments test the ability of the three depolymerizers to rapidly disassemble actin comet tails into dilute buffer. In cells, actin disassembly must occur in the presence of high concentrations of polymerizable G actin. To examine actin comet tail disassembly under polymerizing conditions, we added a source of polymerizable G actin to the depolymerizing mixture. G actin was provided in the form of a profilin G–ATP actin complex that is currently thought to represent the physiological state of polymerizable actin in cells (). Profilin suppresses spontaneous nucleation but has little effect on the ability of actin to add to the barbed ends of existing actin filaments. Because profilin also promotes disassembly and ATP exchange on cofilin–actin–ADP complexes, the presence of profilin should prevent the accumulation of these complexes in our reaction.
Cofilin in the presence or absence of either coronin or Aip1 alone was tested to see whether any of these combinations were sufficient to disassemble actin comet tails in the presence of profilin–actin (). A high concentration of cofilin (10 μM) was used for these experiments to obtain reasonably fast baseline disassembly rates. The addition of increasing amounts of profilin–actin in the disassembly reaction blocked comet tail disassembly by cofilin alone. Disassembly by a combination of 10 μM cofilin and coronin was also highly sensitive to the competing profilin–actin complex. In contrast, the combination of Aip1 and a high concentration of cofilin permitted comet tail disassembly in the presence of the competing polymerizable profilin–actin complex. All three factors together also disassembled comet tails in the presence of profilin–actin, but the rate of disassembly was markedly enhanced (). In the presence of all three factors, as little as 1 μM cofilin was sufficient to disassemble comet tails in the presence of 8 μM profilin–actin at a rate three times faster than a combination of 10 μM cofilin and Aip1 alone. These experiments demonstrate that although cofilin alone is sufficient to disassemble comet tails into buffer, it is not sufficient to disassemble comet tails under more physiological conditions in which a source of polymerizable actin is present. Aip1 appears to be the most important factor conferring resistance to actin monomer. However, comet tail disassembly by cofilin and Aip1 alone requires high concentrations of cofilin. The addition of coronin lowers the cofilin requirement 10-fold even when the reaction is challenged with high concentrations of polymerizable actin. Therefore, coronin appears to either promote cofilin binding to comet tails or alters filament structure, thereby making the filaments more sensitive to cofilin action.
To examine these possibilities, we first tested whether any individual or combination of two out of the three disassembly factors could act in a preincubation step or whether all three factors had to be present at once to disassemble comet tails. To do this, comet tails were assembled in perfusion chambers, and the assembly mixture was replaced with a solution containing only one or two of three disassembly factors. After an incubation period, the preincubation solution was replaced with a solution containing the remaining factors, and comet tail disassembly was monitored by fluorescence imaging. Comet tail disassembly under these conditions was compared with a positive control in which comet tails were preincubated in buffer alone and were chased with all three factors at once. Actin comet tails preincubated in coronin and chased with cofilin and Aip1 disassembled as rapidly as washing into all three factors at once. However, comet tails were stable under all other conditions (). Therefore, coronin binding to the actin comet tail alters the substrate either by promoting cofilin binding or altering actin filament structure, thereby making the filament more sensitive to cofilin action.
To test whether coronin promotes cofilin binding, a point mutation was generated in cofilin by changing Ser108 to Cys to allow labeling with fluorescent dyes. This reagent was then used to measure cofilin binding to actin comet tails with or without a coronin preincubation using fluorescence imaging to quantify the intensity of bound cofilin normalized to actin fluorescence in the comet tail. AlexaFluor488 actin was used to label the comet tails, and AlexaFluor568 cofilin was used to quantify cofilin. Comet tails disassembled using this combination of fluorophores in the presence of AlexaFluor568 cofilin, coronin, and Aip1 with kinetics similar to those detected with wild-type cofilin (unpublished data).
Using 4 μM of labeled cofilin, comet tails bound little cofilin in the absence of a coronin preincubation. However, comet tails bound readily detectable amounts of cofilin after a coronin preincubation (). To quantify this effect, we compared the amount of cofilin bound to comet tails normalized to the amount of actin (). In the absence of coronin preincubation, the amount of cofilin bound to actin comet tails rose steadily with increasing amounts of cofilin and appeared to approach saturation near 10 μM cofilin. Unfortunately, it was not possible to extend the analysis to higher concentrations of cofilin because the comet tails washed out of the perfusion chamber upon fixation. Comet tails bound approximately four to five times more cofilin after a coronin preincubation and appeared to approach saturation near 8 μM cofilin. Despite the high amount of cofilin bound under these conditions, relative to in the absence of coronin, the comet tails were stable, which is consistent with our previous experiment demonstrating that coronin and cofilin together do not disassemble comet tails much faster than cofilin alone (). These results demonstrate that coronin alters the comet tail substrate by promoting cofilin binding.
We repeated the binding experiments using fluorescently labeled actophorin (Ser88 to Cys mutation) from and obtained similar results. Therefore, coronin-dependent increased cofilin binding to comet tails is not peculiar to the point mutation used for labeling, and, in fact, the coronin effect probably extends to all members of the ADF/cofilin family across phyla. The four- to fivefold coronin-dependent increase in cofilin binding is close to the 10-fold reduction in the amount of cofilin required to disassemble comet tails in the presence of coronin and Aip1 relative to cofilin alone. Therefore, coronin's effect on actin comet tail disassembly can largely be attributed to its ability to promote cofilin binding. We also tried to measure cofilin binding in the presence of coronin and Aip1, but comet tails disassembled too rapidly to make such measurements, even at acidic pH.
Coronin could promote cofilin binding either by creating new binding sites or increasing the affinity of cofilin's interaction to existing binding sites. To distinguish between these possibilities, the mutant fluorescently labeled cofilin was used to monitor its dissociation from the comet tail after a preincubation in buffer alone or 1 μM coronin. After this step, the buffer-treated tails were incubated in 6 μM of labeled cofilin, and the coronin-treated comet tails were incubated in 3 μM of labeled cofilin. The chambers were then washed with three chamber volumes of buffer while acquiring a fluorescence time-lapse sequence. Cofilin appeared to dissociate from comet tails with similar kinetics irrespective of a coronin preincubation step. shows representative frames from these experiments after preincubation in buffer alone or coronin. Note that these images are not showing comet tail disassembly; the actin comet tails are, in fact, stable. Rather, they show the loss of cofilin signal from a constant density of actin. The dissociation rates were quantified by measuring the decay of cofilin fluorescence over time (). The results confirm that cofilin dissociates from comet tails in the presence or absence of exogenous coronin with similar, perhaps identical, kinetics. Therefore, coronin appears to create new cofilin-binding sites of similar, if not identical, affinity to those that occur in the absence of additional coronin.
All of the previous assays used actin comet tails as a physiologically relevant actin gel for disassembly. Thus, the generality of the activity of the three factors was tested on pure actin filaments in solution using a sedimentation assay (). 2.5 μM actin was polymerized in vitro, mixed with combinations of the disassembly factors, and the resulting products were separated by ultracentrifugation. This assay separates long filaments (>20 subunits) from fragments and actin monomer. Under starting conditions, nearly all of the actin in a pure actin solution alone pellets, which is consistent with actin's submicromolar critical concentration. Neither 1 μM coronin alone, 0.1 μM Aip1 alone, nor the combination of coronin and Aip1 altered the distribution of actin between supernatant and pellet. Thus, with this assay, Aip1 and coronin in the absence of cofilin do not appear to promote actin depolymerization, which is consistent with our observations of comet tails. 2.5 μM cofilin alone resulted in a slight increase in the amount of actin in the supernatant, which is consistent with reports that stoichiometric cofilin scores only as a weak depolymerizer (; ; ; ). This distribution was not further altered by the addition of either coronin alone or Aip1 alone. In contrast, in the presence of all three factors together (2.5 μM cofilin + 1 μM coronin + 0.1 μM Aip1), more than half of the actin was found in the supernatant. Therefore, the three disassembly factors together are more potent at depolymerizing actin than any factor alone or the combination of any two. These data are consistent with the results from the fractionation scheme using comet tails.
Although coronin in combination with cofilin did not alter the distribution of actin, coronin increased the amount of cofilin bound to filamentous actin (F actin) in the pellet (), which is consistent with results using comet tails. To compare coronin-dependent increased binding on pure actin filaments with that seen on comet tails, the amount of cofilin bound to F actin as a function of increasing cofilin concentration in the presence or absence of coronin was measured in a sedimentation assay. A representative gel from such an experiment is shown in , and a plot from the densitometry of this gel is shown in . Although coronin promotes cofilin binding to pure actin filaments, the effect is not nearly as pronounced as it was on comet tails. With pure actin filaments, coronin increased the amount of cofilin bound to F actin by 1.5 times compared with the four- to fivefold increase detected on comet tails. The quantitative discrepancy could be the result of several factors. One possibility is that the sedimentation assay is simply not very sensitive or quantitative because sedimentation alters the equilibrium. Alternatively, other factors within the comet tail in addition to coronin might influence cofilin binding. Nevertheless, results from the sedimentation experiments qualitatively support the conclusion from comet tails that coronin promotes cofilin binding to F actin.
Cofilin is thought to be the key factor promoting fast depolymerization of actin filaments in cells, but other cellular factors clearly contribute to this reaction. Although cofilin alone is sufficient to depolymerize actin comet tails, it is unable to rapidly disassemble comet tails under more physiological conditions in which depolymerization is challenged by a physiological concentration of G actin presented as a profilin–actin–ATP complex. Similar results were obtained using actin–ATP alone (unpublished data). We identified Aip1 and coronin as two factors in thymus extracts that facilitate cofilin-mediated actin disassembly. Together, Aip1 and coronin accelerate the cofilin-mediated disassembly of comet tails, reduce the amount of cofilin required for this reaction, and permit efficient disassembly of the comet tail in the presence of a source of polymerizable G actin. The combination of these factors provides a means whereby actin comet tail filaments can efficiently depolymerize while suppressing assembly.
Aip1 has already been implicated as a cofactor in cofilin-mediated depolymerization by both biochemical and genetic evidence (; ; ), although the relative importance of Aip1 appears to vary between organisms and cell types. In S2 cells () and in embryos (), the inhibition of Aip1 function leads to severe defects in actin organization and accumulation of excess actin polymer. Deletion of Aip1 in yeast () and () also perturbs actin dynamics and function, but the phenotypes are comparatively less severe. Our biochemical data implicate Aip1 in the disassembly of the highly branched comet tail that is assembled primarily by frequent actin nucleation by the Arp2/3 complex (). Cells can also assemble more highly bundled actin arrays through formin-mediated polymerization (), but the mechanisms that disassemble these arrays are not known. Variations in the severity of Aip1 (as well as coronin and cofilin) deletion between organisms and cell types might reflect differences in the relative importance of Arp2/3 versus formin-mediated actin arrays in different cells and alternative disassembly mechanisms.
In contrast to Aip1, coronin has not previously been implicated in actin depolymerization. However, genetic data and its intracellular localization are consistent with a role in actin turnover. Coronin shows synthetic genetic interactions with nonlethal point mutations in cofilin as well as with mutations in actin that slow ATP hydrolysis and, therefore, stabilize actin filaments (). Coronin preferentially localizes to dynamic actin structures such as yeast cortical actin patches () as well as lamellipodia and phagocytic cups in higher eukaryotes (, ; ). It is also abundant in comet tails (). Our biochemical data suggest that one of the functions of coronin within these structures is to facilitate cofilin recruitment to these structures and, thus, fast turnover.
The combination of Aip1 and coronin lowers the concentration of cofilin required to disassemble actin comet tails at rates comparable with those seen in cells. Cellular cofilin concentrations have only been measured in a few cases, and the reported concentrations vary widely between organisms and cell types. For example, egg extracts contain 3 μM cofilin compared with 30 μM in platelets (). However, total cofilin measured in cells cannot be compared directly with the concentrations we used in , , and because cofilin in cells is probably present partly in the form of cofilin–actin–ADP complexes, which are presumably inactive in depolymerization, as well as in the form of phosphocofilin, which is also relatively inactive (; ). In the future, it will be important to measure the concentrations of free actin, cofilin, coronin, and Aip1 that are available for depolymerization. Our dose-response analysis suggests that low micromolar concentrations of recombinant cofilin are sufficient for fast disassembly in the presence of competing G actin as long as coronin and Aip1 are present in the 10–100 nM range. In yeast, total Aip1 () and coronin () are approximately equimolar with cofilin. Coronin and Aip1 concentrations in other cells have not been measured. From our purification tables, we roughly estimate that our starting thymus cytosol contains at least 0.5 μM Aip1 and 1.4 μM coronin.
The response of disassembly to polymerizable G actin suggests the factor that prevents polymerization of barbed ends in tails is primarily Aip1 acting in cooperation with coronin and cofilin. Aip1 is known to suppress polymerization from barbed ends of cofilin-decorated filaments (; ; ). Others have cast capping protein in this role (), and the role of Aip1 in cofilin-mediated disassembly might simply be to cap barbed ends, thereby suppressing further growth and preventing the reannealing of severed filaments. Although Aip1 suppresses filament elongation, it is not as effective of a capping agent as cytochalasin D or the gelsolin–actin complex. However, it is more effective than either of these agents in facilitating cofilin-mediated actin disassembly, suggesting that Aip1 actively contributes to the disassembly reaction (). Our purification strategy did not reveal a role for capping protein, and, in preliminary tests, we have not observed the potentiation of cofilin-mediated depolymerization by recombinant capping protein. Therefore, like , we favor a more active role for Aip1 in actin disassembly. Our data suggest that Aip1 in conjunction with coronin and cofilin contributes to a rate-limiting transition that converts stable, growing filaments into highly unstable filaments that rapidly disassemble ().
Coronin promotes cofilin binding to actin comet tails as well as pure F actin filaments, and the increased binding might account for most, if not all, of coronin function in actin disassembly. Actin subunits within filaments can deviate from their canonical helical organization () to one in which the subunits are rotated to a lesser extent, resulting in a change in the twist of the filament. Cofilin binds to these imperfections along the filament and, thus, destabilizes it (). Coronin could promote cofilin binding either by increasing the affinity of cofilin for these untwisted sites or by generating additional untwisted sites. We favor the latter because the rate of cofilin dissociation from comet tails was similar in both the presence and absence of exogenous coronin. Coronin had a greater effect on cofilin binding to comet tails than to pure actin filaments, suggesting that filaments within comet tails might be less disordered than pure filaments and therefore resist cofilin binding. If this is true, coronin binding to the comet tail somehow makes it easier for filaments within the tail to alter their twist and bind more cofilin.
Cells contain structurally distinct actin arrays that disassemble at different rates and possibly disassemble through different mechanisms.
near 30 s, whereas the parallel bundles found in filopodia disassemble with a
as long as 25 min (). The biochemical mechanisms underlying these differences are not known. In the future, it will be important to determine the relative contributions of cofilin, coronin, Aip1, and other disassembly factors such as gelsolin to the disassembly of different actin arrays. Understanding the biochemical mechanisms controlling actin disassembly is essential for understanding the spatial organization of the actin cytoskeleton.
Rabbit skeletal muscle actin, bovine Arp2/3 (), and bovine profilin () were purified as described previously. Recombinant human cofilin (a gift from A. Weeds [Medical Research Council, Cambridge, United Kingdom] and A. McGough [Purdue University, West Lafayette, IN]) was purified as described previously (). Purified, recombinant, rhodamine-labeled actophorin-S88C from was provided by D. Mullins (University of California, San Francisco, San Francisco, CA). Actin was labeled with tetramethylrhodamine or AlexaFluor488 -hydroxysuccinimidyl esters as described previously (). A point mutant of human cofilin (Ser108 to Cys) was generated by standard techniques expressed in bacteria and purified in the same way as wild-type cofilin. The mutant cofilin was labeled on the introduced cysteine with AlexaFluor568 maleimide.
Frozen cell pellets from 10 liters of culture were obtained from the National Cell Culture Center. Frozen pellets were thawed at 37°C and lysed with a 2:1 buffer (cell pellet volume of 20 mM Hepes, pH 7.4, 100 mM KCl, 2 mM MgCl, 2 mM EGTA, 2% Triton X-100, 14.3 mM β-mercaptoethanol, and 0.2 mM PMSF). The lysate was centrifuged at 100,000 for 1 h at 4°C, and the supernatant was dialyzed overnight against the lysis buffer lacking Triton X-100. The dialyzed sample was centrifuged at 100,000 for 1 h at 4°C, and the supernatant was frozen in liquid nitrogen and stored at −80°C.
Chemically inactivated were prepared and introduced into perfusion chambers as described previously (). Comet tails were assembled by mixing 20 μl of assay buffer (100 mM Hepes, pH 7.8, 50 mM KCl, 2 mM MgCl, 2 mM ATP, and 14.3 mM β-mercaptoethanol) with 5 μl HeLa extract, 0.1 μM Arp2/3, 4 μM actin, and 1 μM tetramethylrhodamine-actin. Comet tail assembly was allowed to proceed for 8–10 min at room temperature before proceeding with the disassembly reaction.
To examine actin comet tail disassembly in the presence of the profilin–actin complex, equimolar amounts of profilin and G actin were combined to make a stock solution of 50 μM profilin and 50 μM actin in G buffer (2 mM Tris, pH 8, 0.2 mM CaCl, 0.2 mM ATP, and 1 mM DTT) and incubated on ice for at least 1 h before mixing it with the appropriate depolymerizers just before assaying the mixture. Comet tail disassembly was quantified by first thresholding the images from a time-lapse stack two SDs above background fluorescence for each experiment. Background fluorescence was determined from the mean fluorescence intensity of regions within the frame that did not contain any actin comet tails or clouds. The thresholded image stack was then used to generate a binary image for each image from the time-lapse sequence. The binary image was multiplied by the corresponding raw image from the original time lapse using the Logical AND command in MetaMorph to generate a stack of images corrected for background fluorescence.
. The data were converted to an apparent disassembly rate by plotting ln2/
. Comet tail disassembly in vitro under our conditions does not necessarily show exponential kinetics, as is implied through this treatment of the data. However, it does serve as a convenient means to display the data. Therefore, we refer to these values as an apparent off rate.
200 g of frozen bovine calf thymus was thawed overnight at 4°C. The tissue was cut into small pieces and homogenized in buffer A (20 mM Tris, pH 7.4, 25 mM NaCl, 1 mM EGTA, 2 mM MgCl, and 14 mM β-mercaptoethanol) in a Waring-type blender. The homogenate was centrifuged for 30 min at 8,000 , and insoluble material was discarded. Polyethyleneimine was added to the supernatant to a final concentration of 0.05% and stirred for 30 min at 4°C. The slurry was centrifuged at 8,000 for 30 min, and the pellet was discarded. The supernatant was centrifuged at 150,000 at a factor of 133 for 90 min. The supernatant was mixed with 200 ml DE52 (Whatman), stirred for 60 min, and allowed to settle. The liquid was decanted off the beads, and the beads were resuspended in 100 ml of buffer A. The slurry was poured into a column, and the flow through was combined with the decanted solution to generate the DE52 flow through fraction. The DE52 column was washed with an additional 400 ml of buffer A and eluted with an 800-ml gradient to 400 mM NaCl in buffer A to generate the DE52-bound fraction. DE52-bound factors that facilitated cofilin-mediated depolymerization eluted between 80–140 mM NaCl.
The DE52-bound fraction was dialyzed into 20 mM Pipes, pH 6.8, 25 mM NaCl, and 14 mM β-mercaptoethanol (buffer B), centrifuged at 100,000 with a factor of 220 for 1 h, and applied to a 70-ml S Sepharose column equilibrated in buffer B. The column was washed with two column volumes of buffer B and eluted with a 1-liter gradient to 400 mM NaCl in buffer B. The S flow through was designated fraction X and contains Aip1. Factor Y (coronin) elutes from the column near 270 mM NaCl.
Factor X (Aip1) was purified from the S flow through fraction. Solid (NH)SO was added slowly to the S flow through fraction to a final concentration of 2 M. The sample was centrifuged at 15,000 for 30 min, and the pellet was discarded. The supernatant was applied to a 45-ml phenyl Sepharose HP column equilibrated in buffer B containing 2 M (NH)SO. The column was washed with 1 vol of the same buffer and eluted with a 600-ml gradient to buffer B. Factor X eluted from the column near 500 mM (NH)SO. Active fractions were dialyzed into 20 mM Tris, pH 8, 25 mM NaCl, and 14 mM β-mercaptoethanol (buffer C) and were applied to a 6-ml MonoQ column. The column was eluted with a 150-ml gradient to 500 mM NaCl in buffer C. Factor X eluted at 150 mM NaCl. Active fractions were dialyzed against buffer B and applied to a 5-ml hydroxyapatite column equilibrated in buffer B. The column was eluted with a 100-ml gradient to 500 mM sodium phosphate, pH 6.8. Factor X eluted at 175 mM sodium phosphate. Active fractions were concentrated using a centrifugal ultrafiltration device with a nominal cutoff of 30 kD. The retentate was loaded on a 1 × 30-cm Superose 12 gel filtration column. Activity eluted near 65 kD. At this point, factor X was pure and was identified by mass spectrometry.
Gel bands were excised, diced, and subjected to in-gel digestion with 12.5 ng/μl of sequencing grade trypsin (Promega) in 50 mM ammonium bicarbonate overnight at 37°C. Peptides were extracted with 50% acetonitrile (MeCN) and 2.5% formic acid (FA) and were then dried. Peptides were resuspended in 2.5% MeCN and 2.5% FA and were loaded using an autosampler onto a microcapillary column packed with 5 μm of reverse-phase MagicC18 material (200 Å; Michrom Bioresources, Inc.). Elution was performed with a 5–35% MeCN (0.1% FA) gradient over 60 min after a 15-min isocratic loading at 2.5% MeCN and 0.5% FA. Mass spectra were acquired in a linear ion trap mass spectrometer (Finnigan LTQ; Termo Electron) over the entire run using 10 mass spectrometry/mass spectrometry scans after each survey scan. Raw data were searched for fully tryptic peptides against the National Center for Biotechnology Information nonredundant indexed database using Sequest software (Thermo Finnigan) and a mass tolerance of 2 D. After the identification of AIP1 and coronin-1A in the respective bands, a BLASTP search was performed to identify related sequences from . Two such sequences were identified for AIP1 (gi 76620344 and gi 86821067), and four were identified for coronin-1A (gi 27806251, gi 73587271, gi 76639125, and gi 76676005). Mass spectrometry data were then researched simultaneously against either the set of two sequences for AIP1 or the set of four sequences for coronin-1A with trypsin off and a mass tolerance of 2 D.
Cofilin binding to actin comet tails was measured using the mutant cofilin (S108C) labeled with AlexaFluor568 on the introduced cysteine. Comet tails were assembled in perfusion chambers in the presence of AlexaFluor488-labeled actin to mark the comet tails. The chambers were then washed twice in assay buffer and incubated either in assay buffer or 1 μM coronin in assay buffer for 5 min. The chambers were washed twice with assay buffer, incubated with varying concentrations of labeled cofilin for 2 min, and fixed with 1% glutaraldehyde in the same buffer. Fluorescent images were collected as described above for comet tail disassembly except a 40× NA 0.95 objective (Nikon) was used instead of a 20× objective. Fluorescence intensity of the actin and cofilin signals was corrected for background and integrated as described for comet tail disassembly, and the corrected cofilin signal was normalized to the corrected actin signal.
5 μM actin was polymerized in assay buffer for 1 h at room temperature and mixed with combinations of coronin, cofilin, and Aip1 at the concentrations indicated in additional assay buffer to give a final concentration of actin of 2.5 μM. After 15 min at room temperature, the samples were centrifuged at 486,000 for 30 min at 20°C. Equal portions of the supernatant and pellet were separated by SDS-PAGE and stained with Coomassie. Densitometry of gel bands was performed using ImageJ software (National Institutes of Health).
Figures show amino acid sequences of Aip1 (Fig. S1), coronin (Fig. S2), and tryptic peptides identified by mass spectrometry with Xcorr values of >1.8 for doubly charged ions and 2 for triply charged ions. Online supplemental material is available at . |
The protein spectrin is the defining element of a nearly ubiquitous submembrane protein network in animal cells. The spectrin supergene family includes a diverse set of proteins that share two main structural features: the spectrin repeat and a calponin-homology actin-binding domain (for review see ). The “immediate family” of the founding member spectrin includes a group of closely related gene products that assemble as tetramers of α and β subunits, that associate with specific subdomains of the plasma membrane in many cells, and that share the ability to form cross-linked arrays with actin. Most, but not all, of these family members are linked to plasma membrane proteins via the adaptor ankyrin (for review see ).
The functions of several spectrins and ankyrins have been tested in gene knockout experiments. One consistently observed effect is that loss of spectrin or ankyrin leads to a failure of interacting membrane proteins to accumulate at the appropriate site. In the most extreme case, RNAi knockdown of ankyrin-G in cultured bronchial epithelial cells led to the loss of an entire domain of the plasma membrane (). However, in other knockout studies, loss of spectrin or ankyrin led to more subtle effects in which specific interacting membrane proteins were lost from the domain normally occupied by the spectrin cytoskeleton. For example, loss of β spectrin in led to loss of Na,K ATPase from the basolateral domain of epithelial cells (); loss of βIV spectrin in mouse brain led to loss of voltage- dependent sodium channels from axon initial segments and the node of Ranvier (); knockouts of ankyrin-G and -B in mouse brain led to reduced levels of voltage-dependent sodium channels and L1 family cell adhesion molecules (; ; ).
Recent studies also indicate an important role for the spectrin cytoskeleton in muscle (for review see ). Defects in ankyrin-B are associated with human cardiac arryhthmia and sudden cardiac death (). Loss of ankyrin-B leads to a concomitant loss of Na,K ATPase, Na,Ca exchangers and inositol triphosphate receptors and perhaps other proteins from their normal cellular sites (). A similar cardiac defect in humans results from a mutation of the voltage-dependent sodium channel in its ankyrin-G–binding site ().
Although the consequences of loss of function are becoming clear, the cues that trigger assembly of the spectrin cytoskeleton within specialized membrane domains remain unknown. Evidence placing ankyrin upstream of spectrin in the assembly pathway has come from studies of cultured cardiac myocytes from ankyrin-B knockout mice. An engineered ankyrin-B transgene lacking spectrin-binding activity appeared to function identically to the wild-type transgene, except that it failed to recruit β2 spectrin in its normal striated pattern (). From these results, the authors concluded that ankyrin was upstream of spectrin in the spectrin cytoskeleton assembly pathway and that its function was independent of spectrin. However, a study of βIV spectrin knockout mice revealed that βIV spectrin and ankyrin-G are dependent on one another for assembly at the neuronal plasma membrane ().
There are also direct interactions between β spectrin and the plasma membrane. Two sites, one near the N terminus of β spectrin and one near the C terminus, were identified in binding studies with NaOH-stripped membranes from rat brain (; ; ). The discovery of these sites led to the hypotheses that spectrin targeting could be achieved by ankyrin-independent interactions with the plasma membrane and adaptor proteins such as ankyrin could have a secondary role in recruiting other interacting proteins to sites of spectrin assembly to form a unique membrane domain (). Although the aforementioned studies focused on the interactions between the membrane association domains of β spectrin and putative protein receptors, the presence of a pleckstrin homology (PH) domain in membrane association domain 2 raised the possibility of an interaction between spectrin and phosphoinositides. The PH domain of mammalian β spectrins binds to inositol lipids and inositol triphosphate (; ), and a reporter consisting of the PH domain from mouse βI spectrin fused to GFP was efficiently targeted to the plasma membrane in transfected COS cells (). However, the relative importance of these ankyrin-independent binding sites to the initial assembly of spectrin has never been tested in vivo.
Here, we used a genetic approach to examine the relative contributions of specific sites in the β spectrin molecule to its targeting and function in polarized epithelial cells. All of the major functional sites in the mammalian spectrin molecule are conserved in , making it a valid model system in which to address basic questions of spectrin's biological function and mechanism of action. One important advantage of the system is that there are only two spectrin isoforms in and their functions do not appear to overlap (; ). There are only two ankyrin genes in , and one of them (Dank2) appears to be expressed primarily in the nervous system (). Genetic studies in mammalian systems are complicated by the presence of five different β spectrin genes, two different α spectrin genes, three different ankyrin genes, and further variation by alternative splicing (). Previous genetic studies in have demonstrated that mutations in the β and α spectrin genes are lethal in the late embryo and early larva, respectively (; ). The lethal phenotypes were fully rescued by recombinant transgenes in which the coding sequences for α and β spectrin were expressed under control of the ubiquitin promoter.
We tested modified transgene products in which specific functional sites were altered or deleted for their ability to rescue the lethal phenotype of β spectrin mutations and for their ability to be targeted to the basolateral domain of the plasma membrane in copper cells. These cells share several properties with gastric parietal cells in mammalian stomach, including the secretion of stomach acid (). Copper cells from spectrin mutants are amenable to immunohistochemistry, and we previously demonstrated that the basolateral distribution of the Na,K ATPase in these cells was dependent on β spectrin function (). We also characterized the effects of transgene modification on the behavior of spectrin in the salivary gland epithelium. The results support a model in which spectrin targeting occurs independently of ankyrin.
We previously described a myc-tagged β spectrin transgene that rescued the lethal phenotype of mutations in the endogenous β spectrin gene (). Here, we produced modified transgenes in which selected functional domains were either altered or deleted. In one construct (
; and ), the putative ankyrin-binding segment 16 (repeat 15; ) was replaced with segment 13 from α spectrin. The rationale for this in-phase repeat replacement strategy was to eliminate ankyrin-binding activity without perturbing the overall structure of the modified protein. The transgene product (, lane 2) was identical in size to the β spectrin control transgene product (lane 1) in Western blots of total proteins from transgenic flies stained with anti-myc antibody. Another transgene was modified by in-phase replacement of repeats 4–11 of β spectrin with repeats 2–9 of α spectrin (, ). This transgene encoded a myc-tagged product that was also similar in size and abundance to the recombinant β spectrin control (, lane 3). A third transgene was modified by replacing the first codon of the PH domain with a stop codon ( , ). Western blots stained with rabbit anti–β spectrin antibody detected the authentic wild-type β spectrin (266 kD) as well as the expected truncated product (251 kD; , lane 4). Staining of an identical blot strip with the anti-myc antibody detected only the smaller, truncated product (lane 6). A fourth transgene was modified by replacing the tryptophan at codon 2033 with an arginine (, ). The expected full-length product was also detected in blots of these flies stained with the anti-myc antibody (unpublished data).
The ability of these transgenes to rescue mutations in the endogenous X-linked β spectrin gene was tested using two different rescue strategies (depending on the site of transgene insertion; see Materials and methods for details). The expected progeny class ratios were different, but otherwise both strategies simply tested the ability of the rescue transgene to overcome the embryonic lethality of the mutant β spectrin gene. Representative cross results are shown in . The control transgene (on chromosome III) was scored for its ability to replace the function of , a translocation of a wild-type β spectrin allele to chromosome III (). The observed 2:1 phenotypic ratio of sibling/transgene rescue adult males indicates a 100% rescue efficiency by the transgene. In contrast, the truncated transgene lacking the PH domain generally failed to rescue, although rare rescue flies were occasionally observed.
Interestingly, the ΔPH rescue flies were unusually small in size (). Rescue males expressing the control transgene (non–-eye; ) were the same size as wild-type sibling males that carried the -eye–marked translocation (). However, males rescued by the ΔPH transgene were strikingly smaller in size compared with their -eyed siblings (), and they were sterile. Western blots demonstrated that the ΔPH rescue males expressed the myc-tagged truncated product (, lane 7). Only the truncated ΔPH product was detected when blots were probed with the anti–β spectrin antibody (lane 5), confirming the absence of wild-type β spectrin.
and the
transgenes was tested after recombination with the β spectrin mutant chromosome ().
transgene altogether lacked rescue activity (), suggesting that replacement of repeats 4–11 of β spectrin removed an important functional site and/or compromised the structure of the molecule.
transgene exhibited rescue activity, although it was significantly less efficient than the wild-type transgene (17%; ). Many of the rescued progeny were fertile; however, some exhibited minor wing defects and many of them died during eclosion (unpublished data). This result suggested that either ankyrin-binding activity was not essential for spectrin function or ankyrin-binding activity occurred elsewhere in the β spectrin molecule.
Each of the β spectrin transgenes was tested for its effect on the targeting of spectrin, both in the presence and in the absence of wild-type β spectrin. Midguts from late embryos or larvae were dissected and stained with anti-myc antibody and rabbit anti-Scribble as a positive staining control. The myc-tagged control transgene product was previously shown () to have the same distribution as endogenous β spectrin (Fig. S1, available at ).
transgene modification was expected to perturb the association of spectrin with ankyrin and therefore to affect ankyrin-dependent targeting of spectrin to the plasma membrane. However, its distribution in copper cells was indistinguishable from the wild-type protein throughout the septate junction, lateral, and basal domains (). Scribble staining marked the septate junction region (). The βspec distribution was the same regardless of whether it was expressed in a wild-type background () or in the absence of wild-type β spectrin (not depicted). Thus, segment 16 appeared to be dispensable for the targeting of spectrin to the plasma membrane.
In contrast, the βspec product had an abnormal distribution regardless of whether it was expressed in a wild-type background or in a mutant background. Interestingly, βspec was enriched in the septate junction region when wild-type β spectrin was present (). But in the mutant background, there was no apparent pattern of plasma membrane staining in copper cells (). Instead, staining was observed in a speckled pattern in the cytoplasm. The Scribble staining pattern was altered slightly in the β spectrin mutant gut, to the extent that spectrin mutations alter cell shape and tissue organization (; ). However, it was still possible to recognize the comma shapes that are characteristic of the septate junctions in copper cells ().
transgene failed to rescue the lethality of β spectrin mutations, but it delayed lethality until shortly after larval hatching. Staining of the βspec product was reduced somewhat relative to controls but was still clearly detectable throughout the septate junction, the lateral domain, and the basal domain (). Thus, the region encompassing segments 4–11 of β spectrin was important for function but not for polarized assembly.
We also examined the behavior of modified β spectrin transgenes in the salivary gland epithelium. Both the endogenous wild-type β spectrin (not depicted) and the myc-tagged wild-type transgene product () were present throughout the lateral membrane domain, including the septate junction (marked by Scribble staining; ). All of the transgene products that we analyzed, including βspec () and βspec (not depicted), exhibited the same lateral distribution in both wild-type larvae and mutants lacking endogenous β spectrin. Thus, once again, targeting of spectrin to the plasma membrane did not require the ankyrin-binding domain. However, unlike the copper cell, targeting in the salivary gland did not require the PH domain either, suggesting that there are additional cell type–specific targeting mechanisms.
The above results suggested that the targeting of spectrin in copper cells and salivary gland was independent of ankyrin. Further insight into this unexpected observation came from studies of a DAnk1-EGFP fusion protein. The transgene was engineered by splicing the coding sequence of EGFP to the last codon of the DAnk1 coding region. A Western blot of larvae carrying the UAS transgene and heat shock (HS) Gal4 was probed with a rabbit anti-DAnk1 antibody (, right). In addition to the 170-kD DAnk1 band found in control larvae (, left), the UAS-DAnk1-GFP flies expressed an ∼200-kD product that is the expected size of the DAnk1-EGFP fusion. The HS Gal4 driver was chosen for this experiment because, fortuitously, it was constitutively active in both copper cells and salivary gland.
The distribution of DAnk1-EGFP was examined in the copper cells of wild-type and β spectrin mutant embryos. In wild type, the pattern was identical to the antibody-stained distribution of β spectrin, throughout the basolateral domain, including the septate junction (). Thus, there appears to be a 1:1 correspondence between the distribution of ankyrin and β spectrin in these cells. The distribution of DAnk1-EGFP was dramatically altered in copper cells from mutants (), becoming diffusely distributed throughout the cytoplasm with no detectable fluorescence at the plasma membrane. A similar effect was observed in salivary gland cells from mutant embryos (). DAnk1-EGFP codistributed with β spectrin throughout the lateral membrane domain of wild-type embryos (), but its distribution was largely shifted to the cytoplasm of mutant cells. However, unlike copper cells, it was possible to detect some DAnk1-EGFP in the lateral domain of salivary gland cells from mutants, although it was greatly reduced compared with wild type. Thus, DAnk1 targeting to the plasma membrane was almost entirely dependent on the presence of β spectrin.
transgene and its normal basolateral targeting in copper cells suggested that either the mutation did not eliminate ankyrin binding or that ankyrin binding was dispensable for β spectrin function.
rescue larvae (in the absence of endogenous β spectrin). The result () showed that DAnk1-EGFP was displaced to the cytoplasm of copper cells, as observed in the -null mutant.
mutation also had a striking effect on the behavior of DAnk1-EGFP in salivary gland. In fact, most of the lateral membrane distribution of DAnk1-EGFP observed in control larvae () was displaced to the cytoplasm in heterozygotes expressing both the transgene and wild-type β spectrin (). The membrane-associated population of DAnk1-EGFP was reduced further in rescue larvae expressing only βspec, comparable to the level of membrane binding observed in the absence of β spectrin (). Quantitative analysis of >100 cells from each genotype established that mean pixel intensity at cell contacts was substantially reduced in both heterozygotes and in rescued mutants. As the mean cytoplasmic level rose (), the amount of ankyrin at cell contacts was reduced so that it was only slightly greater than the cytoplasmic pool.
mutant transgene had a dramatic effect on the behavior of ankyrin, presumably by disrupting ankyrin binding to spectrin.
We previously demonstrated a striking effect of β spectrin mutations on the behavior of the Na,K ATPase in copper cells (). In wild-type larvae, the sodium pump was concentrated within the basal and lateral domains of copper cells (; two different planes of focus).
transgene ().
transgene (). Many of these larvae survive beyond hatching. Weak plasma membrane staining could be detected in a few cells from 33% of the mutant guts, indicating that the phenotype was not completely penetrant (). But the association of the Na,K ATPase with the plasma membrane was dramatically altered in the majority of cells. Thus, behavior of the Na,K ATPase was more closely associated with the assembled state of β spectrin than with ankyrin.
The results presented here provide several novel insights into the assembly and function of spectrin. There are two general models to explain the assembly of the spectrin cytoskeleton in polarized cells. Both models incorporate ankyrin as an adaptor that couples integral membrane proteins to the spectrin cytoskeleton. In the first case (, left), assembly begins with a protein receptor that recruits ankyrin to a specific region of the plasma membrane. In this model, ankyrin serves two distinct roles: as an adaptor that couples spectrin to a cue for assembly and as an adaptor that links interacting proteins such as the Na,K ATPase and voltage-dependent sodium channels to the preassembled spectrin cytoskeleton. In the “spectrin first” model (, right), ankyrin functions as an adaptor that couples interacting membrane proteins to a preassembled spectrin cytoskeleton (). In this model, the site of assembly is determined directly by spectrin and the role of ankyrin is to couple the diverse membrane proteins that interact with ankyrin to that site.
The results of the present study provide the first direct evidence supporting the spectrin-first model. Ankyrin assembly at the basolateral membrane domain of copper cells was dependent on spectrin. Spectrin in turn was dependent on the PH domain of the β subunit in copper cells and on an as-yet-unidentified signal in salivary gland cells. There are examples of ankyrin-independent assembly of spectrin in other systems: During erythrocyte differentiation, ankyrin assembly occurs after the stable assembly of spectrin (). A related observation is that spectrin assembly appears remarkably normal in erythrocytes that lack band 3, the major membrane receptor for ankyrin in the erythrocyte (; ). Thus, even in the best-characterized membrane model, it has been difficult to ascertain the sequence of events that leads to spectrin assembly. Targeting of the αβ isoforms of spectrin is thought to occur by an ankyrin-independent mechanism (). These spectrins have unusually large and divergent β subunits and are targeted to the apical membrane domain of polarized epithelia in . Together with the current results, it appears that targeting to the plasma membrane is a shared property of spectrins, whether or not they interact with ankyrin.
PH domains have been detected in hundreds of different proteins and in many cases they have physiological roles in binding to phosphoinositides (for review see ). Structures have been determined for spectrin PH domains from mammals () and (), and binding to phosphoinositides has been demonstrated (; ). The PH domain of spectrins does not have the expected lipid specificity of a protein that mediates phosphatidylinositol (PtdIns)-3-kinase signaling (). Although the structure of the spectrin PH domain appears to be compatible with binding to PtdInsP a more likely binding partner in vivo is PtdInsP (; ). The six residues that contact PtdInsP are conserved in the PH domain and out of 37 amino acid identities among PH domains from mammalian βI, βII, βIII, and βIV spectrins, 31 are conserved in . Overall, there was 44% identity between the fly PH domain and mammalian PH domains. For comparison, there was a mean of 62% identity between the PH domains of the four mammalian β spectrin isoforms. Interestingly, in comparisons of full-length sequences, there was greater identity between β spectrin and human βII spectrin (57%) than between human βII and βIV (53%) or between βI and βIV (49%). Therefore, it appears likely that the functions of β spectrin, including the lipid-binding function of the PH domain, are conserved between and mammals.
The PH domain of spectrin may also mediate membrane targeting through interactions with protein receptors. For example, the membrane-binding activity originally described for brain β spectrin was protease sensitive (). Interactions between PH domains and protein ligands such as protein kinase C and G protein βγ subunits have been reported (). However, the interaction between the PH domain of a mammalian β spectrin and βγ subunits was tested and found to be comparatively weak ().
product in a mutant versus a wild-type background. Further experiments will be necessary to determine whether mixed tetramers form and whether other sites in the molecule affect their targeting.
Although the PH domain was required for the targeting of spectrin in copper cells, neither the PH domain nor ankyrin binding were required for targeting in the salivary gland. One obvious candidate to explain the recruitment observed in salivary gland is the ankyrin-independent membrane-binding site identified near the N terminus of mammalian β spectrin (; ). It is also possible that multiple membrane-binding sites contribute to targeting in some cells, even though the PH domain alone appears to be critical in copper cells. To help resolve these questions, it will be important in future studies to identify sites that are sufficient for membrane targeting in different cell types and to produce mutant transgenes in which multiple candidate targeting activities have been knocked out simultaneously.
We expected that loss of ankyrin binding would severely compromise the function of β spectrin.
transgene, and they often had wing phenotypes and appeared less healthy than their wild-type siblings (unpublished data), a surprising number of these flies survived as fertile adults. In contrast, the four loss-of-function mutations that we previously characterized all exhibited embryonic lethality (). The rescue result reinforces the conclusion that spectrin targeting is independent of its interaction with ankyrin and further suggests that some important aspects of spectrin function are independent of its association with ankyrin.
rescue that will require further experiments to resolve. There may be redundant cellular mechanisms that can partially compensate for loss of the adaptor function of ankyrin. The modification of the ankyrin-binding domain of βspec may have selectively blocked its association with DAnk1 but left Dank2 binding intact. Once appropriate tools become available it will be important to test whether DAnk2 has an essential function and whether that function depends on the ankyrin-binding site defined here. There may be residual ankyrin-binding activity in βspec that was below the threshold of detection in our experiments. One reason to consider this possibility emerged from sequence comparisons between fly and human β spectrins (). Current structural models indicate that part of the putative ankyrin-binding site may include part of segment 15 (repeat 14), where there is striking sequence conservation. Future experiments will address whether this conservation represents part of the ankyrin-binding site or whether it is an as-yet-unidentified functional site in the β spectrin molecule. In any case, it's apparent that most of the ankyrin 1–binding activity of spectrin was removed in βspec. Finally, it is also formally possible, although unlikely given the degree of sequence conservation, that spectrins and ankyrins in vertebrates and invertebrates have fundamentally different roles in plasma membrane organization and function.
Another surprising result in the present study was the finding that the behavior of the Na,K ATPase was more closely linked to the behavior of spectrin than to ankyrin. Based on biochemical evidence showing that purified mammalian ankyrin and Na,K ATPase directly interact with one another in vitro (), we previously assumed that any effects of spectrin mutations on the behavior of the Na,K ATPase in vivo were likely to be mediated through ankyrin (). Current evidence raises the possibility that the Na,K ATPase may be linked to spectrin either directly or perhaps by some other indirect mechanism. It is possible that vertebrates and invertebrates evolved independent mechanisms to link the Na,K ATPase to the spectrin cytoskeleton. It was recently suggested that mammalian KCNQ potassium channels and voltage-dependent sodium channels acquired their functional interaction with ankyrin through a process of convergent molecular evolution, after the split between vertebrates and invertebrates (). The conserved ankyrin-binding sequence found in these mammalian proteins is not present in their homologues. That does not appear to be the case with the Na,K ATPase, as the amino acid sequence that mediates interaction with ankyrins in vertebrates () is remarkably conserved in the Na,K ATPase (). Future studies should address the possibility that, even though there is a direct interaction between ankyrin and the Na,K ATPase in vitro, there may be an important functional interaction with mammalian spectrin in vivo that is ankyrin independent.
The greatest sequence identity between and mammalian β spectrins occurs in the first three segments of the protein (), which includes the actin-binding site () and tail-end subunit interaction sites in segments 2 and 3 (). Another site of striking sequence conservation among β spectrins is the partial repeat 18, where α and β spectrin interact to form tetramers (). Spectrin is thought to have evolved from a large single-subunit ancestor by fracturing of the coding sequence at a site within an ancestral structural repeat (). A point mutation in the N-terminal partial repeat of α spectrin (R22S) produced a temperature-sensitive defect in spectrin tetramer formation (). Here, we tested the effect of a comparable mutation in the β subunit that was also identified by its role in human anemia (). The W2033R mutation corresponds to the W2024 position of human erythroid β spectrin, and that residue is also conserved in human βII, βIII, and βIV, but not in βV spectrin. This tryptophan residue was dispensable for β spectrin function in , presumably because it does not affect tetramer formation. Polarized targeting of the transgene product was also unaffected by the mutation (unpublished data). Another tryptophan at position 2061 of human erythroid β spectrin that has been implicated in hereditary elliptocytosis () is conserved in human βII, βIII, and βIV spectrin, but not in β or in human βV spectrin. The importance of these tryptophan residues in nonerythroid spectrin tetramer formation and function has not been tested.
Relatively little is known about the function of repetitive segments 4–14 in β spectrin. This region has more limited sequence conservation than segments near the ends of the molecule ().
. For now, this transgene helps to establish that not all functional defects in β spectrin result in mislocalization.
, are probably identifying functional sites that are responsible for targeting.
The domain swap strategy described here takes advantage of the powerful genetic tools available in to study protein function. This approach is especially well suited to studies of a modular protein such as spectrin and has provided valuable insights into both the function and the targeting of the protein in vivo. Combined with the fact that is a simpler system, having only three spectrin genes and two ankyrin genes, this approach should continue to provide valuable information that will be more difficult to obtain in mammalian systems. One intriguing observation in this study was the effect of the ΔPH mutation on the size of the rare flies that survived to adulthood. This phenotype was reminiscent of mutations that affect PH domain–containing components of the insulin/insulin-like growth factor signaling pathway (). Given the importance of a phosphoinositide-binding PH domain to β spectrin function, it will be interesting to genetically test the possibility that spectrin has a functional interaction with growth factor signaling pathways.
A rabbit anti–β spectrin serum (Kcar) used for immunolocalization experiments and affinity-pure rabbit anti-ankyrin1 (DAnk1) antibody were previously described (). Another rabbit anti–β spectrin serum (337; ) was used on Western blots. Myc epitope–tagged β spectrin transgenes were detected using the 9E10 mouse monoclonal antibody (). Mouse anti-Dlg (discs large) antibody 4F3 was obtained from the Developmental Studies Hybridoma Bank. Rabbit anti-Scribble was a gift from C. Doe (University of Oregon, Eugene, OR; ). Monoclonal antibody α5 against the α subunit of the Na,K ATPase was a gift from D. Fambrough (Johns Hopkins University, Baltimore, MD).
The WUMB–β spectrin transgene was described previously (). All of the following were produced by PCR-based modifications and/or mutagenesis of the WUMB–β spectrin plasmid (as indicated), and the modifications were confirmed by DNA sequencing. Primers were obtained from Operon, and DNA sequencing was performed by the DNA Services Facility at the University of Illinois at Chicago Research Resource Center. All of the transgenic lines described here were produced by standard embryo microinjection into by Genetic Services, Inc.
The four β spectrin mutant alleles that were produced in our genetic screen () were initially maintained as balanced stocks with the compound X chromosome. Subsequently, the mutant X chromosomes were rebalanced using the chromosome to allow identification of mutant embryos.
chromosome. Without rescue, the only surviving F1 males are those that inherited the marked balancer (carrying the wild-type β spectrin gene from their mother). Unexpectedly, all of the mutant transgenes that we tested produced at least a low level of rescue.
transgene on Western blots stained with the anti–β spectrin antibody 337, we found flies that expressed only the rescue transgene (and not the full-length wild-type protein). However, we also found male progeny with the rescue phenotype (i.e., they lacked the marker ), but on Western blots these flies had both the rescue transgene product and the full-length wild-type product (unpublished data). Thus, these were not true rescue flies and ultimately we were able to trace the source of these flies to transmission of wild-type X chromosomes from fathers to sons in the rescue cross. This unusual transmission pattern could arise through the presence of XXY females in the balanced population or through nondisjunction of .
To bypass the false rescue problem, we crossed autosomal transgenes into a background. In this compound X line, X chromosomes are transmitted from mothers to daughters and from fathers to sons, and females transmit Y chromosomes to their sons. females carrying autosomal transgenes to be tested were crossed to
mutant males that also carried a translocation of the X chromosome (including a wild-type β spectrin gene) on chromosome 3: (). These males transmitted the mutant β spectrin gene on X to their sons, and we could score the ability of a test transgene to replace the translocation by the presence of male progeny that lacked the dominant eye phenotype. Progeny that inherited both the test transgene and the translocation were not scored as rescue because they had a eye phenotype. Thus, a transgene with 100% rescue efficiency would yield a 2:1 ratio of eye/non– eye rescue progeny, as observed with the control wild-type transgene.
An EcoRI–NotI fragment of a previously described DAnk1 ankyrin-EGFP construct () was subcloned in the vector pUAST (). An upstream EcoRI fragment was subcloned from DAnk1 cDNA 11 () to complete the coding sequence. Expression of this transgene was driven by the HS Gal4 line 89–2-1 (Bloomington stock 1799), which caused constitutive expression of DAnk1-EGFP in copper cells and salivary gland.
Whole flies were photographed with a camera (Coolpix 5700; Nikon) on a dissecting microscope (MZFL111; Leica). Larval midguts and salivary glands were dissected and stained as previously described () and mounted using Vectashield medium (Vector Laboratories) for viewing at room temperature. Images were captured on a confocal microscope (IX70 or FV500; Olympus) using a 60× Plan-Apo oil-immersion objective (NA 1.4) and Fluoview 2.1 software. Brightness settings were adjusted for control specimens (either wild-type siblings or expression level–matched control transgenes) using the photomultiplier, and settings were kept constant for capturing data for mutants. Images were saved as experiments in Fluoview and were converted to jpeg format by NIH ImageJ. Montages were assembled using Photoshop 6.0 (Adobe), and gamma adjustments after conversion to grayscale were performed for all panels simultaneously. The freehand line tool of NIH ImageJ was used for fluorescence intensity measurements of 10–14 salivary glands of each genotype. Intensity values were plotted as a histogram using Excel (Microsoft) with the control value set to 100.
Adult flies were homogenized in SDS gel sample buffer in a dounce homogenizer and immediately boiled. A clarified supernatant corresponding to 0.8 flies per sample was subjected to electrophoresis, blotted to nitrocellulose, reacted with primary antibody and alkaline phosphatase–coupled secondary antibody, and stained with bromochloroindolyl phosphate as previously described (). Expression of UAS DAnk1-EGFP driven by HS Gal4 was found to be independent of HS (unpublished data). However, the DAnk1 antibody did cross-react with polypeptides in the 70–80-kD range in a HS-dependent manner ().
DNA sequence analysis was performed using the Wisconsin package (), and sequence alignments were performed using Gap. The sequence of β spectrin () was compared with human βspecIΣII (P11277; ), βspecII (Q01082; ), βIII spectrin (NP008877; ), βIV spectrin (NP079489; ), and βV spectrin (NP057726; ). Plotted points in were aligned with the center of each segment.
Fig.
mutants. Online supplemental material is available at . |
Focal adhesions are sites of matrix engagement with cell surface integrin clusters that are linked to the actin cytoskeleton at stress fiber termini through interactions with multiple intracellular proteins, such as talin, vinculin, and paxillin (; ). The signaling and molecular mechanisms leading to focal adhesion assembly are well characterized and involve multiple Rho family GTPases, actin binding proteins, and integrin-matrix binding (). In contrast, relatively little is known about the mechanisms involved in adhesion disassembly, but the involvement of Rho–Rho kinase (ROCK) signaling, calpains, and microtubules have been proposed (; ). In particular, Rho–ROCK promotes focal adhesion disassembly at the cell rear, and inhibition of this pathway produces a striking contractile and/or tail-retraction defect that is associated with decreased myosin light chain (MLC) 2 phosphorylation in various cell types (; ; ; ; ; ).
ROCK-based contractility is not only involved in the disassembly of cell–matrix adhesions during tail retraction but can also disrupt the stability of cell–cell adhesions associated with adherens junctions (). Adherens junctions occur at sites of cell–cell contact in organized epithelial cell monolayers and are formed via the homotypic interaction between E-cadherin on adjacent cells. The cytoplasmic tail of E-cadherin is linked to the actin cytoskeleton through interactions with catenin proteins (α, β, and p120) and actin binding proteins (vinculin). Adherens junctions can be regulated by translational events but are also subject to direct control by posttranslational cellular mechanisms, including their disassembly by the actin cytoskeleton and endocytosis ().
Endocytic dynamics have been shown to coordinate several key intracellular signaling events (; ; ). In this study, we investigated whether endosomal signaling could represent an integral part of the deadhesion process, both in rear cell retraction and adherens junction breakdown. In particular, we have investigated the role of the endocytic receptor Endo180 in these events. Endo180 (also known as CD280; uPARAP) is a 180-kD type I transmembrane receptor comprised of an N-terminal cysteine-rich domain followed by a fibronectin type II (FNII), 8 C-type lectin-like domains, a single transmembrane domain, and a short cytoplasmic domain (; ). Within this cytoplasmic domain, a critical dihydrophobic Leu/Val motif mediates the constitutive recruitment of Endo180 into clathrin-coated pits on the cell surface, which is followed by rapid internalization into intracellular endosomes and efficient recycling back to the cell surface (; ). This trafficking of Endo180 is essential for its function as a collagen internalization receptor in which collagen bound to Endo180 is rapidly taken up into the endosomes and then dissociated from the receptor for delivery to, and degradation in, lysosomal compartments (; ; ; ). In addition to its role in ligand internalization, a promigratory function for Endo180 has also been demonstrated. Cells derived from mice with a targeted deletion in Endo180 and in which Endo180 expression is knocked down by siRNA oligonucleotides both display a reduced migratory capacity. Conversely, ectopic expression of Endo180 in Endo180- negative cell lines results in the acquisition of a polarized phenotype and enhanced cell migration (; ; ). Here, we have further investigated the promigratory function of this receptor and provide a mechanism by which intracellular Endo180 can spatially regulate cell contractility and adhesion dynamics.
The potential involvement of endosomes in the spatial activation of ROCK during rear cell deadhesion was investigated by comparing the effects of ROCK inhibition with the down-regulation of constitutively recycling endocytic receptors. Treatment of MG63 osteosarcoma cells with ROCK inhibitor produced a tail-retraction defect that was associated with a decrease in MLC2 phosphorylation (; and Fig. S1 a, available at ), consistent with its effects in other cell types (; ; ; ; ; ). Three endocytic receptors, Endo180, transferrin receptor, and low-density lipoprotein (LDL) receptor, were targeted with siRNA oligonucleotides (Fig. S1 b). These receptors were chosen as Endo180 drives cell migration through an endocytosis-dependent mechanism () and transferrin receptor enhances cell migration on a transferrin substrate (). Because the LDL receptor (LDLR) has no reported role in cell migration, it was included as a negative control. Endo180 siRNA treatment of MG63 cells resulted in a striking elongated phenotype, indicative of a tail-retraction defect, that was indistinguishable from that produced by ROCK inhibition (). This was not an off-target effect of Endo180 siRNA, as an identical phenotype was observed using alternate oligonucleotides (; and Fig. S1 c). In contrast, no defect in rear cell deadhesion was apparent in transferrin receptor or LDLR siRNA–treated cells (). Further, Endo180 was the only endocytic receptor required for MLC2 phosphorylation ( and Fig. S1 c). Defective cellular contractility and/or rear cell retraction resulting from the down-regulation of Endo180 was not restricted to MG63 osteosarcoma cells but was also confirmed in MDA-MB-231 breast carcinoma, HT-1080 fibrosarcoma, and BE colon carcinoma cells ( and Fig. S2 a).
Endo180 is a coreceptor for the glycosphosphatidylinositol-anchored urokinase-type plasminogen activator (uPA)–uPA receptor (uPAR) complex () and is required for the activation of directional signaling pathways during sensing of a uPA gradient by migrating cells (). Because uPA–uPAR has been reported to regulate Rho–ROCK signaling and the phosphorylation of MLC2 (; ), we considered this to be a potential regulatory component in rear cell deadhesion and contractility promoted by Endo180. To address this possibility, uPAR was targeted using siRNA oligonucleotides in MG63 and BE cells, which express low and high levels of uPAR, respectively (Fig. S1 b and Fig. S2 b). As previously reported (), treatment with uPAR siRNA was effective at decreasing membrane ruffles in BE colon carcinoma cells (unpublished data). However, no tail-retraction defect or reduction in MLC2 phosphorylation was observed in either MG63 cells () or BE cells (Fig. S2 c), indicating that uPA–uPAR does not promote cell contractility associated with rear cell tail retraction and is unlikely to have a role in this particular Endo180-mediated event during random cell migration.
Endo180 is a well-established collagen binding and internalization receptor (; ; ; ). To investigate whether these specific functions of Endo180 played a part in rear cell deadhesion and contractility, the behavior of cells targeted for knock down of Endo180 on both non–collagen- and collagen-based substrata was investigated. MG63 cells treated with Endo180 siRNA oligonucleotides for 72 h formed unretracted tails within 4 h of being seeded onto uncoated glass coverslips or glass coverslips coated with fibronectin, collagen I, or Matrigel (of which collagen IV is a major constituent), whereas on all substrata no tail-retraction defect was observed in control siRNA–treated cells (). Further, MG63 cells grown on uncoated tissue culture plastic or tissue culture plastic coated with fibronectin, collagen I, or Matrigel displayed similar decreases in MLC2 phosphorylation (). These findings are in keeping with a previous report () and new data presented here (Fig. S3, available at ) that Endo180-mediated cell migration is not dependent on collagenous extracellular matrix substrata.
To test the hypothesis that Endo180 regulates ROCK activity, a series of experiments were undertaken to assess whether Endo180 down-regulation could recapitulate the specific cellular and biochemical events associated with ROCK inhibition. First, a comparison using time-lapse microscopy of tail formation after global ROCK inhibition and Endo180 siRNA treatment revealed a similar sequence of cellular dynamics ( and Videos 1–3, available at ). In both cases, cells exhibited a collapsed morphology, an increase in cell body movement (), and impaired translocation. This resulted in the formation of multiple and elongated tails, some of which then became protrusive and displayed localized membrane ruffles (Videos 1–3). In these assays, it was noted that the tail phenotype produced by ROCK inhibition was more extensive than that associated with Endo180 siRNA treatment. This most likely is a consequence of ROCK inhibition at multiple cellular locations, whereas tails resulting from Endo180 siRNA treatment were generally restricted to the rear of migrating cells and could result from the inhibition of spatially regulated ROCK activity. The global inhibition of ROCK also explains the more collapsed phenotype of cells treated with ROCK inhibitor that occurs before their formation of multiple tails (, t = 0).
Second, the signaling pathways downstream of ROCK were investigated. Activation of MLC2 by ROCK occurs through both diphosphorylation at threonine 20/serine 19 and a single phosphorylation event at serine 19. ROCK also phosphorylates LIM kinase (LIMK) 1/2 at threonine 508/505 and myosin phosphatase (MYPT) 1 at threonine 696 (; ). Treatment of MG63 cells with Endo180 siRNA or ROCK inhibitor not only reduced the diphosphorylation of MLC2 (, , and Fig. S1 a) but also reduced the monophosphorylation of MLC2 at serine 19 and decreased the phosphorylation of both MYPT1 and LIMK1/2 (). The reduction in phosphorylation attributed to Endo180 siRNA treatment was not due to an effect on kinase stability, as the levels of serine/threonine kinases involved in the phosphorylation of MLC2 remained unchanged (Fig. S4, available at ).
Finally, the involvement of the GTPase Rho, which functions as the key effector that directly binds and activates ROCK (), was investigated. Previous work established a role for Endo180 in the activation of the other two Rho family GTPases, Cdc42 and Rac (but not Rho), during uPA-mediated “directional” migration (chemotaxis; ). However, the Endo180-dependent signaling pathways that promote “random” cell migration were not examined. To investigate whether ROCK activation by Endo180 occurs through the Rho pathway, levels of active Rho were measured in lysates of MG63 cells treated with Endo180 siRNA. The results of these experiments confirm that Rho activation is a downstream target of Endo180 (). Further, the inability of ROCK inhibitor to block Rho activity () confirms previous reports that ROCK lies downstream of Rho. Interestingly, the inhibition of Rho by TAT-C3 transferase in MG63 cells produced an unretracted tail phenotype similar to that observed in Endo180 siRNA– or ROCK inhibitor–treated cells (compare and ). Collectively, these data suggest the existence of an Endo180–Rho–ROCK–MLC2 signaling pathway that is involved in rear cell retraction.
Spatial regulation of MLC2 has been proposed to occur through localized signals that emanate from different upstream effectors (). It has also been hypothesized that endosomes have the capacity to perpetuate and/or amplify intracellular signaling pathways (; ; ). Consequently, it was important to determine whether Endo180 could spatially localize to activate Rho–ROCK–MLC2 during rear cell deadhesion and tail retraction.
First, live cells were stained at 4°C with Endo180 antibody to assess the cell surface distribution of this receptor. Unlike integrins that cluster at the cell surface of unretracted tails after ROCK inhibition (), no accumulation or clustering of plasma membrane Endo180 at the unretracted tails was observed in ROCK inhibitor–treated cells (). Instead, the receptor remained uniformly distributed across the plasma membrane in punctate structures, consistent with previous reports that 10–30% of cellular Endo180 is localized to the plasma membrane in clathrin-coated pits (). In contrast, immunofluorescent staining and confocal microscopy of Endo180 in permeabilized ROCK inhibitor–treated cells revealed a dramatic accumulation of Endo180 in the majority of unretracted tails (; 79 ± 4%; >50 cells scored in each of three separate experiments). High-magnification and multiple xz and yz confocal imaging of unretracted tails confirmed that this accumulated Endo180 was localized to intracellular endosomes (). In contrast, in cells treated with either ROCK inhibitor or Endo180 siRNA, ROCK itself did not accumulate/relocalize to rear cell adhesion sites and remained diffusely cytosolic (unpublished data). Importantly, a small number (4 ± 1%; >50 cells scored in each of three separate experiments) of untreated migrating cells also displayed very strong localization of Endo180-containing endosomes to cell–matrix adhesion sites at the termini of stress fibers or unretracted tails (, arrowhead). This confirms that accumulation of Endo180-containing endosomes occurs in normally migrating cells and suggests that the relatively small number of normal cells displaying localization of Endo180 to adhesion sites reflects the highly dynamic nature of endosomal trafficking during cell migration.
Second, the Endo180 endosomes were characterized to address two questions. Is the accumulation of Endo180 in the unretracted tails of ROCK inhibitor cells specific to this endocytic receptor, and does the accumulation of Endo180 reflect a change in endosomal trafficking caused by the inhibition of ROCK? For this purpose, the colocalization of Endo180 with transferrin receptor (a constitutively recycling receptor), early endosome antigen 1 (EEA1; a marker for sorting endosomes), and Rab11 (a marker for pericentriolar recycling endosomes; ) was assessed. In untreated cells, Endo180 displayed near total colocalization with transferrin receptor and partial colocalization with both EEA1 and Rab11 (), demonstrating that Endo180, like the transferrin receptor (), is localized to both rapidly recycling EEA1-positive sorting endosomes and slower recycling Rab11-positive pericentriolar endosomes. In ROCK inhibitor–treated cells, both transferrin receptor and EEA1 strongly accumulated with highest concentrations of these endosomal markers observed in the unretracted tails. A similar distribution of Rab11 was not evident in ROCK inhibitor–treated cells. As in untreated cells, Rab11 was uniformly distributed throughout ROCK inhibitor–treated cells, with the highest concentrations accumulated in pericentriolar recycling endosomes. The unretracted tails in ROCK inhibitor–treated cells were quantified by scoring for their accumulation of high concentrations of Endo180 together with high concentrations of transferrin receptor, EEA1, or Rab11. In this analysis, Endo180 displayed an almost total coaccumulation with transferrin receptor (98 ± 2%), a partial coaccumulation with EEA1 (24 ± 7%), and only minimal coaccumulation with Rab11 (4 ± 1%; ), indicating that it is the redistribution of receptor-positive endosomes, rather than the altered internalization/export of receptors, that accounts for the tail localization and that accumulation of endocytic receptors in unretracted tails is not exclusive to Endo180. Despite this lack of exclusivity, it is notable that transferrin receptor also strongly accumulated in the unretracted tails of Endo180 siRNA–treated cells (unpublished data), supporting previous experimental findings () that transferrin receptor has no functional role in the promotion of cell contractility during rear tail retraction. In contrast, further support for the mechanistic regulation of cell contractility by Endo180 was provided by the observation in untreated cells that Endo180-containing endosomes strongly accumulate at adhesion sites and unretracted tails that have a localized high level of MLC2 phosphorylation ().
The ability of Endo180 to regulate the adhesion/deadhesion process and generate contractile signals could emanate from the plasma membrane or internalized endosomes. To address this, studies were undertaken using the Endo180(Ala/Ala) mutant, which is expressed at the cell surface but internalization defective (; ). It has previously been demonstrated that the expression of wild-type Endo180 promotes MCF7 cell migration, whereas Endo180(Ala/Ala) does not (). For these experiments, stable transfectants with equal protein-expression levels of wild-type Endo180 and Endo180(Ala/Ala) were generated (Fig. S5, available at ). As expected from its promigratory function, expression of wild-type Endo180 increased cell spreading and the assembly of new focal adhesions (). In addition, these transfected cells showed enhanced adhesion to a Matrigel substratum (), consistent with a previous report that cells from mice with a targeted deletion in Endo180 show a defect in adhesion to a variety of collagen substrata (). The Endo180(Ala/Ala) mutant contains an intact collagen binding domain and, as expected, supported a level of adhesion to Matrigel similar to that of wild-type Endo180 (). However, Endo180(Ala/Ala) did not promote cell spreading and spatial assembly of new focal adhesions (). Rather, both vector alone and Endo180(Ala/Ala) transfected cells developed an elongated phenotype upon plating (), with tail structures that were reminiscent of those observed in Endo180 siRNA– or ROCK inhibitor–treated cells (compare and ). Moreover, in these cells, talin was seen to accumulate in the tail structures. These data indicate that an uncoupling of adhesion and deadhesion underlies the migration defect observed in cells expressing the internalization-defective Endo180(Ala/Ala) mutant and that although internalization of Endo180 is not necessary for the promotion of initial cell–matrix adhesions, it is required for the correct spatial formation and efficient turnover of cell–matrix adhesions during cell spreading and migration.
Next, several approaches were taken to confirm that Endo180 internalization is required for the generation of contractile signals. First, it was demonstrated that MLC2 phosphorylation generated by serum stimulation of starved cells was significantly elevated in cells transfected with Endo180 compared with that generated by cells transfected with vector alone or Endo180(Ala/Ala) ( and Fig. S5). Second, using a method that has been used to demonstrate the existence of intracellular endosomal signaling events (), cells were stimulated with serum for 10 min and the extracellular stimulus was withdrawn. Upon serum withdrawal, elevated MLC2 phosphorylation levels persisted for at least 2 h in Endo180-expressing cells but rapidly returned to basal levels (within 10 min) in vector alone transfected cells (). Finally, the recycling of endosomal components back to the plasma membrane was inhibited using primaquine () in serum-starved cells. As expected, primaquine resulted in a dramatic decrease in the localization of Endo180 at the cell surface () but did not affect total cellular levels of Endo180 (), indicating that Endo180 had been internalized but not recycled back to the plasma membrane. The resultant fourfold increase in MLC2 phosphorylation in primaquine-treated cells indicated that intracellular accumulation of Endo180 was sufficient to stimulate a contractile response that was blocked by the presence of ROCK inhibitor (). The observation that MLC2 phosphorylation levels remain unchanged in primaquine-treated vector alone transfected cells () provides independent evidence that other recycling receptors, such as transferrin receptor and β integrin (CD29), are not involved in the regulation of this signaling event. The results of these experimental approaches confirm that the internalization of Endo180 into endosomes can generate and sustain ROCK-dependent intracellular contractile signals.
Because Endo180 can promote localized ROCK–MLC2 signaling during tail retraction, we further hypothesized that ectopic expression of this receptor in epithelial cells should be sufficient to promote disassembly of their adherens junctions, as ROCK-based contractility can disrupt cell–cell adhesions (). The expression of Endo180 in MCF7 cells resulted in the loss of E-cadherin from cell–cell junctions, and this was reversed by treatment with ROCK inhibitor (). This effect was not specific to E-cadherin, as a similar redistribution was also observed with immunofluorescent staining of the junctional component α-catenin (unpublished data). Moreover, the decreased stability of MCF7 cell–cell junctions in cells ectopically expressing Endo180 was reversed by Endo180 siRNA treatment (). These findings suggest that in addition to regulating the disassembly of cell–matrix adhesions at the cell rear during cell migration, Endo180 has the capacity to activate ROCK and generate contractile signals that promote the disassembly of adherens junctions at epithelial cell–cell contacts.
Our results point to a role for endosomes in the disassembly mechanism for cell–matrix and cell–cell adhesions. In particular, we have demonstrated that the endocytic localization of the Endo180 receptor activates mechanotransduction pathways that promote cell contractility and adhesion disassembly. A role for Endo180-containing endosomes in adhesion disassembly is complemented by a report that identified dynamin as a downstream target for microtubule-induced focal adhesion disassembly in fibroblasts (). Dynamin provides a ubiquitous molecular mechanism for driving endocytosis via its ability to recruit actin monomers to the neck of clathrin-coated pits, where their polymerization is required to force endosome internalization and propulsion through the dense cytocortex into the cytosol ().
Because the expression of Endo180 is predominantly restricted to fibroblasts and other highly motile cells, including a range of highly invasive cancer cell types (; ), this receptor could represent a highly specific endocytic component involved in focal adhesion disassembly during cell migration. In this respect, we would predict that the endocytosis-driven system of tail retraction promoted by Endo180–Rho–ROCK–MLC2 mechanotransduction acts downstream of the microtubule-mediated activation of dynamin. This hierarchy of molecular events gains additional support from earlier work that placed Rho–ROCK signaling downstream of microtubule-mediated adhesion disassembly during monocyte tail retraction (). Moreover, it is notable in the experiments described here that the localization of Endo180-containing endosomes to the cell rear adhesion sites in cells treated with ROCK inhibitor is reminiscent of the mature integrin adhesions that accumulate in the unretracted tails of Rho-inhibited monocytes (). It is well established that Endo180 functions as a collagen receptor, mediating ligand uptake for delivery to intracellular degradative organelles (; ; ), but we find here that binding of collagen is mechanistically independent of the ability of Endo180 to promote cell migration. This suggests that Endo180 has a constitutive function in promoting cell migration that is unrelated to the binding of exogenously added collagen. Indeed, the ability of Endo180 to couple with the signaling network that drives cell migration was previously implied by its involvement in the activation of Cdc42 and Rac by uPA and efficient sensing of a uPA chemotactic gradient (). The key questions that now need to be addressed are whether Endo180 signals to Rho GTPases via individual or multiple guanidine nucleotide exchange factors and/or other membrane-associated components localized in endosomes and whether the endogenous collagens produced by the Endo180-expressing cells may modulate receptor activity. It certainly remains a distinct possibility that Endo180 promotes uptake of focal adhesion components bound to extracellular collagen during collagen internalization and that this could impact cell migration.
A key finding of the studies described here is that the generation of contractile signals by Endo180 was not elicited from the plasma membrane but rather from Endo180 localized in intracellular endosomes. Further, the differential colocalization of other endosomal markers with Endo180 in ROCK inhibitor–treated cells indicated that it is the Endo180-positive sorting endosomes that preferentially accumulate at the rear cell adhesion sites. Notably, in these tail structures, the extensive colocalization of Endo180 and transferrin receptor is retained. Together with the data showing that there is no accumulation of Endo180 on the plasma membrane of unretracted tails, this suggests that the accumulation of endosomal Endo180 in the tails results from a redistribution of Endo180/transferrin receptor–positive endosomes rather than from altered internalization/export kinetics of Endo180. However, we cannot rule out the possibility that export of Endo180 and transferrin receptor from sorting endosomes in the tails may also be reduced by ROCK inhibition. The colocalization of Endo180 and transferrin receptor in ROCK inhibitor–treated cells demonstrates that the accumulation of endocytic receptors is not exclusive to Endo180. However, the transferrin receptor also showed a strong accumulation in the unretracted tails of Endo180 siRNA–treated cells (unpublished data), and down-regulation of transferrin receptor by siRNA treatment had no effect on tail retraction or phosphorylation of MLC2. Moreover, primaquine treatment of Endo180-negative cells resulted in the intracellular accumulation of transferrin receptor and β integrin but did not enhance MLC2 phosphorylation. As a consequence, we conclude that, at least among the endocytic receptors examined, Endo180 has an exclusive functional role in promoting cell contractility during rear tail retraction. Finally, the observation that ROCK itself did not accumulate/relocalize to rear cell adhesion sites and remained diffusely cytosolic in either ROCK inhibitor– or Endo180 siRNA–treated cells indicates that Endo180 localization at these sites could be the rate-limiting step in the spatial activation of ROCK and has led us to propose that the localization of Endo180-containing endosomes results in the spatial activation of Rho–ROCK–MLC2 to promote adhesion disassembly.
In addition to regulating adhesion of cells to a substratum, ROCK can regulate the integrity of cell–cell adhesion complexes (). Although there are conflicting reports in the literature as to whether ROCK activity promotes an increased (; ; ) or decreased (; ) integrity of intercellular junctions and permeability of cell monolayers, activation of ROCK has been shown to result in the disruption of E-cadherin–containing adherens junctions and the redistribution of junctional components in epithelial cells in culture (; ). The demonstration here that ectopic expression of Endo180 results in a redistribution of E-cadherin in epithelial monolayers and that this redistribution can be reversed by treatment with ROCK inhibitor provides further support for a role of Endo180 as an upstream regulator of ROCK. These data additionally suggest that aberrant expression of Endo180 in epithelial cells may promote the acquisition of a more mesenchymal migratory phenotype, and in this respect it is of interest that the noninvasive MCF7 epithelial cell line is Endo180-negative, whereas the more aggressive BE and MDA-MB-231 tumor lines are Endo180-positive.
These studies, combined with previous work, have demonstrated that Endo180 has a dual function, acting both as a regulator of Rho–ROCK–MLC2 signaling and as a collagen internalization receptor. To date, the most striking in vivo phenotype resulting from manipulating Endo180 expression comes from crossing mice with a mammary tumor–predisposing transgene to mice with a targeted deletion in Endo180. These mice develop tumors at the same rate as those expressing wild-type Endo180 but show an increased collagen deposition in the tumor-associated stroma and a decreased tumor burden (). Given the data presented here, we propose that this phenotype arises not only from the inability of Endo180-null fibroblasts to remodel the collagen-rich extracellular matrix but also because such cells will also have a mechanotransduction defect that will impair their motility within the tumor. Given that these stromal cells are major collagen producers, this impaired motility will contribute to the aberrant accumulation of extracellular collagen. Certainly it will be of interest to determine whether Endo180 plays a similar role in other pathological scenarios where altered fibroblast activity and matrix turnover are associated with disease progression.
Anti-Endo180 mAb A5/158 has been previously described (). B3/25 anti–transferrin receptor mAb was a gift from C. Hopkins (Imperial College London, London, UK). Mouse anti–human LDLR was obtained from Fitzgerald. Mouse anti–human uPAR was obtained from American Diagnostica, Inc. Rabbit anti–human monophospho-MLC2 (Ser19) and diphospho-MLC2 (Thr18/Ser19), LIMK1, and phospho-LIMK1 (Thr508)/LIMK2 (Thr505) were obtained from Cell Signaling. Mouse anti–human MLC (clone MY21), MLCK, talin, and γ-tubulin were obtained from Sigma-Aldrich. Mouse anti–human MYPT1, ROCK1, ROCK2, ZIP kinase, EEA1, and Rab11 were obtained from BD Biosciences. Rabbit anti–human phospho-MYPT1 (Thr696) was obtained from Upstate Biotechnology. Rabbit anti–human MRCKα and MRCKβ were a gift from S. Wilkinson (Institute of Cancer Research, London, UK; ). Mouse anti–human CD29 was obtained from Serotec. Mouse anti–human E-cadherin (clone HECD-1) was obtained from Abcam; secondary antibodies Alexa Fluor 488/555 anti-rabbit Ig and Alexa Fluor 488/555 anti-mouse Ig, Alexa Fluor 488/555/633 phalloidin, and TO-PRO-3 were obtained from Invitrogen. HRP anti-mouse Ig was obtained from Jackson ImmunoResearch Laboratories, and HRP anti-rabbit Ig was obtained from Santa Cruz Biotechnology, Inc. E-cadherin antibody was labeled using Zenon Alexa Fluor 555 mouse IgG labeling kit (Invitrogen). For some experiments, Endo180 directly conjugated to Alexa Fluor 488 (Invitrogen) was used. MG63, MDA-MB-231, HT1080, and BE cells were maintained in DME + 10% FCS. MCF7 cells transfected with vector alone, wild-type Endo180, and Endo180(Ala/Ala) were cultured as previously described ().
Immunostaining and cell surface labeling of Endo180 were performed as described previously (; ). For confocal imaging, cells were fixed, stained, and mounted in Vectashield H-1000 (Vector Laboratories) at room temperature. Images were captured at room temperature with a confocal microscope (TCS SP2; Leica) and Confocal Software (Leica) using 63× (1.40 NA, oil; Leica) or 40× (1.25 NA, oil; Leica) lenses and Immersol 518F oil (Carl Zeiss MicroImaging, Inc.). Images were imported in to Photoshop 8.0 (Adobe) for processing.
Immunoblotting was performed as described previously (). Rho activation assay kit was purchased from Upstate Biotechnology, and assays were performed according the manufacturer's guidelines. The cell permeable TAT-C3 transferase toxin was a gift from G. Mavria (The Institute of Cancer Research, London, UK). Quantification of phosphorylation levels and Rho activation were measured using ImageJ densitometric software. Data is adjusted for loading and normalized to 100% for control levels ± SEM.
Endo180, reversed Endo180 (control siRNA), and uPAR single siRNA oligonucleotides were as described previously (; ). Transferrin receptor, LDLR, and Endo180 were targeted using SMARTpool siRNA oligonucleotides (Dharmacon). siRNA oligonucleotides (20 nmol/ml) were transfected into cells seeded on coverslips or culture dishes (30–50% confluent) with 100 μM Oligofectamine (Invitrogen) in Opti-MEM reduced-serum medium (Invitrogen). Knock down of targeted receptors was assessed by flow cytometry as previously described (; ). Optimal knockdown was obtained 72 h after transfection. The highly specific ROCK inhibitor (S)-(+)-2-Methyl-1-[(4-methyl-5-isoquinolinyl)sulfonyl]homopiperazine, 2HCl (Calbiochem) was diluted in culture media from a stock solution of 10 mM in sterile water. TAT-C3 transferase toxin was diluted in culture media from a stock solution of 7.2 μM and used at a final concentration of 1 μM. To inhibit endocytic recycling, MCF7 cells were starved for 48 h in DME before incubation with 0.6 μM primaquine (Sigma-Aldrich).
MG63 cells were left untreated or treated with siRNA oligonucleotides for 72 h. Cells were seeded on coverslips and allowed to adhere for 1 h. The coverslips were then placed on to counting chambers (Hawksley Technology) and sealed with wax as previously described (). siRNA-treated cells were assayed in fresh growth medium, and ROCK inhibitor–treated cells were assayed in fresh growth medium containing 1 μM ROCK inhibitor, which was added to cells 30 min before the start of image collection. Images of cells were digitally recorded at a time-lapse interval of 1 min for 4 h using an microscope (IX70; Olympus) fitted with humidified 37°C incubation chamber, a 20× lens (0.4 NA, dry; Olympus), and Simple PCI acquisition software (Digital Pixel). Speed of cell body movement (mean ± SEM; > 100 cells) was calculated as previously described for mean cell migratory speed () using Motion Analysis software (Kinetic Imaging Ltd) and Mathematica software (Wolfram Research Ltd).
Flow cytometry was performed as previously described (). Data is presented as relative fluorescent intensity (median fluorescent intensity of antibody binding/median fluorescent intensity of isotype-matched control IgG binding) in which the isotype-matched control IgG binding is set at zero.
To measure adhesion, 10 Calcein AM–labeled cells were added to each well of a Matrigel-coated 96-well plate in growth medium. Cells were left for 1 h at 37°C and washed two times, and adherent cells were counted using a fluorescence plate reader.
Fig. S1 shows a dose-dependent inhibition of MLC2 by ROCK inhibitor, flow cytometric analysis of receptor levels after siRNA treatment, and that the single Endo180 siRNA oligonucleotides and Endo180 SMARTpool oligonucleotides have similar effects on Endo180 knockdown and MLC2 phosphorylation. Fig. S2 shows that treatment of MDA-MB-231 breast cancer cells with Endo180 siRNA oligonucleotides results in a tail-retraction defect similar to that seen in MG63 cells, in BE cells there is efficient knock down of both Endo180 and uPAR after treatment with their respective siRNA oligonucleotides, and knock down of Endo180, but not uPAR, in BE cells results in a reduction of MLC2 phosphorylation. Fig. S3 shows that MCF7 cells expressing Endo180 plated onto either Matrigel, collagen IV, or fibronectin show enhanced migration compared with vector alone transfected cells. Fig. S4 shows that treatment of MG63 cells with Endo180 siRNA oligonucleotides does not affect the expression levels of six MLC regulatory kinases. Fig. S5 shows expression levels of Endo180 and Endo180(Ala/Ala) transfected into MCF7 cells. Videos 1–3 show time-lapse microscopy of control siRNA–, Endo180 siRNA–, and ROCK inhibitor–treated MG63 cells, respectively. Online supplemental material is available at . |
The interplay of mechanical forces between cells and their surroundings regulates crucial processes such as cell shape, differentiation, and protein expression (; ; ; ). Integrins are large transmembrane adhesion proteins that provide the physical link between the extracellular matrix and the contractile cytoskeleton. Integrins are heterodimers, composed of α and β subunits that associate noncovalently to form an extracellular, ligand-binding headpiece, followed by two multidomain “legs,” two transmembrane helices, and two cytoplasmic tails ().
Integrins are not constitutively active. Rather, ligand-binding affinity is regulated via conformational change that can originate from cytoplasmic or extracellular interactions (). For example, ligand binding has been shown to induce the opening of the hinge angle between the βA (also called the I-like or βI domain) and hybrid domains in the integrin headpiece (, ; ). Opening of this hinge has been linked to the switch to high binding affinity (, ; ). The x-ray crystallographic structures of the unliganded αβ and the liganded αβ integrins display the integrin headpiece conformations before and after the ligand-induced transition from the closed to the open βA/hybrid domain hinge, respectively (; ). The α1 and α7 helices of the βA domain are known to be involved in this transition (, ; ).
Because of a lack of dynamic, nonequilibrium information, the sequential details by which structural change in the ligand-binding pocket is allosterically coupled to the remote hinge conformation at the βA/hybrid interface have remained unknown, as have the details of force-induced integrin conformational change, which would enable insights into how integrin function is kinetically regulated during cellular mechanosensing. With no techniques currently available to gain these details experimentally, we used MD and steered MD (SMD) to derive high-resolution, dynamic structural models computationally.
Our simulations are based on the liganded crystal structure of the αβ integrin, which was obtained by soaking RGD peptide into unliganded integrin crystals that were preformed with bent legs and a closed βA/hybrid hinge (). Although comparison of this crystal structure with the unliganded αβ integrin structure reveals ligand-induced conformational changes local to the binding site (Fig. S1, available at ), both structures share the same long-range conformation of bent legs and a closed βA/hybrid hinge. Thus, although unequivocally recognized as a milestone in integrin research, the liganded αβ integrin crystal structure has sparked much debate; does it represent the final, high-affinity conformation (; ; ), or were the long-range, ligand-induced structural changes that underlie integrin activation arrested by conformational constraints (; , ; ; )?
RGD ligands bind to the integrin β subunit via a divalent metal ion located at the top of the βA domain, termed the “metal ion–dependent adhesion site” (MIDAS; ). Two additional ion-binding sites border the βA domain MIDAS on either side, which are termed the “ligand-induced metal-binding site” (LIMBS) and the “adjacent to the MIDAS” (ADMIDAS; ). Although all three sites are occupied by ions when the ligand is bound (; ), only the ADMIDAS is occupied in unliganded αβ crystal structures (, ). Moreover, the ADMIDAS of unliganded αβ crystal structures directly contacts the β6-α7 loop (Met), whereas this contact is lost and the ADMIDAS is shifted inward by ∼4 Å in both liganded β integrin crystal structures (, ; ).
Because integrins in a cellular context bind to extracellular matrix proteins, we replaced the RGD peptide in the liganded αβ integrin crystal structure () with the RGD-containing 10 type III fibronectin (FnIII) module () and equilibrated this complex in a box of explicit water molecules and in the absence of the integrin legs. Domains included were the propeller domain from the α subunit and the βA and hybrid domains from the β subunit (). Based on these simulations, together with equilibrations of the same domains from the unliganded αβ integrin crystal structure (), we propose that the energy barrier to ligand-induced hinge opening in the αβ headpiece is lowered by (a) the MIDAS conformation found in liganded crystal structures (; ), (b) Mg in place of Ca at the LIMBS and ADMIDAS, or (c) a stabilizing contact outside of the RGD-binding pocket between the top of the α1 helix and FnIII, which is reported for the first time in this study. Finally, we show how ligand-mediated mechanical force accelerates hinge opening along the same allosteric pathway. Through comparison with current experimental data, we propose a structural model that provides new insight into existing notions of integrin activation.
xref
sup
#text
Integrin activation is known to have a structural origin (; ), yet the key sequential details of activating conformational change have been controversial. For example, experiments have shown that ligand binding to the integrin head induces the opening of the hinge between the βA and hybrid headpiece domains (, ; ; ; ), but this hinge is closed in the liganded αβ crystal structure (). We replaced the peptide ligand of this structure with FnIII and revealed in MD simulations of the isolated integrin headpiece complex how ligand- induced structural change propagates across the βA domain via the α1 helix, to then be amplified by an increase in the remote βA/hybrid hinge. In separate equilibrations of the same complex, three different structural perturbations at the top of the βA domain induced small, elastic distortions in the α1 helix, and the hinge did not increase substantially in the nanosecond time window. The structural variations that we found to impede hinge opening on the nanosecond timescale are a 2-Å MIDAS shift, Ca in place of Mg at the LIMBS and ADMIDAS, and disruption of an FnIII–α1 helix–binding contact. In equilibration of the unliganded integrin headpiece domains (), maximum fluctuations of the β1−α1 loop and α1 helix were found to occur. Together, these investigations enabled us to deduce the allosteric pathway whereby a small inward (vs. outward) movement of the α1 helix promotes (or impedes) βA/hybrid domain hinge opening. In addition to predicting a role for a new FnIII–β integrin contact located outside of the RGD-binding pocket, these findings offer a structural basis for mAb experimental data that also link an outward α1-helix position with Ca in place of Mg at the ADMIDAS and an inward α1-helix position with hinge opening and integrin activation (, ).
The crystal structures of the unliganded αβ- and the liganded αβ integrins provide stationary endpoints before and after the transition of the integrin headpiece to the open conformation of the βA/hybrid hinge, respectively (; ). Together with experiments that have linked integrin activation with βA/hybrid hinge opening (; ; ), we propose that the unliganded αβ structure represents a low-affinity state of the integrin headpiece (which we shall call “state 1”), and the liganded-αβ structure represents a high-affinity state of the integrin headpiece (hereafter referred to as “state 2”). In this framework, the MD-derived conformations presented in this study are snapshots along the ligand-induced transition from state 1 to state 2 ().
Quite unusual for domain–domain contacts, βA connects to the hybrid domain via two peptide linkers, at both the C and N terminus. In this fashion, floppiness of the βA/hybrid hinge itself is greatly restricted. A key finding of this study is the allosteric link between floppiness and distortion in the βA domain α1 helix and the opening of the βA/hybrid hinge. Although we found that Ca in place of Mg at the LIMBS and ADMIDAS resulted in the outward shift of the α1 helix and no substantial hinge-angle increase on the nanosecond timeframe, experiments have shown that ligand-induced opening of the hinge does ultimately result in Ca in the LIMBS and ADMIDAS (). Importantly, although experimental results were obtained under equilibrium conditions in which the strain built-up by ligand binding had sufficient time to be released by the opening of the βA/hybrid hinge, the nonequilibrium snapshots described in this study probe structural dynamics on the nanosecond timescale and, thus, reveal how the ligand-induced strain either propagates to the βA/hybrid interface and is released by hinge-angle opening or, as a result of conformational variations near the binding site, produces elastic α1-helix distortions. Notably, the liganded αβ integrin crystal structure, which presents the open-hinge conformation with Ca at the LIMBS and ADMIDAS, also displays the inward-shifted α1 helix joined together with the β1−α1 loop, which is not found in integrin crystal structures with closed βA/hybrid hinges (, ). Also unique to the αβ crystal structure is a molecule of glycerol bound directly to the ADMIDAS ion at the top of the α1 helix (), which may lend additional stability to this key regulatory region.
These findings address a long-standing debate regarding a feature of the liganded αβ−integrin crystal structure that surprised the integrin community; a severe bend in the legs (). Because the structure is liganded, it has been considered to be in the active conformation (, ). However, because the structure was obtained by soaking RGD peptides into unliganded integrin crystals that were preformed with bent legs and closed hinges (), it has been suggested that the conformation of this structure is the result of constraint imposed by the preexisting crystal lattice (; ; ). In support of the latter notion, the legless, liganded αβ crystal structure, which presents the same MIDAS⋅⋅⋅HO⋅⋅⋅Asp conformation as the liganded αβ−integrin crystal structure, but was formed by cocrystallization with a ligand, displays the open hinge (). We have shown that once the conformational constraints of the leg domains are lifted, the α1 helix joins together with the β1−α1 loop and moves inward, and the hinge opens spontaneously on the 6-ns timeframe when the liganded αβ integrin crystal structure is occupied with Mg and equilibrated in complex with FnIII. Thus, the liganded αβ crystal structure may depict a nonequilibrium headpiece conformation that was constrained along the transition from state 1 to state 2 at a point after RGD peptide–binding had induced a change in the ADMIDAS and β1−α1 loop and before the strain built-up by this structural perturbation could be relieved by the opening of the βA/hybrid hinge (Fig. S1).
Although the bend in the integrin legs is known to be physiologically relevant (; ), the function of the bend during activation is debated. Currently, two models are widespread; in the “deadbolt” model, activation results from the loss of a constraining contact between the β-tail and βA domains (), whereas in the “switchblade” model, activation results from extension and separation of the legs and the opening of the βA/hybrid hinge (). Because the legs remain bent in the former and become extended in the latter, these models are often presented as incongruent. However, in both paradigms, the conformation with the bent legs and the closed headpiece hinge angle is low affinity (state 1), and activation necessitates disruption of headpiece–tailpiece contacts. Mutational and mAb experiments have shown that the low-affinity state is stabilized by a close association between the α and β cytoplasmic tails (; ), α and β transmembrane helical packing (, ; ), and a close association in multiple points along the α and β legs (; ; ; ). According to the switchblade model, α- and β-subunit contacts stabilize the low-affinity state by constraining the integrin legs in the bent conformation. Recently, electron microscopy studies of the intact, Fn-bound αβ integrin ectodomain were found to display the bent-leg conformation together with a βA/hybrid hinge angle increase of ∼11 ± 4°, relative to the αβ crystal structures (). We have linked an inward shift of the α1 helix with a hinge increase of ∼20° (), indicating that hinge opening by this amount might already be sufficient to induce affinity-regulating conformational change. Aside from the findings by , studies correlating movement in the headpiece hinge with changes in affinity have been done on integrin ectodomains with modified, truncated, or entirely absent legs (, ; ; ). In consideration of these experimental findings, together with our computational snapshots, we propose that the major physiological role of the of the bent-leg conformation is to tune, via domain–domain contacts, the height of the energy barrier that has to be overcome for the βA/hybrid hinge to increase and thereby transition the integrin head to the high-affinity state.
Integrin activation is bidirectional and can originate from either the cytoplasmic or the extracellular end of the molecule (). The transition of integrins from the low- (state 1) to the high-affinity state (state 2) before ligand binding has been referred to as “priming” (). When induced from inside the cell, the structural mechanisms underlying priming have been shown to be generated by conformational change in the cytoplasmic tails, transmembrane helices, and leg domains (; ; ). However, conformational changes in these structural regions distal to the ligand-binding site are not required when integrin priming is induced extracellularly (). In other words, extracellular factors that prime integrins, such as Mn ions, conformation-dependent mAbs that map to the integrin headpiece (, ; Clark et al., 2005), and mutations that induce the shortening of the α7 helix () or the opening of the headpiece hinge () do so by inducing the high-affinity conformation directly. As expected for an allosterically regulated protein, we thus propose that any mechanism which causes the headpiece hinge angle to open can thereby transition the RGD-binding pocket from state 1 to state 2, even in the absence of bound ligand. Hinge opening and the ensuing inward movement of the α1 helix would then constitute the final step in integrin priming.
Finally, we asked how mechanical force might impact the transition between state 1 and state 2, and we found in SMD simulations that force greatly accelerated hinge opening. The force-accelerated transition to the open hinge conformation was accompanied by the same inward movement of the α1 helix that we found to be characteristic of hinge opening under equilibrium conditions (). This finding implies that force accelerates hinge opening along the same allosteric pathway. Moreover, we have now shown that a 2-Å MIDAS shift or Ca in place of Mg at the ADMIDAS can prevent spontaneous hinge opening, but does not inhibit force-induced hinge opening on the same nanosecond timescale. This finding implies that these structural perturbations raise the energy barrier and that force acting along the β-subunit domains lowers the energy barrier to transition the ligand-bound integrin headpiece from state 1 to state 2. Considering the high sensitivity to minor structural perturbations that these findings reveal, the way in which point mutations to this region enhance () or disrupt () the allosteric activation pathway remains enigmatic.
Because ample experimental data, together with the MD simulations presented in this work, demonstrate that ligand binding can induce the transition from the closed to the open hinge conformation, what would be the advantage of accelerating this transition with force? Kinetic measurements have shown that the ligand-induced conversion of αβ integrin extracellular domains to the high-affinity conformation occurs on a timescale of ∼10 s (; ). Yet turnover rates at the timescale of seconds have been reported for constituents in newly formed adhesion sites, and a recent study shows that cells lose their ability to sense the rigidity of their surrounding Fn matrix when the binding of αβ integrins to Fn is blocked (). Thus, considering the shortness of the time window during which cells have to form adhesions that can sustain cell-generated tension, and assuming that integrins are in state 1 (i.e., not primed) when they first bind to matrix-exposed ligands, mechanical force could then play a physiologically important role in up-regulating the kinetics by which ligand-bound integrins transition from the low- to the high-affinity state. Do integrins then form catch bonds with their RGD ligands, bonds that are switched from a short-lived, low-affinity state to a long-lived, high-affinity state under mechanical force? A preliminary comparison between the αβ integrin complex investigated in this study and the catch bond characteristics of the bacterial adhesin FimH when bound to mannose reveals striking similarities () that will be considered in detail elsewhere (unpublished data).
The crystal structures of the αβ integrin in complex with an RGD-containing mimetic ligand (Protein Data Bank [PDB] code 1L5G; 3.2 Å resolution []) and of FNIII from the FNIII tetramer (PDB code 1FNF; 2.0 Å resolution []) were adopted to build the FNIII–αβ starting structure with the program VMD (). First, the RGD tripeptides of the two separate structures were aligned, and then the mimetic ligand from the αβ structure was removed. The resulting protein–protein complex contained slight overlap between the C and D strands of FnIII and the β6-α7 loop of the I-like domain in the integrin β subunit. The steric conflict was readily resolved by fixing the RGD tripeptide and rotating the remainder of the FnIII module around the Cγ atom of the Arg, away from the integrin. The rotation caused an ∼10° decrease in the angle formed by the Cα carbons at the N-terminal end, the Asp, and the C-terminal end. The two backbone bonds that connect the RGD sequence to the rest of the FnIII module were lengthened after the rotation, but a short minimization of 600 steps restored these bonds to their normal length. The key RGD–integrin–binding contacts, the bond between the Asp and the MIDAS cation and the bidentate salt bridge between the Arg and α-Asp, were well maintained. However, the salt bridge observed in the crystal structure between the ligand and α Asp was broken when the peptide was replaced by the FnIII module. To reduce the integrin to a size that can be feasibly simulated, we used only the βA and hybrid domains of the β integrin subunit, the β-propeller domain of the α integrin subunit. The starting structure was solvated in a 116 × 115 × 122 Å TIP3 () water box, resulting in 153,570 atoms. The water molecule manually added to the MIDAS conformation, between Asp and the MIDAS cation, was placed according to the location of the water molecule in the αβ integrin crystal structure (). Seven cations resolved in the headpiece, three resolved in the MIDAS-, LIMBS-, and ADMIDAS- binding pocket motifs, and four resolved in the solvent-exposed β hairpin loops at the bottom of the α-propeller, were occupied by Mn in the original αβ structure and by Mg in these simulations. For comparison, a solvated complex was also built, in which Mg occupied the MIDAS and Ca occupied the other six ion binding sites. This complex is called Ca–Mg–Ca in the text, in reference to the occupancies of the LIMBS, MIDAS, and ADMIDAS, respectively.
All MD simulations were performed with the program NAMD () using the CHARMM27 force fields (). The system was first minimized for three consecutive 2,000-conjugate gradient steps, during which the protein was initially held fixed and water molecules were allowed to move; next, only the protein backbone was held fixed, and finally, all atoms were allowed to move. After the minimizations, the system was heated from 0 to 300 K in 10 ps and subsequently equilibrated under constant pressure and temperature conditions (NPT). The bidentate salt bridge between Arg and Asp of the integrin α domain was held with harmonic constraints until 10 ps into the equilibration and remained stable in all equilibrated structures (). A more detailed explanation of the simulation protocol is described elsewhere (). During equilibrations the complex remained stable, exhibiting an overall Cα RMSD in the head domains (the β-propeller from the α subunit and βA domain from the β subunit) of <2.0 Å.
To open the closed hinge under force external forces were applied by SMD protocols at a constant pulling velocity of 5 and 10 Å/ns (). The spring constant was set to 6 kcal/mol/Å. To examine the mechanical response of the complex in various pulling geometries, we fixed either one or both of the Cα atoms of at the C-terminal residues of the integrin headpiece (α-Arg and/or β-Asp) and attached the spring to a Cα atom at either terminal residue of FnIII (Val or Thr). Generally, the direction of stretching force was chosen along the vector pointing from the fixed atom to the pulling atom. In the special case when both C-terminal ends of the α and β subunit were fixed, the direction of force was pointed from the midpoint of the two fixed atoms to the pulling atom.
Simulations presented in this work lasted 107 ns altogether. Simulations were completed at the National Center for Supercomputing Applications at the University of Illinois and on the Gonzales cluster at the Swiss Federal Institute of Technology. 1-ns simulation required ∼12 h on 128 nodes (IBM )with 1.5-GHz processors (Itanium; Intel), or 24 h on 64 Dlco nodes with 2.4-GHz processors (Opteron; AMD), respectively. All structure alignments were done in VMD () via the backbone atoms of the 6 βA domain β-strands (unless otherwise noted). Hinge angles were measured using Hingefind (). Figures were rendered using VMD.
Fig. S1 shows the headpiece domains of the liganded αβ integrin structure (1L5G.pdb) aligned with those of the unliganded αβ integrin structure via the backbone atoms of the 6 βA domain β-strands. Fig. S2 compares the two previously established conformations of the αA domain MIDAS with the two βA domain MIDAS conformations presented in this study. Fig. S3 depicts the “lever-arm” scenario, in which a 2-Å MIDAS shift causes a characteristic repositioning of the ADMIDAS, β1-α1 loop, and α1 helix that we found to impede hinge angle opening. Fig. S4 considers the FnIII-βA domain contact between Glu and Tyr, together with additional data and sequence analysis. Within the additional data is a repeat equilibration in which Mg occupied all binding sites, the MIDAS⋅⋅⋅HO⋅⋅⋅Asp conformation was maintained, and the hinge angle again spontaneously increased by ∼20°. Fig. S5 describes FnIII domain deformation under force. Videos 1, 3, and 5 show trajectories from the equilibrations of the reference complex, the Ca–Mg–Ca complex, and the unliganded integrin, respectively. In each case, the MD complex is overlaid with the starting crystal structure. Videos 2, 4, and 6 are 180° rotations around the y axis of the final frames from Videos 1, 3, and 5, respectively. Videos 7 and 8 are the trajectory and the 180° rotation, respectively, of the SMD simulation in which the hinge was opened 70° under force. Online supplemental material is available at . |
italic
xref
#text
xref
italic
#text
Various microorganisms tend to cluster together to survive nutrient depletion, forming multicellular communities called biofilms. In such a social community, the benefit of a cellular suicide program seems evident. The self-destruction of virus-infected, damaged, and old cells, which consume dwindling nutrients or spread an infection, contributes to the viability and reproductive success of healthier members of the community harboring similar genomes. has been shown to initiate the formation of biofilms in both natural and laboratory environments, particularly when nutrients are depleted (; ).
In the wild, an individual yeast cell landing on a rotting apple will divide and form a colony until all readily utilizable nutrients are exhausted. When the whole fruit is colonized and the next apple is not in sight, cells cease to proliferate and enter a postdiauxic but still metabolic active phase known as chronological aging (). To ensure survival of the clone, it seems to make sense for old or damaged cells to undergo cell death instead of consuming vanishing resources in a futile attempt to repair themselves. Chronologically and replicatively aged yeast cells die while exhibiting typical features of apoptosis, accompanied by the accumulation of ROS, which is a crucial factor in aging and apoptosis as well (; ; ). The question arises of how it is determined exactly which cells are to die in a chronologically aged population. In principle, this could be just a stochastic marker or, alternatively, a selection based on fitness. Notably, however, only replicatively older cells harboring two or more bud scars kill themselves in times of hardship (), demonstrating that during chronological aging, apoptosis selectively removes older individuals from the population. Furthermore, when dying, aged yeast cells actively stimulate the survival of the clone by releasing defined substances into their surrounding (; ). Interestingly, described specific aromatic alcohols as molecules mediating social interactions between yeast. These alcohols could have putative signaling functions leading to quorum sensing during chronological aging. To summarize, the death of older cells is advantageous for two reasons: first, because it spares nutrients for younger cells; and second, because the older cells release nutrients that can be metabolized by replicatively younger cells.
In addition, apoptosis has been demonstrated to occur during the development of colonies on solid media (). Aged colonies undergo metabolic changes that induce apoptotic death in a spatially restricted fashion, namely in the center of the colony. These colonies can be considered as multicellular organisms that undergo a sort of differentiation coupled to apoptosis. Interestingly, mechanical deletion of the colony center results in decreased survival of the colony margin. Thus, the outer cells profit from cell death occurring in the colony center, again arguing for an altruistic nature of cell death. This is in accordance with the classical definition of altruism, as this behavior increases the fitness of the group relative to other groups while it decreases the fitness of the altruist compared with others within the group (). Similar to yeast colonies, cells in a metazoan multicellular organism undergo cell death to ensure normal development and survival of the whole organism.
Deletion of the yeast caspase gene enhances the resistance against oxidative stress and also delays age-induced cell death (), although caspase-independent apoptosis occurs in yeast as well (; ). The fact that disruption of the apoptotic machinery results in an extended chronological life span is consistent with the idea that apoptosis might have evolved to clean the population of damaged or old cells that would consume resources and, thus, reduce the viability of the clone. However, the existence of a death program may not be advantageous under all conditions. In the wild, a yeast cell washed off a grape by a rain shower gains no advantage by undergoing apoptosis because there is neither food around to economize nor fitter relatives to spare the food for. Consistently, yeast incubated in distilled water have an extended life span as compared with cells cultured in minimal media (). However, it is also possible that nutrient metabolism produces ROS or other molecules that cause apoptosis.
In organisms ranging from yeast to mammals, the link between diet and aging is well established, as calorie restriction extends life span and increases stress resistance (). Pathways that mediate glucose-dependent signaling have recently been shown to interact with age-induced apoptosis in yeast (). Disruption of , the yeast orthologue of AKT1 kinase, which is a crucial regulator of glucose-dependent signaling and aging in and probably also in humans (), leads to an extended life span and delays apoptosis during chronological aging (, ). The decreased activity of Sch9p signals nutrient depletion and might therefore mimic a surrounding of pure water (). In the process of aging, calorie restriction delays apoptotic cell death and prolongs life span.
italic
#text
italic
xref
#text
xref
#text |
In response to environmental stress, cells reprogram their translational machinery and sort mRNAs that are released from polysomes to stress granules (SGs; ; ). This translational arrest is initiated by the phosphorylation of the translation initiation factor eIF2α, which results in a limited availability of the eIF2–GTP–tRNA complex (, ). It is presumed that by this limiting of translation initiation, mRNAs become stalled in 48S complexes, which then aggregate in SGs (). Thus far, there is no evidence for a decay of mRNAs within SGs. Hence, by temporal “storage” of mRNAs in SGs, transcripts can be protected from decay via the exosome or from degradation in processing bodies (PBs) to provide a reservoir of silenced mRNAs available for resuming translation upon stress release (; ; ). Recent studies identified an association of SGs with PBs, suggesting that mRNAs can become reprogrammed for further processing, including decay in PBs (). This model is supported by SG recruitment of several mRNA-binding proteins (RBPs), including TIA-1, TIAR, FMRP, Staufen, and CPEB, which, in nonstressed cells, regulate mRNA translation and/or mRNA turnover (; ; ; ).
In recent studies, we have investigated the role of the Zipcode-binding protein (ZBP) family in posttranscriptional gene regulation. This family of oncofetal proteins comprises a group of three RBPs (Imp1-3 in human) that modulate the localization, translation or stability of their target mRNAs (). In primary neurons, ZBP1 regulates localized translation of the β-actin mRNA in growth cones under control of Src-family kinases (). Translational control via ZBP proteins has also been identified for the insulin-like growth factor (IGF)-II mRNA (; ). Besides the control of mRNA localization and translation, ZBP1 (CRD-BP in mouse) was identified as a regulator of c-Myc, CD44, and βTrCP1 mRNA stability (; ; ). Although, to date, various RBPs have been identified as SG components in different cell lines, little is known about the requirement of these trans-acting factors for mRNA- specific processing in SGs (). We identify ZBP1 as a novel SG component that modulates the turnover of its target mRNAs during the integrated stress response (ISR).
To investigate the role of ZBP1 during the ISR, the subcellular distribution of ZBP1 was analyzed in response to oxidative stress or heat shock. Endogenous or GFP-fused ZBP1 were recruited to SGs traced by TIAR (TIA1-related protein; ; and Fig. S1 A, available at ; ). ZBP1 was also identified as a component of G3BP-induced (Ras-Gap SH3 domain-binding protein) SGs, suggesting that the protein is a ubiquitous component of SGs (; ). However, in contrast to G3BP or TIA-1, which induce SG formation upon overexpression, forced expression of ZBP1 had no influence on the rate of SG assembly (Fig. S1, B and C).
Although ZBP1 was targeted to SGs, the protein remained absent from PBs traced by RFP-DCP1 (Decapping protein 1) or endogenous DCP2, which are both exclusively targeted to PBs in stressed and nonstressed cells (; ). In contrast, various RBPs tested for targeting to SGs or PBs, including PTB, Raver1, KSRP, and hnRNP-U, which, like TIAR, all reside mainly in the nucleus in nonstressed cells, were not localized to either SGs or PBs (Fig. S1 D). This indicated that ZBP1 is targeted exclusively to SGs during the ISR, a feature not shared by RBPs in general.
In nonstressed cells, ZBP1 regulates the fate of target transcripts via an interaction of these mRNAs with the C-terminal RNA-binding KH domains (). The overexpression of GFP-fused deletion fragments revealed that the KH domains of ZBP1 are also required for SG targeting, suggesting that ZBP1 is associated with mRNA when localized to SGs (). Therefore, the subcellular distribution of various transcripts, including the ZBP1-associated RNAs IGF-II, c-Myc, β-actin, and H19, was analyzed by FISH during the ISR (). All ZBP1-specific RNA targets tested, including the nontranslated H19 RNA, were localized to SGs (; ; ). Likewise, several non–ZBP1-associated mRNAs, including GAPDH, were recruited to SGs (; and Fig. S2 A, available at ). The only tested mRNA that barely associated with SGs was the HSP90 mRNA, confirming that mRNAs translated during the ISR weakly associate or rapidly transit SGs (). These observations indicated that mRNA targeting to SGs is a nonspecific event affecting, presumably, most RNAs, whether actively translated or not. If so, one would expect that mRNA targeting to SGs is independent of regulatory cis-elements. This hypothesis was tested by exploring the SG recruitment of reporter mRNAs encoding a firefly luciferase equipped with MS2 repeats fused in proximity to the ZBP1-binding site of the β-actin mRNA (“Zipcode”) or to a Zipcode deletion fragment (; ). Both transcripts were targeted to SGs, as observed upon cotransfection of MS2-binding protein (MS2BP) fused to nuclear localization signal (NLS)–GFP (). Hence, mRNA targeting to SGs was independent of the cis-acting Zipcode. Therefore, we hypothesized that ZBP1 is not essential for SG recruitment of target mRNAs, but serves a function in regulating mRNA retention in SGs. Because this requires an association of ZBP1 with its target mRNAs in SGs, RNA binding was characterized by bimolecular fluorescent complementation (BiFC; ; ). YFP fluorescence in SGs was reconstituted from cotransfected YFP-fused ZBP1 and YFP-fused MS2BP only upon the expression of a reporter comprising Zipcode in proximity to the MS2 repeats (). This indicated that ZBP1 associated with the same transcript as the MS2BP in SGs when the Zipcode was present (). Thus, although SG targeting of reporter mRNAs appeared to be Zipcode independent, ZBP1 remained associated with its target mRNA in SGs.
It appeared plausible that ZBP1 either was essential for the structural integrity of SGs or that it controlled the transit of specific mRNAs to PBs to regulate mRNA decay (). To investigate this, ZBP1 levels were reduced via siRNAs before stress induction (Fig. S3 E, available at ). The assembly and structure of SGs, as assessed by staining for TIAR, revealed no obvious defects upon ZBP1 knockdown (). In agreement with ZBP1 not being essential for SG targeting of mRNAs, the localization of associated mRNAs like c-Myc remained basically unaffected when ZBP1 was depleted (). To investigate a putative role of ZBP1 in preventing rapid SG transit, and thus mRNA degradation of associated mRNAs, decay rates of c-Myc mRNA that is stabilized by ZBP1 via a translation-coupled mechanism in nonstressed cells were tested during the ISR, when translation of c-Myc is prevented because of the sequestering of the mRNA to SGs (). Cellular mRNA levels of c-Myc were assessed by quantitative RT-PCR (qRT-PCR) upon stress induction and normalization to RPLP0 and 18S RNA ratios, respectively (Fig. S3, A and B). The ratios of various mRNAs tested, including c-Myc, increased after the onset of the ISR, indicating that these mRNAs are stabilized in SGs and that transcription continues during stress, possibly at elevated levels (Fig. S3, A and B). To explore mRNA turnover that was unbiased by transcriptional reorganization, the decay of c-Myc mRNA during stress was analyzed after blocking transcription by actinomycin D. Compared with nonstressed cells, the c-Myc mRNA decayed slower; ∼40–50% of the initial mRNA levels remained after 1 h of arsenate treatment. This indicated that at 1 h after stress-induction, most of the mRNA had been targeted to SGs and only modest decay was observed after 1 h of stress induction, suggesting that SG transit of the transcript was slow (; Fig. S3 C).
Hence, we speculated that in SGs ZBP1 prevented the rapid decay of the c-Myc mRNA in a nontranslationally coupled fashion. Therefore, variations in cellular c-Myc and non–ZBP1-associated mRNAs (α-tubulin) were analyzed upon ZBP1 knockdown (). The c-Myc mRNA was selectively destabilized in stressed knockdown cells, whereas the stability of α-tubulin mRNA remained unaffected (). Notably, mRNA decay was normalized to mRNA ratios before stress induction to account for putative effects on mRNA stability upon ZBP1 knockdown. Thus, unbiased of “overall changes” in mRNA ratios, ZBP1 knockdown resulted in elevated c-Myc mRNA decay rates and increased total decay during the ISR (; and Fig. S3 D).
This selective stabilization of mRNAs by ZBP1 should also be observed for other mRNA targets, which, in nonstressed cells, are translationally regulated by ZBP1, as demonstrated for the β-actin and IGF-II mRNAs (; ). Hence, various mRNAs ratios were analyzed in stressed knockdown cells treated with actinomycin D by normalization to input levels. In addition to c-Myc, all tested mRNA-targets of ZBP1, including IGF-II, CD44, and the β-actin mRNA, were selectively destabilized upon ZBP1 knockdown (). This was also observed in the absence of actinomycin D (Fig. S3 E). Because ZBP1 knockdown increased the decay of target transcripts, we predicted that overexpression of ZBP1 would result in increased stabilization. Therefore, mRNA levels of the β-actin and IGF-II mRNA were compared in cells transfected with either GFP alone or GFP-fused ZBP1. The overexpression of GFP-ZBP1 resulted in selectively increased steady-state levels for both target mRNAs during the ISR (). In conclusion, these observations indicated that ZBP1 stabilized associated transcripts recruited to SGs, presumably by retaining target mRNAs in SGs. Thus, ZBP1 modified its function during the ISR; mRNAs such as IGF-II or β-actin that are translationally regulated by ZBP1 without affecting mRNA turnover in nonstressed cells were stabilized by ZBP1 during the ISR.
The role of ZBP1 in regulating mRNA turnover during the ISR was further explored by analyzing the fate of SG-targeted reporter transcripts during a cycle of stress induction and stress release. Compared with controls, the recovery of luciferase activity expressed from reporter transcripts harboring Zipcode was reduced after stress release upon ZBP1 knockdown (). This decrease in luciferase activity correlated with a Zipcode-dependent destabilization of reporter transcripts (). Hence, specific interactions between ZBP1 and target mRNAs in SGs were required to retain, and thereby stabilize, mRNAs in SGs.
To validate these findings, the loss of ZBP1-associated mRNAs from SGs upon ZBP1 knockdown was quantified by FISH. As expected, the relative amount of β-actin and IGF-II mRNA in SGs was selectively reduced in cells treated with ZBP1-directed siRNAs (). Notably, this selective reduction of target mRNAs in SGs was not a result of an “overall decrease” of these mRNAs upon ZBP1 knockdown (Fig. S2 B). Hence, the stabilization of target-mRNAs by ZBP1 apparently required retention of these mRNAs in SGs to prevent rapid decay.
Aiming at investigating SG transit of mRNAs in further detail, the PB localization of ZBP1-associated mRNAs was analyzed by FISH. However, endogenous mRNAs were barely observed in PBs, suggesting that during the ISR, mRNAs shuttled to PBs were rapidly decayed and therefore could not be detected (). Hence, we traced exogenous mRNAs by MS2 tagging that was expected to transiently stabilize RNA, even in PBs. Quantification of the mRNA content in SGs and PBs based on GFP fluorescent intensities normalized to either TIAR or RFP-DCP1 signals revealed a Zipcode-dependent reduction of mRNA levels in both foci upon ZBP1 knockdown (). In contrast, the amount of transcripts lacking Zipcode remained unaffected. Although we cannot exclude that mRNAs lost from SGs were degraded in part by the exosomes, these findings indicated that mRNAs usually retained in SGs by ZBP1 were transferred to PBs once ZBP1 levels were reduced. Rapid decay then presumably prevented the accumulation of tested mRNAs in PBs, suggesting that these foci are not rate limiting for mRNA decay during the ISR.
In conclusion, this work presents experimental evidence demonstrating that interactions between mRNAs and transcript-specific RBPs like ZBP1 can determine the fate of an mRNA during cellular stress. ZBP1 associates with its target transcripts within SGs, and this binding is required for the stabilization of associated mRNAs. Although in nonstressed cells ZBP1 either controls mRNA turnover or translation; the protein adopts a solely stabilizing function in SGs, as it stabilizes transcripts exclusively regulated on the translational level in nonstressed cells, e.g., the β-actin or IGF-II mRNA. This selective stabilization correlates with increased mRNA levels in SGs, and upon ZBP1 knockdown, ZBP1 target mRNAs are selectively reduced in SGs, whereas overall levels of the β-actin or IGF-II mRNAs in nonstressed cells remain unaffected by ZBP1 knockdown. Hence, stabilization of target mRNAs by ZBP1 requires retention of transcripts in SGs to prevent rapid translocation and subsequent decay. No accumulation of endogenous or reporter mRNAs was observed in PBs upon ZBP1 knockdown, suggesting that PBs are not rate limiting for mRNA decay during the ISR, and thus reflect the mRNA content observed in SGs. Therefore, one would assume that SGs serve as “filters” that regulate mRNA turnover during cellular stress. This function requires defined cis–trans interactions between mRNAs and RBPs like ZBP1. These retain, and thereby stabilize, target transcripts in SGs to prevent rapid decay by the exosome or, as previously suggested, in PBs ().
Luciferase reporter constructs fused to MS2 repeats were generated by subcloning from previously described constructs (; ). Plasmids used for BiFC were generated basically as previously described ().
U2OS cells were grown in DME (10% FBS). The assembly of SGs was induced by various stimuli, as follows: the application of sodium arsenate (final concentration of 2.5 mM) for indicated time, heat shock at 42°C for 1 h, or the overexpression of GFP-G3BP. Where indicated, transcription was blocked by 5 μM actinomycin D added 15 min before arsenate application. Plasmids and siRNAs were transfected with Lipofectamine 2000 (Invitrogen). ZBP1-specific siRNAs and controls were as previously described ().
U2OS cells grown on coverslips were treated with 2.5 mM sodium arsenate 48 h after transfection. Fixed cells were processed for immunostaining as previously described (). For indirect immunofluorescence of ZBP1, cells were stained with the following indicated antibodies: anti-ZBP1 (), anti-TIAR (BD Biosciences), anti–hnRNP U (Sigma-Aldrich), anti-HA (Sigma-Aldrich), anti-Flag (Sigma-Aldrich), and anti–human DCP-2 (raised in rabbit by peptide immunization).
FISH was essentially performed as previously described (). Where indicated, cells were fixed after the addition of primary and secondary antibodies, before performing FISH. All probes used were designed, synthesized, and labeled with Cy3-dye according to Singerlab protocols.
At ∼72 h after transfection with siRNAs, U2OS cells were treated with 2.5 mM sodium arsenate, and 5 μM actinomycin D were stated. Total RNA was extracted using TRIZOL (Invitrogen). 2 μg RNA was used for reverse transcription with random primers and MMLV-RT (Promega). qRT-PCR was performed based on SYBRgreen technology using 2xTaq-Master mix (Promega) in a cycler (MX3000p; Stratagene). For all primer pairs, an annealing temperature of 60°C was used. For primer sequences, see the table in Fig. S2 C. Relative changes of mRNA amounts were calculated based on the ΔC method, as described by . For Western blotting, total protein extracts were analyzed by Western blotting for ZBP1 (anti-ZBP) and vinculin (hVin1; Sigma-Aldrich), as described previously ().
Fig. S1 shows selective targeting of RBPs to SGs. Fig. S2 shows SG targeting of mRNAs not associated with ZBP1. Fig. S3 shows selective stabilization of mRNAs during the ISR. Online supplemental material is available at . |
Asymmetric cell division generates distinct progeny from a single cell division, and the two proteins Notch and Numb are critical for this process. In , Numb functions as a negative regulator of Notch (for review see ). Numb protein is asymmetrically localized to one daughter cell in cell divisions that generate distinct progeny. The cell receiving high levels of Numb suppresses Notch signaling, whereas the cell with low levels of Numb maintains Notch activity (; ). Numb and Notch are evolutionally conserved proteins. Two mammalian Numb homologues, Numb and Numblike, have been identified (, ). Gene targeting in mice reveals partially redundant functions for Numb and Numblike; i.e., the compound knockout of Numb and Numblike has a more severe phenotype than knockouts of each gene alone (; ). Data from (; ) and from adult mouse muscle progenitors (i.e., satellite cells; ) support a differentiation-promoting role for Numb/Numblike.
The Notch signaling pathway controls numerous cell fate decisions during development, often by maintaining a more undifferentiated fate. The Notch receptor is a single transmembrane protein that undergoes a complex series of proteolytic processing events. This ultimately leads to the release of the intracellular domain (ICD) of the receptor in response to activation from membrane-tethered ligands of the Delta or Serrate type (). The released Notch ICD translocates to the nucleus, where it interacts with the DNA-binding protein CSL (also termed RBP-Jκ [] and suppressor of hairless in ) to control activation of a specific set of downstream genes, most notably the Hes and Hey family basic helix-loop-helix transcription factor genes ().
Although Numb is known to be a negative regulator of Notch, we describe a more complex relationship between Notch and Numb/Numblike. Unexpectedly, high levels of Notch signaling lead to a reduction of Numb/Numblike protein levels, revealing a reciprocal negative regulation between Notch and Numb/Numblike.
A dose-dependent reduction of the levels of both a truncated membrane-tethered (Notch 1 ΔE) and an intracellular (Notch 1 ICD) form was observed in response to increasing amounts of Numb (). The Notch 1 ΔE protein was cleaved in the presence of Numb, although cleavage appeared to be somewhat reduced at higher Numb levels (). Furthermore, Numb and Numblike negatively regulate Notch signaling both from full-length Notch, Notch 1 ΔE, and Notch 1 ICDd (Fig. S1, available at ). We next studied whether Numb affected Notch intracellular localization. Transfected Numb-HA immunoreactivity was largely confined to intracellular vesicles (), which are likely to be endosomes, based on the codistribution of Numb-HA and Eps15 immunoreactivity (not depicted). In cells where the activation and cleavage of full-length Notch 1 was induced by coculture, the resulting Notch 1 ICD was predominantly localized to the nucleus also in the presence of transfected Numb (). In summary, these data indicate that Numb and Numblike negatively regulate Notch signaling and that Numb does not sequester Notch 1 ICD in the cytoplasm, arguing against a function for Numb in excluding Notch 1 ICD from the nucleus (; ; ).
We next analyzed the effects of Numb on cellular differentiation at various levels of Notch signaling in the myogenic cell line C2C12, in which differentiation to myotubes can be blocked by Notch signaling (; ). Different levels of Notch signaling were accomplished by coculturing cells expressing different levels of Notch receptor and ligand. Thus, regular C2C12 cells or C2C12 cells stably expressing Notch 1 (C2C12-N1) cocultured with regular 3T3 cells yielded low levels of Notch signaling as measured by 12XCSL-luc activation, whereas coculture with Jagged-1 (Serrate-1)–expressing 3T3 cells (3T3-J1) yielded high levels of Notch signaling (). Transfection of Numb exerted a negative effect on Notch signaling in all combinations, although the remaining level of Notch signaling after Numb transfection was considerably higher after coculture with Jagged-1–expressing cells ().
Transfection of Numb into C2C12 cells followed by 3 d under differentiation-promoting conditions resulted in a dramatic increase in the differentiation of Numb-expressing cells at low levels of Notch signaling (); 88 and 95% of the Numb-positive cells were also myosin heavy chain (MHC) positive in C2C12/3T3 and C2C12-N1/3T3 cocultures, respectively, and 92% were positive in C2C12 cells cultured alone (). Similar values (88%) were obtained for Numblike (). The proportion of differentiated cells after Numb and Numblike transfection was considerably higher compared with the blockage of Notch signaling by the γ-secretase inhibitor (GSI) DAPT, which only causes an increase in MHC-positive cells from 32 to 50% (). The addition of DAPT did not further enhance the differentiation-promoting effect of Numb and Numblike (). This argues for a more instructive role for Numb and Numblike in myogenic differentiation rather than only blocking Notch signaling. In keeping with a differentiation-promoting role, we observed that a C2C12 cell line stably expressing Numblike showed accelerated myogenic differentiation as compared with the parental cell line; 2 d after the switch to prodifferentiation conditions, 12.2% of the cells were MHC positive as compared with only 4.1% in the parental cell line (). The stable expression of Numblike was accompanied by somewhat elevated levels of the myogenic differentiation factor MyoD and a robust induction of myogenin during differentiation ().
At high levels of Notch signaling (i.e., in C2C12/3T3-J1 and C2C12-N1/3T3-J1 cocultures), we found no MHC-positive cells even after the transfection of Numb (). Unexpectedly, Numb-expressing cells were also rare in C2C12/3T3-J1 and C2C12-N1/3T3-J1 cocultures (∼1% as compared with C2C12/3T3 and C2C12-N1/3T3; ). The combination of Numb expression and elevated Notch signaling did not increase cell death (Fig. S2 b, available at ). In summary, this indicates that Numb can override the differentiation-inhibiting effects of Notch at lower levels of Notch signaling and promote myogenic differentiation, but, at higher Notch levels, differentiation is blocked, and there is a strong decrease in Numb expression.
We examined the apparent loss of Numb expression at high Notch levels in a C2C12 cell line stably expressing HA-tagged Numb (Numb-HA), which was cocultured with 3T3 or 3T3-J1 cells. The Numb protein level was markedly reduced in 3T3-J1 cocultures but not in 3T3 cocultures or in a mixed lysate of Numb-HA and 3T3 or 3T3-J1 cells cultured separately (). This effect was dependent on Notch signaling, as addition of the GSI L-685,458 to cocultures of Numb-HA and 3T3-J1 cells blocked Numb down-regulation (). In contrast, Numb mRNA levels were not reduced (Fig. S2 c). Transfection of Notch 1 ΔE led to a similar down-regulation of Numb in the cell line stably expressing Numb-HA (). To learn whether down-regulation required the activation of Hes 1 and Hey 1, we transiently expressed Notch 1 ICD, Hes 1, or Hey 1 from adenoviral vectors in the stable Numb-HA () and Nbl-HA () cell lines. Robust down-regulation of Numb was observed by Notch 1 ICD but to a much lesser extent by the canonical Notch downstream genes Hes 1 or Hey 1 ().
To further investigate the Notch-mediated down-regulation of Numb and Numblike, we speculated that the PEST domain may be of importance, as PEST domains have been shown to be involved in protein turnover and have been implicated in proteasome-mediated degradation (). The Numb protein contains two PEST domains, and Numblike contains only one, and, as Numblike in all assays behaved similarly to Numb, we generated a Numblike construct lacking the PEST domain (amino acids 260–273; ) and produced stable cell lines with HA-tagged Numblike () or PEST-deficient Numblike (Nbl-HA and Nbl-HA–ΔPEST, respectively). Like Numb, Nbl-HA protein was reduced by high levels of Notch signaling through coculture with Jagged-1–expressing cells, whereas the mixing of Nbl-HA and Jagged-1 cells immediately before lysis did not reduce Nbl-HA levels (). In contrast, Nbl-HA–ΔPEST protein levels did not change after coculture (). As for Numb, L-685,458 blocked the 3T3-J1 coculture–induced reduction on Nbl-HA protein levels, whereas Nbl-HA–ΔPEST protein levels were unaffected (). No substantial differences in the levels of Nbl-HA or Nbl-HA–ΔPEST mRNA were observed (Fig. S2, d and e), nor was mouse Numb mRNA expression increased by the expression of Notch 1 ICD (Fig. S2 f). Nbl-HA–ΔPEST was equally efficient as Nbl-HA in accelerating C2C12 myogenic differentiation, and 88% of Nbl-HA–ΔPEST-expressing C2C12 cells were MHC positive 3 d after the differentiation switch. Nbl-HA–ΔPEST was more efficient than Nbl-HA in negatively regulating Notch signaling as measured by 12XCSL-luc activation (Fig. S3 a, available at ). This is in keeping with the increased stability of Nbl-HA–ΔPEST and its resistance to degradation by Notch.
To address whether the down-regulation of Nbl-HA was a result of increased protein turnover or reduced synthesis, we performed pulse-chase experiments in the stable cell lines. After a 1-h [S]methionine pulse 16 h after coculture with Jagged-1–expressing cells, there was no difference in the synthesis rate between Nbl-HA and Nbl-HA–ΔPEST (chase = 0 h), but, after 4 h of chase, considerably less Nbl-HA was observed as compared with Nbl-HA–ΔPEST (). This suggests that degradation rather than the synthesis rate is affected in the Notch-mediated down-regulation of Nbl-HA. Addition of the proteasome inhibitor MG132 abrogated the Notch 1 ICD–mediated down-regulation of Nbl-HA and, in fact, increased levels to more than what was observed in the absence of Notch signaling (, compare the first and third lanes). No effect of MG132 was observed with Nbl-HA–ΔPEST (). In conclusion, these experiments argue that the PEST domain is important for the Notch-mediated reduction of Numblike protein levels and that Numblike protein degradation is proteasome dependent.
The negative regulation of Numb and Numblike by Notch demonstrated in this study may play a role in stabilizing the cell fate switch by an asymmetric cell division, which generates two distinct cells: one daughter cell that receives high levels of Numb and, therefore, down-regulates Notch signaling and a second daughter cell with no or low Numb and continued Notch signaling (for review see ). In , we propose that in the latter cell, down-regulation of Numb/Numblike by Notch may assure that Numb levels are kept low and, thus, safeguard the outcome of the cell fate switch resulting from Numb segregation. Such a mechanism would reduce the requirements on the asymmetric segregation machinery to perfectly distribute Numb to only one daughter cell, as small amounts of Numb segregating to the Notch-signaling cell would be eliminated. It may also be particularly important to reduce Numblike protein levels in this cell, as Numblike appears not to be asymmetrically localized ().
To test the model, we wanted to learn whether the level of Notch signaling was inversely correlated with Numb levels in a tumor cell line because a correlation between reduced Numb protein levels and elevated Notch expression is frequently found in breast tumors (; ). The human ovarian carcinoma cell line SKOV-3 was selected for analysis because it contains detectable levels of both Notch 1 ICD and Numb (). Treatment with L-685,458 resulted in reduced levels of Notch 1 ICD and, importantly, elevated levels of endogenous Numb (), suggesting that a reduction of Notch activity leads to enhanced Numb levels. Treatment with L-685,458 also reduced the proliferative rate in the SKOV-3 cells (). To test the converse situation (i.e., the experimental elevation of Notch), we introduced Notch 1 ICD by in ovo electroporation into the developing chick central nervous system. Areas of Notch 1 ICD overexpression showed reduced levels of Numb protein (, arrowhead), whereas areas of low Notch ICD expression contained higher levels of Numb (, arrow). In conclusion, these data support the idea that Notch-mediated down-regulation of Numb can be observed in vivo.
3T3, 293T, and C2C12 cells were grown in DME containing 10% FCS. Transfections were performed using LipofectAMINE Plus reagent (Invitrogen) or FuGene 6 reagent (Roche) according to the manufacturer's instructions. The luciferase assay is described in the supplemental material (available at ). High-titer stocks of adenoviruses for Notch 1 ICD, Hes 1, and Hey 1 were used for infection of Numb-HA or Nbl-HA stable C2C12 cells, and, in all cases, at least 70% of the cells expressed the protein 24 h after infection with minimal cell toxicity.
C2C12 cells grown on 10-cm plates were transfected with 4 μg DNA using LipofectAMINE Plus reagent. For coculture experiments, transfected C2C12 cells were seeded with approximately the same number of 3T3 or 3T3-J1 cells and grown for 24–48 h in DME containing 10% FCS before analysis. Differentiation experiments were performed by plating transfected C2C12 cells at high density onto coverslips precoated with gelatin and laminin in DME containing 10% FCS. Equivalent amounts of 3T3 or 3T3-J1 cells were added several hours later. Culture medium was changed the following day to DME containing 2% horse serum. Cultures were fixed 3 d later in 4% PFA and subjected to immunocytochemistry.
Cultures were washed in PBS, harvested, resuspended in 50–200 μl whole cell extraction buffer (20 mM Hepes, pH 7.8, 420 mM NaCl, 0.5% NP-40, 25% glycerol, 0.2 mM EDTA, 1.5 mM MgCl, 1 mM DTT, 1 mM PMSF, and complete protease inhibitors [Roche]), and incubated on an end-to-end rotator for 30 min at 4°C. The lysate was centrifuged for 30 min at 12,000 rpm, and protein concentration in the supernatant was determined by Bradford analysis (Bio-Rad Laboratories). Western blotting was performed as described in the supplemental material.
RNA was isolated from cultured cells using the RNeasy Mini Kit (QIAGEN). Reverse transcription was performed on 2.5 μg of total RNA using oligo-dT and Superscript II reverse transcriptase (Invitrogen). Real-time PCR was performed in accordance with the manufacturer's instructions using a rapid thermal cycler system (LightCycler; Applied Biosystems). A mastermix containing nucleotides, Taq polymerase, SYBR green, and buffer was mixed with primers and cDNA.
C2C12 cells stably expressing Nbl-HA or Nbl-HA–ΔPEST were cocultured with 3T3-J1 or 3T3 cells. At 16 h of coculture, the growth medium was replaced by serum-free DME without methionine and cysteine containing 30 μCi [S]methionine. After a 1-h pulse, the cells were either harvested (0 h) or washed, and the medium was replaced with normal growth medium for a 4-h chase. Cells were lysed in radioimmunoprecipitation buffer (150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, and 50 mM Tris-HCl, pH 8.0, supplemented with protease inhibitors [Complete; Roche]). Cell lysates were centrifuged 14,000 for 20 min at 4°C for the removal of insoluble material followed by preclearing with Sepharose G beads. HA-tagged Nbl or Nbl-ΔPEST were immunoprecipitated using a monoclonal HA antibody (HA11; BioSite) and captured by Sepharose G beads. The immunoprecipitated proteins were eluted by Laemmli sample buffer and separated by SDS-polyacrylamide gel electrophoresis.
Fig. S1 shows that Numb and Numblike down-regulate Notch signaling. Fig. S2 shows the detection of Numb protein and the role of Notch and Numb for apoptosis. Fig. S3 shows that Numblike-HA–ΔPEST down-regulates Notch-induced reporter gene activation more efficiently than Numblike-HA. Supplemental material provides details about the generation of Numb, Numblike, and Notch DNA constructs as well as generation of the anti-Numb antiserum and the sources of commercial antibodies. Online supplemental material is available at . |
Insoluble deposits of β-amyloid (Aβ), in the form of senile plaques, and of tau, as neurofibrillary tangles, have long been accepted as the primary histopathological markers of Alzheimer's disease (AD). Although initial research focused on the role of Aβ and tau individually, recent evidence, including data demonstrating that amyloid pathology can up-regulate tau pathology (; ), defines a signaling pathway that leads from Aβ through tau (; ; ). Regrettably, the key steps within this pathway remain poorly understood.
A promising new focus of investigation has been the role that nonfibrillar forms of Aβ, and to a lesser extent tau, play in AD. Soluble forms of Aβ are more potent than fibrillar forms at eliciting cellular responses, such as increased apoptosis () and decreased synaptic plasticity (). In fact, studies of transgenic animal models and AD patients have shown that cognitive deficits and synaptic loss correlate with soluble Aβ, rather than senile plaques (; ), suggesting that AD is initiated well before extracellular Aβ deposits are evident.
Neuronal microtubules serve as highways for axonal transport and, by extension, are critically involved in supporting synaptic integrity and neuronal viability. The loss of axonal microtubules is a hallmark of AD, and a longstanding question has been whether their loss or the accumulation of insoluble tau filaments and Aβ plaques causes neurodegeneration. To shed light on this issue, we have used cultured neuronal and nonneuronal cells to model effects of various forms of Aβ on microtubules. Remarkably, we found that brief exposure of cells to submicromolar levels of prefibrillar Aβ42 caused massive and rapid tau-dependent disassembly of microtubules. Similar results were obtained for prefibrillar Aβ40, albeit at much higher concentrations, but microtubules in either tau-expressing or -deficient cells were relatively resistant to fibrillar Aβ. Collectively, these results highlight the most dramatic, rapid, and sensitive link between Aβ and tau described to date, identify microtubules as primary, tau-dependent targets of Aβ, and suggest that nonfibrillar Aβ and tau underlie the detrimental neurodegeneration observed in AD before the accumulation of fibrillar forms in senile plaques and neurofibrillary tangles.
Coexpression of tau-CFP and YFP-tubulin in CV-1 African green monkey kidney cells, which do not express endogenous tau, allowed effects of various forms of Aβ on tau and tubulin distributions to be monitored in live cells by time-lapse fluorescence microscopy. Aβ is known to transition gradually from monomers to oligomers, protofibrils, and finally to highly stable fibrils (). Because we did not observe any consistent differences in behavior between freshly solubilized Aβ42, which is predominantly monomeric, versus Aβ42 enriched in octamers and larger oligomers that are recognized by a specific antibody (), we refer to these forms of Aβ42 collectively as “prefibrillar.” It must be noted, however, that oligomers were readily detectable in freshly solubilized Aβ42 (Fig. S1, available at ). Within 30 min to 3 h after adding 0.1–3 μM prefibrillar Aβ42 to culture media, tau dissociated from microtubules, which completely disassembled soon thereafter ( and Video 1). Prefibrillar Aβ40 was also capable of inducing tau-dependent microtubule disassembly in CV-1 cells, but at a minimum concentration of 3 μM (Video 2). In contrast, microtubules remained intact for >3 h when CV-1 cells expressing YFP-tubulin, but not tau-CFP, were exposed to as much as 3 μM prefibrillar Aβ42 ( and Video 3). Similarly, microtubule integrity was unaffected in cells expressing tau-CFP after >2 h of exposure to as much as 3 μM fibrillar Aβ42 ( and Video 4). Microtubule stability before and after exposure of cells to prefibrillar Aβ42 was also assessed using a biochemical assay that partitions unassembled and polymerized tubulin into Triton X-100–soluble and –insoluble fractions, respectively (). Prefibrillar Aβ42 led to a dramatic redistribution of tubulin into the unassembled fraction within 2 h of treatment (). This result was reinforced by quantitation of micrographs depicting Aβ42-induced microtubule loss in cells that did or did not express tau-CFP (). We thus conclude that tau makes CV-1 cell microtubules hypersensitive to prefibrillar Aβ peptides, particularly Aβ42, but not to fibrillar Aβ42.
Microtubule disassembly was also observed in primary rat cortical neurons after exposure to prefibrillar Aβ42. Discrimination between polymerized and disassembled tubulin was not evident from immunostaining of the treated neurons. Nevertheless, the immunofluorescence images show that before prefibrillar Aβ42 addition, the cells displayed numerous neuritic projections (). After only 30 min of cellular exposure to 1 μM prefibrillar Aβ42, the neurites displayed swollen varicosities, and at later time points the cells appeared to lose the majority of their neuritic projections. Similar cultures were examined by electron microscopy (). Neurites in control cells typically contained densely packed microtubules arranged in parallel. In contrast, neurites in cells treated with prefibrillar Aβ42 contained fewer, less organized microtubules and conspicuous swellings that were filled with membrane-bound organelles and were virtually devoid of microtubules. Similar findings have been reported for primary cortical neurons exposed to 5 μM oligomeric Aβ40 for 3–6 h (; ).
The biochemical assay for unassembled and polymerized tubulin () was used to monitor the effects of prefibrillar and fibrillar Aβ42 on microtubule integrity in cultured hippocampal neurons (). About 90% of the tubulin was polymerized in cells that were not exposed to Aβ42, but only 45% of the tubulin was polymerized after 2 h of exposure to 1 μM prefibrillar Aβ42. Some microtubule loss (65% polymerized) was observed after a comparable exposure to fibrillar Aβ42, but only at the much higher total peptide concentration of 3 μM.
Tau was found to be required for microtubule disassembly in primary hippocampal neurons induced by prefibrillar Aβ42. Neurons were treated with siRNA to reduce tau expression to trace levels ( and Fig. S2, available at ). After 2 h of exposure to 2 μM prefibrillar Aβ42, the level of polymerized tubulin remained nearly unchanged in tau-deficient neurons (∼20% soluble tubulin), whereas the tau-expressing neurons again showed an increase in soluble tubulin (∼60% soluble tubulin). It was not possible to determine how much tau was microtubule bound or soluble in these experiments, because the tau was quantitatively solubilized by the Triton X-100 under conditions in which the polymerized tubulin was resistant to extraction (). Thus, the results for tubulin demonstrate that the endogenous tau in neurons, like transfected tau in CV-1 cells, makes microtubules acutely sensitive to prefibrillar Aβ42.
Treatment of tau-expressing neurons with prefibrillar Aβ42 under conditions that induced microtubule disassembly did not cause increased AD-like tau phosphorylation at any of several sites (). This was found by immunoblotting using phosphorylation-sensitive monoclonal anti-tau antibodies: PHF-1, AT180, and tau-1. Although many additional AD-like phosphorylation sites remain to be examined, these data suggest that conversion of tau to an AD-like phosphorylation state does not underlie the release of tau from microtubules and subsequent microtubule disassembly induced by prefibrillar Aβ42.
The specificity of tau for microtubule loss in cells treated with prefibrillar Aβ42 was shown by expressing GFP-tagged MAP2c in CV-1 cells. MAP2c is a neuron-specific microtubule-associated protein with a microtubule binding domain ∼70% identical to that of tau's (). 3 h of exposure to prefibrillar Aβ42 did not cause any apparent loss of microtubule integrity in cells. A region of tau responsible for conferring sensitivity to Aβ42 was mapped using a combination of tau/MAP2c chimeric proteins and a CFP-tagged tau fragment. Only cells expressing “tau chimera,” a GFP-tagged protein comprising the microtubule binding domain of MAP2c flanked by the N-terminal arm and C-terminal tail of tau ( and Videos 5–7, available at ) responded to the addition of prefibrillar Aβ42. Similar activity was observed when the tau projection domain-CFP, which did not localize on microtubules, was expressed in cells that subsequently were treated with prefibrillar Aβ42 ( and Video 8). The qualitative results shown in and Videos 1 and 4–8 were confirmed by quantitation of fluorescence micrographs for microtubule-containing cells before and after 3 h of exposure to prefibrillar Aβ42 (). The N-terminal half of tau therefore responds to prefibrillar Aβ42 and does not have to target to microtubules to exert its effects. Furthermore, the closely related neuronal microtubule protein, MAP2c, cannot substitute for tau at promoting microtubule disassembly in cells exposed to prefibrillar Aβ42.
The nature of the functional connection between Aβ and tau has been one of the most enduring and intractable mysteries in AD research, and solving this mystery is bound to open potential new avenues of early detection and therapeutic intervention for AD. The results presented here represent the swiftest and most deleterious tau-dependent effects of Aβ that have yet been described. When considered together with recently published studies demonstrating localization of oligomeric Aβ in AD brain at sites distinct from classic amyloid plaques (), a critical role for oligomeric Aβ in memory loss (), and colocalization of Aβ with tau in tangle-bearing AD neurons in vivo (), the present results suggest a mechanism by which Aβ and tau conspire coordinately to compromise neuronal function. Diminished microtubule integrity could hinder essential microtubule functions, such as axonal transport, which is required to maintain synapses, and delivery of exocytotic membranes to the cell surface to repair plasma membrane holes () known to be induced by oligomeric Aβ (). The biochemical steps that underlie tau-dependent microtubule poisoning by prefibrillar Aβ remain unknown, but direct binding of tau to Aβ has been reported () and thus represents one possible step. Nevertheless, we have been unable to demonstrate specific coimmunoprecipitation of tau and Aβ out of tau-expressing cells exposed to prefibrillar Aβ42.
That the combination of prefibrillar Aβ and nonfilamentous tau were able to elicit such a dramatic disruption of microtubules supports the hypothesis that fibrillar forms of tau and Aβ are at least somewhat neuro-protective, because they sequester more dangerous, nonfibrillar forms of Aβ and tau (). The fact that tau is required for Aβ-induced microtubule loss could explain, at least in part, why neurons, the principal tau-expressing cell type, are the cellular targets for destruction in AD. Moreover, the model presented here does not preclude other toxic functions of prefibrillar or fibrillar Aβ or filamentous tau, such as tau-dependent degeneration of cultured neurons induced by fibrillar Aβ40 () or toxicity related to intracellular tau filament accumulation (). Nevertheless, the rapid, tau-dependent destruction of microtubules that we observed to be induced by submicromolar concentrations of prefibrillar Aβ42 suggests that this process is one of the seminal events in AD pathogenesis at the cellular level.
Tau-5, Tau-1, and R1 anti-tau antibodies were gifts from L. Binder (Northwestern University Medical School, Chicago, IL), and PHF-1 anti-tau was provided by P. Davies (Albert Einstein College of Medicine, New York, NY). AT180 anti-tau (Pierce Chemical Co.), DM1A anti-α-tubulin (Sigma-Aldrich), C4 anti-actin (Chemicon), 6E10 anti-Aβ (Signet; recognizes all forms of Aβ40 and Aβ42), fluorescently tagged goat anti–mouse IgG and goat anti–rabbit (Southern Biotechnology Associates, Inc.) and HRP-labeled goat anti–mouse IgG (KPL) were acquired from the indicated commercial sources. I-11 is one of the Glabe laboratory's rabbit polyclonal antibodies against oligomeric Aβ ().
CV-1 (African green monkey kidney) cells were cultured in DME (Invitrogen) supplemented with 10% Cosmic Calf Serum (Hyclone) and 50 μg/ml gentamycin. Cells were transiently transfected using Fugene (Roche) or a nucleofector (Amaxa) with cDNAs for the longest human isoform of tau or the projection domain (amino acids 1–248) of tau linked at their C terminals to ECFP or EYFP (CLONTECH Laboratories, Inc.); with YFP–α-tubulin (CLONTECH Laboratories, Inc.); or with MAP2c or MAP2c/tau chimeras () tagged at their N termini with GFP, which were gifts from S. Halpain (The Scripps Research Institute, La Jolla, CA). For nucleofection, program A-033 and solution V were used according to the manufacturer's instructions. Primary cortical neurons were purchased from Genlantis and cultured according to their guidelines. Primary hippocampal neurons () were grown for at least 8 d before Aβ treatment. The tau siRNA (SMARTpool; Dharmacon) and control scrambled siRNA (Nonspecific duplex II; Dharmacon) were transfected into primary hippocampal neurons by nucleofection using the rat neuron solution (Amaxa) and program G-13 (). Cells were cultured for 4 d after nucleofection and were then exposed to Aβ.
Previously described methods were used to synthesize () and resuspend () Aβ42 and Aβ40. The Aβ was added to cells cultured in serum-free DME to final concentrations from 0.1 to 5 μM. Prefibrillar Aβ was used in the first and second days after resuspension, whereas fibrillar Aβ was used after at least 7 d of stirring.
Live cell imaging and immunofluorescence microscopy were performed as previously described () on an Axiovert 100 (Carl Zeiss MicroImaging, Inc.) equipped with 63× 1.4 NA planapo and 25× 0.8 NA plan-neofluar objectives (Carl Zeiss MicroImaging, Inc.), a CARV spinning disk confocal head (BD Biosciences), an X-Cite 120 illuminator (EXFO Photonic Solutions), a cooled charge-coupled device (Orca ER; Hamamatsu), and OpenLab software (Improvision) for image acquisition and processing. For live cell time-lapse imaging, the cells were maintained on the microscope stage in DME at 37°C in an atmosphere of 95% air and 5% CO. The supplemental time-lapse videos were produced by first using the public domain software, ImageJ to pseudocolor eight-bit grayscale images stacks to eight-bit cyan, green, or yellow image stacks, and then using QuickTime Pro 7 and Keynote 3 (Apple) to produce self-playing videos compressed with the H.264 codec. For electron microscopy, primary cortical and hippocampal neurons were grown on glass coverslips, treated with Aβ42, and fixed in 2.5% glutaraldehyde plus 0.5% tannic acid in 0.1 M cacodylate buffer, pH 7.4. Cells on coverslips were dehydrated and capsule embedded in EPON, and the glass coverslip was removed from the EPON by alternating liquid nitrogen and warm water submersion of the capsule. Sectioned samples were viewed on an electron microscope (JEM 1010; JEOL) at 80 kV, and images were captured using a 16-megapixel cooled charge-coupled device (SAI-12c; Scientific Instruments and Applications, Inc).
Nucleofection was used to express fusion proteins of CFP, GFP, or YFP coupled to tau, the N-terminal tau arm, MAP2c, or MAP2c/tau chimeras in CV-1 cells growing on glass coverslips. Cultures that either were or were not exposed to prefibrillar or fibrillar Aβ42 for 3 h were fixed and permeabilized for 5 min in −20°C methanol and stained for immunofluorescence with anti–α-tubulin followed by TRITC-labeled goat anti–mouse IgG. For each coverslip, six randomly chosen fields of view were photographed separately in both the TRITC channel and the CFP, GFP, or YFP channel using the 25× objective. Typically, 40–50% of the total cells expressed the transfected protein. Next, without knowing the identity of the sample, an observer counted the total cells and microtubule-containing cells in one anti-tubulin field and then counted the total number of transfected cells and microtubule-containing transfected cells in the same field. This process was repeated for the remaining fields of the coverslip, and the results from all six fields, which comprised ∼500 total cells, were merged into a single dataset. Each such experiment was performed in triplicate, and the net results were graphed in and as the mean ± SD of the percentage of microtubule-containing transfected and nontransfected cells for each experimental condition. For , pairwise comparisons were made of transfected versus nontransfected cells at 0 and 3 h of Aβ exposure and of nontransfected cells at 0 versus 3 h of Aβ exposure.
CV-1 cells and primary hippocampal neurons were treated with 1–3 μM prefibrillar or fibrillar Aβ42. Cells were washed briefly with PBS and extracted with PHEM buffer (60 mM Pipes, pH 6.9, 25 mM Hepes, 10 mM EGTA, and 2 mM MgCl) with 10 μM taxol and 0.2% Triton X-100 for 5 min. The buffer was collected and centrifuged for 5 min at maximum speed in a table top centrifuge (model 5415; Eppendorf), and the supernatant was removed and mixed with 1/5 volume of 6× sample buffer for SDS-PAGE to generate a Triton-soluble fraction. An equivalent volume of PHEM buffer and 6× sample buffer was added to the dish, and this sample was added to the pellet from the spin to create a Triton-insoluble fraction. Equal volumes of Triton-soluble and -insoluble fractions, which contained soluble and polymerized tubulin, respectively (), were then analyzed by immunoblotting with anti–α-tubulin. Quantitation of scanned immunoreactive bands was performed using ImageJ.
Fig. S1 shows dot blots of prefibrillar and fibrillar Aβ42 and control proteins. Fig. S2 shows a Western blot of cell lysates from the scrambled siRNA- and tau siRNA–treated hippocampal neurons. Video 1 shows that microtubules disassemble in tau-expressing CV-1 cells exposed to prefibrillar Aβ42. Video 2 shows that microtubules disassemble in tau- expressing CV-1 cells exposed to prefibrillar Aβ40. Video 3 shows that tau is required for prefibrillar Aβ42 to induce microtubule disassembly. Video 4 demonstrates that fibrillar Aβ42 does not induce microtubule disassembly in tau-expressing cells. Video 5 shows that prefibrillar Aβ42 does not induce microtubule disassembly in MAP2c-expressing cells. Video 6 demonstrates that prefibrillar Aβ42 does not induce microtubule disassembly in cells expressing MAP2c chimera. Video 7 shows that prefibrillar Aβ42 induces microtubule disassembly in cells expressing tau chimera. Video 8 demonstrates that prefibrillar Aβ42 induces microtubule disassembly in cells expressing the N-terminal tau arm. Online supplemental material is available at . |
The tumor syndrome von Hippel-Lindau (VHL) disease is caused by heterozygous germline inactivation of the tumor suppressor gene, which resides on chromosome 3p25 (). The cardinal feature of this hereditary cancer syndrome is the development of multiple vascular tumors called hemangioblastomas in the central nervous system and retina combined with clear cell carcinoma of the kidney and pheochromocytoma. VHL disease is an autosomal-dominant disorder, and tumor development in VHL disease is linked to somatic inactivation of the remaining wild-type allele, leading to loss of the wild-type gene product, VHL protein (pVHL). In the kidney, this event not only precipitates the development of clear cell carcinoma but is also associated with the growth of premalignant renal cysts (; ). Restoration of pVHL expression is sufficient to suppress kidney tumor formation by pVHL-defective renal carcinoma cells in vivo, suggesting that tumorigenesis is a direct effect of the loss of both alleles (; ). Despite recent advances in our understanding of pVHL function in tumor formation (; ), the pathogenesis of cystic kidney disease in VHL patients remains unknown.
Recently, the molecular pathogenesis of other cystic kidney diseases has been linked to the monocilia of kidney cells (). Cilia are highly conserved organelles that project from the surfaces of many cells (). The essential structure of renal monocilia consists of nine peripheral microtubule doublets forming the axoneme and surrounded by a membrane lipid bilayer that is continuous with the plasma membrane. The ciliary axoneme emerges from the basal body, a microtubule-based structure that also functions as the spindle-organizing center in mitosis. Cilia are sensory organelles (; ), and it has been demonstrated that renal monocilia are involved in mechanosensation (; ,). The assembly and maintenance of cilia are mediated by intraflagellar transport (IFT), a bidirectional microtubule-based transport system.
In this study, we demonstrate that pVHL localizes to the monocilia of kidney cells and controls ciliogenesis. Furthermore, we show that pVHL is essential for the oriented growth of microtubules toward the cell periphery, a prerequisite for the formation of cilia. Moreover, pVHL interacts with the Par3–Par6–atypical PKC (aPKC) polarity complex, suggesting that pVHL may connect Par3–Par6–aPKC polarity proteins to microtubule capture and ciliogenesis. Our results uncover a novel role for pVHL that links the pathogenesis of premalignant renal cysts in VHL disease with the role of kidney cell monocilia in cystogenesis.
We examined the localization of pVHL in polarized kidney cells using different anti-pVHL antisera. Renal tubular epithelial cells (MDCK clone II) were grown on cell culture inserts for a minimum of 5 d after confluence to allow complete epithelial polarization. We observed specific staining in the cytoplasm of the cells and some nuclear staining, as described previously (; and unpublished data). In addition, strong pVHL staining in monocilia was detected by several pVHL antibodies (). Cilia were identified with an antiacetylated tubulin antibody, which is a marker of the ciliary axoneme (, a and b). pVHL staining was completely blocked by adding an excess of recombinant pVHL peptide, confirming the staining specificity (). The same pVHL localization was observed using anti-pVHL antibody and anti–rabbit AlexaFluor-conjugated antisera but omitting the antiacetylated tubulin antisera, excluding cross-reactivity or bleed-through of the fluorescent label. No immunofluorescence was detected when secondary antibodies were used alone (unpublished data).
Double and triple labeling of native human respiratory epithelial cells revealed that pVHL is also present in motile cilia of respiratory epithelial cells (Fig. S1, available at ). pVHL was detected in the ciliary axoneme (costained with antibodies against acetylated tubulin) and basal bodies (costained with antibodies against γ-tubulin). Immunoelectron microscopy of respiratory epithelial cells revealed that pVHL staining was confined to protein complexes associated with the ciliary microtubules (Fig. S1 c). These data suggested that pVHL may be important for the formation or maintenance of cilia and prompted us to examine the effect of deletion on ciliogenesis.
To address a possible role for pVHL in the formation or maintenance of cilia, we examined ciliogenesis in pVHL-deficient and in wild-type kidney cells. The A498 renal cell carcinoma (RCC) cell line contains a single VHL allele with a frameshift mutation at codon 142, leading to the expression of a defective C-terminally truncated pVHL (). Lentiviral vectors containing V5-tagged human pVHL cDNA or an empty cassette (control) were used to transduce pVHL-deficient A498 RCC cells. The reexpression of pVHL was confirmed by immunoblotting with anti-V5 antibody (). VHL-defective and -positive cells were grown 10 d after confluence to allow epithelial polarization and cilia formation. Under the chosen conditions, VHL-negative cells did not assemble cilia. Even after >14 d after confluence, control cells transduced with empty lentivirus did not form cilia (visualized with antiacetylated tubulin antibodies; , b and c). In contrast, the lentivirally mediated reexpression of pVHL resulted in the formation of intact monocilia at the apical surface of the cells, suggesting that pVHL expression is essential for cilia formation. Ciliogenesis was quantified by blinded counting of cilia in two independent experiments (). Similar to wild-type cells, pVHL-reexpressing cells showed a ciliary localization of pVHL, as demonstrated by the costaining of pVHL with the cilia marker protein acetylated tubulin (). Reexpressed pVHL could also be stained with anti-V5 antibody, confirming specific staining in monocilia (Fig. S2 a, available at ).
Next, we engineered a cell line that expressed VHL under the control of a tetracyclin-dependent promoter. VHL-negative A498 cells were lentivirally transduced to express a tet repressor together with a tet-dependent VHL construct. Incubation of the cells in the presence of doxycycline resulted in pVHL expression already at very low doxycycline concentrations (). Several independent clones of cells were generated, plated at high density, and grown on cell culture inserts in the absence and presence of 500 ng/ml doxycycline for 5 d. As demonstrated in the previous paragraph, VHL-negative cells did not show cilia formation under the conditions chosen (unpublished data). In contrast, doxycyclin-treated cells displayed the formation of monocilia (∼10% of cells). These monocilia again stained positive for pVHL (as demonstrated by stainings with an anti-V5 antibody; ).
To further study a role for pVHL in the formation of cilia, we next searched for short hairpin RNAs (shRNAs) to interfere with VHL expression in renal cells. To test the efficacy of candidate shRNAs, the cDNA of mouse (coding sequence plus 3′ untranslated region [UTR]) was cloned into a bicistronic luciferase vector to fuse this cDNA with the coding sequence of luciferase. In this system, the activity of the luciferase is a quantitative parameter of RNA degradation mediated by cotransfected shRNAs. Coexpressed firefly () luciferase served as a control to normalize for transfection efficiency, expression level, and cell number. Several shRNAs were synthesized based on publicly available prediction programs (; ). These reagents were then tested by sequentially measuring the activities of firefly and luciferases in a 96-well format (Fig. S3 a, available at ). shRNA#3 resulted in an almost 70% knockdown of the reporter construct (Fig. S3 a) and efficient knockdown of the mouse pVHL (Fig. S3 b). We used shRNA#3 cloned into a lentivirus vector to selectively knockdown VHL expression in mouse inner medullary collecting duct 3 (mIMCD3) kidney cells. shRNA expression was monitored by simultaneous coexpression of GFP from the same construct (Fig. S3 c). This strategy allows the identification of cells that effectively express shRNA. Interestingly, this level of knockdown was not associated with a marked change in the growth rate of the cells. Cell growth and formation of a closed monolayer was indistinguishable in cells with and without VHL knockdown, suggesting that low levels of pVHL are sufficient to prevent tumorlike cell growth. The lentivirally mediated expression of shRNA#3 against VHL but not of a control scrambled shRNA inhibited cilia formation, supporting the concept that pVHL controls ciliogenesis. Identical results were obtained with shRNA#2 (unpublished data). Importantly, although cilia formation was greatly attenuated in VHL knockdown cells, ciliogenesis was not entirely abrogated in this assay (some cilia could form at later stages).
Next, we examined the mechanism by which pVHL may affect ciliogenesis. The essential structure of cilia consists of nine peripheral microtubule doublets forming the axoneme and surrounded by a membrane lipid bilayer that is continuous with the plasma membrane (). In microtubule sedimentation experiments, we found that pVHL associates with microtubules (), and immunoprecipitation experiments revealed that pVHL interacts with β-tubulin (). Thus, it appears that pVHL is associated with microtubules. But how could this protein play a role in ciliogenesis? One possible explanation is that pVHL may have a direct effect on microtubule stability, as has been described previously by . We tested this possibility by treating VHL-positive and -negative cells with 20 μM of the microtubule-depolarizing drug nocodazole for 20 min, staining the cells with antiacetylated tubulin antibody, and checking for the integrity of microtubules by fluorescence microscopy. We were particularly interested in an effect on the peripheral microtubule network. However, we could not find an obvious defect in microtubule stability in VHL-negative cells or rescue from microtubule instability in pVHL-reexpressing cells (unpublished data).
Alternatively, pVHL might influence microtubule growth rates or their directionality and organization. To test this possibility, we examined microtubule growth and directionality using end-binding protein 1 (EB1) tagged with GFP in VHL-positive and -negative cells by high speed time-lapse videos. EB1 and its homologue EBP-2 have been shown to decorate the plus ends of growing microtubules (; ) and, thus, can be used to dynamically monitor microtubule formation (). GFP-EB1 decorates the microtubules of cilia (), suggesting that this protein is a suitable reagent to study the microtubule formation required for ciliogenesis. Targeting to cilia is specific to GFP-EB1 and could not be demonstrated with GFP alone (Fig. S2 b). However, because VHL-negative cells do not form cilia, studies addressing the microtubule formation required for ciliogenesis have to be performed in the cytoplasm.
We reasoned that monitoring microtubule formation at the cell cortex may allow us to address the mechanism that leads to the ciliogenesis defect in VHL-negative cells. In fact, this is what we found. Using life confocal microscopy, we observed multiple small dots of fluorescence moving at the cell periphery in both VHL-positive and -negative cells ( and Videos 1 and 2; available at ). The number of forming microtubules at the cell cortex did not differ in VHL-negative cells (). To study the direction and growth rate of microtubules, fluorescent GFP-EB1 dot movements were tracked over 10 s in high speed time-lapse videos using MetaMorph software; microtubule growth rates and direction were determined from these data. Examples of microtubule growth tracks are shown in . To exclude the possibility that differentiation plays a major role in the effect of pVHL on microtubule growth, the experiments were performed in nonpolarized, nonconfluent cells (EB1-GFP–expressing A498 cells rescued with V5.lacZ or V5.VHL). The analyzed region of interest was chosen between the centrosome and the cell membrane. In undifferentiated cells, the centrosome is located near the nucleus, whereas in fully differentiated cells, the centrosome localizes to the apical membrane, where the cilium originates. Only undifferentiated cells with centrosomes close to the nucleus were taken for the analyses, excluding the possibility that polarization, reorientation of centrioles, changes in microtubular polarity, or marked differences in cell cycle progression are responsible for any of the effects observed.
As stated in the previous paragraph, we found no obvious difference in microtubule growth between VHL-positive and -negative cells. However, we noticed that in wild-type cells, the direction of growth of newly formed microtubules at the cell periphery is coordinated toward the outer plasma membrane, whereas in VHL-deficient cells, growth directions appear to be less coordinated (, a and b). To perform a statistical analysis of the directionality of microtubule growth in regions of interest in different cells, the microtubule growth directions were expressed as the deviation from a calculated sum vector of all growth directions in one particular experiment. The summary of five independent experiments revealed a statistically significant difference in deviation from the sum vectors in VHL-negative cells (), demonstrating that VHL deficiency affects the coordinated growth of microtubules. These data suggest that the deficiency in ciliogenesis in VHL-negative cells may be a result of the uncoordinated growth of microtubules.
Members of the family of Ras-related Rho GTPases have emerged as key regulators controlling microtubule–cortex interactions and the coordinated growth of cortical microtubules (; ). Three critical effectors of the GTPase-mediated microtubule control have been identified, including Par6 (), IQGAP1 (; ), and the mammalian homologue of Diapharous (mDia1; ). Interestingly, coimmunoprecipitation experiments revealed a specific interaction of pVHL with Par6 (). pVHL precipitated the Par3–Par6–PKCζ protein complex in human embryonic kidney (HEK) 293T cells (). Immunoprecipitation of endogenous pVHL from the mouse kidney demonstrated that pVHL is in a complex with Par3, Par6, and aPKC in vivo (). Moreover, aPKC (PKCζ) colocalized with pVHL in the monocilia of MDCK cells (). Collectively, these data suggest that Par3– Par6–aPKC and pVHL may operate in the same pathway to regulate the cortical growth of microtubules and the formation of cilia.
Recently, the Par3–Par6–aPKC polarity proteins have been shown to localize to cilia and interact with the anterograde IFT motor kinesin-2, linking polarity proteins with IFT, microtubules, and the formation of cilia (). However, the exact mechanisms of how ciliogenesis, IFT, and polarity proteins may be intertwined remained elusive. We now show that pVHL also localizes to cilia, interacts with Par3–Par6–aPKC, and controls ciliogenesis. Although we failed to detect kinesin-2 components in the complex (unpublished data), our data demonstrating a critical role for pVHL in controlling the growth direction of microtubules suggest a common function of pVHL, polarity proteins, and IFT in controlling microtubular orientation and dynamics. This adds another aspect to an important role for pVHL in regulating the microtubule cytoskeleton in cells. Previous studies showed that pVHL stabilizes microtubules () and influences microtubule dynamics at the periphery of living cells (). We now demonstrate that pVHL is critically involved in regulating the orientation of microtubular growth at the cell periphery. Because pVHL also interacts with Par6, which is a key regulator of microtubule–cortex interactions and the coordinated growth of cortical microtubules (), it is conceivable that pVHL and Par6 are in a common pathway to regulate microtubule orientation. Well-regulated microtubule orientation is a prerequisite for ciliogenesis, so the function of pVHL in controlling cilia formation may be the result of a common function of pVHL, polarity proteins, and IFT in controlling microtubule dynamics in the cell. Future studies will have to address the question of how pVHL, polarity proteins, and IFT regulate microtubule dynamics independently of ciliogenesis.
These data assign a novel role for the tumor suppressor pVHL and have important implications for the understanding of VHL disease. A critical role of pVHL in regulating ciliogenesis has also been documented by others (; ) and could explain why VHL patients can develop polycystic kidney disease. Although much has been learned about the function of pVHL in tumorigenesis at the molecular level, the pathogenesis of premalignant kidney cysts in VHL patients has remained elusive (). Thus, our finding that pVHL plays a critical role in ciliogenesis sheds new light on the pathogenesis of premalignant kidney cysts in VHL patients.
HA- and FLAG-tagged human VHL constructs were provided by S.A. Karumanchi (Harvard Medical School, Boston, MA). Human VHL was cloned into pLenti6.V5/Dest and pLenti4/TO/V5/Dest (Invitrogen) using GATEWAY cloning technology. Mouse VHL (coding sequence and 3′ UTR) was PCR cloned from the full-length clone IRAV-6402536 (Open Biosystems) into a modified GATEWAY pENTR1A vector, and recombination was performed into pcDNA3.1.nV5.Dest to obtain V5.mVHL. GATEWAY vectors and lentiviral constructs were obtained from Invitrogen or were provided by T. Tuschl and M. Landthaler (Rockefeller University, New York, NY), D. Trono (University of Geneva, Geneva, Switzerland), and L. Naldini (University of Torino, Torino, Italy). EB1-GFP was provided by Y. Mimori-Kiyosue (KAN Research Institute, Kyoto, Japan; ). Site-directed mutagenesis was performed using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). All plasmids were verified by automated DNA sequencing. Antibodies were obtained from Sigma-Aldrich (anti-FLAG, antiacetylated tubulin, and anti-PKCζ), Santa Cruz Biotechnology, Inc. (anti-myc, anti-HA, anti-VHL pAb, and anti-Par6), Oncogene Research Products (anti-VHL mAb), Serotec (anti-V5 mAb), Covance (anti-HA pAb), and Roche Biochemicals (anti-HA mAb).
HEK 293T cells were cultured in DME supplemented with 10% FBS. For transfection experiments, cells were grown until 60–80% confluence and transfected with plasmid DNA using a modified calcium phosphate method as described previously ().
Coimmunoprecipitations were performed as described previously (). In brief, HEK 293T cells were transiently transfected and lysed in a 1% Triton X-100 lysis buffer (1% Triton X-100, 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 50 mM NaF, 15 mM NaPO, 2 mM NaVO, and protease inhibitors) for 15 min on ice. After centrifugation at 15,000 for 15 min and ultracentrifugation at 100,000 for 30 min (both at 4°C), cell lysates containing equal amounts of total protein were precleared with protein G–Sepharose and incubated for 1 h at 4°C with the appropriate antibody followed by incubation with 40 μl protein G–Sepharose beads for ∼3 h. The beads were washed extensively with lysis buffer, and bound proteins were resolved by 10% SDS-PAGE. For precipitation of endogenous proteins, mouse kidneys were perfused in situ with ice-cold PBS and homogenized in 1 ml of lysis buffer (20 mM Tris-HCl, pH 7.5, 1% Triton X-100, 25 mM NaF, 12.5 mM NaPO, 0.1 mM EDTA, 50 mM NaCl, 2 mM NaVO, and protease inhibitors). After centrifugation to remove cellular debris, the supernatant was subjected to an ultracentrifugation at 100,000 for 30 min followed by extensive preclearing with protein G–Sepharose. Immunoprecipitation was performed as described previously (). Coprecipitating proteins were detected after 12% SDS-PAGE of the precipitates and immunoblotting.
Freshly isolated mouse trachea was cut into small pieces and immediately transferred to 2.5% PFA in PBS, pH 7.2, for 30 min, washed three times for 10 min in 50 mM ammonium chloride in PBS, and permeabilized with 0.15% Triton X-100 in PBS for 5 min. Incubation with rabbit anti-VHL pAbs for 4 h was followed by washing in PBS four times for 10 min and overnight incubation with the secondary antibodies coupled to 5-nm gold particles (diluted 1:3 in PBS; GE Healthcare). Unbound antibodies were removed by several washings with PBS, and the cells were fixed with glutaraldehyde (2.5%; 50 mM cacodylate buffer, pH 7.2) for 30 min at 4°C. Thereafter, cells were postfixed with 2% OsO for 60 min at 4°C, rinsed with water, dehydrated with ascending alcohol concentrations and propylene oxide, and processed for embedding in Epon. Ultrathin sections were cut with a microtome (Reichert-Jung) and examined with an electron microscope (EM 10A; Carl Zeiss MicroImaging, Inc.).
shRNAs were designed based on the prediction of publicly available prediction programs (), which are summarized in . shRNAs were cloned into the transient micro-RNA expression vector pcDNA6.2-GW/emGFP/miR (Invitrogen), which coexpresses the shRNA surrounded by miR-155–flanking sequences together with emGFP. To monitor the efficiency of shRNA-mediated knockdown, we created a luciferase reporter construct using psicheck2 (Promega) in which the coding sequence and the 3′ UTR of VHL were fused to the coding sequence of luciferase as an artificial 3′ UTR. In addition to luciferase, this construct expresses firefly luciferase for internal control. 50 ng of the reporter plasmid was cotransfected with 50 ng of the respective pcDNA6.2-GW/emGFP/miR shRNA construct into HEK 293T cells in a 96-well format using LipofectAMINE 2000 (Invitrogen) as a transfection reagent. luciferase and firefly luciferase activities were measured by a dual-luciferase reporter assay system (Promega) in a luminometer (Mithras LB940; Berthold) 24 h after transfection. Transfections and measurements were performed in triplicate. Selected hairpins were GATEWAY cloned into pLenti4/V5/TO/Dest for stable lentiviral expression in mIMCD3 cells.
For fast live cell imaging, cells were seeded on custom-built 35-mm glass-bottom dishes and analyzed the next morning at subconfluent stages. Fluorescence images (512 × 512 pixels) were recorded with a confocal slit scanning microscope (LSM5 LIVE; software 4.0; Carl Zeiss MicroImaging, Inc.) with a C-Apo 63× NA 1.4 oil immersion objective (Carl Zeiss MicroImaging, Inc.) on a heating stage at 37°C in nonperfused condition for 30 s. The scanning speed was set to 0.12 s per image, corresponding to a pixel time of 232 μs; a delay of 500 ms was used between each image. The usual pixel size was 0.2 μm in x/y, and the confocal pinhole was set to achieve an optical slice thickness of 0.9 μm. EB1-GFP was excited at 488 nm, and fluorescence emission was collected above 505 nm. For quantification of the growth direction of microtubules, tracking paths were analyzed for 10 s using MetaMorph software (Universal Imaging Corp.). Tracking paths were measured in one to two square fields (256 μm) positioned in the cytosol in a total of 10 independent experiments (VHL negative and positive). As a measure for directed or nondirected movement, the deviation of the individual growth angles from the mean angle was calculated in each square. For better visualization of EB1-GFP dots in the videos, γ adjustment was performed. Statistical analysis was performed for VHL-positive and -negative cells using the two-tailed test.
Videos 1 and 2 show representative time-lapse tracings of EB1-GFP fluorescence in VHL-positive (Video 1) and -negative (Video 2) A498 RCC cells. For the mode of data acquisition, refer to the previous section. Fig. S1 shows that pVHL localizes to cilia in respiratory epithelial cells. Fig. S2 shows that cytosolic GFP does not enter the ciliary compartment. Fig. S3 shows that the knockdown of pVHL inhibits the formation of monocilia. Online supplemental material is available at . |
Ror2 is a member of the Ror family of receptor tyrosine kinases, which possess characteristic structural domains, i.e., extracellular Frizzled-like cysteine-rich domain (CRD), cytoplasmic tyrosine kinase domain, and proline-rich domain (PRD; ). Ror2 is expressed primarily in neural crest–derived cells and mesenchymal cells during mouse embryogenesis (), and plays crucial roles in developmental morphogenesis (; ). -deficient mice exhibit skeletal, genital, and cardiovascular abnormalities (; ), presumably caused by partially disrupted convergent extension (CE) movements during gastrulation. In addition, CAM-1, which is the orthologue of Ror2, has been implicated in cell migration, asymmetric cell division, and axonal extension during embryogenesis (). However, the molecular mechanisms by which Ror2 and/or CAM-1 regulate cell migration during embryogenesis remain poorly defined. Importantly, it has recently been shown that Ror2 acts as an alternative receptor or coreceptor for Wnt5a (; ). In fact, Ror2 mediates Wnt5a signaling by activating the Wnt–JNK pathway and/or inhibiting the β-catenin–TCF pathway.
Wnt proteins constitute a large family of cysteine-rich secreted glycoproteins that play crucial roles in various developmental processes and tissue homeostasis in the adult (). Binding of Wnt proteins to their cognate Frizzled receptors elicits several distinct signaling pathways, including the canonical β-catenin–TCF and noncanonical planar cell polarity (PCP)–CE pathways (; ; ). According to conventional classification, Wnt5a is a representative noncanonical Wnt signaling protein. Indeed, previous studies have shown that loss or gain of Wnt5a function results in dysregulated CE movements in vertebrates (; ).
We show that Ror2 is required for filopodia formation and Wnt5a-induced cell migration. Irrespective of Wnt5a stimulation, ectopic expression of Ror2 can induce filopodia formation by actin reorganization via coupling with the actin- binding protein filamin A (FLNa), and this requires the PRD of Ror2. Intriguingly, disruption of filopodia formation by disrupting the expression of either Ror2 or FLNa inhibits Wnt5a-induced cell migration, indicating the critical role of Ror2-mediated filopodia formation in Wnt5a-induced cell migration. However, Ror2-mediated filopodia formation is not sufficient for Wnt5a-induced cell migration. In fact, Dishevelled proteins (Dvls), which are regulatory cytoplasmic proteins mediating both canonical and noncanonical Wnt signaling (), are also required for Wnt5a-induced cell migration, but not Ror2-mediated filopodia formation.
We first examined the role of Ror2 in Wnt5a-induced cell migration using mouse embryonic fibroblasts (MEFs) from wild-type () and -deficient () mice. Treatment of MEFs with conditioned medium (CM) containing Wnt5a (Wnt5a CM), but not control CM (prepared from culture of mock-transfected L cells) or DME, resulted in a drastic increase in cell motility (). Interestingly, Wnt5a-induced cell migration was impaired in MEFs compared with MEFs. In contrast, Wnt3a could induce considerable cell migration of both and MEFs at comparable levels (), demonstrating that Ror2 is indeed required for Wnt5a-induced cell migration. We next addressed the important question of which domains or functions of Ror2 are required for Wnt5a-induced cell migration. Because MEFs are a somewhat heterogeneous population, we used L cells stably expressing similar amounts of wild-type (WT) or various Ror2 mutants (). Ror2DK lacks intrinsic tyrosine kinase activity, and Ror2ΔCRD, Ror2ΔC, or Ror2Tc lack the extracellular CRD, the cytoplasmic C-terminal region containing PRD, or most of the cytoplasmic region of Ror2, respectively. As shown in , L cells expressing Ror2WT or Ror2DK, but not mock-transfected L cells, exhibited drastically enhanced cell motility after Wnt5a stimulation, indicating that tyrosine kinase activity of Ror2 is dispensable for Wnt5a-induced cell migration. Importantly, no obvious cell migration was observed after Wnt5a stimulation of L cells expressing Ror2ΔCRD, Ror2ΔC, or Ror2Tc (). The results indicate that both the CRD (binding site for Wnt5a) and PRD of Ror2 are required for Wnt5a-induced cell migration.
Surprisingly, we observed that ectopic expression of Ror2 in various cell types (human embryonic kidney [HEK] 293T, MCF7, B16BL6, and L cells) induced remarkable filopodia formation in the absence of Wnt5a stimulation (, , , and not depicted). Ror2 was colocalized with actin at the filopodia (). We next examined which domains within Ror2 are responsible for Ror2-mediated filopodia formation. Ror2WT and a series of Ror2 mutants () were each fused to GFP, and expressed in HEK293T cells, followed by rhodamine-phalloidin staining. HEK293T cells expressing Ror2WT, Ror2DK, or Ror2ΔCRD exhibited drastic formation of the filopodia at the site where Ror2 and actin were colocalized (), indicating that neither tyrosine kinase activity nor the CRD of Ror2 is required for Ror2-mediated filopodia formation. In contrast, cells expressing Ror2Tc or Ror2ΔC, both of which lack the cytoplasmic PRD, failed to form filopodia (). Cell surface expression levels of Ror2WT and the respective Ror2 mutants on transient or stable transfectants are generally comparable as assessed by cell surface biotin-labeling experiment (Fig. S1, available at ; and not depicted). Similar results were obtained when Ror2 and its mutants were expressed in MCF7, B16BL6, or L cells (unpublished data). These results indicate that the cytoplasmic region of Ror2, particularly the PRD, is responsible for Ror2-mediated filopodia formation. Importantly, Ror2-mediated filopodia formation by itself failed to induce cell migration in the absence of Wnt5a (Videos 1 and 2).
Although ectopic expression of Ror2WT, Ror2DK, or Ror2ΔCRD in L cells, lacking the expression of endogenous Ror2, could induce filopodia formation irrespective of Wnt5a (), Wnt5a stimulation of cultured fibroblasts (e.g., NIH3T3 cells) expressing Ror2 endogenously also induced apparent filopodia formation (not depicted). Importantly, MEFs, but not MEFs, clearly exhibited Wnt5a-induced filopodia formation (), indicating that Ror2, indeed, plays a crucial role in Wnt5a-induced filopodia formation. On the other hand, Wnt3a stimulation of and MEFs resulted in the formation of lamellipodia (or pseudopod)-like structures, but not filopodia (Fig. S2, available at ).
We next examined the role of the cytoplasmic region of Ror2 in filopodia formation. We performed yeast two-hybrid screening using the cytoplasmic region of Ror2 as bait to identify potential associating molecules. We identified FLNa, an actin-binding protein that plays important roles in actin cytoskeletal reorganization, cell migration, and various aspects of signal transmission (; ). We thus examined the association of Ror2 with FLNa in HeLa and MCF7 cells that express Ror2 endogenously. Anti-Ror2 immuno-precipitates prepared from either HeLa or MCF7 cells also contained FLNa, indicating that endogenous Ror2 and FLNa, indeed, associate in these cells (). Consistent with the result, Ror2-GFP, but not Ror2ΔC-GFP, expressed ectopically in MCF7 cells was found essentially colocalized with FLNa at the filopodia (), although expression of FLNa was rather concentrated at the roots of filopodia (Video 3, available at ).
FLNa consists of an N-terminal actin-binding domain and 24 tandem repeats (FLN repeats) of ∼96 amino acids each, and dimerizes via the FLN repeat 24 (). FLNa has been shown to interact with various proteins involved in cell motility and signaling via its FLN repeat 15–24 (; ). Thus, we performed pull- down assays using either GST-FLNa/15-19 or GST-FLNa/20-24 bound to glutathione–Sepharose, and found that GST-FLNa/20-24, but not GST-FLNa/15-19, could associate with Ror2WT expressed in HEK293T cells (unpublished data). To determine the cytoplasmic domains within Ror2 that are required for its association with FLNa, we performed a pull-down assay using GST-FLNa/20-24 bound to glutathione–Sepharose. Ror2WT expressed in HEK293T cells was found coprecipitated with GST-FLNa/20-24, whereas Ror2ΔC was not (). This indicates that the cytoplasmic C-terminal region of Ror2, containing the PRD, is required for its association with FLNa in vitro. We also investigated which domains within FLNa are required for its association with Ror2. GST-FLNa/20-21, but not GST-FLNa/21-22, GST-FLNa/22-23, or GST-FLNa/23-24, could associate with Ror2 in a manner similar to GST-FLNa/ 20-24 (), showing that FLN repeat 20-21, in particular repeat 20, is responsible for its association with Ror2. Conversely, FLNaΔ20 lacking FLN repeat 20, expressed in HEK293T cells, was incapable of associating with GST-Ror2 (). These results indicate that the PRD of Ror2 and FLN repeat 20 of FLNa are required for the association of Ror2 and FLNa, respectively.
Next, we used two human melanoma cell lines, M2 and A7, to examine the role of FLNa in Ror2-mediated filopodia formation. M2 lacks FLNa expression, whereas A7 is a derivative of M2 stably transfected with the FLNa cDNA. Expression of Ror2-GFP in A7, but not in M2 cells, resulted in considerable formation of the filopodia at the site where Ror2 and actin were colocalized (). Furthermore, filopodia formation in A7 cells was not induced by expression of Ror2ΔC, which fails to associate with FLNa (). In M2 cells, coexpression of FLNaWT and Ror2 resulted in the considerable formation of filopodia at the site where FLNa and Ror2 colocalize, whereas coexpression of FLNaΔ20 and Ror2 failed to induce filopodia (). The results indicate that Ror2-mediated filopodia formation requires the presence of Ror2-associated FLNa.
Because Ror2 protein was expressed endogenously in M2 and A7 cells (, insert), we evaluated the abilities of M2 or A7 cells to induce cell migration after Wnt5a stimulation. We observed that Wnt5a could induce significant cell migration of A7, but not M2, cells (). The result indicates the critical role of Ror2/FLNa-mediated filopodia formation in Wnt5a-induced cell migration. At present, it remains unclear about the mechanism underlying Ror2/FLNa-mediated filopodia formation. Further study will be required to clarify this issue.
Because previous studies have shown that Dvl proteins are required for noncanonical PCP–CE movements (), we next examined, using siRNA-mediated gene knockdown, whether the Dvl proteins (Dvl2 and Dvl3) might also be required for Ror2-mediated filopodia formation and/or Wnt5a-induced cell migration. To this end, we used L cells where expression of Dvl2 and Dvl3, but not Dvl1, were detectable ( and not depicted). Intriguingly, in L cells expressing Ror2WT, suppression of Dvl2 and Dvl3 expression resulted in the drastic inhibition of Wnt5a-induced cell migration, but not of Ror2-mediated filopodia formation (). This indicates that the Dvl proteins are required for Wnt5a-induced cell migration, but not for Ror2-mediated filopodia formation. It remains unclear whether or not Dvls are involved in Wnt5a–Ror2-mediated signaling pathways, leading to cell migration. It will be important to elucidate how Dvls regulate Wnt5a-induced cell migration in vitro.
Plasmids for mouse Ror2 and its derivative mutants were constructed by subcloning the respective Ror2 cDNAs (Ror2 WT, DK, ΔCRD, ΔC, or Tc cDNAs) into pcDNA, pEGFP, pECFP, and pGEX vectors, respectively (; ; ). Plasmid for YFP-actin was provided by K. Mizuno. Human FLNa and its Δ20 mutant lacking aa 2,151–2,257 were subcloned into the pEYFP- and pMYC-C1 vectors. Plasmids expressing the series of GST-FLNa fusion proteins were constructed by subcloning cDNAs encoding the following FLNa fragments into the pGEX vector, respectively: 15–19 (aa 1,640∼2,150), 20–24 (aa 2,151∼2,647), 20–21 (aa 2,151∼2,330), 21–22 (aa 2,236∼2,424), 22–23 (aa 2,331∼2,539), and 23–24 (aa 2,425∼2,647). The siRNA plasmids, pSUPER-Dvl2 and -Dvl3, were constructed as previously described (). The control siRNA plasmid containing a nonspecific siRNA sequence (GTACCGCACGTCATTCGTA) was used. An anti-Ror2 antibody was prepared as previously described (). Antibodies against Flag (M2; Sigma-Aldrich), Myc (9E10; Santa Cruz Biotechnologies, Inc.), FLNa (FLMN01; Abcam), Dvl2 (H-75; Santa Cruz Biotechnologies, Inc.), Dvl3 (4D3; Santa Cruz Biotechnologies, Inc.), and α-tubulin (Ab-1; Calbiochem) were purchased commercially.
L cells were cultured in DME containing 5% FCS.
and MEFs prepared from mouse embryos (E13.5) were cultured in DME containing 10% FCS. Human melanoma cell lines, M2 and A7, were maintained as previously described (). To establish L cells stably expressing Ror2 or its mutants, transfected cells were selected with G418 (1 mg/ml) and single cell clones were isolated and screened for Ror2 expression by immunoblot analysis. Cells were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instruction. For electroporation, cells were suspended in 400 μl of OPTI-MEM medium (Invitrogen) containing the respective plasmids and electroporated at 250 V/975 μF, using Gene Pulser II (Bio-Rad Laboratories, Inc.). Wnt5a, Wnt3a, and control (neo) CM were harvested from confluent monolayers of Wnt5a/L, Wnt3a/L, and neo/L cells () that had been cultured for 72 h in serum-free DME, respectively.
Cells were lysed in lysis buffer (50 mM Hepes, pH 7.4, 150 mM NaCl, 0.5% Nonidet P-40, 5% glycerol, 1 mM MgCl, 1 mM MnCl, 20 mM NaF, 1 mM NaVO, 1 mM dithiothreitol, 0.25 mM PMSF, and 10 μg/ml leupeptin). Whole-cell lysates were subjected to analyses by immunoprecipitation and immunoblotting as previously described (). Whole-cell lysates prepared from HEK293T cells expressing Ror2-Flag (WT or ΔC) or Myc-FLNa (WT or Δ20) were incubated with purified-GST or -GST fusion proteins bound to glutathione–Sepharose beads. Pelleted beads were subjected to SDS-PAGE, followed by immunoblot analyses.
Cells cultured on coverslips were fixed and stained with the respective antibodies, as previously described (). To visualize F-actin, fixed cells were stained with rhodamine-phalloidin (Invitrogen). Fluorescent images were obtained at room temperature using an inverted microscope (Axiovert 200M) equipped with a laser scanning confocal imaging system (LSM510) and a 63×/NA 1.4 oil immersion objective lens (Plane Apochromat; all from Carl Zeiss MicroImaging, Inc.). Images were processed using Photoshop CS (Adobe).
MEFs (2 × 10 cells) or L cells (6 × 10 cells) suspended in 100 μl DME were loaded into the upper well of the Transwell chamber (8-μm pore size; Costar) that was precoated on both sides with 0.1% gelatin for 2 h at 37°C. The lower well was filled with 600 μl of CM or DME. For the migration assay using M2 and A7 cells, the underside of a Transwell chamber was precoated with 0.1% gelatin, and 3 × 10 cells in 100 μl DME containing 0.1% BSA were loaded into the upper well. The lower well was filled with 600 μl of CM containing 0.1% BSA. After incubation for 7 h for MEFs and L cells or 3 h for M2 and A7 cells, the membrane was fixed in 3.7% formaldehyde. Nonmigrating cells on the top of the membrane were removed by wiping and rinsing, and migrating cells on the lower face of the membrane were stained with DAPI and counted at room temperature under an Axiovert 200M inverted microscope equipped with a LSM510 laser scanning confocal imaging system and a Plan-Neofluar NA 0.3 10× objective lens (Carl Zeiss MicroImaging, Inc.).
Fig. S1 shows surface expression of WT and mutant Ror2 proteins on HEK293T cells. Fig.
MEFs. Videos 1 and 2 show the time-lapse fluorescence of B16F1 cells expressing YFP-actin alone (video 1) or YFP-actin and Ror2-CFP (Video 2). Video 3 shows the time-lapse fluorescence of M2 cells coexpressing Ror2-CFP and YFP-FLNa. Online supplemental material is available at . |
During development, melanocyte precursors migrate from the neural crest toward the epidermis, where they are arrested when contacting keratinocytes. Differentiated human melanocytes remain strictly localized at the basement membrane and cannot survive within the upper epidermal layers unless transformed in nevi or melanomas. Keratinocytes control the normal growth of melanocytes to maintain a lifelong stable ratio, and they regulate the expression of cell surface molecules (). They also produce growth factors and cytokines that may act as paracrine factors to regulate the phenotype of melanocytes, including interleukin-1β (IL-1β), TNF-α, stem cell factor (SCF), and EGF ().
CCN3 (nephroblastoma overexpressed) is a matricellular protein that shares, with five other family members, structural modules of insulin-like growth factor–binding domains, von Willebrand factor type C repeats, thrombospondin type 1 repeats, and secreted regulatory factors containing cysteine knot motifs for dimerization (). Depending on the cell type and tissue context, matricellular proteins participate in diverse functions, such as cell adhesion, proliferation, differentiation, and survival ().
In a search for the molecular players involved in cross talk between melanocytes and keratinocytes, we compared gene expression profiles of melanocytes grown in monoculture with melanocytes grown under the control of keratinocytes. We found CCN3 protein being up-regulated in melanocytes after coculture with keratinocytes. The CCN3 protein was secreted into the culture medium and affected two fundamental features of melanocytic physiology: it inhibited the proliferation of melanocytes and was required for the 3D organization of the melanocyte network on the basal membrane of human skin equivalent. Furthermore, CCN3 stimulated the adhesion of melanocytes to basal membrane collagen IV but not to dermal collagen I, as confirmed by the siRNA-mediated down-regulation of CCN3 and two-photon (2P) microscopy. We then identified discoidin domain receptor 1 (DDR1), a receptor tyrosine kinase, as being responsible for the adhesive properties of CCN3.
To investigate how keratinocytes change the phenotype of melanocytes, we cocultured pure populations of melanocytes and keratinocytes and conducted global gene expression analyses of monoculture- versus coculture-grown melanocytes. The gene was found to be consistently up-regulated in cocultured melanocytes (). Keratinocytes in culture or human skin did not express CCN3, whereas melanocytes constitutively expressed it at low levels. After melanocytes were cultured with keratinocytes, CCN3 was strongly expressed in the cytoplasm of melanocytes and was secreted into the culture medium (). As demonstrated by immunofluorescent staining of normal human skin, CCN3 was expressed at the basal layer of the epidermis, where melanocytes are positioned (Fig. S1, available at ). Keratinocyte-derived culture supernatants stimulated CCN3 expression, but the up-regulation of CCN3 by melanocytes was strongest when keratinocytes were in direct cell–cell contact (, first to third lanes). Coculture with control epithelial cells had no effect on CCN3 secretion by melanocytes (, fourth to seventh lanes).
Keratinocytes commonly express several growth factors and cytokines that may change the phenotype of melanocytes. TNF-α and IL-1β stimulated the expression of CCN3 (). IL-1β was constitutively expressed by keratinocytes (unpublished data), and TNF-α was found to be undetectable in the culture supernatants derived from keratinocytes without stimulation (). Therefore, we chose to further investigate the effect with IL-1β. Melanocytes began to increase CCN3 secretion 8 h after stimulation by IL-1β, and it continued for 48 h (). To investigate whether IL-1β produced by keratinocytes contributes to the induction and secretion of CCN3, we performed immunodepletion of IL-1β in coculture medium using neutralizing antibodies (). The depletion of IL-1β decreased CCN3 in cocultures. However, this inhibiting effect was only partial (20% reduction), suggesting that other keratinocyte-derived factors are involved in the mechanism of CCN3 production by melanocytes.
Because CCN3 has antiproliferative activity in fibroblastic, glioma, and Ewing's sarcoma cells (; Fu et al., 2004; ), we sought to determine whether CCN3 inhibits the growth of melanocytes. A lentiviral vector (si--C) designed to knockdown CCN3 in melanocytes demonstrated a considerable decrease in protein production compared with an empty vector (H1UG-1), a one-pair mismatch (si--Cm), and two related siRNA (si--A and -B) vectors in conditioned media () and lysates (not depicted). Melanocytes transduced with si--C showed increased growth rates compared with cells transduced with control vectors (). The difference in growth rates between CCN3 knockdown (si--C) and control cells (si--Cm) was significant (P = 0.0095) on day 4 after coculture, when the medium from si--Cm contained more CCN3 than si--C (Fig. S2 A, available at ). They also showed a notable decrease in attachment to collagen type IV, which is present in the basement membrane ( C and S2 B) but not to type I collagen present in the dermis () or laminin, which is another component of the basement membrane (Fig. S2 C, left). This result suggested that CCN3 modulates collagen type IV adhesion of melanocytes.
The melanocytes in mouse skin are localized in the dermis, suggesting that mouse melanocytes have different regulatory mechanisms from humans. The behavior of melanocytes in the skin reconstructs resembles that in vivo (; ). We further tested the proliferation and localization of melanocytes within human skin reconstructs (). To identify melanocytes in sections of skin reconstructs and to determine their localization by immunohistochemistry, we used a melanocyte-specific marker, HMB-45 (, left). Staining for collagen IV demonstrated the formation of an intact basement membrane within the 14-d skin formation period (, right). Knockdown of CCN3 in melanocytes increased their numbers and changed their localization by migrating either upward to the suprabasal layers of the epidermis or downward into the dermis. To better determine the cell number and localization of melanocytes within the basement membrane zone, we performed 2P microscopy on the skin reconstructs. The lentiviral vectors contained GFP as a marker for melanocytes. Second harmonic generation (SHG) signals served as a sensitive indicator of collagen I to separate the epithelial layer from underlying stroma (). A combination of 2P-excited GFP and SHG imaging of skin reconstructs showed that the knockdown of CCN3 in melanocytes not only increased their numbers within the reconstructs but also shifted their localization (). Individual melanocytes separated from the basement membrane, migrated into the epidermis, and invaded the dermis (). In contrast, melanocytes transduced with control vector remained confined to the basement membrane zone.
Melanocytes transduced with an adenoviral vector to overexpress the CCN3 protein () were growth inhibited (). Although the CCN family members CCN1 and CCN2 have been shown to induce apoptosis (; ), apoptosis as measured by caspase 3 levels was not changed in CCN3-overexpressing melanocytes (). There was no increase in the sub-G1 population of melanocytes overexpressing CCN3 either as tested by FACS (unpublished data). Moreover, the cyclin kinase inhibitor p21 was up-regulated (), suggesting that the decrease in proliferation is caused by growth inhibition and not cell death. Similar inhibition was seen when recombinant CCN3 () protein was added to culture medium (). Melanocytes overexpressing CCN3 increased attachment to collagen IV () but not to collagen I () or laminin (Fig. S2 C, right). Melanocytes overexpressing CCN3 were firmly localized to the basement membrane zone of human skin reconstructs. Sections of the reconstructs showed tightly aligned melanocytes at the epidermal/dermal interface as assessed by melanocyte marker HMB-45 and by collagen IV staining (). 2P microscopy of skin reconstructs confirmed that the dendrites of control melanocytes remained separate from the basement membrane; however, those overexpressing CCN3 were localized at the border between the epidermis and dermis (). A limitation of the adenoviral vector system is that it is not integrated into the host genomic DNA, and its gene expression is extinguished through divisions of host cells. Therefore, GFP-positive cells were not observed by 2P imaging as frequently as those using a lentiviral vector system. The number of melanocytes identified by HMB-45 staining decreased when they overexpressed CCN3 (unpublished data). These data demonstrate that melanocyte-derived CCN3 inhibits growth to maintain normal homeostasis and secures the attachment of melanocytes to the basement membrane.
Because matricellular proteins themselves have only weakly adhesive functions (), we compared the expression profile of melanocytes overexpressing CCN3 with that of control cells by microarray analysis. DDR1 was up-regulated, as verified by Western blotting in two melanocyte cultures when CCN3 was overexpressed (). When melanocytes were transduced with siRNA against CCN3 (si--C), DDR1 expression was down-regulated. DDR1 is a tyrosine kinase receptor for several collagens, particularly collagen IV (). Down-modulation of DDR1 with an siRNA, as confirmed by Western blotting (), showed decreased adhesion to collagen IV () similar to those from siRNA . Consistently, adhesion to collagen I was not up-regulated (). 2P imaging of skin reconstructs showed that the knockdown of DDR1 in melanocytes shifted their localization; the proportion of cells at the basement membrane zone to total cell number was particularly decreased compared with the control (). To test whether DDR1 is essential for the regulation of melanocyte adhesion to basement membranes by CCN3, we overexpressed CCN3 in melanocytes transduced with si--C. CCN3 overexpression in melanocytes transduced with si--C recovered neither adhesion to collagen type IV nor normal localization in skin reconstructs (Fig. S3, available at ), confirming that up-regulation of the adhesion of melanocytes to the basement membrane by CCN3 is mediated through the collagen IV receptor DDR1. Knockdown of DDR1 in melanocytes did not increase their number in skin reconstructs (). Our results suggest that CCN3 regulates melanocyte growth through a mechanism that is distinct from adhesion. It is possible that CCN3 up-regulates DDR1 expression through the activation of p53 because p21 is a downstream target of p53, was up-regulated in CCN3-treated cells, and DDR1 is also one of the transcriptional targets of this tumor suppressor ().
Melanocytes appear to have a contingency mechanism that is essential for their survival and secures continuous attachment to the basement membrane of the skin. The primary mechanism for attachment was thought to be through integrins (), of which the laminin-binding integrin α6β1 was the main candidate (; ). Down-modulation of α6 integrin in melanocytes does not alter their localization in skin reconstructs (unpublished data), suggesting that α6 integrin is not essential for anchorage under homeostatic conditions. Because expression of the α6-integrin subunit is down-modulated by ultraviolet irradiation (), the melanocytes must have developed alternative mechanisms to maintain localization at the basement membrane. Our study indicates that CCN3 production by melanocytes after their contact with keratinocytes up-regulates the DDR1 adhesion receptor for collagen IV and influences melanocyte localization, contributing to the homeostasis in skin. When the proinflammatory cytokine IL-1β produced by keratinocytes up-regulates CCN3 in melanocytes, their normal localization in the skin is secured through DDR1-mediated adhesion to collagen type IV. Knockdown of DDR1 did not affect melanocyte proliferation in skin reconstructs, suggesting that there must be other downstream effectors of CCN3 that are responsible for the growth inhibitory effect of CCN3. Such a mechanism remains to be elucidated. CCN3 can bind to αvβ3 (), a multiligand-binding integrin, but the β3 subunit is not expressed by normal melanocytes (; ). CCN3 can also bind to Notch (); however, Notch signaling is not activated in melanocytes (unpublished data). In other cell types, CCN3 can be up-regulated by basic FGF (bFGF; ), which stimulates melanocyte growth but does not modulate adhesion. Growth inhibition and basement membrane localization conferred by CCN3 are important, if not essential, functions for maintaining melanocyte homeostasis in the normal skin. Our findings suggest that CCN3 dysregulation may be the first step toward disrupting the normal balance between melanocytes and keratinocytes. Therefore, clarifying CCN3's role in melanocyte pathogenesis and melanoma is an important goal for further work.
Normal human keratinocytes, melanocytes, and fibroblasts were isolated from neonatal human foreskins. Keratinocytes were cultured in EpiLife medium supplemented with human keratinocyte growth supplement (Cascade Biologics, Inc.). Melanocytes were cultured in MCDB153 (Sigma-Aldrich) supplemented with 2% FBS, 10% chelated FBS, 2 mM glutamine, 20 pM cholera toxin (Sigma-Aldrich), 1.5 nM recombinant human bFGF (Sigma-Aldrich), 100 nM recombinant human endothelin-3 (Peninsula Labs), and 10 ng/ml recombinant human SCF (Sigma-Aldrich). Fibroblasts were cultured in DME with 10% FBS. For cocultures, melanocytes were cultured with keratinocytes at a 1:5 ratio in EpiLife medium for 2 d. As a control, monocultured samples (melanocytes and keratinocytes at a 1:5 ratio) were cultured separately for 2 d. For gene expression comparison of monocultures with cocultures, melanocytes transduced with GFP were cultured with keratinocytes and isolated by FACS. As required, cells were counted or lysed for protein and RNA extraction.
Cocultures were treated with 2 μg/ml human IL-1β–specific goat IgG (R&D Systems) to neutralize IL-1β. Control cultures were treated with 2 μg/ml nonspecific goat IgG.
The human cDNA was amplified using the primers 5′-GACGGGTACCTGAGCATGCAGAGTGTG-3′ and 5′-CTTGTCTAGAAGGTTACATTTTCCCTCTGG-3′ and were subcloned into pAdTrack-CMV at KpnI–XbaI sites. Recombination between pAdTrack-CMV–CCN3 and pAdEasy was performed in to generate the CCN3 adenoviral vector (). The mammalian expression vector H1UG-1 derived from the FG12 lentiviral vector () was used to produce lentiviral RNAi vector. DNA sequences encoding siRNA targeting and mRNA were cloned into BamH1 and XhoI sites under control of the HuH1 promoter. The original DNA sequences encoding the siRNAs targeting mRNA were as follows: si--A (5′-GCTGCAAATTCCAGTGCACCT-3′), si--B (5′-GTTGAGGTGCCTGGAGAGT-3′), and si--C (5′-GTGTCAACTGCATTGAACA-3′). One (si--C) out of three siRNA vectors displayed high efficiency CCN3 knockdown in melanocytes and was selected for the creation of a mutated (indicated in bold) control siRNA sequence (si--Cm, 5′-GTGTCAACTTCATTGAACA-3′). The DNA sequences encoding the siRNAs targeting mRNA were si--B (5′-GAGGAGCTGACGGTTCACC-3′) and si--C (5′-GATCTGGTTAGTCTTGATT-3′). The lentivirus was produced by cotransfection of human embryonic kidney 293T cells with four plasmids: a packaging defective helper construct, a Rev plasmid, a plasmid coding for a heterologous envelope protein, and the H1UG-1 vector construct harboring the selected siRNA sequence.
Total RNA was isolated using the total RNeasy kit (QIAGEN). Human Genome U133A arrays were used for mRNA expression profiling according to the manufacturer's instructions (Affymetrix, Inc.). A laser scanner (GeneArray; Hewlett-Packard) was used to analyze gene chips, and the expression levels were calculated using Microarray Suite software (Affymetrix, Inc.). Gene expression values were normalized to the mean value of all genes in each experiment.
Total RNA was reverse transcribed into first-strand cDNA for quantitative RT-PCR. The gene specific primers were designed as CCN3 (5′-GAACCGTCAATGTGAGATGC-3′ and 5′-ACAGAACCTGGGCTTGTAGG-3′) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH; 5′-ATGGAAATCCCATCACCATCTT-3′ and 5′-CGCCCCACTTGATTTTGG-3′). ABsolute QPCR SYBR Green Mixes (ABgene) were used with 1 ng/ml cDNA and with 70 nM of primers for the evaluation of GAPDH and CCN3 expression. A negative control without the cDNA template was run with each assay. Amplifications were performed in an ABI Prism 7000 Sequence Detection System (Applied Biosystems). Thermal cycler conditions were 95°C for 15 min and 40 cycles of 15 s at 95°C followed by 1 min at 60°C. All experiments were performed in triplicate, and a mean value was used for the determination of mRNA levels.
values were exported to Excel (Microsoft) for analysis.
method according to the manufacturer (Perkin-Elmer). All samples were normalized to the relative levels of GAPDH.
The CCN3 coding sequence was cloned into the pGEX4T1 vector. Expression of the recombinant GST-CCN3 protein was induced by adding 0.1 mM IPTG to the bacteria cultures when they reached 0.7–0.9 OD at 600 nm. After centrifugation, pellets were resuspended in 50 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl, and proteinase inhibitors (complete cocktail [Roche], 200 mM PMSF, 10 mM TLCK, 200 mM benzamidine, and 10 mM TPCK), and 300 μg/ml lysozyme was added. Lysis was performed for 20 min on ice. Triton X-100 was then added to 1%, and lysates were sonicated on ice. After centrifugation, supernatants were incubated with glutathione–Sepharose beads (GE Healthcare) in PBS for 1 h at 4°C on a rotating wheel. For GST-CCN3, PBS was complemented with 5% fat-free milk and 0.5 mM ATP. Beads were then washed several times with PBS-proteinase inhibitors. Recombinant proteins were recovered by three elutions of 1 h on ice with 20 mM glutathione, 100 mM Tris, pH 8.0, and 120 mM NaCl. Fractions were pooled, dialyzed overnight at 4°C against 10 mM NHHCO, and lyophilized. Quantification was performed by SDS-PAGE and Coomassie blue coloration of the gel.
For Western analyses to detect CCN3 or DDR1 expression, cells were washed with PBS and harvested in radioimmunoprecipitation buffer. To detect secreted CCN3, proteins from conditioned medium corresponding to 2 × 10 melanocytes were collected with heparin–Sepharose beads after overnight incubation at 4°C as described previously (). Samples were separated on 4–12% Bis-Tris gels, transferred to polyvinylidene difluoride membranes, and probed with K19M, antifibronectin (Transduction Laboratories), anti–β-actin (Sigma-Aldrich), anti-p21 (BD Biosciences), anticaspase 3 (Cell Signaling Technology), or anti-DDR1 antibodies (C-20; Santa Cruz Biotechnology, Inc.). To detect the signal, HRP-conjugated secondary antibody was added followed by ECL (GE Healthcare). The immunoblot images were acquired with a scanner (Perfection 636U; Epson) and quantitated using Image Beta 4.02 software (Scion Corp.). For immunofluorescence, cells were fixed, permeabilized, and incubated sequentially with mouse anti-TYRP1 (Signet), FITC-conjugated goat anti–mouse IgG, K19M anti-CCN3 rabbit serum, and AlexaFluor594-conjugated goat anti–rabbit IgG. Immunofluorescent staining was also performed on serial sections of human foreskins with K19M and anti-pmel 17/HMB-45 (DakoCytomation) as primary antibodies. For immunohistochemistry, sections were stained with mouse anti–human type IV collagen (Calbiochem) or HMB-45 using standard procedures. Biotin-labeled anti–mouse secondary antibodies were applied, and signal was detected using the ABC kit (Vector Laboratories). After adding 3-amino-9-ethyl carbazole, samples were counterstained with Mayer's hematoxylin (Sigma-Aldrich). The immunostaining images were taken by an upright microscope (E600; Nikon) with 20/40× objectives. A digital camera (SPOT RT Slider; Diagnostic Instruments) was used to acquire the pictures. Photoshop 6.0 software (Adobe) was used for contrast and brightness adjustment.
For cell growth experiments, melanocytes were plated in quadruplicates in 24-well plates at a density of 5.66 × 10 cells/cm. Cell growth was monitored by counting cells in five random high power fields. For [H]thymidine uptake assays, recombinant adenovirus-transduced melanocytes were seeded in quadruplicates in 96-well plates at 5,000 cells/well 48 h after transduction and cultured in 200 μl of medium. After 24 h, cells were pulsed with 1 μCi [H]methyl thymidine and harvested after 12 h for counting.
For adhesion assays, melanocytes were suspended in serum-free MCDB153 (6 × 10 cells/ml) and transferred in triplicate to either CytoMatrix cell adhesion strips coated with human collagen type IV, human laminin (Chemicon), or 96-well plates coated with 10 μg/ml bovine type I collagen (Organogenesis) and incubated for 90 min at 37°C. After washing to remove unattached cells, the attached cells were stained with 0.2% crystal violet. The cell-bound stain was solubilized, and the optical density (570 nm) was determined.
Reconstructions of normal human skin were prepared as previously reported (). 3 ml of fibroblast-containing bovine type I collagen (7.5 × 10 cells/ml) was added to each insert of tissue culture trays (Organogenesis) and allowed to constrict in DME with 10% FBS for 7 d at 37°C. For epidermal reconstruction, keratinocytes were mixed with melanocytes at a ratio of 5:1 in keratinocyte serum-free medium (Invitrogen) containing 2% dialyzed FCS, 60 μg/ml bovine pituitary extract (Invitrogen), 4.5 ng/ml bFGF, 100 nM human endothelin-3, and 10 ng/ml human SCF. A total of 5 × 10 cells were seeded on each contracted collagen gel. Cultures were kept submerged in medium containing 1 ng/ml EGF for 2 d, 0.2 ng/ml EGF for another 2 d, and were raised to the air–liquid interface via feeding from below with high calcium (2.4 mM) medium. After 14 d, skin reconstructs were either directly analyzed using a 2P microscope or fixed with 4% PFA and embedded in paraffin for subsequent sectioning and staining.
2P imaging was performed with an upright multiphoton microscope (Ultima; Prairie Technologies) attached to a microscope (BX-61; Olympus) fitted with 20/40× water immersion objectives (Olympus). This arrangement was combined with a diode pumped wideband mode-locked titanium-sapphire femtosecond laser (Chameleon; Coherent). Components of the extracellular matrix (e.g., collagen) were detected by SHG signals (). In all of the experiments, the samples were exposed to a wavelength of 920 nm. The wavelengths emitted by the GFP (515 nm) and the extracellular matrix (460 nm) were distinguished using a filter cube (Dichroic). Z stacks of a series of x-y planes at a resolution of 2 pixels/μm in step size 2 μm were captured using Photonics photomultiplier tubes (R3896; Hamamatsu) with amplifiers and View acquisition software (Prairie Technologies). Volocity software (Improvision) was used to generate x-z sections and to render 3D reconstructions of the skin. To assess localization of melanocytes in skin reconstructs, five fields (×200) were randomly selected in each reconstruct and scored by counting GFP-positive cells on x-y planes at 24-μm intervals. Distribution (percentage) = number of melanocytes on each plane/total number of melanocytes on all planes × 100. All experiments were performed three times using melanocytes derived from three different donors. The data were analyzed by test (two-tailed distribution and two-sample unequal variance) and expressed as the mean ± SD. Each figure shows one representative experiment.
Fig. S1 shows CCN3 expression in human skin. Fig. S2 shows that adhesion on laminin is not affected by CCN3 modulation. Fig. S3 shows that the overexpression of CCN3 does not restore the localization of melanocytes transduced with si-DDR1. Online supplemental material is available at . |
Clathrin-coated vesicles (CCVs) are important transport intermediates in all eukaryotic cells. Their coats consist of two major components: clathrin, which provides a stabilizing scaffold, and heterotetrameric adaptor protein (AP) complexes, which attach the clathrin to the membrane and select the vesicle cargo (). There are at least two AP complexes associated with CCVs: AP-1, which functions in transport between the TGN and endosomes (although there is some question about directionality), and AP-2, which functions in clathrin-mediated endocytosis (; ; ).
Although it was originally assumed that clathrin and AP complexes were all that was necessary to make a CCV, it is now apparent that CCV formation is much more complex. New components of the machinery are continually being discovered, including various “alternative adaptors” and proteins that contribute to other stages of the CCV cycle (). Thus, although the most abundant components of CCVs are known, many questions remain regarding the initiation of vesicle formation, cargo selection, budding, scission, uncoating, and transport. Clearly, a complete knowledge of the protein composition of CCVs would greatly advance our understanding of clathrin-mediated trafficking.
In recent years, organelle proteomics has emerged as a powerful tool to guide cell biological research (; ; ), and two studies have so far been published on the CCV proteome. and prepared CCV-enriched fractions from rat brain and liver, respectively, and identified proteins by tandem mass spectrometry (MS/MS). In both studies, an impressive degree of CCV enrichment was achieved (73–89% vesicle homogeneity, as judged by electron microscopy), and numerous proteins were identified. However, neither study could distinguish which of the identified proteins were true constituents of CCVs and which were copurifying contaminants.
Because it is impossible to prepare completely pure CCVs, the challenge becomes finding unbiased criteria that allow one to identify genuine CCV components. With such criteria at hand, the purity of the preparation is no longer critical. Here, we introduce a novel criterion: the dependence of a protein on clathrin to be present in a CCV fraction. By pairing cell biological tools with state-of-the-art quantitative proteomics techniques, we develop a strategy for identifying bona fide CCV proteins from human tissue culture cells.
A CCV-enriched fraction was prepared from HeLa cells using an established protocol (), and the same procedure was performed on cells that had been depleted of clathrin heavy chain (CHC) by siRNA knockdown. Such cells have been shown to contain no detectable clathrin-coated budding profiles or vesicles (), so the prediction is that the “mock CCV” fraction from these cells should be devoid of CCVs but should still contain proteins that contaminate CCV preparations from control cells. Thus, a comparison between the two fractions should reveal genuine CCV proteins as those present only (or mainly) in the control CCV fraction and contaminants as those equally present in both fractions ().
To validate the approach, control and mock CCV fractions were analyzed by Western blotting (). All of the known CCV components that we tested, including coat proteins (e.g., CHC, AP-1 and AP-2 subunits, and epsinR) and cargo proteins (e.g., cation-independent mannose 6-phosphate receptor [CIMPR] and transferrin receptor), were enriched in our CCV preparations compared with whole cell homogenate (the homogenate lanes contain approximately four times as much protein as the CCV lanes) and were absent or depleted from mock CCVs. In contrast, known contaminants (e.g., elongation factor 2) were equally present in control and mock CCVs. Thus, our comparative proteomics approach should allow us to distinguish bona fide CCV constituents from irrelevant proteins that copurify. Interestingly, we consistently observed differences in the degree of enrichment and depletion of AP-1 and AP-2 subunits. AP-1 subunits were more highly enriched in CCVs relative to cell homogenate, and they were also more strongly depleted from the mock CCVs, suggesting that our preparation favors intracellular, nonendocytic CCVs.
As a first step toward analyzing the CCV proteome, control and mock CCV fractions were separated by 1D SDS-PAGE and stained with Coomassie blue (). The two fractions have similar protein patterns, indicating that CCV-enriched preparations from control cells are heavily contaminated with non-CCV material. However, four high molecular weight proteins appear to be depleted from mock CCVs (, arrows). These proteins were analyzed by MALDI-TOF (matrix-assisted laser desorption/ionization time-of-flight) and identified as CHC (which comigrates with glycogen debranching enzyme), CIMPR, the β1 subunit of AP-1, and transferrin receptor. Thus, the 1D SDS-PAGE data confirm the results of the Western blot, but they offer insufficient resolution to analyze more minor constituents.
To identify additional CCV components, control and mock CCV fractions were compared by 2D DIGE (). The two fractions were labeled with different fluorescent dyes, pooled, and analyzed in single 2D gels. A representative gel is shown in , with the control CCV fraction in red and the mock CCV fraction in green. The spot pattern was highly reproducible and allowed a clear distinction between proteins present in both fractions (yellow spots) and proteins depleted from the mock CCVs (red spots), which are likely to be genuine CCV components. Proteins were excised from the gels and identified by liquid chromatography (LC)–MS/MS (). Several prominent red spots were found to correspond to known clathrin coat components (clathrin heavy and light chains, β1, μ1A, β2, epsinR, and CVAK104) and fusion machinery (-ethylmaleimide–sensitive fusion protein [NSF]), lending further support to the validity of the approach. Again, we consistently observed more AP-1 than AP-2, as indicated by the relative intensities of the β1 and β2 spots.
Several other proteins implicated in membrane traffic were also identified as red spots. These include five sorting nexins, Snx1, Snx2, Snx5, Snx6, and Snx9, none of which had been identified in either of the two previous proteomic analyses of CCVs. Snx1 and Snx2 are putative components of the mammalian retromer complex, which functions in retrograde traffic from endosomes to the TGN (). Another retromer component, mVPS35, also appears as a red spot. mVPS35 was also found in the proteomic analyses of brain and liver CCVs (; ), but at the time, the physiological relevance of this observation was unclear.
An inherent disadvantage of 2D gels is the poor resolution of hydrophobic proteins. Hence, it is likely that integral membrane proteins, including CCV cargo, are strongly underrepresented in our DIGE analysis. To overcome this limitation, as well as to increase the sensitivity of our screen, we used iTRAQ (isobaric tags for relative and absolute quantification), a gel-free quantitative proteomics method (). Control and mock CCVs were digested with trypsin and labeled with iTRAQ tags, which bind to free amines. The labeled peptides were then pooled into a single sample, and the peptide mixture was fractionated by cation exchange chromatography and analyzed by LC-MS/MS. The tags are chemically similar and isobaric (i.e., have the same mass), so identical peptides that originally came from different samples and therefore have different tags coelute from the LC and are simultaneously analyzed by precursor ion scanning. However, fragmentation of the iTRAQ tags during MS/MS results in different signature peaks. Integration of these peaks allows the determination of the relative abundance of a given peptide in the original samples. In turn, this allows the quantification of the parent protein from which the peptide was derived.
Using this approach, 522 proteins were identified and quantified from control and mock CCV fractions (Table S1, available at ). To determine which proteins were depleted from mock CCVs, the mean relative abundance of proteins in control and mock CCVs was calculated and expressed as a ratio (control/mock CCVs). High ratios correspond to proteins depleted from mock CCVs and thus to candidate CCV proteins. Proteins were ranked according to this ratio in descending order.
A plot of normalized ratio over rank () reveals that the majority of proteins did not change substantially between mock and control CCVs; in fact, 378 proteins have ratios between 0.5 and 1.5 (i.e., changes in relative abundance of 50% or less). However, ∼10% of the proteins were significantly depleted from the mock CCVs (ratios between ∼2.0 and 12.2; , left). Although the differences between control and mock CCVs are often not as great as predicted by our other data, the curve has the expected biphasic shape, and the figure of 10% depleted proteins agrees well with our 1D and 2D gels.
To determine a useful (albeit arbitrary) cutoff, we chose the lowest ranking (i.e., least depleted) AP-1 or AP-2 subunit, as APs are known CCV proteins that show clear depletion from mock CCVs. This was the α subunit of AP-2 (rank 53; ratio 2.03). By prediction, all proteins with higher ranks (i.e., ranks 1–52) should correspond to bona fide CCV proteins (). Indeed, out of the top 53 proteins, about half are established CCV components, including clathrin heavy and light chains, subunits of the AP-1 and AP-2 complexes, cargo molecules, alternative adaptors, and other machinery. Of the remaining proteins, four could be classified as likely false positives, based on their known functions in RNA binding or metabolism. The others are novel candidate CCV components.
The iTRAQ analysis may also have produced a few false negatives. The control/mock ratios of known CCV proteins are generally lower than we expected from our other data, suggesting that there may be additional genuine CCV components among proteins with ranks between 54 and ∼100. However, because of overlap with background proteins, such proteins were only included in the list of identified CCV proteins (Table S2, available at ) when they showed clear depletion from mock CCVs by Western blotting and/or 2D DIGE.
How do our data compare with the two other proteomic analyses of CCVs? The studies on brain and liver both focused on optimizing the CCV preparation method to minimize the number of false positive identifications (; ). In both cases, the yield and purity of the CCV fractions were higher than we achieved using HeLa cells, but because neither study used unbiased criteria to identify genuine CCV proteins, it is difficult to compare their results quantitatively with ours. Nevertheless, as expected, the CCVs from all three sources show an overlapping composition, including clathrin heavy and light chains, adaptors, and abundant accessory factors such as NSF and cyclin G–associated kinase (). Some of the promising candidate CCV proteins identified in the two previous studies are not depleted from our mock CCV fraction, including myoferlin, various annexins, and Vac14, suggesting that they may in fact be contaminants (Table S1). There are also several hits from the two previous studies that are clearly physiologically relevant, such as epsin, Eps15, Numb, and the asialoglycoprotein receptor, which we did not find in the present study, either because they are cell type specific or because they are associated with endocytic CCVs, which are underrepresented in the HeLa cell CCV fraction.
One notable feature of the HeLa cell CCV preparation is the high AP-1/AP-2 ratio. AP-1 is more enriched over whole cell homogenate than AP-2, it is more abundant than AP-2, and it is more strongly depleted from the mock CCV fraction than AP-2. There are probably at least two reasons for this phenomenon. First, although HeLa cells appear to contain at least as many plasma membrane as intracellular CCVs (based on immunofluorescence and electron microscopy; ), we have found that most of the plasma membrane–associated clathrin remains tenaciously attached to cell remnants upon homogenization (). Thus, it is likely that the majority of the AP-2–containing CCVs are discarded in our first low-speed centrifugation step. Second, many of the smooth vesicles that contaminate our CCV fraction are derived from the plasma membrane (e.g., the transferrin receptor is a major component of our mock CCVs), and because AP-2 is recruited onto the plasma membrane even in the absence of clathrin (), it is likely to be associated with these contaminating vesicles. Together, these observations probably explain why AP-2 subunits cluster near the lower end of the top-ranking proteins in our iTRAQ analysis (, ranks 39, 41, and 53), whereas subunits of the AP-1 complex are clustered among the highest-ranking proteins (ranks 1, 6, 8, 20, and 26), showing similar ratios to clathrin heavy and light chains (ranks 2, 12, and 18).
This differential depletion of AP-1 and AP-2 can be exploited to interpret the results of the iTRAQ analysis and to assign other proteins in to either the AP-1 or the AP-2 pathway. The highest-ranking component of AP-2 is β2 (rank 39). Thus, as a rough approximation, proteins with ranks higher than 39 should be on the AP-1 pathway, and those that fall betweens ranks 39 and 53 should be on the AP-2 pathway. Supporting this notion, essentially all proteins known to function with AP-1 that were identified here are among the top 35 ranking proteins, including epsinR, PI 3-kinase C2α, and both mannose 6-phosphate receptors. The results are less clear-cut for the AP-2 cluster because it partially overlaps with the AP-3 cluster (see below).
The clustering approach predicts that other high-ranking proteins in , such as syntaxins 6 and 7, are also associated with TGN/endosomal CCVs. It suggests that some proteins with known functions in endocytosis, such as dynamin-2 and Snx9, may also function in intracellular CCV trafficking, a possibility that is supported by immunofluorescence studies (). The presence of two lysosomal enzymes, DNase II and cathepsin Z, in the AP-1 cluster may provide some insights into the directionality of AP-1–mediated trafficking. Lysosomal enzymes travel from the TGN to lysosomes via endosomal intermediates and are not recycled. Because DNase II and cathepsin Z cocluster with AP-1, our results support a role for AP-1 in forward transport.
AP-1 and AP-2 belong to a family that includes two other complexes, AP-3 and AP-4, neither of which are enriched in CCVs purified from brain (; ), although several studies suggest that AP-3 participates in clathrin-mediated trafficking (; , 2003). Subunits of both AP-3 and AP-4 complexes were identified in the iTRAQ analysis, with consistently higher ratios for the AP-3 subunits than for the AP-4 subunits (). Western blots of control and mock CCVs show that both complexes are enriched in CCV fractions, but that only AP-3 is depleted from the mock CCVs (). Thus, it appears that AP-3 indeed plays a role in CCV trafficking, whereas AP-4 is a contaminant. This example illustrates that neither the presence nor the enrichment of a protein in a CCV preparation is sufficient evidence for its specific association with CCVs and further highlights the discriminating power of our comparative proteomics approach.
Subunits of two other protein complexes involved in membrane traffic, retromer and BLOC-1, were also identified as candidate CCV components. Western blots probed with antibodies against the retromer-associated sorting nexin Snx1 and the retromer subunit mVPS26 show similar levels of enrichment in the CCV fraction over the whole cell homogenate and similar levels of depletion from the mock CCV fraction, confirming the iTRAQ data (). BLOC-1 consists of at least eight subunits (), seven of which were found in the iTRAQ analysis, with ratios ranging from 1.25 to 2.36. Western blots probed with an antibody against the eighth subunit, pallidin, show that this protein is also moderately depleted from mock CCVs (). Thus, all eight of the known BLOC-1 subunits behave like bona fide CCV components. BLOC-1 has recently been shown to interact both physically and genetically with AP-3 (), and the similar ratios of BLOC-1 and AP-3 subunits in the iTRAQ analysis suggest that BLOC-1 may depend on AP-3 for its association with CCVs.
Among the membrane proteins that showed more than twofold depletion in the mock CCV fraction are three post-Golgi SNAREs: syntaxins 6, 7, and 8. SNAREs are essential for the fusion of vesicles with their target organelles, and they are frequently used as markers to define membrane compartments, but little is known about how they traffic through the secretory and endocytic pathways. showed by Western blotting that several post-Golgi SNAREs are highly enriched in CCVs from rat liver, and the identification of three syntaxins in our iTRAQ analysis suggested that many SNAREs may use clathrin-mediated transport. To test this hypothesis, Western blots of control and mock CCV fractions were screened with a panel of 16 SNARE antibodies (). SNAREs involved in traffic between the ER and Golgi (syntaxin 17 and Sec22) or in exocytosis (syntaxins 2, 3, and 4 and SNAP23) showed little or no enrichment in the CCV fraction and were unaffected by CHC knockdown. However, all of the SNARES that have been localized to the TGN, endosomes, and/or lysosomes, including syntaxins 6, 7, 8, and 16; vti1 a and b; VAMPs 3, 4, and 7; and SNAP29, were found to be enriched in control CCVs and depleted from mock CCVs.
One of the most strongly depleted proteins in our iTRAQ analysis of mock CCVs is CHC22, a homologue of conventional CHC that is predominantly expressed in muscle (). The two CHC isoforms are 85% identical, but it is unlikely that CHC22 would be targeted by our 21-base siRNA, because its mRNA contains eight mismatches. Although it has been proposed that CHC22 is not associated with CCVs (), our observations suggest that it is in fact a CCV component.
The proteomic analysis also identified five novel proteins of unknown function in the CCVs. Features of these proteins suggest that they too are bona fide CCV components. Two of the proteins, gi38570101 and gi15214676, have domains predicted to interact with rabs, which are known regulators of membrane traffic. Another protein, D19, is an SH3 domain–containing protein related to intersectin, a protein involved in endocytosis. The other two proteins, TPD52L1 and TPD52, are members of the D52 family of tumor proteins, which has been implicated in the secretory and endocytic pathways (). Thus, all of these proteins may perform important regulatory roles in clathrin-mediated trafficking, and we are currently investigating their functions.
Apart from identifying novel CCV proteins, we have developed a method whose scope can be expanded to investigate the function of individual CCV components. For example, one could knock down specific adaptors to see which cargo proteins are lost from CCVs (). Similarly, knockdown or overexpression of accessory proteins may affect CCV composition, and this could be monitored by iTRAQ. Such an approach is not limited to studies on CCVs. Other organelles could also be analyzed by 2D DIGE or iTRAQ, using either siRNA knockdowns or HRP-induced compartment ablation to generate mock organelle fractions. Thus, the comparative proteomics approach described in the present study could be used to gain insights not only into CCVs but into other parts of the cell as well.
siRNA duplexes against target cDNAs were purchased from Dharmacon. CHC knockdown was performed using a custom-made duplex described by . HeLa cells were transfected using Oligofectamine (Invitrogen) in the presence of 10% fetal calf serum; the final concentration of siRNA was 10 nM. For effective knock down, two sequential transfections were performed on days 1 and 3. Experiments were performed 2 d after the second transfection (day 5). CCV-enriched fractions and mock CCVs were prepared from control and CHC knockdown HeLa cells as described by , with minor modifications to the protocol.
SDS-PAGE and Western blotting were performed according to standard protocols. Detailed descriptions of 2D DIGE analysis, iTRAQ analysis, mass spectrometric identification of proteins, and a list of antibodies used in this study are provided in the supplemental text and Table S3 (available at ).
The supplemental text provides a detailed explanation of iTRAQ and DIGE and a protocol for the preparation of CCVs. Table S1 provides a complete list of proteins identified and quantified by iTRAQ. Table S2 provides a summary of CCV proteins identified in this study by different methods. Table S3 provides a list of antibodies used in this study. Online supplemental material is available at . |
The presence of a nucleus requires that a diverse set of macromolecules must be efficiently transported across the nuclear envelope (NE). The sole mediators of this exchange are nuclear pore complexes (NPCs), which span pores in the NE to connect the nuclear and cytoplasmic compartments. Transport of macromolecules across the NPC depends on dynamic interactions between transport cargoes, their cognate-soluble transport factors, and NPCs (). Many transport factors belong to a related family collectively termed karyopherins (Kaps; also called importins, exportins, and transportins). Kaps bind to specific import (NLS) or export (NES) signals in their cargoes (). Unexpectedly, few Kaps are essential, as there appears to be a significant degree of functional redundancy amongst family members (). On import, a Kap–NLS cargo complex diffuses from the cytoplasm to the NPC; transient binding and unbinding with a particular set of NPC proteins (FG-Nups) is central to all proposed models for how transport complexes traverse the NPC (; ).
Once in the nucleus, import Kap–cargo complexes are dissociated by RanGTP. Ran is maintained in its GTP-bound form in the nucleus by a nuclear GDP/GTP exchange factor, RanGEF. Conversely, in the cytoplasm, RanGTP is hydrolyzed to RanGDP by a cytoplasmically restricted GTPase-activating protein, RanGAP. In this way, cells maintain Ran in its GTP-bound form in the nucleus and limit its GDP-bound form to the cytoplasm. This RanGTP/RanGDP gradient is an essential indicator for the directionality of nucleocytoplasmic transport, and possibly the only directional cue for many Kap-mediated transport pathways. With the cargo delivered, Kaps and Ran are then recycled via a nested series of reactions and translocations (for review see ).
Most kinetic studies of nuclear transport have been performed in vitro using permeabilized cell systems. More recently, the interplay between Kaps, NLS-bearing cargoes, and Ran has been modeled in silico from data collected in vivo from mammalian cells. Expanding on an earlier study (), fitted import rate data using a systems analysis including >60 separate parameters. Their findings suggest that the maximum flux of the NPCs in the cell was ∼500 molecules/NPC/s (at least for the flux of Ran across the NE), and that the NPC is not the rate-limiting factor for nuclear transport.
The yeast represents an excellent organism with which to examine the mechanism of nuclear translocation, as it is possible to make systematic alterations in components of its nucleocytoplasmic transport machinery in vivo. However, we have been limited in our ability to study nucleocytoplasmic transport quantitatively in yeast by two factors: the lack of a method to accurately quantitate import rates in single living yeast cells, and the inability to accurately quantitate the concentrations of key players in the import reaction in those individual cells. Therefore, we defined a model import pathway (Kap123p-mediated import of ribosomal proteins), and devised high-resolution quantitative single-cell assays to measure the effectiveness of that import pathway. We determined the import rate of Kap123p (as well as other Kaps), as a function of intracellular concentrations of Kap123p and its cargo. Our results indicate that simple concentration and binding-constant relationships between Kap123p, its cargo, and NPCs determine the rate of import; surprisingly, it is the inefficient formation of the Kap–cargo complex in the cytoplasm, rather than limitations in the NPC or the Ran gradient, that restricts import rates in vivo.
To quantitate Kap-mediated import, we developed a nuclear import assay (based, in part, on a previous method; ) that facilitates rapid, semiautomated cell-by-cell quantitation of import with high spatial and temporal resolution (; for a detailed description of these methods see ). Our model cargoes were NLSs fused to either GFP or a GFP carrying a C-terminal copy of a single PrA repeat. These fusion proteins were small enough to diffuse rapidly across the NPC; hence, in the absence of active import, they equilibrate between the nucleus and cytoplasm within minutes (). Transport was stopped by the addition of metabolic energy poisons, which destroy the RanGTP/GDP gradient (). Re-import of NLS-GFP was observed seconds after a sample of cells had been washed free of poison and resuspended in glucose-containing media on a microscope slide, allowing the Ran gradient to reform. Import assays were performed in a strain background that contained CFP-tagged Htb2p (a histone) and CFP-tagged Tpi1p (a glycolysis enzyme), which demarked the nucleoplasm and cytoplasm, respectively, whereas Htb2-CFPp also served as an internal calibrant for NLS-GFP concentration. An automated spinning-disk microscopy system was used to acquire confocal images at 15-s intervals over a 10-min time course, by which time the NLS-GFP had fully returned to its steady-state distribution. At each time point, images were taken at three focal points throughout the cells of both the GFP and CFP signals ().
The NLS-GFP concentration in a population of importing cells was calibrated using quantitative Western blotting on a sample from this population, which compared the abundance of NLS-GFP to that of the internal standard, Htb2-CFPp (). Once the quantitative blotting had determined the mean NLS-GFP molar amount for cells in an import assay and correlated this to the mean cellular NLS-GFP fluorescence in that same assay, we could then use a given cell's actual fluorescence to calculate the abundance of NLS-GFP cargo in that cell (). We estimated the probable error of these single-cell cargo abundance values to be ±30%.
We needed to measure the size and shape of the assayed cells because a cell with a smaller nucleus would appear to import faster (simply because it has less volume to fill), just as a cell with a greater number of NPCs could potentially import cargo faster. To measure each cell's volumetric statistics, a confocal image series was acquired through the cells, immediately after the time course. From these data, nuclear and cytoplasmic volumes could be directly measured, whereas the area of the nuclear envelope was interpolated from the 3D data using standard isosurface location algorithms (). From similar volumetric measurements of commercially prepared spherical beads (2.5 μm diam), we estimated the uncertainty in our volume measurements to be ±10% and in our surface area measurements to be ±6%. Because it has been shown in yeast that the density of NPCs is a relatively constant 12 NPCs/μm throughout the cell cycle (), we could convert the surface area of the nuclear envelope to an estimate of the number of NPCs in each cell.
These fluorescence microscopy, concentration calibration, and cellular morphometric measurements were combined to give plots of cargo import over time in single cells, which fit single exponential relationships (; average for all single-cell fits, 0.95). The initial rate of cargo import (i.e., import rate at = 0, where net passive diffusion of naked NLS-GFP across the NPC is negligible) could be estimated in units of cargo molecules/NPC/s for each and every cell assayed. Each time course assayed ∼30 cells within a single microscope field, and multiple assays were repeated until the data from ∼100 cells had been collected.
Import rates were spread over wide ranges because the NLS-GFP expression constructs were cloned in multiple copy vectors, conferring a random number of copies of the gene to each cell with a commensurate random expression level of the fusion protein. We used this variability to examine the relationship between cargo concentration and import rate (). Each cell's initial import rate measurement and initial cytoplasmic NLS-GFP concentration is displayed as a single point in a scatter plot (gray dots). To better visualize trends in this population dataset, the cargo concentration range was split into several statistical bins, and the mean import rate and cargo concentration within each bin was calculated (blue squares; error bars are the SD of the mean). This moving-average analysis determined that initial import rates of NLS-GFP cargoes followed a simple linear relationship in respect to the available cytoplasmic concentrations of the fluorescent cargo (). We term the slope of this line an effective import rate, being a measure of how quickly a given cargo is imported by its available transport pathways, in units of cargo molecules per NPC per second per micromolar concentration of cargo.
We began our investigations with one particular model transport pathway, that of Kap123p, chosen for several reasons. First, it is the most abundant yeast Kap (; ; this study), mediating the import of ribosomal proteins (; ), histones (; ), and the mRNA export factor Yra1p (). Second, deletion of Kap123p leads only to a mild growth defect, so we can work with yeast without deleterious effects on cell growth and nuclear transport as a whole. Third, it is partially redundant with other Kaps (; ; ), which accounts for its nonessential nature, and is a matter of considerable interest, being typical of the partial redundancy often observed between import pathways.
The NLS of the ribosomal protein Rpl25p () has been used previously as the model cargo for Kap123p (). To test whether this NLS is largely dependent on Kap123p, we used our import assay to quantitatively compare import of the Rpl25NLS between wild-type and cells (). In wild-type cells, these initial import rates ranged from 5 to 200 cargo molecules/NPC/s/cell, with a mean of 62, whereas these rates varied from 1 to 11 cargo molecules/NPC/s/cell with a significantly lower mean of 3 in cells. By normalizing for the different concentrations of NLS-GFP cargo in each cell (as previously described; ), we determined that Rpl25NLS cargo was imported at a rate of 1.2 ± 0.1 cargo molecules/NPC/s/μM cargo in wild-type cells; a rate that was ∼17 times more rapid than in the absence of Kap123p. Thus, in wild-type cells, Rpl25NLS-GFP-PrAp is primarily imported by Kap123p, with only a residual 6% of its import proceeding through alternative pathways. Hence, in cells with Kap123p present, we could reasonably ignore this low residual transport component.
To test whether other NLSs that are recognized by Kap123p behave the same as Rpl25p, we compared the import of Rpl25NLS-GFPp with GFP carrying the NLSs from other Kap123p import cargoes. First, we identified numerous Kap123p-binding proteins, all of which were known to be targeted to the nucleus, and thus likely to contain NLSs (Supplemental materials and methods and Figs. S1 and S2, available at ); as expected, the majority of these were either ribosomal proteins or proteins involved in the ribosomal assembly process (Table S3; ; ). We constructed NLS-GFP fusion proteins from five proteins of varied function chosen amongst this Kap123p-binding group (). The distributions of all but one NLS-GFP fusion protein was significantly affected by Kap123p deletion, but most were still strongly localized to the nucleus, indicating that a significant proportion of their import likely goes through pathways mediated by Kaps other than Kap123p (Fig. S3). One of the NLSs identified was that of another ribosomal protein (Rps1b), which was as profoundly mislocalized in the absence of Kap123p as Rpl25NLS-GFPp. Throughout this study, Rps1bNLS-GFPp transport was quantitated in addition to Rpl25NLS-GFPp and demonstrated essentially identical behavior in both wild-type and yeast (). Thus, Rpl25p (and likely other ribosomal proteins) is an appropriate model cargo for our studies.
We compared the import of ribosomal NLS-GFP cargoes with that of similarly sized GFP cargoes carrying other published NLSs (); the NLSs of Nab2p, which is carried predominantly by Kap104p (), and of Pho4p, which is a cargo of Kap121p (). Individual cell data from these two other cargoes also produced smooth import curves (), which again gave rise to simple linear relationships between import rate and cargo concentration (). These linear relationships suggested that the import system was never saturated with NLS-GFP cargo, even at cargo concentrations exceeding 100 μM, which is some 20-fold higher than the estimated natural cargo concentrations (). Import of Rpl25NLS by Kap123p was found to be significantly faster than that of the two other pathways, being ∼5-fold faster than Nab2NLS/Kap104p and ∼10-fold faster than Pho4NLS/Kap121p. Kap123p, thus, appears to be a significantly more effective importer than the other Kaps tested here.
Based on in vitro data, it has been suggested that a large variety of Kaps can replace Kap123p for import of ribosomal proteins (). Indeed, the results in indicate that Kaps other than Kap123p are able to import Rpl25NLS-bearing cargoes, although not nearly as rapidly. Therefore, we tested whether other Kaps could substitute for Kap123p's import of Rpl25NLS cargo in vivo if expressed at comparable levels (). Rpl25NLS-GFPp–expressing cells were transformed with vectors overexpressing HA tagged versions of the four Kaps examined in this study. From quantitative Western blots (; ) performed on strains containing genomically tagged versions of these Kaps, we measured the natural abundances of Kap95p, Kap104p, Kap121p, and Kap123p to be 60,000, 12,000, 18,000 and 100,000 copies/cell, respectively. These data are consistent with codon bias data indicating that Kap123p is the most highly expressed (i.e., abundant) of all Kaps. Although Western blot analysis demonstrated that all four HA-tagged Kaps were expressed at significantly higher levels than endogenous Kap123p, only overexpression of Kap121-HAp (1.8-fold above the natural Kap123p level) was able to compensate for loss of Kap123p in the import of Rpl25NLS-GFPp (), in agreement with previous studies showing that Kap123p was partially redundant with Kap121p (; ; ). Thus, our results confirm that, in vivo, ribosomal proteins are not cargos general to most Kaps, but, instead, are only imported by particular Kaps. This contrasts with results obtained for homologous mammalian Kaps in vitro ().
As increasing amounts of Kap121p augmented the import rate of the model ribosomal cargo, we tested the hypothesis that Kap123p's effective import rate may be caused by its high cellular abundance. To do this, we manipulated the concentration of either Kap123p or Kap121p and measured how such concentration changes affected Rpl25NLS import. Either Kap123p or Kap121p was expressed as a CFP-tagged fusion protein in cells coexpressing Rpl25NLS-YFPp to allow simultaneous quantitation of both Kap and cargo in individual cells (). Because and expression plasmids were separate 2μ multiple-copy vectors, each cotransformed cell produced random amounts of the fluorescently labeled Kap and cargo, presumably proportionate to the amount of each plasmid they received. The Kap-CFPp and Rpl25NLS-YFPp quantities were found to vary independently of each other, resulting in cells with a wide range of either protein (). With two independent concentration variables, each potentially affecting import-rate, we needed to analyze larger numbers of cells than was practical with the quantitative import assay. Hence, we chose to measure import rate using the nuclear-to-cytoplasmic ratio (N/C) of cargo, as this value could be obtained relatively rapidly. We analyzed dependence of the steady-state N/C ratio of Rpl25NLS-YFPp on the cytoplasmic concentrations of either Kap121-CFPp or Kap123-CFPp.
Around the physiological concentration of Kap123p (∼5 μM), we observed an approximately linear relationship between its cytoplasmic concentration and the import of Rpl25NLS-YFPp (). At concentrations of Kap123-CFPp in excess of 15 μM, the N/C ratio of Rpl25NLS-YFPp plateaued with a half-maximal Kap123p concentration of ∼7 μM.
Interestingly, Kap121-CFPp displayed a relationship in respect to Rpl25NLS import that was practically indistinguishable from Kap123p-CFPp (). These highly similar import curves were observed despite findings that Kap123p and Kap121p prefer a significantly different subset of Nups to mediate their exchange across the NPC (; ; ; ).
Hence, we measured the dissociation binding constants (
) between Kaps and NLSs, using recombinant purified proteins and an in vitro binding assay ( and ; see Materials and methods).
values of 82 and 94 nM, respectively; this is consistent with both Kaps having similar import kinetics. We also measured binding between Pho4NLS and Kap121p to be similar to Rpl25NLS binding, at 93 nM, whereas Nab2NLS binding to Kap104p was significantly tighter at 17 nM.
values (unpublished data). Binding controls found no quantifiable binding between nonspecific Kap–cargo pairs, such as Pho4NLS-Kap123p (unpublished data).
of the Rpl25NLS interaction with Kap123p and Kap121p. We mutagenized lysine residues 21 and 22 in the Rpl25NLS sequence to alanines, which reduced the binding strength of the NLS–Kap interaction with both Kap121p and Kap123 ∼2.8-fold, and consequently reduced the steady-state NLS-GFP N/C distributions by ∼3.5-fold ().
between Kap and cargo gives a proportionate decrease in cargo import.
Certain observations from our studies led us to propose that binding between NLS-GFP cargo and Kaps is highly competitive with its cytoplasmic environment (see Discussion). To test this hypothesis, we examined the effect of adding bacterial cytosol, absent of any natural Kap cargoes, to the NLS–Kap interactions. Increasing concentrations of bacterial cytosol were found to effectively compete with the binding of Pho4NLS to Kap121p (). 1 mg/ml cytosol disrupted ∼50% of the specific Kap–NLS binding, and 10 mg/ml cytosol was sufficient to disrupt >90% of the import complexes. This result is unlikely to be a consequence of specific competition, because there are no NLSs in , which is a prokaryote.
We showed that import fits well to single exponential relationships, the initial rates of which are linearly related to cytoplasmic cargo concentration (). Around physiological levels, the Kap concentration is linearly related to the import rates. However, at still higher concentrations of Kap the import rates ultimately saturate, following Michaelis–Menten–like curves (). In addition, we find that reducing the affinity of a cargo for its Kap proportionally reduced its import (). Considering these observations, nuclear import of our small NLS-GFP cargoes appears to be well represented by a pump–leak model, such as those used to describe a variety of biological pumps (e.g., ion channels). These pumps actively transport their cargo across a membrane, against a tendency for that cargo to leak back across the membrane through channels (). We therefore fitted our import data to a simple pump–leak model (). From this model, we obtained quantitative estimates of the kinetic parameters of nuclear import in living yeast cells.
In building our model, we assumed a constant Kap concentration in the cytoplasm, consistent with our observations that Kap distributions were not altered by even the highest obtainable Kap or cargo concentrations, and that Kap distributions did not change during our import assays (unpublished data). This assumption was further supported by the reported rapid translocation of naked Kaps (). Thus, the recycling system for Ran and Kaps, as previously modeled (; ), was regarded as an essential but imperturbable and very rapid background process. We also assumed instantaneous mixing in the nucleus and cytoplasm, justified by the small size of yeast cells and the estimated 100-ms intracellular diffusion rates (; ), as compared with the several minutes each cargo required to return to steady-state ().
illustrates the components of our pump–leak model. In this model, NLS-GFP cargo binds its cognate Kap (), competing with other unlabeled or “bulk” cargoes (iii) and nonspecific cytosolic proteins (iv and v). Because it was not possible to quantitate the degree of nonspecific competition (see Discussion), we assumed a value for this nonspecific competition that was common to all Kap–cargo pairs; namely, an interaction of 500 nM
with 1-mM competitors. The magnitude of this competition is the minimum capable of simulating the linearity of the observed relationships (), but is physiologically reasonable, given the high concentration of total protein in the cytoplasm ( ). Also, we cannot state whether this competition is indicative of binding to the NLS-GFP, to the Kap, or to both, but the resulting import complex concentrations will be the same in either case (unpublished data); hence, we ignore nonspecific protein binding to NLS-GFP ().
of 20 nM for these cargoes (), except for Kap104p, for which we assumed a competitor concentration of 0 μM, as it was observed to bind very few cargos (Fig. S1, available at ). All three binding processes () were assumed to be at equilibrium.
= [A][B]/[AB], were solved for [AB] in terms of total Kap and binding protein concentrations (denoted by subscript “” in their variable names), resulting in the quadratic equations shown in (bottom). Thus, the concentration of NLS-GFP–Kap import complexes () was calculated by simultaneous solution of these equations, using the measured and estimated values for the dissociation constants and the total cytoplasmic abundances of each reactant, as listed in .
Next, we considered the import processes. Import through NPCs and subsequent nuclear dissociation of import complexes was modeled as a saturable first-order process () because we had observed saturation of import by Kap that followed a Michaelis–Menten–like relationship (). Here, the kinetic constant is analogous to the Michaelis–Menten's , and the saturation constant is analogous to . Because nonspecific proteins likely interact only transiently with Kaps, we assumed such complexes () were not involved in import. Saturation of import was assumed to be caused by the total concentration of import-competent complexes (i.e., + ), although at nonsaturating conditions (i.e., low Kap concentrations) we modeled import of complexes as a pseudolinear system. Pseudolinear rate constants () were calculated for each cell importing each model import cargo in and from their initial import rates and their expected intracellular NLS-GFP–Kap concentration at that time point (see values in ).
Lastly, we considered the relevant passive diffusion processes. Naked cargo is free to leak through NPCs via diffusion, where the rate of passive flux is proportional to the difference between the nuclear and cytoplasmic concentrations of cargo, with an NPC permeability constant of (). At steady-state, cargo maintains a constant N/C ratio, which can be shown to be proportional to the ratio + (). Having estimated as described in the previous paragraph, we calculated , simply using the final N/C ratio for each cell in these import assays. Estimated and values were combined with the measurements and estimates of the other cellular and biochemical parameters listed in , creating a parameter set that was used to simulate import of NLS-GFP cargoes in single cells over time. These simulated curves closely fit the measured import data ().
h
a
v
e
u
s
e
d
o
u
r
s
i
n
g
l
e
-
c
e
l
l
a
s
s
a
y
t
o
a
n
s
w
e
r
s
e
v
e
r
a
l
k
e
y
q
u
e
s
t
i
o
n
s
c
o
n
c
e
r
n
i
n
g
t
h
e
n
a
t
u
r
e
o
f
i
m
p
o
r
t
,
n
a
m
e
l
y
,
w
h
e
t
h
e
r
d
i
f
f
e
r
e
n
t
i
m
p
o
r
t
p
a
t
h
w
a
y
s
h
a
v
e
i
n
h
e
r
e
n
t
l
y
d
i
f
f
e
r
e
n
t
r
a
t
e
s
;
w
h
e
t
h
e
r
i
m
p
o
r
t
b
y
t
h
e
N
P
C
i
s
s
a
t
u
r
a
b
l
e
i
n
v
i
v
o
;
w
h
y
s
o
m
a
n
y
p
a
r
a
l
l
e
l
p
a
r
t
i
a
l
l
y
r
e
d
u
n
d
a
n
t
i
m
p
o
r
t
p
a
t
h
w
a
y
s
e
x
i
s
t
;
a
n
d
w
h
y
h
i
g
h
c
a
r
g
o
c
o
n
c
e
n
t
r
a
t
i
o
n
s
a
r
e
u
n
a
b
l
e
t
o
s
a
t
u
r
a
t
e
t
h
e
i
r
i
m
p
o
r
t
s
y
s
t
e
m
s
i
n
t
h
e
c
y
t
o
p
l
a
s
m
.
W
e
f
i
t
t
e
d
t
h
e
r
e
s
u
l
t
s
o
f
o
u
r
s
t
u
d
i
e
s
t
o
a
s
i
m
p
l
e
p
u
m
p
–
l
e
a
k
m
o
d
e
l
o
f
t
h
e
i
m
p
o
r
t
s
y
s
t
e
m
t
o
e
x
a
m
i
n
e
h
o
w
t
h
e
s
e
d
a
t
a
c
o
m
b
i
n
e
t
o
p
r
o
v
i
d
e
t
h
e
o
v
e
r
a
l
l
m
e
a
s
u
r
e
d
i
m
p
o
r
t
r
a
t
e
s
.
All plasmids used in this study are listed in Table S1 (available at ) and were constructed by standard recombinant DNA methods () using the restriction sites described in the Supplemental materials and methods. All strains listed in Table S2 were made in DF5 (), with the exception of Kap-GFP strains ().
Cells transformed with NLS-GFP plasmids were, for some experiments, simultaneously transformed with KAP-HA or -CFP expression plasmids. The trace amounts of copper in synthetic selective minimal media (US Biological) facilitated the promoter of the Kap expression plasmids to express their fusion proteins several fold above their wild-type levels (as measured from Western blots that compared the signals from the plasmid-borne Kaps with those of genomically tagged versions). Yeast were grown to mid-log phase in selective minimal media and harvested by centrifugation. Cell pellets were reserved for concentration calibration (see import assay). Confocal image sections of several fields of cells in growth media were acquired at 0.4-μm increments, using the microscope setup described in Microscopy. Cell-by-cell image analysis measured subcellular volumes, nuclear envelope area, and fluorescent fusion protein concentrations for a statistically significant number of cells, essentially as described for import assays.
Cells were grown as described for steady-state analysis and then treated essentially as described by , using metabolic energy poisons to stop import, before measuring the rate of re-import of fluorescent cargoes. Modifications were made to data collection and analysis of this yeast import assay, such that import-rates in single cells could be quantitated in units of cargo molecules/NPC/s. First, advanced automated microscopy was used to observe import in single cells over time, while 3D reconstruction of confocal images was used to measure subcellular volumes. Second, a quantitative Western blotting technique was used to calibrate fluorescence measurements to actual subcellular concentrations of NLS-GFP cargo. For a detailed description of these methods, see our recent survey of nuclear import methods in yeast ().
Images were collected at room temperature (air conditioned to ∼23°C) with a cooled charge-coupled device camera (Orca ER; Hamamatsu) attached to a microscope (Axiovert 200; Carl Zeiss MicroImaging, Inc.) fitted with a spinning disk (UltraView; Perkin-Elmer) confocal imaging head and using a 100× objective lens (NA 1.45). CFP and YFP were excited with HeCad 442-nm or Argon 514-nm lasers, respectively, and separate images of each were obtained using a standard CFP/YFP dichroic with separate excitation/emission filter sets (Chroma Corp.). These same CFP/YFP optics were used for import assays where NLS-GFP needed to be imaged along with Htb2-CFPp and Tpi1-CFPp; no bleed-through of fluorescence was seen between GFP and CFP image channels. For other experiments, GFP was excited with the 488-nm line of a Krypton–Argon laser, using a dedicated 488-nm dichroic and standard GFP excitation/emission filters (Chroma Corp.). The system was controlled with MetaMorph imaging software (Universal Imaging Corporation). All images were background subtracted and controlled for uniform field illumination before analysis. For presentation, a maximum intensity projection of all focused image planes is shown, and noise in the CFP images was reduced with the adaptive Weiner filter (MatLab).
Nab2NLS-GFPp, Pho4NLS-YFPp, and Rpl25NLS-GFPp were expressed as 6xHis, C-terminal fusion proteins in BL21DE3 Gold (Novagen). Protein production was induced at OD of 0.6–0.8 with 1 mM IPTG (Roche) for 4–5 h at 30°C. Cells were harvested at 4°C in PBS (50 mM NaHPO, pH 7.0, and 300 mM NaCl), containing the protease inhibitors 0.1 mM EDTA, 1/100 Solution P (0.5 mg/ml Pepstatin A and 20 mg/ml PMSF in ethanol), 1/500 Protease Inhibitor Cocktail (Sigma-Aldrich), and 1/100 Protease Arrest (G-Biosciences). Cells were lysed by four passages through a microfluidizer (Microfluidics). Cell debris was removed by centrifugation, and fusion proteins were bound, washed, and eluted from TALON resin (BD Biosciences) via their 6xHIS tags at 4°C, as per the manufacturer's recommendations (buffers used after binding did not contain either EDTA or the Protease Arrest cocktail). Purified fusion proteins were concentrated with 10,000 MWCO centrifugal concentrators (Millipore). Protein concentrations were determined by the Bradford method ().
GST-Kap104p, GST-Kap121p, and GST-Kap123-HAp were expressed in BL21DE3 pLYS . Proteins were expressed and purified exactly as previously described (). When removal of the GST was required, the Thrombin Cleavage Capture kit (Novagen) was used according to the manufacturer's instructions.
Binding assays were performed in TB-T (20 mM Hepes-KOH, pH 7.5, 110 mM KOAc, 2 mM MgCl, 0.1% Tween-20, 1 mM DTT, and 1/100 Solution P). A custom high-titer anti-GFP rabbit polyclonal antibody () was affinity purified and conjugated to Sepharose by standard methods (). For each experiment, 0.5 μg of purified NLS-GFP-HIS was incubated per microliter (bed volume) of anti-GFP–conjugated Sepharose, generating immobilized NLS-GFP-HIS resin. 10-μl aliquots of this resin were dispensed into 10–25 ml of TB-T plus 2.5% milk, containing a particular concentration of purified Kap; volumes were adjusted so that the Kap was always in molar excess to the NLS. NLSs were allowed to reach binding equilibrium with the Kap overnight at 4°C, with mixing. The resin, with bound Kap, was harvested and transferred to minicentrifugation columns. Binding buffer was removed by centrifugation, and the resin was washed with 500 μl of ammonium acetate buffer (0.1 M NHOAc, 0.1 mM MgCl, and 0.02% Tween-20). All proteins were eluted from the resin by incubation with ammonium hydroxide solution (0.5 M NHOH and 0.5 mM EDTA). Evaporated protein samples were prepared for SDS-PAGE, and proteins were observed via Coomassie blue staining. The total background-subtracted pixel intensities of bands from scanned gels were measured using OpenLab (Improvision) image analysis software. Dissociation constants were calculated from the resulting binding curves by fitting the data to the predicted simple bimolecular equilibrium relationship.
Provided as online supplemental material are descriptions of the plasmids and strains used throughout this work. Also provided, are the raw data of overlay assays, performed to identify Kap123p-binding proteins, and the identification and distribution of NLS-GFP fusion proteins created from these Kap123p binding proteins. Fig. S1 shows Kap123p interacts with a distinct subset of small nuclear proteins. Fig. S2 shows that the identities of Kap123p-interacting proteins were determined from overlay assays. Fig. S3 shows that steady-state distributions of ribosomal NLS-GFP cargoes are significantly altered in yeast. Online supplemental material is available at . |
Macroautophagy, which is usually referred to as autophagy, is responsible for degradation of the majority of intracellular proteins in mammalian cells, particularly during starvation-induced proteolysis. Cytosolic proteins, organelles, and selected regions of the nucleus can be cleared by autophagy to ensure cellular homeostasis and remove damaged or unwanted products (). Cytoplasmic constituents are first enclosed by a double-membrane autophagosome; next, the outer membrane of the autophagosome fuses with the lysosome, with the consequent destruction of the cargo and the inner membrane of the autophagosome by hydrolytic enzymes.
The autophagic pathway has been dissected at the molecular level in yeast, in which 27 genes, referred to as , have been found to be required for autophagosome formation (; ). Two ubiquitin-like conjugation systems are involved in the process: one mediates the covalent attachment of Atg12 to Atg5, and the other mediates the conjugation of Atg8 to phosphatidylethanolamine (). The resulting products, Atg12–Atg5 and Atg8-phosphatidylethanolamine, are essential for proper autophagosome formation. Only a small fraction of the Atg12–Atg5 complex localizes at the autophagic isolation membrane throughout elongation, and the complex dissociates when the phagosome is complete. On the other hand, Atg8 remains on the mature autophagosome constituting the recognized consensus marker of this structure. These two conjugation systems are highly conserved in mammalian cells. -deficient stem cells have been instrumental for demonstrating that the Atg12–Atg5 complex localizes at autophagosome precursors and plays an essential role in autophagosome formation (). The production of -deficient mice has revealed that autophagy is required for neonatal survival before maternal feeding (). The analysis of endogenous LC3, which is one of the mammalian homologues of Atg8, and the production of transgenic mice expressing a fluorescent autophagosomal LC3 have shown that the regulation of autophagy is organ dependent and that the role of autophagy is not restricted to the starvation response ().
A pivotal role for autophagy in growth control has emerged from studies on Beclin1, the mammalian homologue of . The Beclin1 gene is deleted in several types of breast cancer that are unable to activate autophagy () and functions as a haploinsufficient tumor suppressor gene (). In addition to involvement in growth control, programmed autophagy plays a role in differentiation (; ) and tissue remodeling (). Moreover, autophagy may be artificially induced by radiation and treatment with certain drugs, including rapamycin (), ceramide (), arsenic trioxide (), and tamoxifen (); the final outcome of such induction is still a matter of debate ().
The calpain family of cysteine proteases is composed of both ubiquitous and tissue-specific isoforms that share homology in their protease domain and are calcium dependent (). The best-characterized ubiquitous calpains are the isoforms μ- and m-calpain, which are also known as calpain 1 and calpain 2, respectively. Both contain an 80-kD catalytic subunit and share a common 28-kD regulatory subunit commonly known as calpain 4 that is required for proper activity. Hereafter, calpain 4 will be referred to as calpain small 1 (CAPNS1) in accordance with the nomenclature reported at the calpain website (). A homologue to CAPNS1, CAPNS2, has been described previously (). This protein appears to be a functional equivalent to CAPNS1 in vitro; however, its in vivo function and tissue distribution are still controversial. Interestingly, ubiquitous calpains are associated with the endoplasmic reticulum and Golgi, a likely reservoir for autophagosome membranes (), where the essential autophagy regulatory complex PI3K–Beclin also localizes ().
Calpain is required for normal embryonic development. Indeed, the targeted disruption of CAPNS1 is embryonic lethal at days 10 and 11 as a result of severe defects in vascular development (). Calpain is a regulator of adhesive complex dynamics in adherent cells (); it plays an important role in osteoclastic bone resorption () and is required for phagocytosis in human neutrophils ().
We have become interested in calpain function because of our observation that the product of GAS2, a gene specifically induced at growth arrest, is in fact an inhibitor of millicalpain and that its overexpression sensitizes cells to apoptosis in a p53-dependent manner (). More recently, we have shown that calpain is also involved in NF-κB activation and in its relative prosurvival function in response to ceramide, in which calpain deficiency strengthens the proapoptotic effect of ceramide (). In this study, we further explore the involvement of calpain in the apoptotic switch and find that in calpain-deficient cells, autophagy is impaired with a resulting dramatic increase in apoptotic cell death.
We have previously reported that calpain plays a crucial role in cell death regulation in response to ceramide. Indeed, we demonstrated that the lack of calpain activity results in a considerable increase in apoptotic cell death (). To further investigate the involvement of calpain in the modulation of alternative types of cell death, in this study, we address the role of calpain in autophagy. LC3 is the only available specific marker to detect autophagosomes in mammalian cells ().
To directly tackle the question of whether calpain is required in autophagy, we analyzed endogenous LC3 in response to ceramide and rapamycin by means of immunofluorescence in wild-type and CAPNS1 mouse embryonic fibroblasts (MEFs). CAPNS1 MEFs, which are also known as calpain 4 MEFs, were derived from CAPNS1 (calpain 4) knockout mice and were shown to lack calpain activity (). Wild-type and CAPNS1 MEFs were induced with each stimulus for 3 h or were left untreated. Afterward, the cells were fixed and subjected to immunofluorescence analysis with an anti-LC3 antibody. As shown in (top), wild-type MEFs are faintly labeled by anti-LC3 antibody and show a clear induction of autophagosome formation after rapamycin or ceramide treatment. On the other hand, CAPNS1 MEFs show a higher diffuse background that does not substantially vary after stimulation with rapamycin or ceramide (, bottom).
In a parallel experiment, rapamycin and ceramide were used to stimulate wild-type and CAPNS1 MEFs, and the lysates were analyzed by Western blotting to detect both cytoplasmic LC3-I and its proteolytic derivative LC3-II that preferentially associates with autophagosomal membranes (). In agreement with the immunofluorescence data, both rapamycin and ceramide trigger an increase of LC3II with respect to LC3I in wild-type cells, whereas the ratio between the two LC3 forms is not considerably altered by both stimuli in CAPNS1 MEFs ().
The fate of endogenous LC3 upon autophagy induction was also addressed in human U2OS cells. U2OS cells were transfected with a scrambled siRNA (siRNA control) or with a CAPNS1-specific siRNA. 72 h later, autophagy was induced by a 3-h incubation with ceramide, and the cells were either fixed and analyzed by immunofluorescence or lysed and subjected to Western blot analysis. As shown in , CAPNS1 depletion prevents the formation of bona-fide autophagosomes upon induction with ceramide. Accordingly, the increase in the endogenous membrane-bound LC3II after ceramide treatment is clearly defective after CAPNS1 silencing ().
To further assess the requirement of calpain for the autophagic process, we followed the increase of lysosomal activity occurring after autophagy induction by means of LysoTracker staining and FACS analysis. Wild-type and CAPNS1 MEFs were incubated with rapamycin for 24 h, trypsinized, stained with LysoTracker green, and analyzed by FACS. The results obtained in a typical experiment are reported in (A–D). A net increase in the percentage of cells that are highly labeled with LysoTracker occurs after rapamycin treatment of wild- type MEFs () as compared with control untreated cells (), clearly showing that rapamycin results in an increase in lysosomal activity. On the contrary, in CAPNS1 MEFs, rapamycin is ineffective in inducing an increase in LysoTracker fluorescence intensity (, compare B with D). The mean values of the percentage of cells highly labeled by LysoTracker obtained in four independent experiments are reported in . To determine whether the increase in lysosomal activity was dependent on macroautophagy induction, 3-methyladenine (3MA) was added just before stimulation with rapamycin. 3MA clearly inhibits the increase in LysoTracker staining, indicating the specificity of the induction (). Next, other autophagic stimuli, namely etoposide, starvation in Earle's balanced salt solution (EBSS) or serum-free medium, and ceramide, were used to induce autophagy in wild-type and CAPNS1 MEFs using the same type of assay. As shown in , all four stimuli induce autophagy-dependent lysosomal activation after 3 h in wild-type MEFs but not in CAPNS1 MEFs; moreover, the increase in LysoTracker labeling is dramatically reduced upon the addition of 3MA just before induction ().
To further investigate the requirement of calpain in autophagy, we monitored long-lived protein degradation upon amino acid starvation both in wild-type and CAPNS1 MEFs according to standard protocols (). Amino acid starvation induces an increase in bulk protein degradation in both cell lines. However, such an increase is considerably lower in calpain-deficient cells as compared with wild-type cells (). A reduction in protein degradation occurs in the presence of the inhibitor 3MA in both cell lines, indicating that although considerably reduced, autophagy may still occur in CAPNS1 MEFs as a consequence of amino acid starvation.
cells (). Similar results were obtained by studying long-lived protein degradation upon serum starvation (), confirming the importance of calpain in activating protein degradation through the autophagic process.
Electron microscopy was used to directly analyze and quantify autophagosomes in wild-type and CAPNS1 MEFs grown in control serum-containing medium and after autophagy induction with rapamycin or amino acid–free medium (EBSS). The morphology of early autophagic vacuole (AV [AVi]) and late AV (AVd) profiles of wild-type MEFs are shown in (A and B, respectively). The quantitative results obtained are represented in .
These electron microscopy experiments demonstrate that in CAPNS1 MEFs, autophagosome formation is completely abolished in response to rapamycin. In the case of EBSS-induced autophagy, CAPNS1 MEFs do show autophagosome formation but at a substantially reduced level compared with wild-type MEFs. In this last case, as shown in Fig. S2 (available at ), the morphology of the autophagosomes is not altered in CAPNS1 MEFs with respect to wild-type MEFs. This led us to conclude that the mechanics of autophagosome formation are still functional in CAPNS1 MEFs, however severely impaired in their efficiency. The diminished accumulation of autophagosomes in CAPNS1 MEFs maintained in EBSS medium may therefore suggest that the dramatic reprogramming of cell functions occurring upon amino acid deprivation may also involve the activation of some autophagic pathways not strictly requiring calpain function. Alternatively, because EBSS possibly represents a stronger and broader stimulus in respect to rapamycin, residual calpain activity caused by some other calpain isoforms may be sufficient to trigger autophagy, albeit at a reduced level.
This evidence prompted us to monitor Tor activity in both cell lines before and after rapamycin treatment using S6 kinase phosphorylation as readout. The results shown in Fig. S3 (available at ) clearly indicate that Tor is active in both cell lines, where it is inhibited with similar efficiency upon rapamycin addition. Further studies are required to define the key regulatory elements targeted by calpain for the efficient induction of autophagosome formation.
A widely used method to study autophagy is monitoring the redistribution of the overexpressed autophagosome marker LC3 from a mostly diffused localization to a punctuated pattern upon the addition of an autophagic stimulus. Therefore, in parallel to the studies on endogenous LC3, we monitored autophagosome formation after overexpressed GFP-LC3 localization in wild-type and CAPNS1 MEFs. 16 h after transfection, the cells were preincubated with the macroautophagy inhibitor 3MA or solvent alone, and ceramide was added for 3 h to induce autophagy (). The cells were subsequently fixed to analyze overexpressed LC3 by fluorescence microscopy. As shown in , in wild-type MEFs, LC3 produces a mostly diffuse staining in the absence of stimulus and after 3MA treatment. After induction with ceramide, LC3 accumulates in the autophagosomes, whose formation is prevented by 3MA pretreatment, indicating the specificity of autophagy induction. However, in CAPNS1 MEFs, LC3 accumulates in specific bodies even under basal conditions; incubation with the autophagy inhibitor 3MA in this case does not prevent the formation of LC3 spots. In addition, the number of cells with intensely stained bodies does not substantially increase after the addition of ceramide (). Each experiment was repeated at least four times, obtaining reproducible results. For each experiment, the percentage of cells with LC3 bodies was counted, and the mean value ± SD is reported in the figure. At least 200 cells were analyzed for each independent experiment.
To further monitor autophagy induction, wild-type and CAPNS1 MEFs overexpressing GFP-LC3 were challenged with several stimuli that were previously used to induce autophagy, namely rapamycin (), amino acid–free medium (), serum starvation (), and etoposide (). All of the treatments triggered autophagosome formation in wild-type MEFs () while leaving unchanged the LC3 staining pattern in CAPNS1 MEFs (not depicted). Altogether, these results further suggest that autophagy is impaired in cells lacking calpain activity, where overexpressed LC3 constitutively accumulates into specific bodies.
To verify whether the constitutive accumulation of ectopic LC3 into the observed specific bodies in CAPNS1 MEFs was a peculiarity of the cell line used or a consequence of calpain inactivation, we monitored the effect of calpain silencing by siRNA on autophagosome formation in the human osteosarcoma cell line U2OS. An -specific siRNA that has been previously described () was used as a tool to prevent autophagy, allowing the distinction between true autophagosomes and other undefined structures containing ectopic LC3. U2OS cells were transfected with a scrambled siRNA (siRNA control) or with a CAPNS1-specific siRNA in combination with siRNA or an ineffective siRNA as a control.
48 h later, the cells were transfected with GFP-LC3, and, after an additional 16 h, autophagy was induced by a 3-h incubation with ceramide. The fixed cells were finally analyzed by fluorescence microscopy. As shown in , LC3 appears diffuse in control U2OS cells and, after induction with ceramide, becomes associated with autophagosomes, which are not formed by specifically inhibiting autophagy through siRNA. On the other hand, CAPNS1-silenced U2OS cells show LC3-positive bodies in the absence of stimulation, and the distribution of GFP-LC3 is not altered by silencing. In addition, the LC3 staining pattern is not considerably affected by ceramide ().
As shown in , in the cells transfected with the control plasmid, HcRed-LC3 appears diffuse before induction and forms autophagosomes after ceramide treatment as expected. However, siCAPNS1-GFP–transfected cells show a punctuated LC3 pattern both in the presence and absence of ceramide stimulation. Collectively, the data obtained in U2OS cells are in agreement with the data obtained comparing wild-type with CAPNS1 MEFs, demonstrating that in CAPNS1-depleted cells, overexpressed LC3 accumulates in specific bodies independently of autophagy induction.
To investigate the identity of LC3-containing structures in CAPNS1-depleted cells, we analyzed whether overexpressed LC3 would eventually colocalize with specific endosomal markers both in control and in CAPNS1-silenced U2OS-LC3 cells. 48 h after silencing with control or CAPNS1 siRNA, autophagy was induced by ceramide together with pepstatin to inhibit lysosomal activity and freeze any eventual transient colocalization event. Thereafter, the cells were fixed and subjected to immunofluorescence against the markers LAMP-2, EEA1, and transferrin receptor and were analyzed by a confocal microscope. Representative images are shown in . The colocalization of LC3 with the endosomal markers was quantified using ImageJ software (colocalizer plug-in). The Pearson correlation coefficient (R) is reported near each merged image. Altogether, the data indicate that overexpressed LC3 bodies are enriched in endosome markers specifically in calpain-depleted cells.
To precisely define the identity of the LC3-positive bodies that accumulate in calpain-deficient cells, immunoelectron microcopy studies were performed. U2OS cells were transfected with GFP-LC3 and siRNA specific for CAPNS1 as described in the first section of Results and were fixed in PFA. Double labeling with GFP and LAMP-2 antibodies allowed us to detect GFP-LC3 bodies and also mark endosomes in the same experiment. As expected, GFP-LC3 partly accumulates in specific bodies, as shown in (A and B). The gold particles indicating the presence of GFP-LC3 bodies (, large gold dots) were associated with endosome-like structures, some of which were also positive for LAMP-2 (; small gold dots are indicated by arrowheads). The morphology and LAMP-2 labeling pattern of GFP-LC3 structures corresponded to that of early endosome-like vesicles, clearly confirming that GFP-LC3 bodies accumulating in CAPNS1-depleted cells are not autophagosomes.
We have previously reported that ceramide triggers apoptosis and induces an NF-κB–dependent prosurvival pathway through calpain. Accordingly, in calpain-deficient cells, apoptosis is enhanced (). In this study, we have demonstrated that calpain is required to induce autophagy in response to ceramide, etoposide, and amino acid or serum starvation. Recent studies indicate the existence of a negative cross-regulation between autophagy and apoptosis (; ; ). However, there are examples of a mutual requirement between the two pathways (; ). Therefore, we decided to define the effect of the calpain-related autophagy block on apoptosis induction in the cellular systems used for this study.
Wild-type and CAPNS1 MEFs were incubated with the autophagy inducers etoposide, ceramide, serum-free medium, EBSS, or vinblastine for 20 h. Afterward, the cells were double labeled with annexin V and propidium iodide (PI) and were analyzed by FACS. The results obtained in a typical experiment of EBSS induction are reported in (A–D). The histogram presented in summarizes the results of four independent experiments for each specific stimulus. A dramatic increase in apoptosis can be seen in calpain-deficient cells after induction with all of the stimuli, with the exception of vinblastine. Vinblastine is a microtubule-depolymerizing agent, and, although it can trigger an autophagic response, it has inhibitory functions on late autophagic events (). Interestingly, we have previously reported that the toxicity of taxol, an indirect inhibitor of autophagy by virtue of its block of microtubule depolymerization, is almost comparable in wild-type and CAPNS1 MEFS (). Collectively, the aforementioned data highlight the cytoprotective potential of autophagy in MEFs and strongly argue for a strict correlation between the impairment of autophagy in calpain-deficient cells and induction of the apoptotic switch.
To further confirm this hypothesis, a similar approach was followed using human U2OS cells silenced either with , CAPNS1-specific siRNA, or a combination of the two. 48 h after silencing, the cells were shifted to amino acid–free medium and incubated for a further 20 h to trigger autophagy. Cell death was then quantified by means of PI and annexin staining followed by FACS analysis. (A–F) presents the results of a representative experiment and indicates that both and CAPNS1 silencing sensitize U2OS cells to apoptosis induced by amino acid starvation. The mean values obtained in four independent experiments are reported in . The effect of CAPNS1 silencing is the most severe, suggesting that CAPNS1 depletion in addition to the block of autophagy also prevents other antiapoptotic mechanisms such as NF-κB activation ().
In this study, we demonstrate that calpain regulates macroautophagy in two cellular systems lacking calpain activity: MEFs derived from CAPNS1 knockout mice and CAPNS1-silenced human cells. In addition, we shed some light on the relationship between apoptosis and autophagy.
Calpains are activated by several stimuli that trigger macroautophagy, including starvation (), ceramide (; ), etoposide (), and arsenic trioxide ( ). Milli- and microcalpain, which both require CAPNS1 for activity, are localized at the endoplasmic reticulum and Golgi, where the autophagic machinery works (; ).
In both calpain-null systems used in this study, deficient autophagy was shown by four different approaches: analysis of exogenous and endogenous LC3 by immunofluorescence and Western blotting, quantification of lysosomal induction, long-lived protein degradation assays, and electron microscopy. Altogether, the data presented indicate that in CAPNS1-deficient cells, the autophagic program is not efficiently activated. As a result of this defect, CAPNS1-deficient cells are more sensitive to apoptosis induced by several autophagic stimuli, including ceramide, etoposide, and starvation.
In light of our findings connecting calpain and autophagy, it will be interesting to investigate this relationship in vivo under physiological and pathological conditions as well as in response to chemotherapy. Previously reported data lend support to such a possibility. The dual role of autophagy has been seen during cancer progression, where, at the initial stages, autophagy prevents tumor growth (), whereas at advanced stages, it might favor tumor cell survival (). An example of this is calpain 9, which appears to be a tumor suppressor of gastric cancer, whereas calpain activation has been shown to be involved in cell transformation and invasion (; ).
Most interestingly, this parallel is maintained in other degenerative pathologies. Autophagy is protective during the initial stages of neurodegeneration (; ), whereas it becomes deleterious at later stages (). A similar dual role could be played by calpain in neuronal diseases, where it may be protective at the initial stages of neurodegeneration () and hyperactivated and detrimental at advanced stages (). A detailed investigation of the molecular cross talk between these two systems in living organisms might be instrumental for performing appropriate strategies of molecular intervention.
The precise molecular network linking autophagy, apoptosis, and other types of cell death is still unresolved (; ). Autophagy is protective against low radiation damage () and against mutant huntingtin-induced cell death (); it is essential for maintaining cell survival after growth factor withdrawal in BaxBak cells (). Furthermore, the inhibition of macroautophagy triggers apoptosis (; Gonzalez-Polo et al., 2005). On the other hand, autophagy is coupled to cell death in BaxBak MEFs () and in tamoxifen-treated MCF7 cells (). In addition, autophagy and apoptosis are positively interconnected in several systems, including development (), rat retinal tissue development (), and peripheral nerves of adult rats after damage (). Collectively, our data further support the argument that macroautophagy is a survival strategy activated both in response to serum and amino acid deprivation. In addition, we have shown that macroautophagy is activated and also plays a protective role upon treatment with specific drugs, such as etoposide and ceramide; thus, the knockdown of autophagy could be a strategy for improving chemotherapy protocols.
Ectopic LC3 is a powerful tool to study autophagy (). In this study, we confirmed that in MEFs and U2OS human osteosarcoma cells, the formation of LC3 autophagosomes is efficiently inhibited by treatment with the autophagy inhibitor 3MA and by siRNA-mediated depletion of the essential autophagy gene , respectively. However, in CAPNS1 MEFs and siCAPNS1-depleted cells, we observed that ectopic LC3 constitutively accumulates in the absence of any specific autophagic stimulus into cytoplasmic endosome-like bodies enriched with endosomal markers. Therefore, we hypothesize that such LC3 bodies could represent a default or salvage lysosomal pathway for protein degradation with slower clearance kinetics, which becomes manifest in a calpain-deficient cellular context. A similar pathway may be predicted for the endogenous proteins that accumulate in diseases coupled to autophagy defects.
Further studies are required to analyze the detailed mechanisms involving calpains in the formation of autophagosomes and in lysosomal-mediated protein degradation pathways. Given the strategic localization of milli- and microcalpain at the endoplasmic reticulum, it would be tempting to speculate that calpain may play a role in the delivery of membranes to the autophagosome. Alternatively, calpain could modulate one or more key components of the signaling networks, ultimately leading to autophagosome formation. The complete block of autophagosome formation occurring in rapamycin-treated CAPNS1 MEFs favors such a possibility.
C2 ceramide, bafilomycin, pepstatin A, 3MA, rapamycin, and etoposide were purchased from Sigma-Aldrich. LipofectAMINE reagent and Oligofectamine were purchased from Invitrogen. SMARTpool for CAPNS1 silencing was obtained from Dharmacon. Micro- and millicalpain siRNAs were purchased from Santa Cruz Biotechnology, Inc.
HcRed-hLC3 and pGFP-hLC3 were constructed by inserting human LC3 in HcRed and pGFP plasmid, respectively. The plasmid pGFP-rLC3 was a gift from T. Yoshimori (National Institute of Genetics, Shizuoka, Japan). The plasmid siCapns1-GFP was obtained by subcloning a double-stranded oligonucleotide containing a sequence complementary to CAPNS1 siRNA into commercial p-superior–GFP plasmid according to the manufacturer's instructions. All constructs were checked by sequence analysis.
Wild-type and CAPNS1 MEFs () were gifts from P.A. Greer (Queen's University, Kingston, Ontario, Canada). U2OS cells and the aforementioned mouse fibroblasts were grown in DME supplemented with 10% FCS.
Standard protocols for immunoblotting and immunofluorescence were used. Rabbit serum raised against LC3 was provided by T. Yoshimori. Monoclonal antibody against LAMP-2 was purchased from BD Biosciences. LAMP-1–specific antibody was purchased from Santa Cruz Biotechnology, Inc., and monoclonal anti-CAPNS1 was purchased from Sigma-Aldrich. Monoclonal antitransferrin receptor (OKT9) was previously described (). Band quantification from Western blots was performed using the Image 1.63.sea program (Scion). Secondary antibodies for immunofluorescence were FITC conjugated, and the cells were mounted using Mowiol medium. Analysis and acquisition were performed using a confocal laser-scanning microscope (Axiovert 100 M; Carl Zeiss MicroImaging, Inc.) with a 63× NA 1.25 or 100× NA 1.30 oil objective (Leica) at room temperature. Images were imported using LSM-510 software (Carl Zeiss MicroImaging, Inc.). ImageJ software (National Institutes of Health) was used to quantify the colocalization observed by immunofluorescence.
Stable transfections of U2OS cells with HcRed-LC3 were performed by the calcium phosphate method using standard procedures. U2OS, CAPNS1, and control mouse fibroblasts at 60–80% confluency were transiently transfected or oligofected using FuGene (Roche) or Oligofectamine (Life Technologies) according to the manufacturer's instructions.
Results are expressed as means ± SD of at least three independent experiments performed in triplicate or quadruplicate unless indicated otherwise. Statistical analysis was performed using a test, with the level of significance set at P < 0.05.
To label lysosomes, LysoTracker green (Invitrogen) was used at a final concentration of 75 nM. The cells induced for 3 or 24 h with the stimuli were trypsinized, resuspended in phenol-free DME, 10% FCS, and 75 nM LysoTracker green, incubated for 30 min at 37°C, and analyzed with the CellQuest program (BD Biosciences) by FacsCalibur (Becton Dickinson).
Cells were labeled in complete medium in the presence of [C]valine for 18 h, washed, and incubated with cold medium for 1 h to allow the degradation of short-lived proteins. Then, after extensive washing, the medium was replaced with EBSS for 4 h in the presence or absence of 10 mM 3MA. Finally, the proteins were precipitated with TCA, and radioactivity was measured. Protein degradation was calculated as the percentage of TCA-soluble counts on total radioactivity.
Monolayers were fixed in 2% glutaraldehyde in 0.2 M Hepes, pH 7.4, for 2 h and were scraped off the dish during fixation. The pellets were embedded in Epon using routine procedures. Approximately 60-nm sections were cut and stained using uranyl acetate and lead citrate and were examined with an electron microscope (JEM 1200EXII; JEOL). The number of AVi and AVd profiles was counted under the microscope by systematically screening the sections at 12,000× using grid squares as sampling units, as previously described ().
For immunogold electron microscopy, U2OS cells were fixed in 4% PFA in 0.2 M Hepes, pH 7.4, for 2 h at room temperature and were further processed as previously described (). The cell sections were analyzed by immunogold electron microscopy with rabbit anti-GFP and mouse anti–LAMP-2 antibodies followed by the secondary antibodies goat anti–rabbit conjugated to 10 nm gold and goat anti–mouse coupled to 5 nm gold.
Translocation of phosphatidylserine to the cell surface was monitored by using an annexin V–FITC apoptosis detection kit (Sigma-Aldrich). Cells were trypsinized, washed in PBS, and resuspended in binding buffer (10 mM Hepes/NaOH, pH 7.4, 140 mM NaCl, and 2.5 mM CaCl). Cell density was adjusted to 2–5 × 10 cells/ml. 1 μl of recombinant human annexin V–FITC/a and 2 μl PI were added to 100 μl of cell suspension; the mixture was briefly mixed and incubated for 10 min at room temperature in the dark. Afterward, 400 μl of binding buffer was added to the cells that were then analyzed by FACScan (Becton Dickinson). 15,000 events were collected in list mode fashion, stored, and analyzed by CellQuest software (BD Biosciences).
Fig. S1 shows that the depletion of microcalpain impairs macroautophagy. Fig. S2 shows that the fine ultrastructure of autophagosomes is not altered in CAPNS1 MEFs. Fig. S3 shows that rapamycin inhibits Tor activity in wild-type and CAPNS1 MEFs. Fig. S4 shows that LC3II is associated with membranes in CAPNS1 MEFs. Fig. S5 shows that the induction of LysoTracker-labeled bodies is reduced in CAPNS1 MEFs. Online supplemental material is available at . |
The transcription factor NF-κB is homo- or heterodimers formed from a multigene family that encodes five structure-related proteins: p50 (NF-κB1), p52 (NF-κB2), p65 (RelA), c-Rel (Rel), and RelB. p50/p65 heterodimer is the predominantly, although not exclusively, detectable form of NF-κB in various cells. Normally, NF-κB is sequestered in the cytoplasm in an inactive form by binding to the IκB inhibitors. Activation of NF-κB requires IκB kinase (IKK) to mediate IκB phosphorylation, an event leading to IκB degradation and consequently freeing NF-κB to translocate into the nucleus for regulating the transcription of its target genes. The IKK complex contains two catalytic subunits, IKKα and -β, and a regulatory subunit, IKKγ. The classical manner for NF-κB activation is mainly dependent on the IKKβ subunit induction (; ).
Induction of the IKK–NF-κB pathway has been observed under various cellular stresses. One important role of NF-κB activation in these biological processes is to modulate the cellular apoptotic response (; ; ; ; ; ). Many antiapoptotic genes, such as Bcl-XL, XIAP (X chromosome–linked inhibitor of apoptosis), IAP1 and -2, c-FLIP, and Bfl-1/A1, have κB elements in their promoter or enhancer regions and therefore are inducible by NF-κB to protect cells from apoptosis under diverse stimulations (; ). In addition, functional suppression of the JNK cascade, a key intrinsic cell death machinery programming cell apoptotic response to environmental changes (; ; ), has recently been proposed as a key mechanism for the antiapoptotic action of NF-κB under multiple cellular stresses, including the transformation conditions (; ; ). NF-κB suppresses the JNK cell death pathway either through the transcriptional up-regulation of a set of its targeted genes, such as the caspase inhibitor XIAP, the zinc-finger protein A20, or GADD (growth arrest and DNA damage inducible) 45β, which can act as the blockers of the JNK cascade (; ; ), or through the transcriptional suppression of GADD45α/γ, a potent activator for the JNK upstream kinase MKK4/JNKK1 (; ).
Although antiapoptosis represents a fundamental role of NF-κB in cellular stress responses, NF-κB is also capable of mediating a proapoptotic response in certain circumstances (; ; ; ). It has been shown that UVC and some anticancer drugs (daunorubicin/doxorubicin) induce NF-κB, especially the p65/RelA subunit, to recruit histone deacetylases to the promoter regions of some NF-κB–dependent antiapoptotic genes, actively suppress the expression of these genes, and promote cell death under these stress conditions (). In the case of UVB radiation, NF-κB is induced to selectively up-regulate the expression of the transcription factor and tumor suppressor Egr-1, which in turn transcriptionally activates GADD45α to trigger cell apoptosis (). Fas and FasL induction is also implicated in the NF-κB–mediated cell apoptotic process (, ). Therefore, molecular mechanisms underlying the proapoptotic action of NF-κB may be diverse, depending on the nature of the stimuli. Notably, to date, both anti- and proapoptotic effects of NF-κB are shown to rely on the p65/RelA subunit, which contains a transcriptional activation domain toward its C terminus (; ; ; ). Little is known about the role of another ubiquitously expressed subunit, p50, which lacks the transcriptional activation domain and, thus, does not have the intrinsic ability to drive transcription, like its p65 counterpart, in the course of the NF-κB–relevant biological processes.
Arsenic is a kind of environmental carcinogen involved in the incidence of multiple human cancers (; ; ; ). Meanwhile, arsenic-containing compounds have long been used as a therapeutic regimen for the treatment of human leukemia (). How arsenic engages in the promotion of oncogenesis or performs an anticancer effect still remains an enigma. With regard to its effects in tumorigenesis, cell transformation can only be observed with exposure to lower concentrations of arsenite. In contrast, high doses of arsenite stimulation appear to induce the cytotoxic effect (, ; ; ). Induction of JNK activation has been proven to be involved in both arsenite-induced killing and cell transformation (, ). However, the roles of the IKKβ–NF-κB signaling pathway in the cellular arsenite response remains controversial, depending on the doses and cell types used (; ).
Here, we show that IKKβ–NF-κB can be induced by a higher concentration of arsenite to transduce a cell apoptotic signal through up-regulation of GADD45α and, subsequently, activation of the MKK4–JNK cell death pathway. The NF-κB activity in arsenite response is specifically linked to the p50 but not the p65/RelA subunit, which mediates the function of increasing GADD45α protein stability through prevention of its ubiquitination and proteasome-dependent degradation. Our results, together with other reports (, ; ; ), indicate a dual role of NF-κB on the regulation of the intrinsic JNK cell death cascade. Most important, for the first time, to the best of our knowledge, we suggest a new model of NF-κB for regulating cellular apoptotic response through an action independent of its transcriptional activity, which is mediated by the p50 but not the p65 NF-κB subunit.
IKKβ is the dominant kinase responsible for NF-κB activation under various conditions (). To characterize the roles of the IKKβ–NF-κB pathway in the arsenite response, mouse embryonic fibroblasts (MEFs) derived from wild-type (WT) or IKKβ gene knockout (IKKβ) mice were exploited, and their responses to two doses (10 and 20 μM) of arsenite stimulations were compared. As shown in , exposure to 10 μM of arsenite did not cause detectable cytotoxic effects to both types of MEFs within 24 h. However, a considerable increase in cell death was observed for WT MEFs at 24 h under the treatment of 20 μM arsenite, but no increase in cell death for IKKβ MEFs was observed under the same conditions, assayed by both the trypan blue staining and flow cytometric analysis (). The difference of cell death in response to the 20 μM arsenite stimulation between WT and IKKβ cells was more obvious at 48 h after the treatment, when >54% of cell death was detected for WT MEFs, versus only ∼10% for IKKβ MEFs (). Accordingly, the induction of the cleavage of caspase3 and poly (ADP-ribose) polymerase (PARP), two indicators of apoptosis (), was readily detectable in WT MEFs but substantially reduced in IKKβ MEFs (). The same cells were also exposed to other cytotoxic stimuli, including UVB, benzo-[a]pyrene-7,8-diol-9,10-epoxide (B[a]PDE), and hydrogen peroxide. In contrast to this observation, IKKβ MEFs showed more sensitivity to apoptosis under these stress conditions (). These results suggest that the apoptotic-resistant phenotype of IKKβ MEFs is specifically exhibited under the arsenite stress.
To confirm that the IKKβ-dependent pathway is explored to transduce a cell death signal in arsenite response, we first showed that efficient induction of IKKβ phosphorylation (), NF-κB component (p50 and p65) nuclear translocation (), and expression of two NF-κB–targeted genes, c-Myc and Cox-2 (), were readily detectable in WT cells but not in IKKβ cells under 20 μM arsenite exposure, indicating that arsenite stimulation has the effect of activating the NF-κB pathway, and such an induction is dependent on the existence and activity of IKKβ. We next stably expressed a dominant-negative IKKβ mutant (IKKβ-KM; ; ; ) in WT MEFs to block the IKKβ–NF-κB pathway (). Meanwhile, an HA-tagged IKKβ construct was introduced into IKKβ MEFs to reconstitute the IKKβ–NF-κB pathway. As expected, stable expression of IKKβ-KM considerably suppressed the arsenite-induced cell death () and PARP cleavage () in WT MEFs; however, stable overexpression of HA-IKKβ in IKKβ-null MEFs () remarkably sensitized these cells to the arsenite-induced cell death (). Collectively, we propose that induction of the IKKβ–NF-κB pathway programs a cellular apoptotic response to arsenite stimulation.
Our previous studies have demonstrated that induction of JNKs contributes greatly to arsenite-induced cell apoptosis (). We thus tested whether induction of IKKβ–NF-κB under arsenite stress has any relevance to JNK activation. As shown in , induction of marked phosphorylation of JNK and its substrate c-Jun in WT MEFs was only observed under 20 μM arsenite stimulation, a dose with an obvious cytotoxic effect on MEFs (). The time course–dependent experiment indicated that the arsenite-induced JNK activation in WT MEFs occurred as early as 6 h and was sustained up to 24 h after the treatment (). In contrast to the observation for WT cells, no obvious induction of JNK phosphorylation was found in IKKβ MEFs in both time- and dose-dependent experiments (). However, effective induction of JNK phosphorylation was still observed in IKKβ MEFs under the UVB radiation (), suggesting that genetic ablation of IKKβ selectively affected the arsenite-induced JNK response. Again, reconstitution expression of IKKβ in IKKβ MEFs restored arsenite-induced activation of JNKs and c-Jun (), whereas overexpression of IKKβ-KM in WT MEFs substantially suppressed arsenite-induced JNK phosphorylation (). These results indicate that arsenite-induced JNK activation is dependent on the IKKβ–NF-κB signaling pathway.
To further confirm that IKKβ–NF-κB–mediated JNK activation is attributable to the arsenite-induced proapoptotic effect, a dominant-negative mutant for a JNK-specific upstream kinase MKK7 (DN-MKK7; ) was stably transfected into WT MEFs to block JNK induction (). As shown in , ectopic overexpression of DN-MKK7 in WT MEFs partially reduced arsenite-induced JNK pathway activation, associated with a remarkable attenuation of cell death in these transfected cells.
The requirement of the JNK pathway in mediating arsenite-induced apoptosis was also indicated by the observation that arsenite-induced cell death was blocked in the JNK1 and -2 double gene knockout MEFs (; ), associated with a considerable suppression of the inducible cleavage of caspase3 and PARP in these MEFs (). Similar results were also found in JNK1 and JNK2 MEFs (unpublished data). Disturbance of mitochondrial function has been shown to be involved in JNK-mediated apoptosis under certain stresses (; ; ). We then examined whether this also occurred in arsenite response. As shown in , a substantial down-regulation of the antiapoptotic protein Bcl-2, activation of the proapoptotic protein Bid (indicated by its inducible cleavage), and the release of cytochrome from the mitochondria into the cytoplasm were observed in WT MEFs upon arsenite treatment, but these events were not detectable in JNK1/2 MEFs (). No obvious changes for other mitochondrial components, including Bcl-XL, Bax, Bak, and Smac, were found during the arsenite treatment (). We thus proposed that disturbance of the mitochondrial function might contribute, at least in part, to JNK-mediated apoptosis in arsenite response.
MKK4/JNKK1 and MKK7/JNKK2 are the two upstream kinases required for the full activation of JNKs (; ). We next determined whether NF-κB activates JNKs via modulation of these two kinases' induction under arsenite stimulation. As shown in , phosphorylated MKK7 was present in both WT and IKKβ MEFs without marked induction before and after the treatment with either 10 or 20 μM arsenite. However, after treating with 20 μM arsenite, a potent induction of MKK4 phosphorylation was found in WT MEFs, but this was not detectable in IKKβ MEFs. The time course–dependent experiment indicated that the dynamics of arsenite-induced MKK4 phosphorylation was consistent with that of the JNK induction in WT MEFs (compare with ), whereas no inducible activation of MKK4 was observed in IKKβ MEFs at the indicated time points (). Furthermore, inhibition of the IKKβ–NF-κB pathway by IKKβ-KM remarkably suppressed arsenite-induced MKK4 and JNK phosphorylation in WT MEFs, whereas reexpression of IKKβ in IKKβ MEFs restored the response of arsenite-induced MKK4 phosphorylation in these cells (). These data indicate that the IKKβ–NF-κB pathway signals to activate MKK4 under the arsenite response.
To determine whether the induction of MKK4 accounts for the arsenite-induced JNK activation and cell apoptosis, DN-MKK4 was stably introduced into the WT MEFs (). As indicated in , expression of DN-MKK4 inhibited arsenite-induced JNK activation and considerably reduced cell death. Based on these results, we conclude that targeting induction of MKK4 links IKKβ–NF-κB on JNK functional activation in arsenite response.
The GADD45 family proteins (GADD45α, -β, and -γ) were proven as binding partner and activator of MEKK4 MAP3K in vitro and led to the activation of the downstream target JNKs (). We thus tested whether this family protein can act as the downstream effecter of IKKβ–NF-κB for mediating JNK activation in arsenite response. We repeatedly found that arsenite stimulation selectively up-regulated GADD45α protein levels in WT cells. However, no obvious alteration of the expression of GADD45β and -γ was observed under the same conditions (). Arsenite-induced GADD45α up-regulation was impaired in IKKβ MEFs () and was restored in IKKβ cells by reconstitution of IKKβ (). In addition, ectopic overexpression of IKKβ-KM in WT MEFs blocked arsenite-induced GADD45α accumulation (). These results indicate that arsenite stimulation up-regulates GADD45α through the IKKβ–NF-κB–dependent pathway.
To test whether GADD45α up-regulation is responsible for JNK activation in arsenite response, we first transiently transfected an HA-tagged GADD45α construct into WT MEFs and showed that overexpression of HA-GADD45α actually induced both MKK4 and JNK phosphorylation in the transfected cell population in the absence of arsenite stimulation (). We also designed a pair of siRNAs that targeted two different regions on the GADD45α mRNA. Stable transfection with a combination of these two siRNAs into WT MEFs nearly abolished arsenite-induced GADD45α up-regulation, accompanied by the blocking of arsenite-induced MKK4 and JNK phosphorylation () and a substantial decrease of arsenite- induced cell death (). Therefore, we conclude that up-regulation of GADD45α is responsible for triggering JNK activation and cell apoptosis under arsenite stress.
We have demonstrated that the NF-κB–GADD45α–MKK4–JNK pathway is responsible for mediating arsenite-induced cell apoptosis. p50 and p65 are two predominant NF-κB components expressed in a variety of cell types (; ) and have different roles in regulating the biological effects of the IKKβ–NF-κB signaling pathway under certain stimulation conditions (; ; ). To further clarify the roles of these two components in arsenite responses, MEFs from p50 and p65 gene knockout animals were exploited (). Interestingly, we found that genetic ablation of p65 did not affect arsenite-induced GADD45α up-regulation, MKK4 phosphorylation, and JNK activation (). These cells also showed sensitivity to arsenite-induced cell death (). However, all of the events of arsenite-induced GADD45α up-regulation, MKK4–JNK activation, and cell apoptosis were considerably suppressed in p50 MEFs (). Furthermore, reconstitution of p50 MEFs with p50 transfection () restored arsenite-induced GADD45α up-regulation and MKK4–JNK activation (), accompanied by the sensitivity of p50(p50) cells to arsenite-induced apoptosis (). These data disclose a property of the p50 NF-κB component in the induction of the GADD45α–MKK4–JNK cell death pathway under arsenite stress.
A recent study showed that UV radiation induces transcriptional up-regulation of GADD45α through an NF-κB–dependent pathway (). Unlike the p65 component, the p50 subunit of NF-κB does not possess the transcriptional activity for lack of the transactivation domain (; ). This implies that the induction of GADD45α observed in the arsenite response may be mechanically different to that in the UV radiation. Supporting this prediction, the RT-PCR assay showed that GADD45α mRNA was constitutively expressed in the resting WT MEFs and its expression levels did not exhibit detectable change over 8 h of arsenite treatment; however, up-regulation of GADD45α protein levels was detected as early as 2 h and, more significant, at 4 h after arsenite stimulation (). As a control, we observed that the up-regulation of GADD45α mRNA and its protein levels under UVR exhibited the consistent time course–dependent response (). This finding suggests that the induction of GADD45α in arsenite response might occur at the posttranscriptional level, most possibly by the modulation of the protein stabilities. To test this hypothesis, WT MEFs were treated with MG132, an established inhibitor of protein degradation by disruption of the proteasome system (). As indicated in , similar to that of the arsenite stimulation, MG132 treatment resulted in a remarkable accumulation of GADD45α proteins. Meanwhile, when the culture was withdrawn from MG132 and subjected to the treatment of cyclohexamide (CHX), a protein synthesis inhibitor, to block the de novo production of proteins, the preaccumulated GADD45α by MG132 could undergo a gradual degradation with the reassembly of the proteasome system (). MG132 treatment also resulted in GADD45α protein accumulation in IKKβ and p50 MEFs, although arsenite failed to accumulate this protein in both cells (). These results together demonstrate that the expressed GADD45α is actively subjected to a proteasome-dependent degradation process in resting cells, and such a process is blocked in arsenite response via the IKKβ–p50–dependent pathway.
Ubiquitination is the common mechanism for dictating proteins into proteasome-dependent degradation (). We next determined whether arsenite induces GADD45α accumulation in WT MEFs by suppressing its ubiquitination. GADD45α was immunoprecipitated with specific anti-GADD45α antibodies, and its ubiquitination status was analyzed with anti-ubiquitin antibodies. As shown in (left), only a weak ubiquitin signal was detected in the sample of untreated WT cells. This result was predictable, as there was little GADD45α protein present in these resting cells because of its rapid degradation. However, a strong ubiquitin signal was detected in the lane for MG132-treated cells, indicating that the accumulated GADD45α protein by MG132 is highly ubiquitinated. Interestingly, the ubiquitin signal for GADD45α in the arsenite-treated cells was almost as weak as that in the untreated control sample, although its amount in these cells is the same as that in the MG132-treated cells, suggesting that the accumulated GADD45α by arsenite stimulation is less ubiquitinated. The effect of arsenite on suppressing GADD45α ubiquitination was also manifested in the MG132/arsenite double-treated cells, in which the ubiquitin signal for GADD45α was substantially decreased as compared with that in cells treated with MG132 alone. The ubiquitination status of GADD45α was also analyzed in IKKβ and p50 MEFs (, middle and right). As expected, MG132 treatment also accumulated large amounts of highly ubiquitinated GADD45α in both cells. However, the effect of arsenite on suppressing GADD45α ubiquitination, which was observed in WT MEFs, was not exhibited in both gene knockout cells, indicated by the same high levels of GADD45α ubiquitination in MG132/arsenite and MG132-treated cells. These results demonstrate that arsenite stimulation provokes a response to suppress GADD45α ubiquitination, depending on the IKKβ–p50 signaling pathway.
To further clarify whether the action of IKKβ–p50 on preventing GADD45α degradation under arsenite stimulation is through preventing this protein from ubiquitination or by mediating a process of deubiquitinating the ubiquitinated proteins, we designed an experimental system in which the MEFs were pretreated with MG132 to accumulate a certain amount of GADD45α proteins in vivo. After the pretreatment, MG132 was removed from the medium and the protein synthesis inhibitor CHX was added to the cells alone or in combination with arsenite. In this way, we were able to analyze the effect of arsenite on the dynamic degradation of the preubiquitinated GADD45α proteins during the course of the reassembly of the proteasome system in vivo. As shown in , the preaccumulated GADD45α proteins disappeared in all tested cells within 12 h after removal of MG132 in the absence of arsenite. With arsenite stimulation, GADD45α proteins remained stable at 12 h after withdrawal of MG132 in WT and p65 MEFs. However, almost no GADD45α proteins were detectable at this time point in either IKKβ or p50 MEFs under the same conditions. Again, reintroduction of IKKβ and p50 into the according gene knockout cells restored the effect of arsenite on preventing GADD45α degradation. In addition, suppression of p50 expression by its specific siRNA (unpublished data) partially disrupted the effect of arsenite in WT MEFs. Based on these results, we propose that IKKβ–p50 might exert an effect for mediating the deubiquitination of GADD45α under the arsenite response, and this might contribute, at least in part, to the phenomenon of the arsenite-induced GADD45α up-regulation.
The IKKβ–NF-κB signaling pathway transmits signals essential for cell survival in a variety of physiological and pathological processes (). Deregulation of this signaling pathway has been directly implicated in the evasion of the apoptotic responses of many human cancers. As a result, targeted inhibition of the IKKβ–NF-κB pathway has been proposed as a strategy for the development of new anticancer drugs (). Interestingly, it has recently been demonstrated that this signaling pathway can also be induced to exert proapoptotic effects in response to some apoptotic inducers (, ; ; ). Although this finding raises caution regarding the rationale of the proposed use of IKKβ–NF-κB inhibitors in combination with other anticancer drugs in a clinical setting, it also provides an alternative for targeting IKKβ–NF-κB for cancer therapy and chemoprevention. Therefore, identification of the conditions and the according mechanisms for converting the antiapoptotic role of this signaling pathway to the proapoptotic action in cancer cells will be of medical significance.
Arsenite was regarded as both a carcinogen and tumor-therapeutic agent because of its ability to mediate cellular apoptotic or transformation effects under different conditions (; ; ). Our previous studies have disclosed that a low dose of arsenite (1.25–5 μM) is capable of promoting the cell cycle by the induction of cyclin D1 expression through the IKKβ–NF-κB pathway (). However, the precise mechanism of arsenite on the tumor therapeutic effect is not well clarified. In this study, we demonstrate that at a higher dose (20 μM) arsenite can induce an activity of the IKKβ–p50 NF-κB complex to program cell apoptosis via triggering the GADD45α–MKK4–JNK cell death cascade (). These novel findings demonstrate that IKKβ–NF-κB has the intrinsic competence to directly turn on the cell death program through the activation of JNK cascade under stress conditions and add a new content to the cross talk between NF-κB and JNK signaling pathways in cell life and death decisions (; ; ). As suppression of the JNK activation by NF-κB is essential for cancer cell survival (; ), our finding that arsenite is capable of subverting the inhibitory effect of NF-κB on JNK activation may be of medical importance. It merits further elucidation as to whether such a mechanism underlies the effect of this reagent in cancer therapy.
How NF-κB can be induced to functionally suppress or activate the JNK cell death pathway is an interesting question. Previous reports (; ) and the data in this study suggest that the GADD45 protein family (GADD45α, -β, and -γ) may serves as a key modulator for functionally linking IKKβ–NF-κB to the JNK pathway. In this model (), by differential regulation of GADD45 family members (up- or down-regulated), NF-κB can be induced to suppress or activate the JNK cell death pathway, thereby exerting its pro- or antiapoptotic effects according to the nature of the stresses. Our results have disclosed the predominant role of GADD45α in arsenite-induced cell death response. In addition, previous data have shown that suppression of GADD45α and -γ contributed to cancer cell survival, whereas induction of GADD45β antagonized TNFα-induced killing (; ; ). These results suggest that GADD45α and -γ mainly contribute to the NF-κB–mediated cell death effect, whereas GADD45β appears to involve NF-κB–dependent cell survival.
Interestingly, we noticed that all of the previous studies regarding the regulatory effect of NF-κB on GADD45 expression are exclusively dependent on the transcriptional activity of NF-κB, such as the transcriptional suppression on GADD45α and -γ by constitutively active NF-κB in cancer cells and the transcriptional induction of GADD45β by NF-κB in response to TNFα stimulation. In addition, NF-κB activity in these responses is mediated by the cooperative action of the p65 and p50 subunits (; ). In contrast, we demonstrate that the role of NF-κB on mediating the arsenite-induced GADD45α up-regulation is by preventing this protein from the ubiquitin–proteasome–dependent degradation, instead of the transcriptional induction of the GADD45α gene. Furthermore, this effect of NF-κB only relates to the p50 but not the p65/RelA subunit. Therefore, our data, for the first time, has provided a strong evidence for a new function of the NF-κB p50 subunit, which is independent of the classical NF-κB transcriptional activity but relates to the posttranslational modification of a protein.
GADD45 is originally described as a p53-dependent and stress-inducible gene that also can be regulated by many other transcription factors (; ; ). In addition, recent evidence suggests that protein ubiquitination represents another way for GADD45 regulation (). Here, we showed that arsenite-induced GADD45α accumulation through preventing this protein from ubiquitination-dependent degradation, further emphasizing the importance of this kind of mechanism on the functional control of GADD45 family members.
The mechanism responsible for the p50-mediated GADD45α modification is currently unknown. Clearly, it has some relevance to the presence of IKKβ protein, indicated by the similar manner of the changes on ubiquitination, degradation, and accumulation of GADD45α proteins under arsenite stimulation in IKKβ and p50 MEFs (). How IKKβ functionally links to the NF-κB p50 subunit in arsenite response remains unclear. However, we have observed that a transient interaction between IKKβ and p50 can be induced by arsenite in WT MEFs in the coimmunoprecipitation assay, and the time course–dependent response of the IKKβ–p50 complex formation (between 1 and 4 h after arsenite stimulation; ) is consistent with the appearance of the accumulated GADD45α proteins (2–4 h after arsenite stimulation; ). In contrast, there is no obvious interaction between IKKα and p50 under the same conditions (). This result provided us with an important clue suggesting that the inducible interaction between IKKβ and p50 by arsenite might confer a novel property to this putative complex for suppressing GADD45α ubiquitination. Because arsenite stimulation increased the ubiquitination of total cellular proteins (unpublished data) while selectively decreasing the GADD45α ubiquitination level, we proposed that the most probable action of IKKβ–NF-κB p50 is to target GADD45α-specific E3 ubiquitin ligase or the cellular deubiquitination enzymes. The functional link between IKKβ and p50 and the possible mechanism responsible for the IKKβ–p50–containing complex on regulating GADD45α ubiquitination and degradation are currently under investigation.
In summary, this study demonstrates for the first time that the IKKβ–NF-κB signaling pathway can transduce the apoptotic signal through eliciting the intrinsic GADD45α–MKK4–JNK cell death route and provides a novel scenario for the implication of the cross talk between these two master pathways in the cell life and death decisions encountering diverse stresses. Furthermore, we disclosed a new function of NF-κB p50 subunit, which can act as a critical regulator of the ubiquitin–proteasome–dependent modification of GADD45α. Our study, together with other recent findings, clearly suggests the existence of different mechanisms underlying the NF-κB pathway. Elucidation of these issues should shed important insight on the understanding of how NF-κB integrates diverse stimuli to generate a unified outcome suitable for a specific situation, a question raised in the current field of NF-κB investigation.
The plasmid expressing the kinase mutant of IKKβ (Flag-IKKβ-KM) was a gift from H. Nakano (Juntendo University, Tokyo, Japan; ). The plasmid expressing HA-tagged full-length GADD45α (HA-GADD45α) was described in a previous study (). The plasmid containing p50 cDNA was provided by J. Ye (Louisiana State University, Baton Rouge, LA). The plasmids expressing dominant-negative MKK4 (HA-DN-MKK4) or dominant-negative MKK7 (HA-DN-MKK7) and the full-length IKKβ (HA-IKKβ) were described in a previous study (). The antibodies against phospho-IKKα/β, IKKα, IKKβ, phospho-JNK, JNK, phospho-MKK4, MKK4, phospho-MKK7, MKK7, phospho-c-Jun, c-Jun, caspase3, PARP, Bcl-2, Bcl-XL, Bax, and Bak were purchased from Cell Signaling Technology. The antibodies against GADD45α, -β, and -γ; c-Myc; p65; p50; Bid; and the agarose-conjugated anti-p50 antibody were obtained from Santa Cruz Biotechnology, Inc. Anti–cytochrome and Smac antibodies were purchased from BD Biosciences. Anti-HA and ubiquitin antibodies were purchased from Upstate Biotechnology; anti-FLAG and anti–β-actin antibodies were obtained from Sigma-Aldrich; and anti-Cox2 antibody was purchased from Cayman Chemical. MG132 and CHX were purchased from Calbiochem.
IKKβ MEFs were a gift from M. Karin (University of California, San Diego, La Jolla, CA; ). The p50 and p65 MEFs and their corresponding WT MEFs were provided by J. Ye. The JNK1/2 MEFs were provided by K. Sabapathy (National Cancer Center, Singapore; ). The WT and the gene knockout MEFs were maintained in DME (Calbiochem) supplemented with 10% FBS, 1% penicillin/streptomycin, and 2 mM -glutamine (Life Technologies) at 37°C. Cell transfections were performed with Lipofectamine reagent (Invitrogen) according to the manufacturer's instruction. For stable transfection, cultures were subjected to either hygromycin B or G418 drug selection, and cells surviving from the drug selection were pooled as stable mass. These stable transfectants were cultured in the selective drug-free medium for at least two passages before used for according experiments. For transient transfection, cells were harvested at 36 h after transfection for immunoblot analysis.
The two sequences, 5′-gatcctgccttaagtcaa-3′ and 5′-agtcgctacatggatcagt-3′, on GADD45α mRNA were selected by the siRNA Target Finder (; Ambion) as siRNA target sites and were expressed by using the GeneSuppressor system (Imgenex). The constructs containing the corresponding scrambled target sequences were used as controls. The established constructs were either separately transfected or cotransfected into WT cells for stable expression.
The arsenite-induced cell death was determined by trypan blue exclusion assay and flow cytometric analysis after propidium iodide staining of the nuclei. For the live cell microscopy, the images were taken with an inverted microscope (CKX41; Olympus) equipped with the achromatic objectives (10×; NA 0.25; working distance 8.8 mm) and the digital camera (DP12-2; Olympus) at room temperature. The images were analyzed using Photoshop (Adobe).
Whole cell extracts were prepared with the cell lysis buffer (10 mM Tris-HCl, pH 7.4, 1% SDS, and 1 mM NaVO). Cytoplasmic and nuclear proteins were prepared with Cellytic nuclear extraction kit (Sigma-Aldrich) following the manufacturer's protocols. The mitochondrial proteins were prepared with the Mitochondria isolation kit for mammalian cells (Pierce Chemical Co.) following the manufacturer's protocols. Protein concentrations were determined by the protein quantification assay kit (Bio-Rad Laboratories). 30 μg of proteins were resolved by SDS-PAGE, probed with the indicated primary antibodies, and incubated with the AP-conjugated second antibody. Signals were detected by the enhanced chemifluorescence Western blotting system as described in our previous reports (; ). The images were acquired by scanning with the phosphoimager (model Storm 860; Molecular Dynamics) at room temperature.
For ubiquitination studies, cells were treated with 10 μM MG132, 20 μM arsenite, or the combination of these two reagents for 12 h and then lysed in the cell lysis buffer (1% Triton X-100, 150 mM NaCl, 10 mM Tris, pH 7.4, 1 mM EDTA, 1 mM EGTA, 0.2 mM NaVO, 0.5% NP-40, and complete protein inhibitors mixture tablet) on ice. 0.5 mg total lysate was precleared by incubation with Protein A/G plus-agarose (Santa Cruz Biotechnology, Inc.) and incubated with 2 μg anti-GADD45α monoclonal antibody for 2 h at 4°C. 40 μl Protein A/G plus-agarose were added into the mixture and incubated with agitation for an additional 4 h at 4°C. The immunoprecipitated samples were washed with the cell lysis buffer and subjected to the Western blot assay with the anti-ubiquitin antibody. To detect putative p50 binding proteins, cell lysates from arsenite-treated WT cells were incubated with agarose-conjugated anti-p50 antibody, and the immunoprecipitated samples were subjected to the Western blot assay with anti-IKKα and -IKKβ antibodies, respectively.
Total RNA was extracted with Trizol reagent (Invitrogen), and cDNAs were synthesized with ThermoScript RT-PCR system (Invitrogen). Two oligonucleotides (5′-atgactttggaggaattctcg-3′ and 5′-cactgatccatgtagcgacct-3′) were used as the specific primers to amplify mouse GADD45α cDNA. The mouse β-actin cDNA fragments were amplified by the primers 5′-gacgatgatattgccgcact-3′ and 5′-gataccacgcttgctctgag-3′. |
Granzymes are cytotoxic serine proteases used by cytotoxic lymphocytes (CLs) to destroy virus-infected and malignant cells. They are delivered into the cytoplasm of the target cell by the pore-forming protein perforin. Once inside the target cell, granzymes cleave specific proteins and trigger apoptosis (for review see ).
Humans have 5 granzymes (A, B, H, K, and M), whereas mice have 10 granzymes, having duplicated the ancestral granzyme B/H gene. In the absence of reported human granzyme deficiency, understanding of granzymes has been built by experimental approaches involving knockout mice or the use of purified granzymes on cells or cell extracts. For example, loss of granzyme A (GrA) or granzyme B (GrB) in mice is not catastrophic, but loss of both results in a severe immunological defect. Numerous studies using purified human and rodent granzymes have identified substrates, regulators, and trafficking pathways, whereas granzyme- deficient mice are widely used to probe the contribution and importance of granzymes to immunity ().
GrA and GrB are the major and most-studied granzymes. GrA cleaves substrates after basic residues and induces caspase-independent apoptosis (for review see ). GrB cleaves after acidic residues, primarily Asp, and induces apoptosis by directly or indirectly activating caspases, and by cleaving other proapoptotic proteins such as Bid or ICAD/DFF45 (for review see ). Studies using CL from knockout mice confirm that GrA and GrB trigger distinct apoptotic pathways and are thus complementary ().
The high sequence homology and conserved primary cleavage specificity of human and mouse granzymes has led to widespread and interchangeable use of human and mouse enzymes in experimental systems, usually without side-by-side comparisons being made. Because GrA is a unique gene in both species, it is reasonable to assume an orthologous relationship between mouse GrA (mGrA) and human GrA (hGrA), but recent structural data show that they have distinct extended substrate specificities (), raising the possibility that they possess distinct functions. The situation with GrB is even more complicated. hGrB has multiple paralogues in the mouse and rat, suggesting that the roles fulfilled by a single enzyme in humans may be split among several in rodents (; ).
Without detailed comparative knowledge of the characteristics of human and mouse granzymes, data from mice cannot be extrapolated with confidence to humans. Here, we show that the cytotoxicity, substrate preferences, and inhibitor interactions of hGrB and mGrB are substantially different. We also extend the study of to show that hGrA and mGrA have markedly different cytotoxic potential. Such differences probably reflect the two species' response to differing immune evolutionary pressure.
Granzyme cytotoxicity can be studied in vitro by adding purified perforin and granzyme to cultured cells and monitoring cell survival (Fig. S1, available at ). Keeping the concentration of perforin constant, granzyme cytotoxicity is quantitated using a dose–response approach, in which the amount of granzyme required to kill 50% of the target cells (EC50) is measured (). Initial experiments using this system with either human or mouse cells showed that hGrB is substantially more cytotoxic than mGrB.
This apparent difference in cytotoxicity between mGrB and hGrB was confirmed using several different targets, as cell type–dependent differences in susceptibility to granzymes have been reported (). In one human cell line and three mouse cell lines examined in detail, hGrB was consistently more cytotoxic than mGrB (). With the exception of YAC-1, cells were 10–60-fold more sensitive to hGrB (). Killing of mouse YAC-1 cells required similarly high levels of hGrB or mGrB, suggesting that a key substrate is low or absent in these cells. mGrB was also 30-fold less effective in killing mouse lymphoma cells ().
To test whether poorer uptake of mGrB explains its inefficient killing, we compared accumulation of fluoresceinated mGrB and hGrB in Jurkat cells (). No impairment in the rate or extent of accumulation of mGrB versus hGrB was noted in these cells (Fig. S2, available at ) or in the mouse P815 and lymphoma cell lines used in this study. The differences in cytotoxicity between mGrB and hGrB also remained when the bacterial porin, streptolysin O (SLO) was used instead of perforin (). SLO admits proteins directly to the cytoplasm via transient plasma membrane pores () and provides a perforin-independent mechanism for granzyme uptake, as shown by its ability to deliver mutant granzymes that cannot be delivered by perforin (). The fact that mGrB killing is as inefficient in SLO as it is in perforin suggests that there is no restriction in perforin-mediated delivery of mGrB. Thus, differences in uptake or delivery of hGrB and mGrB do not explain their markedly different cytotoxic potential.
To investigate the basis for the differences in cytotoxicity between mGrB and hGrB, we compared their activities on peptide substrates and on the natural substrates procaspase 3 and Bid (; ). Combinatorial peptide substrate analysis has identified the peptide IleGluXxxAsp↓ as an optimal recognition and cleavage sequence for hGrB and rat GrB (; ). This appears in human and mouse Bid, and both are cleaved at this site by hGrB (). Although mGrB cleaves Bid (), the cleavage site has not been determined. A similar P4-P1 cleavage sequence is present on procaspase 3, but the downstream (P′) sequence differs from Bid. The nomenclature of cleavage site positions of a protease substrate follows . The protease cleaves the peptide bond between P1 and P1′. Adjacent residues in the N-terminal direction are numbered P2, P3, P4, etc. On the carboxyl side of the cleavage site, residues are numbered P2′, P3′, P4′, etc. The P1 residue is accommodated in the S1 pocket of the catalytic cleft, the P2 residue is accommodated in the S2 pocket, and so forth.
mGrB and hGrB both cleaved the small peptide substrate Boc-AlaAlaAsp-thio benzyl ester (AAD-sbzl) with similar efficiency (, right), consistent with their known Aspase activities, indicating that their apparent functional distinctions are not due to differences in ability to hydrolyse a P1(Asp)-P1′ bond. As expected, hGrB also efficiently cleaved IleGluThrAsp-NA (IETD-NA; , left). In contrast, mGrB hydrolyzed IETD-NA 30-fold less efficiently than hGrB (, left), suggesting that it cannot comfortably accommodate this peptide in its substrate binding pocket.
The markedly inferior ability of mGrB to cleave IETD-NA also suggested that it would not cleave full-length Bid or procaspase 3 efficiently. To test this, we produced S-labeled mouse Bid and mouse procaspase 3 and incubated the protein with hGrB or mGrB. Under these conditions, complete cleavage of Bid was achieved in 25 nM hGrB, whereas 2 μM mGrB failed to completely cleave Bid (, right). Thus, mGrB cleaves Bid 80–100-fold less efficiently than hGrB. In contrast, the two granzymes cleaved procaspase 3 with similar efficiency (, left), suggesting that other factors such as P′ residues also influence cleavage ( and ).
Previous studies using hGrB have indicated that Bid cleavage is a key event in GrB-mediated apoptosis (; ; ). GrB initiates procaspase 3 activation by cleaving it once, but full activation requires a second cleavage, autocatalytically or via another caspase. This is illustrated in and Fig. S3 (available at ) by the inclusion of the caspase inhibitor z-VAD in the reactions. In vivo this second cleavage step is prevented by inhibitor of apoptosis protein (IAP) bound to caspase 3 and will not occur unless IAP is displaced by Smac/Diablo released from mitochondria by Bid (; ). Thus, in the absence of Bid, hGrB cannot drive full caspase 3 activation.
To examine the importance of Bid in mGrB-mediated apoptosis, we compared the ability of hGrB and mGrB to kill mouse cells lacking Bid. As reported previously (), hGrB efficiently killed mouse lymphoma cells containing Bid (EC50 2.5 U/ml), but was >1,000-fold less cytotoxic toward lymphoma cells lacking Bid (, left). In contrast, mGrB killed the Bid+/+ and Bid−/− cells with similar efficiency, indicating that Bid cleavage is not essential for mGrB cytotoxicity (, right). These observations cannot be explained by differences in procaspase 3 levels between these lines or by differences in the ability of mGrB or hGrB to cleave procaspase 3 either in vitro or in cell extracts (compare and Fig. S3). It is also unlikely that mGrB can circumvent Bid by directly and fully activating procaspase 3, as z-VAD blocks the second cleavage event when procaspase 3 is exposed to mGrB ( and Fig. S3). Collectively, the aforementioned results indicate that mGrB and hGrB have similar but distinct substrate specificities and that they trigger death differently.
The granzymes were used to probe a library in which the fourth residue of the nonamer had been fixed as Asp. As shown in , phage enriched by hGrB treatment displayed a P4-P3-P2-P1-P1′-P2′-P3′ consensus sequence, [I/V][G/E]ADVLV. This closely matches the P4-P1 sequence previously identified by combinatorial peptide analysis (; ) and validates the phage display approach. In contrast, treatment with mGrB yielded the consensus sequence [I/L]X[F/Y]DXGV. Specifically, the preferences at the P4, P2, and P2′ positions are different between mGrB and hGrB. mGrB shows an equal preference for Ile or Leu at P4 and an aromatic residue at P2, whereas hGrB strongly prefers Ile at P4 and has a relaxed requirement at P2. mGrB also has a very strong preference for Gly at P2′, whereas hGrB has no obvious preference at this position.
To confirm that mGrB can cleave a substrate with the aforementioned characteristics, we synthesized a quenched-fluorescence substrate peptide based on one of the phage sequences that met the consensus [L/I]X[F/Y]DXGX. This peptide abz-LEYDLGALK(dnp)S was cleaved by both mGrB and hGrB, but kinetic analysis showed that it is a 370-fold better substrate for mGrB (, compare specificity constants). These results show that the consensus substrate sequences for hGrB and mGrB differ at the key P4, P2, and P2′ positions and indicate structural differences in the corresponding pockets in the catalytic clefts of the two enzymes.
As shown in , mGrB cleaves Bid poorly but procaspase 3 relatively well. The substrate phage display results suggest that mGrB requires Gly at the P2′ to efficiently cleave substrates, but both mouse and human Bid have Glu at this position. To determine whether P2′ Gly enhances cleavage by mGrB, we synthesized a peptide substrate based on the GrB cleavage site in mouse Bid (IEPD↓SESQ) and a derivative with P2′ Gly (IEPD↓SGSQ). Both hGrB and mGrB cleaved the wild-type (wt) Bid substrate, but hGrB cleaved it >40-fold more efficiently (). This is consistent with the 30-fold difference in the ability of the two proteases to cleave IETD-NA (). In contrast, mGrB cleaved the P2′ Gly substrate sevenfold better than the wt Bid substrate, whereas hGrB showed a threefold improvement over cleavage of the wt sequence, consistent with the slight overrepresentation of Gly at P2′ in the phage display data (). Thus, mGrB prefers P2′ Gly in natural substrates, and its poor cleavage of Bid can be explained partly by lack of Gly at this position. Given the conservation of the P4-P1 sequence in Bid and procaspase 3 (IEXD), these results also explain why procaspase 3 is apparently a better substrate of mGrB than Bid (): mouse procaspase 3 has a P2′ Gly at its GrB cleavage site.
The inability of mGrB to cleave after IETD suggested differences between mGrB and hGrB in the substrate binding cleft, presumably in the pockets that interact with the side chains of the P4, P3, or P2 residues of a substrate. Inspection of the structure of hGrB bound to the tetrapeptide aldehyde inhibitor IEPD-CHO () revealed that Asn218 in the S3 pocket forms hydrogen bonds with the P3 Glu of the substrate, whereas Tyr174 in the S4 pocket undergoes hydrophobic interactions with P4 Ile of the substrate (). These two residues are conserved in rat GrB and have been shown to influence substrate specificity (). In contrast, they are not conserved in mGrB (, Arg174 and Lys218). Thus, we predicted that mutating mGrB toward hGrB by replacing Arg174 with Tyr and Lys218 with Asn would enable it to more efficiently cleave IETD and Bid.
/
values on both substrates similar to those of hGrB (compare with ). In addition, the mutant mGrB showed enhanced ability to cleave Bid, together with a slight increase in the ability to cleave procaspase 3 (). This translated into an increased ability to kill the Bid+/+ lymphomas (). Interestingly, the mutated mGrB also showed an increased ability to kill the Bid−/− lymphomas (), suggesting that rather than being simply converted to hGrB selectivity, it has gained the ability to cleave hGrB-specific targets (e.g., Bid) while retaining the ability to cleave mGrB-specific targets. Finally, mutant mGrB showed 100-fold enhanced ability to kill P815 cells, with similar efficiency to hGrB (, left).
hGrB is regulated by the intracellular protease inhibitor, PI-9, a member of the serpin superfamily (; ). The serpin inhibitory mechanism requires recognition of an exposed reactive center loop (RCL) as a potential substrate by a cognate protease, followed by cleavage between the P1 and P1′ residues. This triggers conformational change in the serpin, distorting the active site of the protease and trapping an intermediate consisting of the protease bound to the serpin (for review see ). hGrB cleaves the PI-9 RCL after Glu, consistent with its preference for acidic residues, and is very rapidly trapped ().
Mice have seven PI-9 homologues (), but the counterpart of PI-9 is not immediately obvious. The closest homologue, SPI6, has not been tested with mGrB but inhibits hGrB 20-fold slower than PI-9 (). When overexpressed, SPI6 protects dendritic cells () and T cells () from CL attack and increases the survival of CD8 memory cells (). Loss of SPI6 results in diminished survival of CTL (). However, none of these studies has demonstrated that SPI6 directly or efficiently inhibits mGrB. To determine whether SPI6 is a mouse GrB inhibitor, we identified the cleavage site for mGrB in the SPI6 RCL and measured the kinetics of the interaction, assessed the ability of SPI6 and mGrB to interact in CL extracts, and tested whether SPI6 protects cells against death mediated by purified mGrB and perforin.
The hGrB–PI-9 interaction requires cleavage of the PI-9 RCL at Glu340 (; ), and mGrB also cleaves PI-9 at this point (unpublished data). N-terminal sequencing of the 6-kD fragment generated by treatment of SPI6 with mGrB or hGrB showed that cleavage occurs at Cys339, adjacent to the predicted P1 Cys340 (). This is a previously unrecognized cleavage specificity of GrB and broadens the range of potential substrates.
Although cleavage at Cys339 might allow capture of some mGrB by SPI6, the kinetics of the interaction will indicate whether mGrB is effectively regulated by SPI6 in vivo. Accordingly, we measured the rate constant (
) and stoichiometry of inhibition of SPI6 and mGrB complex formation (). The results strongly suggest that SPI6 is a poor inhibitor of mGrB.
indicates that complex formation is very slow. SPI6 is even less effective in inhibiting hGrB (). To put these results into context, PI-9 inhibits hGrB with equimolar stoichiometry and a rate of 1.2 × 10 Ms (). Furthermore, the inhibition of mGrB by SPI6 is slower than the inhibition of hGrB by the viral serpin CrmA (), which does not block GrB-mediated apoptosis in vivo.
Another signature of a serpin–protease interaction is the formation of SDS-stable complexes. For example, when human CLs are lysed, GrB is released from granules and rapidly binds cytosolic PI-9, and because GrB is in excess, essentially all of the PI-9 moves into complex (, left; ). Similarly, in extracts of mouse CLs probed with anti-mGrB or anti-SPI6 antibodies, a complex containing mGrB and SPI6 was evident (, right). However, less than half of the SPI6 moved into complex, consistent with a less efficient interaction between this serpin and GrB, and indicating that the slow in vitro kinetics are not due to lack of a SPI6 cofactor present in CLs. To confirm that complex formation is GrB dependent, extracts were prepared from CLs taken from mice lacking GrB. No complex was evident in these cells (unpublished data).
If mouse and human GrB are orthologous, PI-9 and SPI6 should provide cross-species protection from GrB-mediated cell death. To test this, we challenged mouse P815 cells expressing SPI6 with hGrB and human SKW6.4 cells expressing PI-9 with mGrB (). As expected, PI-9 protected SKW6.4 cells against hGrB (, top left), increasing the EC50 ∼20-fold. SPI6 protected P815 cells against mGrB (, bottom right), increasing the EC50 >14-fold. Furthermore, the inactive SPI6T327R hinge mutant did not prevent mGrB-mediated death, indicating that the classic serpin inhibitory mechanism is involved in SPI6 cytoprotection (unpublished data). However, PI-9 did not protect cells against mGrB (, top right), nor did SPI6 effectively protect cells against hGrB (, bottom left).
Finally, the humanized mutant mGrB almost completely lost its ability to be regulated by SPI6, exhibiting a similar cytotoxicity profile to hGrB on SPI6-expressing cells (, right). This is consistent with the alterations to its active site cleft and substrate interactions and emphasizes the high specificity of a cognate serpin–protease interaction. From the aforementioned experiments, we concluded that mGrB is indirectly inhibited by SPI6 and that PI-9 and SPI6 are not functionally interchangeable. These results confirm that mGrB and hGrB have distinct substrate specificities and indicate that they are regulated differently in vivo.
The functional differences that we identified between mGrB and hGrB raised the possibility that such interspecies differences would also be evident in other granzymes. Although mouse and human GrA cleave the peptide substrate benzyloxy-carbonyl-Lys-thiobenzyl ester with similar efficiency (; ), structural and substrate preference analysis of GrA has identified differences between the mouse and human enzymes in the active site cleft (). This study identified an optimal P4-P1 substrate for hGrA as VANR↓, whereas the optimal mGrA substrate was identified as GYFR↓. Mutation of the hGrA S4 and S3 pockets converted it to a mouse-like GrA ().
To determine whether these structural differences translate into differences in cytotoxic potential, we compared the ability of mGrA and hGrA to kill human or mouse cells in the presence of perforin. We used two different death assays: our standard MTT (3-[4,5-Dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) test, which measures loss of mitochondrial respiration (), and Cr release, which measures loss of membrane integrity (). As shown in , mGrA efficiently killed mouse and human cells, but hGrA did not. mGrA-induced death was perforin dependent, as cells exposed to mGrA alone survived (). The inability of hGrA to kill cells was not related to the source of the protease, as both recombinant and native forms were used. All of the hGrA used was in the physiological 60-kD dimeric form, which could be reduced to the 30-kD monomeric form (). Monomeric hGrA also failed to kill cells (unpublished data).
GrA and GrB are believed to be the primary cytotoxins in human and mouse CLs. The evidence presented here clearly demonstrates substantial interspecies distinctions that suggest caution when using mouse models to elucidate the molecular and pathophysiological functions of these proteases in humans. For example, loss of GrB may be much more devastating to humans than the mild phenotype of GrB-deficient mice would suggest, given that hGrB is more efficient than mGrB and that hGrA is apparently nonlethal and could not compensate for GrB absence.
Such interspecies differences between granzymes are illustrated by the GrB–Bid interaction, which indicates that distinct death pathways are initiated by hGrB and mGrB (). Our results show that hGrB is a much more effective killer than mGrB because it can cleave and activate Bid, whereas mGrB cleaves Bid poorly and does not require it for initiating apoptosis. The requirement for Bid in hGrB-mediated killing is not an artifact arising from its use on mouse target cells, as Bcl-2 overexpression in human cells blocks hGrB cytotoxicity but can be relieved by Bid coexpression (). The failure of hGrB to kill in the absence of Bid indicates that it lacks the capacity to independently and fully activate effector caspases, as suggested previously (; ). Its ability to cleave other substrates must also be insufficient to cause death. In contrast, mGrB probably directly activates another effector caspase, such as procaspase 7, or it relieves IAP inhibition of caspase 3 independently of Bid. It may also cause caspase-independent death by cleaving substrates such as ICAD or as-yet-undefined mitochondrial targets ().
The differences between hGrB and mGrB are also emphasized by our ability to humanize mGrB and improve its capacity to cleave Bid. However, humanized mGrB remains relatively inefficient compared with hGrB, indicating that additional changes to the mGrB substrate binding pocket are required to fully accommodate Bid. Supporting evidence for differences between mGrB and hGrB can also be found in previous studies. For example, synthetic catalytic inhibitors designed to hGrB do not inhibit mGrB (); the human adenoviral protein L4-100K inhibits hGrB but not mGrB (); mGrB does not process human Bid (); and mGrB does not require Bax or Bak for cytotoxicity, whereas hGrB does (; ).
Similarly, hGrA and mGrA do not appear to be functionally equivalent, as we have shown that mGrA is cytotoxic and that hGrA is not. This is supported by several other lines of evidence. Modeling of mGrA on the structure of hGrA has identified differences in their active site clefts, substantiated by differences in peptide substrate preference and the conversion of hGrA to an mGrA-like protease by mutation of the S4/S3 pockets (). Early work showed that mGrA is cytotoxic when produced with perforin in rat basophilic leukemia cells (), and many studies on the mechanisms of GrA-mediated death have used CTL and target cells from mice (, ). In contrast, recent studies using specific inhibitors show that hGrA is a minor contributor to CL cytotoxicity (). Such findings do not necessarily overturn the idea that hGrA engages apoptotic pathways by cleaving specific subcellular proteins and generating reactive oxygen species (; ), but they suggest that it does so much less efficiently than mGrA and that it is not sufficient to induce cell death.
The differences between hGrB and mGrB also extend to their regulation, as indicated by differences between PI-9 and SPI6. PI-9 is an efficient hGrB inhibitor that protects CLs and bystanders from misdirected GrB (; ; ), and SPI6 is a putative counterpart of PI-9 (; ; ). Mice lacking SPI6 have fewer CTL, and remaining CTL have fewer granules and less evidence of cytoplasmic mGrB (). Crossing SPI6-deficient mice with mGrB-deficient mice rescues the CTL defect, but because the latter mice also lack several mGrB paralogues (), it cannot be assumed that SPI6 directly controls mGrB. Although the accumulating evidence is consistent with a cytoprotective role for both PI-9 and SPI6 in CLs, our present work strongly suggests that the question of the mechanism of cytoprotection by SPI6 remains open, as it is not an effective mGrB inhibitor. Given that other mGrB paralogues are cytotoxic, four possibilities exist: the primary target of SPI6 is a cytotoxic mGrB paralogue; SPI6 simultaneously controls multiple mGrB paralogues through a one-size-fits-all RCL; mGrB or paralogues activate a downstream apoptotic protease, which is efficiently inhibited by SPI6 (and not activatable by hGrB); or SPI6 protects granule integrity independently of granzyme inhibition (in which case, PI-9 would be predicted to control GrB and independently protect granule integrity). We consider the last two possibilities more likely because they explain why the rat, which has a more human-like GrB (), has an orthologue of SPI6 rather than PI-9.
Among mammals, the multiple granzyme and serpin genes found in rodents appear to be the exception rather than the rule (), and it is evident that GrB and PI-9 paralogues emerged after divergence of Lagomorpha and Rodentia. A priori, the disappearance of the PI-9 RCL sequence from rodents suggests a fundamental change to the structure of mGrB, which we have confirmed by demonstrating that mGrB is catalytically distinct from hGrB and is not controlled by PI-9. Conversely, we have shown that hGrB is not controlled by SPI6. Although mGrB has changed, the GrB cleavage site in mouse Bid is conserved and is accessible to hGrB. Thus, it can be surmised that mGrB has lost cytotoxicity rather than hGrB gaining it. Conservation of the mouse Bid cleavage site suggests that another protease attacks it or that it serves another function in the molecule.
Why have these changes occurred in rodents? Given its role in immunity, the simplest explanation is that mGrB has evolved to target a virulence factor from a particular pathogen or to avoid inhibition by a specific pathogen protein. Specific targeting of virulence factors has been demonstrated for another leukocyte protease, neutrophil elastase (). Duplication of mGrB to generate paralogues may have allowed compensating cytotoxic mechanisms to evolve and/or multiple virulence factors to be targeted. The former possibility is supported by the reported cytotoxic functions of GrC and GrF (). The appearance of cytotoxic mGrB paralogues may have then driven the evolution of cognate PI-9 paralogues.
Although functioning primarily as a cytotoxin, GrB probably has other roles. For example, in humans it is found in testis and has been implicated in reproduction (). It may also be involved in extracellular matrix remodeling (). With multiple substrates and roles, it is easy to envisage that particular functions of GrB may have devolved to separate mGrB paralogues in rodents. For example, mouse granyzme N appears to be restricted to the testis and may have assumed the testicular role of hGrB ().
In closing, we have shown that mGrB is structurally and enzymatically distinct from hGrB and has a different killing mechanism. We have also shown that human GrA is not cytotoxic, whereas mGrA is substantially cytotoxic. Such species differences may also extend to granzymes M and K, but appropriate comparisons have yet to be made. It may be that the overall killing efficiency of human and mouse CLs is comparable and that the loss of GrB cytotoxicity in the mouse has been offset by a gain in GrA cytotoxicity or the evolution of GrB paralogues. Nevertheless, use of a granzyme from one species on cells or extracts from another species may produce misleading results because key substrates are missing or are inefficiently cleaved because recognition sites are not conserved. Use of mice to test inhibitors of human granzymes will fail because differences in active site topology of the mouse enzymes will prevent inhibitor binding. Likewise, use of mice to study viral infection may be compromised, as conserved virus inhibitors (e.g., SPI2/CrmA, Bcl-2 homologues, and AdL4-100K homologues) will interact differently with or fail to control mouse granzymes. Finally, use of mouse models to study the impact of granzymes on tumorigenesis is also problematic because the species-specific death pathways and substrates outlined here imply that tumors will subvert the immune system differently in mice and humans by up- or down-regulating different proteins.
Recombinant serpins were produced in as described previously (, ). Recombinant granzymes were also produced from (, ). To improve expression levels, the wt mGrB cDNA was modified by QuikChange site-directed mutagenesis (Stratagene) to remove a cryptic polyadenylation site, using the oligonucleotide 5′-CACTGTGAAGGAAGTATAATCAATGTCACTTTGGGGGC-3′ and its complement. The S4/S3 subsites in mGrB were mutated using the oligos 5′-GAGTCCTACTTTAAAAATTATTACAACAAAACCAATCAG-3′ and its complement (R174Y) and 5′-TCCTATGGATATAATGATGGTTCACCTCC-3′ and its complement (K218N). Activity of recombinant GrB was assessed by cleavage of the peptide substrate benzyloxycarbonyl- AlaAlaAsp-thiobenzyl ester (, ). The specific activity of hGrB was 90 U/μg, and the specific activity of mGrB was 120 U/μg. The proportion of active enzyme in each batch of purified protein was estimated by the ability to complex with PI-9 or SPI-6. Routinely, >90% of granzyme in a batch could complex with the appropriate serpin. Native human GrA was from lymphokine-activated killer cells. Activity of GrA was assessed by cleavage of the substrate benzyloxycarbonyl- Lys-thiobenzyl ester.
Quenched fluorescence substrates were synthesized and used as described previously (). Activity of granzymes on the peptide substrates Boc-AAD-thiobenzyl ester (BIOMOL Research Laboratories, Inc.) and IETD-NA (Calbiochem) was assessed as described previously (). Inhibition of granzymes by serpins was measured using standard approaches (; ). Granzyme cleavage sites in SPI6 and PI-9 were identified by N-terminal sequencing ().
S-labeled mouse procaspase 3 or mouse Bid was produced from cDNAs in the expression vector pSVTf via in vitro transcription and translation (). Cleavage by granzymes was assessed by adding a specified amount of protease directly to the translation mix and incubating for 30 min at 37°C. The caspase inhibitor z-VAD was obtained from Calbiochem. Products were separated by SDS-PAGE and visualized by fluorography.
Construction of substrate libraries based on bacteriophage T7 used the T7Select1-1b phage display system (Novagen). Each library was made by synthesizing a degenerate oligonucleotide, annealing it to complementary half-site oligonucleotides, ligating the resulting heteroduplex to vector arms, and adding it to a T7 phage packaging extract. The half-site oligonucleotides were 5′-GCCGCCTGGAGTGAGAG-3′ and 5′-AGCTTAGTGATGGTGATGGTGATG-3′. One library was constructed using the degenerate oligonucleotide 5′-AATTCTCTCACTCCAGGCGGC(NNK)CATCACCATCACCATCACA-3′ (where N represents any nucleotide and K represents T/C). This added a randomized nonameric peptide and a hexahistidine tag to the C terminus of the 10B coat protein. The complexity of this “random” library was 7 × 10 plaque-forming units. Another library was constructed using 5′-AATTCTCTCACTCCAGGCGGC(NNK)GAC(NNK)CATCACCATCACCATCACA-3′. This encoded an aspartate near the middle of the randomized peptide. The complexity of this “fixed” library was 4 × 10 plaque-forming units. Each library was amplified to >10 plaque-forming units. About 10 plaque-forming units of amplified phage were bound to nickel-chelated Sepharose beads at 4°C. Unbound phage was removed by washing the beads in phosphate-buffered saline containing 850 mM NaCl and 0.1% (vol/vol) Tween 20. After two further washes in phosphate-buffered saline containing 1 mM MgSO, the suspension was split into two parts (treatment and control; control treatments were performed to assess the extent of phage release from beads in the absence of protease). 200 nM granzyme was added to the treatment tube, and both tubes were incubated overnight at 37°C. Plaque-forming units in the supernatants representing cleaved or released phage were then counted and amplified to form sublibraries for the next round of selection. Phage remaining bound to the beads was eluted with 0.5 M imidazole, and plaque-forming units were counted to assess cleavage efficiency. After several selection rounds, individual plaques were chosen at random for sequence analysis. Phage DNA was amplified by PCR using dedicated primers (T7Select cloning kit; Novagen). Sequencing of PCR products using the same primers was performed using the Big Dye 3.1 kit (GE Healthcare).
The sequencing results were analyzed to determine the statistical distribution of each amino acid at each position of the nonamer (), allowing for the redundancy of the code, the fact that only 32 out of 64 codons are represented by NNK, and the exclusion from the analysis of any sequences encoding a stop codon in the nonamer. In a binomial distribution of amino acids, Δσ yields the difference of the observed frequency from the expected frequency in terms of standard deviations:where is the number of times amino acid occurs in the selected sequences, is the theoretical probability of amino acid occurring, and is the total number of sequences analyzed.
Human YT and Jurkat cells were maintained as described previously (). Eμ-myc/bid−/− and Eμ-myc/bid+/+ mouse B cell lymphomas () were obtained from R. Johnstone (Peter MacCallum Cancer Institute, Melbourne, Australia). Human SKW6.4 and mouse EL4 and YAC-1 cells were maintained in Dulbecco's modified Eagle medium containing 10% heat-inactivated fetal calf serum, 2 mM glutamine, and 55 μM β-mercaptoethanol. P815 cells were maintained in RPMI 1640 medium containing 10% heat-inactivated fetal calf serum, 2 mM glutamine, and 55 μM β-mercaptoethanol. Isolation and culture of mouse splenocytes followed standard procedures.
To generate SKW6.4 cells stably expressing PI-9, a PI-9 expression vector () was cotransfected by electroporation with a plasmid expressing neomycin resistance, and selection was performed in 1 mg/ml G418 (Sigma-Aldrich). Clones were established by limiting dilution and screened for PI-9 expression by indirect immunofluorescence and immunoblotting. FLAG-tagged SPI6 and SPI6T327R (containing an inactivating mutation in the proximal hinge) cDNAs were generated by PCR from the appropriate templates using the primer pair 5′-CGGGATCCATGGACTACAAAGACGATGACGATAAAGGGAATACTCTGTCTGAAG-3′ and 5′-GCTCTAGATTATGGAGATGAGAACCTGCCACA-3′. Products were cloned into the BamHI and XbaI sites of the expression vector pEF-PGKpuropA (). 1.25 × 10 P815 cells were electroporated at 330 μF and 350 V with either pEFpuro/FlagSPI6 or pEFpuro/FlagSPI6T327R vectors linearized with SalI. Transfected pools were grown for 3 d in 1.1 μg/ml puromycin, and subsequently 45 SPI6 and 28 SPI6T327R lines were cloned by limiting dilution in puromycin selection. Clones were scored for relative SPI6 expression by indirect immunofluorescence using the α-FLAG M2 antibody. Expression was confirmed by immunoblotting using rabbit antiserum (R25) raised to recombinant SPI6. This antiserum does not recognize the highly related mouse serpins (Serpinb9b, Serpinb9e, Serpinb6, Serpinb6b, and Serpinb1) and can be considered specific for SPI6.
Cell death mediated by granzymes and recombinant perforin was assessed as previously described (; ). Recombinant perforin was produced using a baculovirus expression system (). Recombinant SLO was prepared and used according to .
For analysis of serpin complexes, cells were lysed in either 1 volume of Laemmli sample buffer (LSB) or 1/2 volume of NP-40 lysis buffer (50 mM Tris HCl, pH 8.0, 10 mM EDTA, and 1% [vol/vol] Nonidet P40), followed by incubation at 37°C for 10 min, after which 1/2 volume of 2× LSB was added. Viscosity was reduced by mechanically shearing the DNA using a needle and syringe. For analysis of caspase 3 cleavage, cells were lysed in NP-40 lysis buffer containing 1 μg/ml pepstatin, 1 μg/ml leupeptin, 1 μg/ml aprotinin, and 10 μg/ml PMSF, and a postnuclear supernatant was prepared by centrifugation at 16,000 . Antibodies to PI-9 (R15) and hGrB (2C5 hybridoma supernatant) were used for immunoblotting at 1:2,000 and 1:100, respectively. A rat monoclonal antibody against mGrB was obtained from eBioscience (clone 16G6) and was used at 1:1,000. Goat anti-actin sera (Santa Cruz Biotechnology, Inc.) was used at 1:1,000. Rabbit antisera to SPI6 (R25) and hGrA (R045) was raised against recombinant protein purified from following standard procedures and was used at 1:1,000. Rabbit antisera to caspase 3 (Cell Signaling Technology) was used at 1:100. HRP-conjugated secondary antibodies was used at 1:5,000 and detected using enhanced chemiluminescence.
Images of blots were initially captured on x-ray films, which were subsequently scanned into Photopaint (Corel). Adjustments to brightness or contrast of digital images were applied to the whole image. No nonlinear adjustments were made.
Fig. S1 shows a comparison of methods used to assess cell survival after GrB treatment. Fig. S2 shows a comparison of uptake of mGrB and hGrB into target cells. Fig. S3 shows GrB-mediated procaspase 3 cleavage in Bid+/+ or Bid−/− mouse B cell lymphoma extracts. Online supplemental material is available at . |
Ubiquitin, especially monoubiquitin, has been a well-documented sorting signal for both endocytosed plasma membrane (PM) proteins and intracellular resident proteins trafficking from the TGN to endosomes or lysosomes (). This has been extensively studied in yeast, where ubiquitination serves an important sorting signal to determine protein targeting to either the lumen or the limiting membrane of the vacuole (lysosome; ; ; ; ). The role of ubiquitin in sorting proteins from the Golgi to endosomes/lysosomes in mammalian cells is not as clear.
Either blocking ubiquitin modification or adding a single ubiquitin to cargo proteins diverts them away from their normal cellular destination. For example, blocking ubiquitination of the vacuolar carboxypeptidase S or adding a ubiquitin to the hydrolase dipeptidylaminopeptidase B in yeast diverts these proteins to the limiting membrane or lumen of the vacuole, respectively (, ). Likewise, the ubiquitination status of the yeast GAP1 and Fur4 permeases determines their fate during sorting at the Golgi (; ; ). In both yeast and mammalian cells, most of the ubiquitinated proteins that reach the vacuole/lysosome interior are either PM proteins, which then undergo degradation (; ; ), or hydrolytic enzymes activated inside the vacuolar lumen (). If proteins are not ubiquitinated, they are either recycled back to the PM () or delivered to the limiting membrane of multivesicular bodies (MVBs) or the vacuole/lysosome (). Therefore, mono- or multimonoubiquitination plays a critical role in targeting cargo proteins to their proper cellular destination (; ; ), hence deciding their fate (degradation vs. recycling).
The monoubiquitin sorting signal can be recognized and transmitted by several proteins involved in both the endocytic and biosynthetic pathways through direct physical interaction. These proteins often possess ubiquitin-binding motifs/domains (“ubiquitin receptors”), such as ubiquitin-interacting motif (UIM; ), GAT (GGA and Tom homologue), UBA, UEV, VHS, and CUE (). The UIMs in epsin and eps15/eps15R play an important role in internalization and the early stages of endocytosis of PM proteins, such as the EGFR (; ), whereas the UIMs of Vps27/Hrs and Hse1/STAM are associated with sorting function at the early and late endosome (; ; ). Interestingly, several UIM-containing endocytic proteins (e.g., epsin, eps15, and Hrs) are themselves ubiquitinated outside of the UIM sequence itself (; ; ). At least in some cases, this ubiquitination appears to involve Nedd4 family members. Nedd4 is a ubiquitin ligase comprised of a C2 domain, 3–4 WW domains that bind PY motifs (L/PPxY) on target proteins, and a ubiquitin ligase Hect domain (). However, it is not yet clear how Nedd4 proteins are recruited to these endocytic proteins, which do not possess PY motifs. Moreover, the precise role of ubiquitination of these UIM-containing endocytic proteins in cargo sorting and transport is not known.
The Golgi-localized, γ-ear–containing, ADP ribosylation factor–binding proteins (GGAs) are primarily associated with the TGN, and play a role in lysosomal and endosomal sorting (; ; ). For example, they sort mannose 6-phosphate receptors from the Golgi to the lysosome by binding to the cytoplasmic dileucine motifs of the receptor (; ; ). GGA3 was reported to also function as an adaptor to sort the internalized EGFR to endosomes (; ). In yeast, it was shown that GGA binds to the ubiquitinated Gap1 permease, diverting it from the TGN to the vacuole/lysosome (; ). In both mammalian and yeast cells, ubiquitin was shown to bind to the GAT domain of the GGAs (; ; ). Thus, monoubiquitin is a universal sorting signal not only at the cell surface or in endosomes but also at the Golgi.
Lysosomal-associated protein transmembrane 5 (LAPTM5) is a multispanning transmembrane protein that resides in the late endosome/lysosome and is expressed primarily in hematopoietic cells (; ). We isolated LAPTM5 in a screen for Nedd4-WW domain–interacting proteins, and detected three PY motifs and a UIM in that protein (unpublished data). The function of this lysosomal protein is so far unknown. LAPTM5 is related to LAPTM4 (also known as MTP), a lysosomal protein that is widely expressed in many cell types (, ; ). LAPTM4 was shown to confer multidrug resistance by transporting a range of small molecules, including nucleosides, nucleobase analogues, antibiotics, anthracyclines, ionophores, and steroid hormones (; ). It is believed to function in sequestration of these small molecules into the lysosome, thus, protecting the cell from their harmful effects. LAPTM4 has two splice isoforms, LAPTM4α and -β (). LAPTM4α was shown to possess two Tyr-based motifs (YxxΠ) that are responsible for its localization to the lysosome (). How LAPTM5 is sorted to the lysosome has not been elucidated and was the focus of our studies.
We show that LAPTM5 is sorted from the Golgi to the lysosome by association via its PY motifs with Nedd4. The Nedd4–LAPTM5 complex recruits GGA3, and ubiquitinated GGA3 binds the UIM of LAPTM5. The Nedd4-triggered LAPTM5–GGA3 association then promotes translocation of LAPTM5 from the Golgi to the lysosome. Interestingly, this translocation does not require LAPTM5 ubiquitination.
Our earlier work identified LAPTM5 in a screen for proteins that interact with the second WW domain of Nedd4-1 (Nedd4; unpublished data). The intracellular domains of LAPTM5 contain three PY (L/PPxY) motifs; an LPAY motif located between transmembrane (TM) domains 3 and 4, and LPSY and PPPY motifs in the cytoplasmic C terminus (). We also identified a UIM in this protein (). The presence of the three PY motifs in LAPTM5, our identification of Nedd4 as a binding partner for LAPTM5, and the observation that LAPTM5 is localized to late endosomes (LEs) or lysosome (), prompted us to investigate whether Nedd4 may be involved in sorting of LAPTM5 to the LE/lysosomes.
To test whether these three PY motifs are responsible for Nedd4 binding, all three Tyr residues in the PY motifs were mutated to alanines (3YA; ), and wild-type (WT) LAPTM5 or its 3YA mutant were tested for binding to a catalytically inactive Nedd4(CS) in a coimmunoprecipitation (coIP) assay. A catalytically inactive Nedd4(CS) was used to prevent a possible degradation of LAPTM5 in the event that it is a substrate for Nedd4. As shown in (top), after ectopic expression in human embryonic kidney 293T (HEK293T) cells, LAPTM5(WT) was able to coIP with Nedd4(CS), and this binding was lost in the LAPTM5-3YA mutant. Of the three PY motifs, the third one (PY) appears most important for binding to Nedd4, with the first PY motif (PY), but not PY, contributing to the interaction as well (Fig. S1 A, available at ). These results demonstrate that Nedd4 binds to the PY motifs of LAPTM5.
To examine whether Nedd4 binding leads to ubiquitination of LAPTM5, lysates from HEK293T cells expressing HA-tagged LAPTM5 together with His(x8)-tagged ubiquitin were boiled in SDS (to dissociate other putative interacting proteins), precipitated with Ni–Agarose beads (which bind His), and subjected to immunoblotting with anti-HA antibodies to detect ubiquitinated LAPTM5. As seen in , LAPTM5(WT) was ubiquitinated in cells (that express endogenous Nedd4), and this ubiquitination was dependent on the presence of its PY motifs, as it was eliminated in the LAPTM5-3YA mutant. In agreement with the binding data (Fig. S1 A), PY appears most important for LAPTM5 ubiquitination (Fig. S1 B). These results suggest that LAPTM5 ubiquitination is mediated by Nedd4. Indeed, overexpression of a catalytically inactive Nedd4(CS), together with LAPTM5, inhibited this ubiquitination (), likely in a dominant-negative fashion. Most of the LAPTM5 in these assays appeared to be either mono- or diubiquitinated. Collectively, these results indicate that Nedd4 is likely the ubiquitin ligase that ubiquitinates LAPTM5.
Because LAPTM5 was previously shown to localize to lysosomes (), we next tested whether this localization is dependent on Nedd4-mediated interaction with LAPTM5. Thus, DsRed-tagged LAPTM5(WT) was transfected into HEK293T cells, and its localization was assessed by confocal microscopy. As seen in , at 14 h after transfection, LAPTM5(WT) was localized to the Golgi, as determined by colocalization with giantin. By 18 h, some of it has translocated to vesicles that costain with the lysosomal marker LAMP1 (), and by 24 h, most of it was localized to the lysosome (, , and Table S1, available at ). This was further validated by immunoEM analysis (), which demonstrates localization of LAPTM5 in LAMP2-containing vesicles (lysosomes). In contrast, the mutant LAPTM5-3YA, which cannot bind Nedd4, was retained in the Golgi and appeared unable to translocate to the lysosome even after 24 (, , and Table S1) or 48 h (not depicted) after transfection. It was also not localized to EEA1- (early endosomes), Hrs- (sorting endosomes), or Rab11 (recycling endosome)-containing vesicles ( and Fig. S2). The ability of the LAPTM5-3YA mutant to localize early after transfection to the Golgi suggests that this mutant is properly folded and able to enter the Golgi from the ER (, immunoEM), much like the WT LAPTM5. Rather, it is the sorting step from the Golgi to the lysosome that is defective in the 3YA mutant, most likely caused by its inability to bind Nedd4.
To further test for the role of Nedd4 in the Golgi to lysosome sorting of LAPTM5, we knocked down endogenous Nedd4 in HEK293T cells with RNAi (). As seen in , , and Table S1, knockdown of endogenous Nedd4 resulted in retention of LAPTM5 in the Golgi. Interestingly, this impaired Golgi to lysosomal sorting was observed despite normal expression of endogenous Nedd4-2, which was unaffected by our knockdown of endogenous Nedd4 (Nedd4-1; ). This suggest that Nedd4, rather than Nedd4-2, is involved in lysosomal targeting of LAPTM5.
Ubiquitination has been well documented as a sorting signal for delivery of biosynthetic proteins to the lumen of the yeast vacuole (). Its role in sorting mammalian transmembrane proteins is less clear. Because we found that Nedd4 binding to LAPTM5 is required for its sorting to the lysosome (), and LAPTM5 is ubiquitinated by Nedd4 (), we next tested if LAPTM5 ubiquitination is involved in its sorting to the lysosome. We, thus, generated a mutant LAPTM5 with all of its 10 cytoplasmic Lys residues mutated to Arg (10KR; ) to render it ubiquitination-impaired. The ubiquitination status of this mutant was then analyzed after its cotransfection with His-tagged ubiquitin, as described in . As expected, the LAPTM5-10KR mutant failed to become ubiquitinated (), although it was still able to bind Nedd4 () because its PY motifs are intact.
To analyze the requirement of LAPTM5 ubiquitination for its sorting from the Golgi to the lysosome, we tested the ability of the ubiquitination-impaired LAPTM5-10KR mutant to translocate to the lysosome from the Golgi. We first verified that LAPTM5-10KR fused to DsRed is not itself ubiquitinated because of the added DsRed moiety (), and then we used it for immunolocalization studies. As seen in and quantified in and S1, most of the LAPTM5-10KR mutant was localized to the Golgi at 18 h after transfection and to the lysosomes at 24 h after transfection, quite similar (although not identical) to WT LAPTM5 ( and ). Moreover, HA-tagged LAPTM5-10KR was also able to sort to the lysosome (Fig. S3, available at ). These results suggest that ubiquitination of LAPTM5 is not necessary for its translocation to the lysosome.
The GGA proteins bind mannose 6-phosphate receptors at the Golgi and play an essential role in lysosomal sorting (; ). Therefore, we first tested whether GGA3 could bind to LAPTM5. Cell lysates expressing either T7-tagged Nedd4 alone or HA-tagged LAPTM5 were immunoprecipitated and subjected to immunoblotting with anti- GGA3 antibody to detect if endogenous GGA3 could coimmunoprecipitate in the complex. As shown in , neither protein alone could bind GGA3 (lanes 2 and 3). However, GGA3 was immunoprecipitated in a complex with LAPTM5(WT) and Nedd4(WT), but only poorly with the catalytically inactive Nedd4(CS; , lane 4 and 5). Furthermore, only a higher form of endogenous GGA3 was found in the complex, although a doublet is seen in the lysate (, lane 6). Immunoprecipitating endogenous GGA3 followed by immunoblotting with anti-ubiquitin antibody demonstrated that this upper band was ubiquitinated GGA3 (not depicted), in accord with earlier work, which reported that GGA3 undergoes monoubiquitination (). Collectively, these data suggest that GGA3 is found in a complex with LAPTM5 and Nedd4 (but not Nedd4[CS]), and that GGA3 in this complex is most likely the ubiquitinated form of the protein.
The GGA proteins contain various domains that are involved in protein–protein interactions (; ; ). Recently, the GAT domain of GGA3 was reported to bind ubiquitin (; ). Thus, we tested whether ubiquitinated LAPTM5 can bind to the GGA3-GAT domain. We first confirmed that endogenous GGA3 binds ubiquitin by demonstrating that G protein–conjugated ubiquitin (prot-G-Ub) is able to precipitate GGA3 (). To confirm that the GGA3-GAT domain is responsible for ubiquitin binding, we expressed wt and GAT domain mutant GGA3 (GGA3-L276A) in HEK293T cells and repeated the precipitation assay with prot-G-Ub. As seen in , GGA3-L276A was no longer able to bind ubiquitin, as previously reported (; ). However, it is interesting that more of the lower band of GGA3 binds to ubiquitin (), which is different from the form that is in a complex with LAPTM5 and Nedd4 ().
To understand how GGA3 forms a complex with LAPTM5 and Nedd4, we expressed WT and the GGA3-L276A mutant, together with a combination of WT and mutant LAPTM5 and Nedd4, and subjected the lysate to coIP with LAPTM5. If ubiquitinated LAPTM5 binds to the GAT domain of GGA3, only WT GGA3 should complex with LAPTM5 and Nedd4. However, both WT GGA3 and the GGA3-L276A mutant were able to bind to the Nedd4–LAPTM5 complex (), suggesting that GAT function as a ubiquitin-binding motif is not necessary for the association with this complex. Furthermore, it suggests that ubiquitination of LAPTM5 may not contribute to the formation of the complex. This is consistent with our data ( and ) that show efficient sorting of LAPTM5-10KR to the lysosome.
Because GGA3 can coprecipitate with the Nedd4–LAPTM5 complex, we tested the possibility that GGA3 itself may become ubiquitinated by Nedd4. demonstrates that, indeed, endogenous GGA3 can become ubiquitinated by coexpressed Nedd4, but not by the catalytically inactive Nedd4(CS).
As mentioned above, only the higher form of GGA3 was immunoprecipitated in a complex with LAPTM5 and Nedd4 (), suggesting that GGA3 in that complex is most likely the ubiquitinated form. LAPTM5 contains a putative UIM at its C terminus (). To test if this putative UIM binds ubiquitin and ubiquitinated GGA3, cell lysates expressing either HA-tagged LAPTM5 or T7-tagged Nedd4 were precipitated with prot-G-Ub. Nedd4 did not bind to this immobilized ubiquitin (). However, LAPTM5 was able to bind ubiquitin (), binding that was not seen in its 3YA mutant (). These results suggest that binding of Nedd4 to LAPTM5 affects the ability of LAPTM5 to bind ubiquitin. To further test this, we incubated cell lysates expressing LAPTM5 together with either WT or catalytically inactive Nedd4(CS) with prot-G-Ub, and immunoblotted the precipitate with antibodies to LAPTM5 or Nedd4 (i.e., to their tags). (top) shows that ubiquitin was able to bind to LAPTM5 only in the presence of WT Nedd4, but not Nedd4(CS). This suggests that active Nedd4 is needed to allow binding of LAPTM5 to ubiquitin.
We next tested the role of the LAPTM5-UIM in ubiquitin binding, as well as the dependency of binding on the levels of Nedd4 expressed in cells. As shown in , LAPTM5 was able to bind ubiquitin substantially better when Nedd4 was overexpressed (compare lane 2 to 7). Mutant LAPTM5 truncated at amino acid 248 at the C-tail (ΔC, ) almost completely lost its ubiquitin-binding capacity (, lanes 5 and 10), suggesting that the UIM, which resides in this region, may be important for binding to ubiquitin. Because this C-terminal region also contains a potential PDZ-binding sequence (SxV), we made two independent mutants of LAPTM5; a UIM mutant (UIMm; ) known to be defective in ubiquitin binding (), and a Val to Ala substitution of the last residue of LAPTM5 (LAPTM5-VA; ), which should preclude binding to PDZ domains. We then tested the ability of these mutants to bind ubiquitin. Our results show that the VA mutant had a slightly reduced ubiquitin-binding ability (, lane 6 and 11). The UIMm mutant, however, revealed a dramatic decrease in ubiquitin binding (), suggesting that the UIM of LAPTM5 plays a key role in binding ubiquitin.
As detailed above, we observed that the LAPTM5-10KR and -3YA mutants, both unable to become ubiquitinated, showed a different pattern of subcellular distribution 24 h after transfection; the 10KR mutant was primarily translocated to the lysosome (similar to WT LAPTM5), whereas the PY motif mutant was retained in the Golgi ( and ). We thus tested whether these two mutants showed a difference in GGA3 binding. Cell lysates expressing Nedd4 together with a combination of either WT or mutant LAPTM5 were immunoprecipitated with Nedd4 antibodies and subjected to immunoblotting with anti-GGA3 antibody to detect if endogenous GGA3 coIPs with the complex. As seen in , the LAPTM5-10KR mutant was able to complex with GGA3 in the presence of Nedd4, much like WT LAPTM5 (, lane 1 and 4). However, the LAPTM5-3YA mutant was unable to bind GGA3 (, lane 2), suggesting that interactions with Nedd4, rather than the ubiquitination status of LAPTM5, is important for GGA3 binding. Furthermore, the UIM of LAPTM5 was responsible for GGA3 interaction with the complex because GGA3 binding was completely abolished in the UIM mutant, LAPTM5-UIMm (, lane 3). Importantly, this same UIM mutant that cannot bind ubiquitin or GGA3 was also unable to sort LAPTM5 from the Golgi to the lysosome (, , and Table S1). Because the mutations in the UIM were close to the second PY motif (LPsY) of LAPTM5, we tested the ability of this UIMm mutant to bind Nedd4. As seen in , the UIMm mutant was still able to bind Nedd4, similar to WT LAPTM5 (lane 2 and 4), suggesting that Nedd4 binding to LAPTM5 does not require the UIM. Collectively, these data suggest that binding of Nedd4 to LAPTM5 promotes the UIM of LAPTM5 to bind ubiquitinated GGA3. This then allows LAPTM5 to translocate from the Golgi to the lysosome.
The ability of the LAPTM5-UIM to bind ubiquitinated GGA3 ( and ), and the inability of the LAPTM5-UIMm mutant (which cannot bind ubiquitinated GGA3) to translocate to the lysosome (), suggest that GGA3 may be responsible for LAPTM5 sorting and translocation. To test this, RNAi for GGA3 was used to knockdown endogenous GGA3 in HEK293T cells cotransfected with LAPTM5 (and expressing endogenous Nedd4). As seen in (ii and iii) and , and S1, knockdown of GGA3 led to a substantial retention of LAPTM5 in the Golgi and impairment of lysosomal sorting.
To further substantiate these results, we performed immunoEM analysis to test for colocalization of LAPTM5 with GGA3. As seen in (i), LAPTM5 and GGA3 are detected together in vesicles that emanate from the TGN and lysosomes.
The aforementioned studies were performed in HEK293T cells. Although these cells express endogenous Nedd4 and GGA3, they do not express endogenous LAPTM5, which is primarily a hematopoietic protein. To ensure that the Nedd4/GGA3-dependent sorting of LAPTM5 from the Golgi to the lysosome was not an artifact of ectopic expression of LAPTM5 in HEK293T cells, we tested this sorting in DC2.4 cells, a cell line derived from dendritic cells of the immune system (). As seen in Fig. S4 (available at ), DC2.4 cells express small amounts of endogenous LAPTM5. They also express endogenous Nedd4 (unpublished data). We next tested Golgi to lysosomal transport in these cells after transfection of DsRed-tagged WT or the 3YA mutant of LAPTM5. demonstrates that WT LAPTM5 expressed in DC2.4 cells is sorted from the Golgi to the lysosome, as seen in HEK293T cells. This is also validated by demonstrating colocalization of LAPTM5 with LAMP1 at the membrane of lysosomes using immunoEM (, i). In contrast, this sorting is attenuated in mutant LAPTM5 that cannot bind Nedd4 (the 3YA mutant); by 24 h after transfection, the proportion of cells translocated from the Golgi to the lysosome was much greater in the WT relative to the 3YA mutant ( and S1). These results suggest that, much like our observation with HEK293T cells, lysosomal sorting of LAPTM5 is also seen in dendritic cells that normally express endogenous LAPTM5, and this sorting likely involves Nedd4 binding as well. Moreover, the segregation of LAPTM5 to GGA3-containing TGN/vesicles and MVBs is also seen in the DC2.4 cells (, ii).
In this paper, we demonstrate that the transmembrane lysosomal protein LAPTM5 is sorted from the Golgi to the lysosome with the aid of Nedd4 and GGA3, a process not requiring LAPTM5 ubiquitination. We propose that LAPTM5, Nedd4, and GGA3 form a complex, whereby ubiquitinated GGA3 binds the UIM of LAPTM5, facilitating transport of LAPTM5 from the Golgi to lysosomes (). It is possible that Nedd4 dissociates from this complex once GGA3 is ubiquitinated (possibly by Nedd4) and able to bind LAPTM5, allowing the GGA3–LAPTM5 complex to be sorted into GGA3-containing vesicles that ultimately arrive at the MVBs/lysosomes ().
LAPTM5 and LAPTM4α and -β belong to a family of lysosomal membrane proteins (; ). Although the ubiquitously expressed LAPTM4 proteins are able to transport small molecules, such as nucleosides, into the lysosome (, ), the function of the hematopoietic-expressed LAPTM5 is not yet known. Moreover, the two family members (LAPTM4 and LAPTM5) may use different sorting signals for lysosomal targeting; LAPTM4α and -β have two Tyr-based sorting signals (YxxΦ) at their C termini, which, together with a di-Leu motif, are responsible for lysosomal targeting (). This is similar to other lysosome-associated membrane proteins, in which Tyr-based or di-Leu sorting signals were shown to interact with adaptor protein complexes (AP-1, -3, and -4; ; ; ; ). As we show, LAPTM5 has adapted a different sorting strategy; it utilizes its PY motifs (which bind Nedd4-WW domains) and UIM region (which binds ubiquitinated GGA3) to facilitate its sorting to lysosomes. It is interesting that two of the three PY motifs in LAPTM5 are also conserved in LAPTM4α and -β, but it is not yet known whether they also contribute to lysosomal targeting in these proteins.
Based on our work here, we propose that Nedd4 plays a crucial role in targeting LAPTM5 to lysosomes because a mutant that cannot bind Nedd4 (LAPTM5-3YA) is retained in the Golgi, and knockdown of Nedd4 leads to the same Golgi-retention defect. Interestingly, although Nedd4 can ubiquitinate LAPTM5, LAPTM5 ubiquitination is not essential for its sorting because the ubiquitination-impaired LAPTM5-10KR mutant can still bind GGA3 and be sorted to lysosomes. This suggests that LAPTM5 ubiquitination may have other, yet unknown, function, perhaps recruiting other proteins that contain ubiquitin-binding domains, enhancing function of the UIM, or regulating LAPTM5 stability. In contrast, GGA3 cannot bind the LAPTM5-3YA mutant, and this mutant is retained in the Golgi, suggesting that GGA3 can only interact with LAPTM5 in the presence of Nedd4. Moreover, this interaction requires the presence of active Nedd4, because overexpression of a catalytically inactive Nedd4(CS) precludes the association of GGA3 with the Nedd4–LAPTM5 complex. Exactly how Nedd4 regulates the interaction between LAPTM5 and GGA3 is currently unknown, but it is intriguing that the presence of active Nedd4 is required for the LAPTM5-UIM to more strongly bind ubiquitin. Thus, it is possible that Nedd4 binding to LAPTM5 ensures that the UIM of LAPTM5 can properly bind ubiquitinated GGA3 (). Moreover, we show that Nedd4 is also able to ubiquitinate GGA3, thereby further promoting the association of GGA3 with the LAPTM5-UIM.
Based on our observation that GGA3 bearing a mutation in its GAT domain, GGA3-L276A (which we show cannot bind ubiquitin), is still able to interact with LAPTM5, we believe that ubiquitination of LAPTM5 is not required for GGA3 binding, an observation that is in accord with our findings of proper lysosomal sorting of LAPTM5 lacking its ubiquitination sites. Thus, our work describes a novel mode of recruiting GGA3 to a protein complex in the Golgi; an interaction of ubiquitinated GGA3 with the UIM of LAPTM5 (). Moreover, this recruitment involves the ubiquitin ligase Nedd4, placing this E3 ligase in a complex with GGA3. Although we do not know which physiological E3 ligases are responsible for GGA3 ubiquitination, we show here that Nedd4 can ubiquitinate GGA3 in cells, suggesting that Nedd4 may be one such E3 ligase.
Our observation of the abolishment of binding of the GGA3-GAT domain mutant L276A to ubiquitin, although in agreement with previous work (; ), is at odds with recent reports (; ) that demonstrate, using structural determinations, that a second site (site 1) in the GAT domain of GGA3 provides stronger affinity of interactions with ubiquitin than site 2, to which L276 belongs. As discussed elsewhere (), ubiquitin binding is strongly affected by the mode of presentation and by the oligomerization state of its interacting partners, which could provide a possible explanation for the discrepant results.
It had been previously documented that the presence of the UIM in several endocytic proteins (e.g., epsin, eps15, Hrs) not only allows binding of ubiquitin to the UIM itself, but also promotes ubiquitination of these proteins on regions outside the UIM (; ). Moreover, this ubiquitination often involves Nedd4 family members. However, it is not known how Nedd4 proteins are recruited to proteins such as epsin, eps15, or Hrs, which do not possess PY motifs. In the case of LAPTM5, we show here that it contains both UIM and PY motifs, allowing it to directly bind Nedd4, to become ubiquitinated by Nedd4, and to bind ubiquitin (attached to GGA3) via its UIM. GGA3 itself was recently shown to become monoubiquitinated ().
As already stated, the function of LAPTM5 is, thus far, unknown. Given its selective expression in immune cells and its lysosomal localization and similarities to LAPTM4 proteins, it is tempting to speculate that it may be involved in lysosomal movement in immune cells, and possibly in transport of molecules destined to be released from the lysosome once at the plasma membrane.
HA-LAPTM5 was inserted into pCDNA3. To generate DsRed-LAPTM5, LAPTM5 cDNA was cloned into the pDsRed C1 vector (Invitrogen) downstream of and in frame with the DsRed fluorescent protein coding sequence. Mutations in LAPTM5, depicted in , were generated by site-directed mutagenesis using PCR, verified by sequencing and subcloned into pCDNA3. GST-WW2 constructs were previously described (). His-tagged ubiquitin construct was obtained from D. Bohmann (University of Rochester Medical Center, Rochester, NY; ) and Myc-tagged WT and mutant GGA3 (L276A) were provided by J. Bonifacino (National Institutes of Health, Bethesda, MD). To knock down Nedd4-1, the oligo GATGAAGCCACCATGTATA was synthesized and inserted into the pSUPER-EGFP vector, as previously described (). For GGA3 knockdown, the oligo AAACGGCTTCCGCATCCTC was used as reported earlier () and inserted into the pSUPER-EGFP as previously described (). Polyclonal antibody against LAPTM5 was generated using GST fusion proteins containing the C terminus of LAPTM5 (amino acid sequence between 225 and 260). Polyclonal anti–Nedd4-Hect domain antibodies (specific for Nedd4-1) were described earlier (). Anti–Nedd4-2 antibodies were generated against a GST fusion protein encompassing a regions that is unique to Nedd4-2. Both anti–Nedd4-1 and -2 antibodies were affinity purified before use. The anti-GGA3 antibody was obtained from Santa Cruz Biotechnology, Inc. Rabbit anti-HA antibody for immunofluorescence was purchased from Sigma-Aldrich.
The immature murine DC2.4 cells () provided by K. Rock (University of Massachusetts Medical Center, Worcester, MA) were maintained in complete RPMI 1640 medium containing 10% fetal bovine serum, 2 mM -glutamine, 100 μM nonessential amino acids, 100 U/ml penicillin, 100 μg/ml streptomycin, and 50 μM β-mercaptoethanol. Transfections were performed using ESCORT V transfection reagent (Sigma-Aldrich) according to the manufacturer's protocol.
Cryosectioning and immunolabeling were performed as previously described (), with the following changes. 24 h after transfection, cells were fixed for 1 h in 4% paraformaldehyde, followed by overnight fixation in 8% paraformaldehyde. 75-nm-thick sections were collected on formvar-coated nickel grids, at −120°C. The grids were blocked in a solution of 5% fish skin gelatin for 30 min and incubated with mouse anti-HA antibodies (1:25) and either rat CD107a (1:20) or rabbit LAMP2 (1:20; Sigma-Aldrich), giantin/furin (1:20; Abcam), or GGA3 (1:40; Abcam), followed by five washes in PBS and appropriate combination of 6 nm goat anti–mouse IgG/IgM and 10 nm goat anti–rabbit IgG, or with 15 nm goat anti–rat IgG (1:20; Electron Microscopy Sciences). The images were taken using a transmission EM (Tecnai 20; FEI) at 80 kV equipped with a camera (Dualview; Gatan), and processed using acquisition software (DigitalMicrograph; Gatan).
HEK293T (or DC2.4 cells) cells expressing either WT or mutant LAPTM5 cells were lysed with 1 ml of lysis buffer (50 mM Hepes, pH 7.5, 150 mN NaCl, 1% Triton X-100, 10% glycerol, 1.5 mM MgCl, 1.0 mM EGTA, 10 mg/ml leupeptin, 10 mg/ml aprotinin, and 1 mM PMSF) and cleared by centrifugation at 14,000 rpm for 10–20 min. The cleared supernatants were used for pull-down and coIP experiments. For pull-down experiments, 500 μg of cell lysates were incubated with 50 μg GST or GST-Nedd4-WW2 (GST-WW2) protein on glutathione–Sepharose beads for 2 h at 4°C. Beads were washed twice with 1 ml HNTG (20 mM Hepes, pH 7.5, 150 mM NaCl, 10% glycerol, and 0.1% Triton X-100) and twice with lysis buffer. Bound proteins were eluted from the bead with 1× SDS-PAGE sample buffer, resolved by SDS-PAGE, and transferred to nitrocellulose membrane. Bound LAPTM5 was identified with anti-HA antibody (1:10,000; Babco), followed by secondary antibodies and ECL detection (GE Healthcare). To analyze the ability of the UIM of LAPTM5 to bind ubiquitin, 500 μg HEK293T cell lysates expressing either WT or mutant LAPTM5 that lacks the UIM (LAPTM5-UIMm) were incubated with ubiquitin immobilized on protein G beads (pro-G-Ub; 20 μg), the beads were washed, and bound LAPTM5 was detected with anti-HA antibody. For coIPs, HEK293T cell lysates expressing transfected WT or mutant HA-LAPTM5 together with T7-Nedd4 (CS; 500 μg each) were incubated with anti-T7 antibody for 1 h at 4°C, followed by the addition of 10 μl of protein G–Sepharose for an additional 1 h. After bound proteins were washed with lysis buffer (two times) and HNTG (three times), the coimmunoprecipitated LAPTM5 proteins were detected with anti-HA antibody (1:10,000). To identify GGA3 in the Nedd4–LAPTM5 complex, HEK293T cells were transfected with T7-Nedd4 (WT or CS mutant), Myc-GGA3 (WT or the L276A mutant that cannot bind ubiquitin) and HA-LAPTM5, and their lysates were subjected to IP with HA (LAPTM5) antibodies and immunoblotting with anti-Myc antibody (1:10,000) to detect bound GGA3.
HEK293T cells were cotransfected with HA-LAPTM5 or DsRed-LAPTM5 (WT or mutants) and His-tagged ubiquitin (His-Ub), and where indicated, cells were also cotransfected with T7-tagged Nedd4 (WT or catalytically inactive CS mutant). Cells were lysed in lysis buffer supplemented with 50 μM LLnL (-acetyl-Leu-Leu-norleucinal; Sigma-Aldrich). Ubiquitinated (histidinated) proteins were then precipitated with Ni–Agarose beads for 4 h at 4°C and blotted with anti-HA antibodies to detect ubiquitinated LAPTM5, or where applicable, with anti-GGA3 antibodies to detect ubiquitinated GGA3. To ensure ubiquitination of LAPTM5, and not of associated proteins, the cell lysates were treated with 1% SDS and boiled for 5 min. These boiled lysates were then diluted 11 times with lysis buffer to dilute the SDS before their use in the Ni-bead precipitation step. Complexes with Ni–Agarose bead were washed with lysis buffer (two times) and HNTG (three times) and immunoblotted with anti-HA antibodies to detect LAPTM5, followed by anti–mouse horseradish peroxidase secondary antibodies and ECL detection. Equal amounts of different lysates (∼2 mg) were used for the precipitations.
Fig. S1 describes the contribution of the individual PY motifs of LAPTM5 to binding to Nedd4 and to LAPTM5 ubiquitination. Fig. S2 shows lack of colocalization of LATPM5(3YA) with EEA1, Rab11, or Hrs. Fig. S3 shows lysosomal localization of HA-tagged LAPTM5(10KR). Fig. S4 shows endogenous expression of LAPTM5 in DC2.4 cells. Online supplemental material is available at . |
Many cell surface proteins are attached to the membrane by a glycophosphatidylinositol (GPI) anchor, which consists of a conserved central structure () with variable carbohydrate and lipid peripheral components (). GPI anchors can determine protein functional specificity, just as switching a transmembrane (TM) domain for a GPI anchor can result in novel function caused by association with new signaling elements located in a shared membrane microdomain (, ).
Membrane rafts, originally defined by their insolubility in cold, nonionic detergents such as Triton X-100 (), are small, heterogeneous aggregations of cholesterol and sphingolipids on the cell surface (; ) that concentrate GPI-anchored proteins, but also contain other proteins. Although the existence of membrane rafts in vivo has been questioned (), recent studies using a variety of methods have provided evidence for raftlike membrane microdomains (; ; ; ; ; ). Such microdomains may act as signaling scaffolds, determining the identity of a subset of signaling elements, as proteomic analyses have found a high concentration of such proteins in purified rafts (; ), with GPI-anchored proteins involved in activating this signaling (; ). The existence of heterogeneous raft populations has been inferred from studies showing that different GPI-anchored proteins exist in separate rafts (; ; ). External rafts with different proteins may each have a defined set of associated cytoplasmic proteins, whereby aggregation of GPI-anchored proteins by external domain self-binding or by multivalent ligand binding could cluster specific rafts, resulting in downstream signaling ().
Carcinoembryonic antigen (CEA), and the closely related CEACAM6, are GPI-anchored, cell surface glycoproteins that block cellular differentiation () and inhibit the apoptotic process of anoikis (; ), effects that appear to be caused by the activation of specific integrins (; ). CEA is up-regulated in many human malignancies (; ), implying a similar role in human cancer, whereas the TM-anchored CEACAM1 (CC1) may act as a tumor suppressor (; ).
Most CEA family members mediate intercellular adhesion by antiparallel self-binding (), which, together with parallel binding on the same cell surface (), may result in clustering of rafts containing CEA (). Deletion of the last two thirds of the CEA N-terminal domain (ΔNCEA) abrogates its adhesive ability, which leads to a loss of differentiation-blocking activity (). The method of membrane anchorage determines CEA family member activity, as genetically fusing the GPI anchor of CEA to CC1's external domain creates a differentiation-blocking molecule, whereas a chimera consisting of the external domain of CEA attached to the TM domain of CC1 does not block differentiation (). The fact that GPI-anchored neural cell adhesion molecule (NCAM) does not block differentiation, but can be converted to a differentiation-blocking molecule, denoted NCB (previously “NC blunt”), by swapping its GPI anchor for that of CEA, suggests that the CEA GPI anchor harbors the specificity for the differentiation-blocking function and that the external domains merely function to cluster the molecules, and thus, the associated rafts ().
Based on the aforementioned model, it should be possible to inhibit the biological functions of CEA (and, by implication, that of any GPI-anchored molecule whose function is regulated by a similar mechanism) by interfering with clustering. This has been achieved for CEA by mutating regions in its N-terminal external domain responsible for self-binding or by the addition of peptides or monovalent mAbs that target these regions (). We test a second strategy which exploits the specificity of the CEA GPI anchor; if “shank-defective” or “shankless” CEA GPI anchors that were incapable of self-association and clustering were introduced, they could occupy the same rafts as CEA, and thus, possibly interfere with its clustering. We show that nonfunctional ΔNCEA inhabits the same membrane microdomains as NCB, as both have the same GPI anchor, but not those of NCAM, and is capable of completely inhibiting NCB's CEA-like differentiation-blocking activity.
To test the hypothesis that the functional specificity of GPI anchors could be exploited to specifically inhibit the activity of GPI-anchored proteins, cells expressing a functional GPI- anchored protein were supertransfected with a shank-defective molecule with the same GPI anchor, and assessed for effects on function. The former functioning molecule was NCB, which has NCAM self-binding external domains linked to the CEA GPI anchor (); the defective molecule was ΔNCEA, which has the same GPI anchor, but external domains that are defective in self-binding (; ). Because ΔNCEA cannot bind to the external NCAM domain of NCB (), this combination allowed a study focused on the potential interaction between their GPI anchors.
ΔNCEA was stably cotransfected into NCB transfectants of rat L6 myoblasts, which are blocked for myogenic differentiation because of the expression of NCB. ΔNCEA was present on the cell surface of the double transfectants at slightly higher levels than NCB, as seen by FACS () and Western blot (unpublished data). As a control for specificity of effects, double transfectants stably expressing molecules with different GPI anchors were used, i.e., ΔNCEA or CEA with CEA GPI anchors, and NCAM with the NCAM GPI anchor. Similar expression levels were also obtained for these transfectants ().
CEA and NCAM appear to exist in separate membrane regions, potentially explaining their opposite biological effects (). If the GPI anchor alone determines cell surface localization, then molecules with the same GPI anchor should exist in close proximity, whereas those with different anchors should not. Thus, NCAM and NCB would be expected to have different cell surface distributions, with NCB showing a distribution similar to that of CEA. To test this hypothesis, we examined whether ΔNCEA existed in close proximity on the cell surface to NCB, but not to NCAM, using confocal microscopy to examine the cell surface localization of these proteins. The relative surface distribution of NCAM and NCB compared with ΔNCEA was determined after indirect immunofluorescent staining. ΔNCEA showed substantial, although incomplete, colocalization with NCB, whereas ΔNCEA and NCAM showed essentially no colocalization (). Because the incubations were performed at room temperature, the antibodies used for detection may have caused partial clustering of the proteins. This, however, should not affect the heterophilic association in question, as clustering of rafts containing both proteins should not change the final amount of colocalization seen. Indeed, fixing the cells before antibody incubation to avoid clustering resulted in very similar patterns of colocalization to what is shown in (not depicted). This therefore suggests that the GPI anchor of CEA is sufficient to determine cell surface localization of a protein.
To verify these results, L6 cotransfectants were treated at 4°C (to limit protein diffusion) with the chemical cross-linker DTSSP. Nonreducing Western blots demonstrated similar cross-linking patterns for NCB and NCAM, consisting of dimers, trimers, and higher molecular weight complexes, both alone and in the presence of ΔNCEA (). To determine the cellular distribution of ΔNCEA relative to NCAM and NCB, immunoprecipitation (IP) studies of extracts from cross-linked cells expressing similar amounts of these proteins were performed. The cross-linking approach was undertaken, rather than using detergent lysis because of the potential effects of detergents on membrane raft structure. IP of extracts from untreated cells did not result in any coIP (), confirming the expected antibody specificity. However, IP with an anti-CEA mAb of extracts of DTSSP-treated cotransfectants resulted in the coIP of a considerable amount of NCB, but, importantly, not of NCAM (). Similarly, IP with an anti-NCAM mAb of extracts of cross-linked cotransfectants showed coIP of ΔNCEA only in the case of NCB, but not of NCAM (). The low proportion of coimmunoprecipitated protein can likely be explained by the lack of interaction between the external NCAM and CEA protein domains, the requirement for close (<12 Å) apposition to be cross-linked, the presence of large levels of monomeric proteins even after cross-linking (), and the incomplete colocalization seen by confocal microscopy (). These results demonstrate that proteins with GPI anchors of the same type can exist in close proximity, providing a rationalization for specific interference with protein function.
Having demonstrated specific colocalization of ΔNCEA and NCB on the cell surface, the effect of this defective protein on NCB's ability to block differentiation was examined. NCB levels in the ΔNCEA coexpressing L6 transfectants were actually higher than those in NCB-only transfectants, thus, validating comparisons between NCB alone and in the presence of ΔNCEA (; FACS means of 124 vs. 63, respectively). NCB completely blocked differentiation, whereas coexpression of ΔNCEA with NCB resulted in an almost complete restoration of differentiation, with a fusion index of 78% of that seen for ΔNCEA alone ( and ; P < 0.0001). As a control, coexpressing NCAM had no effect on the differentiation block imposed by CEA (), despite the differentiation-enhancing effects of NCAM (). To confirm this result, up-regulation of myosin, a biochemical differentiation marker, was examined. ΔNCEA induced myosin production in two independent populations of NCB-expressing cells, as shown by Western blot, whereas NCB alone showed no myosin expression (), confirming the previous results. Because of the length of the differentiation assay (10 d total), it was possible that a loss of NCB expression caused the differentiation restoration in these cotransfectants. However, no decrease in NCB levels was seen in differentiated ΔNCEA + NCB cultures, as Western blots showed higher expression levels in both cell populations than the NCB-alone transfectants for which differentiation was blocked (). Thus, ΔNCEA expression interfered markedly with the differentiation-blocking function of NCB, presumably via their common feature, the GPI anchor.
ΔNCEA releases NCB's block of differentiation, suggesting that it is interfering with downstream signaling by NCB. CEA signaling has been found to involve activation of the integrin α5β1 in rat myoblast and human colonic cell lines () and the integrin αvβ3 in neuronal cells (unpublished data). We assessed NCB signaling by incubating single-cell suspensions, prepared from exponential cultures, with plates coated with the ECM components fibronectin (Fn), vitronectin (Vn), and collagen I. Either CEA or NCB expression increased binding to both Fn and Vn, relative to L6 parental cells (; P < 0.004). Cells expressing ΔNCEA + NCB (and ΔNCEA alone) showed no such increase, demonstrating a complete loss, in the presence of ΔNCEA, of the NCB-mediated increase in ECM binding (P < 0.0001). As a control, no difference in binding to collagen I was seen between any of these cell lines (). In addition, LR-73 (LR) transfectants were tested for binding to Fn, and these cells showed a similar loss of NCB-mediated effects upon coexpression of ΔNCEA (; P < 0.01). The total cell levels of α5 and β1 integrins in the L6 transfectants were assessed by Western blot (), and cell surface levels were assessed by FACS (α5 only; not depicted) and showed only minor differences between transfectants, confirming that the ability of cells to adhere to Fn, rather than changes in integrin surface expression level, was the source of the observed difference (; unpublished data). Thus, the ability of ΔNCEA to interfere with the NCB-mediated differentiation blockage is correlated with interference of enhanced integrin–ECM interaction promoted by NCB.
Signaling by GPI-anchored proteins requires intact membrane rafts (). One possible mechanism for the effects of ΔNCEA on NCB functional properties could be by expulsion of NCB from rafts. When this possibility was examined, however, NCB remained primarily insoluble in cold Triton X-100 after coexpression of ΔNCEA in either L6 or LR cells (, A and B, respectively). Complete cellular lysis was demonstrated by the fact that the integrin α5 chain, an integral membrane protein, was localized in the soluble fractions. As confirmation, isopycnic sucrose gradient ultracentrifugation, where raft-associated proteins migrate to the lower density regions of the gradient, was performed on cold Brj-98 lysates of L6 transfectants. Again, no obvious difference was noted between NCB alone and NCB coexpressed with ΔNCEA (), as almost all of the NCB was present in the low-density fractions in both cases. Under these conditions, the α5 integrin chain showed partial raft association for both transfectants, demonstrating that an alteration of α5 localization was not responsible for the lack of NCB function. The distribution of ΔNCEA was also found to be essentially the same as that of NCB, as expected for two GPI-anchored proteins. As controls, the α2 integrin chain was found solely in higher density fractions, whereas the raft lipid GM1 was entirely in the low- density fractions. Thus, NCB retained membrane raft association in the presence of ΔNCEA, so that this could not explain the loss of NCB function.
The proteins of the CEA family mediate intercellular adhesion, as does NCAM (), by external domain self-binding. Such self-binding is required for the differentiation-inhibitory activity of CEA, presumably to affect raft clustering (; ). If ΔNCEA interfered with NCB clustering, one might predict a reduction in the ability of NCB to mediate intercellular adhesion. NCAM was used as a control, as it inhabits different rafts from ΔNCEA (). NCAM and NCB were expressed at very similar levels, with and without ΔNCEA, on the surface of LR cells, thus allowing for quantitative comparisons in adhesion between populations (). A significant reduction in the strength of NCB-mediated adhesion occurred in the presence of ΔNCEA, as shown by a reproducible decrease of ∼20% in the number of aggregated cells in suspension after 2 h (P < 0.001), a difference that was not seen for NCAM-mediated adhesion (). This was accompanied by a decrease in the size of aggregates in NCB-expressing cells as a result of ΔNCEA coexpression (P < 0.0001), which, again, was not seen for NCAM (). Thus, introducing the same functional GPI anchor with a defective shank led to a specific reduction in the strength of intercellular adhesion by NCB. Effective intercellular adhesion by GPI-anchored proteins is believed to involve the formation of large, zipperlike structures through the aggregation of multiple proteins and rafts, creating stabilized platforms (). The ability of ΔNCEA to interfere with NCB-mediated adhesion is thus consistent with models invoking interference with NCB clustering.
One mechanism whereby ΔNCEA could interfere with NCB clustering is by altering the structure of the rafts it is associated with. Therefore, the size of the rafts that NCB occupied was approximated by lysing the cells under conditions identical to those used for isopycnic separation on sucrose density gradients and separating the lysate by velocity sedimentation through a uniform 12.5% sucrose solution. Under these conditions, NCB was almost entirely raft associated (); therefore, this technique should provide a measure of the size of the rafts inhabited by these proteins. The fractions, which were collected from the top (fraction #1), were assessed by Western blot for protein localization, using equal volumes of each fraction. NCB was found to be shifted to fractions farther from the top when ΔNCEA was coexpressed, indicating that it was present in larger complexes under these conditions (). The distribution of NCAM, on the other hand, was found to be similar whether ΔNCEA was present or not (), demonstrating that the size of the NCAM complexes was not altered in the presence of ΔNCEA. The distribution of ΔNCEA was very similar to that of NCB in shifting to larger complexes when coexpressed with NCB, while remaining in smaller complexes when coexpressed with NCAM (unpublished data). A significant (P < 0.05) difference in NCB distribution () relative to NCAM distribution () upon coexpression with ΔNCEA was demonstrated by densitometric analysis of three independent experiments. This suggests that the presence of ΔNCEA specifically alters the rafts containing NCB. To confirm that this was a raft-specific effect, cells were pretreated with methyl-β-cyclodextrin (MβCD) to sequester cholesterol and disrupt raft structure. Initially, sucrose gradient ultracentrifugation was performed on lysates of these treated cells, to confirm the disruption of the rafts. The distribution of NCB demonstrated that this treatment partially disrupted the rafts, as a portion of the NCB was now present in higher density fractions (; compare to ). When these samples were tested for the size of the complexes that NCB was localized to, it was found that NCB, both alone and coexpressed with ΔNCEA, remained in the first few fractions after velocity sedimentation (). Thus, treatment with MβCD abrogated the difference seen for NCB complex size after ΔNCEA coexpression, confirming that the difference seen was a raft-mediated effect. Although cellular lysis with detergents at low temperatures can affect raft structure (), the fact that the ΔNCEA-dependent sedimentation difference is seen for NCB, but not for NCAM transfectants, suggests that it represents a valid increase in raft size. This would indicate a dilution of the NCB concentration in membrane rafts, as an increase in the size of a raft containing the same number of NCB molecules would cause a relative concentration decrease. This would thus reduce the incidence of cis-interactions between the proteins, which are necessary for clustering, explaining the decrease in intercellular adhesion (), and the loss of biological function ( and ).
As the mechanism of inhibition of NCB function by ΔNCEA appears to involve interference with clustering, NCB function should be restored by artificial clustering with antibodies. Antibody cross-linking of cell surface proteins induces signaling events, including restoring the defective differentiation-blocking function of ΔNCEA () through integrin activation manifested by increased cellular binding of Fn (unpublished data). ΔNCEA and NCB coexpressed with ΔNCEA both appear to be nonfunctional because of defects in protein clustering, so clustering of NCB with antibodies should have a similar effect to what has previously been seen for ΔNCEA. To test if NCB retained the potential to modulate ECM binding, in spite of the deactivating effects of coexpressed ΔNCEA, cells in monolayer culture were treated with mAbs directed against the NCAM external domains of NCB, along with secondary antibodies to enhance clustering, and binding of soluble Fn was measured. Several mAbs were used, including J22, which binds to internal CEA domains and, as such, remains capable of clustering ΔNCEA; D13, which is a control mAb that has an epitope in the region deleted from ΔNCEA; and 123C3, which binds to the NCAM external domains of both NCAM 125 and NCB. As expected, cross-linking ΔNCEA with J22, but not with D13, resulted in a significant increase in bound Fn (). Similarly, cross-linking of NCB, alone and in the presence of ΔNCEA, increased bound Fn (). Cross-linking NCAM, which does not normally modulate integrins, with the NCAM-specific antibody did not lead to an increase in bound Fn levels, demonstrating the specificity of this effect. The lack of difference in Fn binding between NCB and parental cells, unlike that seen in , in monolayer culture is likely caused by the intact ECM surrounding the cells in this assay. This would provide the ligands for integrins that have previously been activated, so that these integrins would not bind to the Fn added to the culture medium. Thus, NCB, in the presence of ΔNCEA, remained capable of altering Fn interaction after antibody cross-linking, which is consistent with the hypothesis that a defect in NCB clustering is created upon introduction of ΔNCEA.
#text
ΔNCEA is a CEA deletion mutant that has the last 75 amino acids of the N domain deleted, such that it is no longer biologically active (). The NCAM splice variant used in this study, p125, is a human GPI-anchored NCAM isoform containing the muscle-specific domain (). NCB is a chimera of the NCAM p125 external domain genetically fused to the CEA GPI anchor signal sequence (). The mAbs J22 and D14 bind to internal CEA domains (), whereas the epitope of D13 is in the portion of the CEA N domain that is deleted in ΔNCEA, and rabbit polyclonal anti-CEA binds to all CEA external protein domains. The mAb 123C3 (Santa Cruz Biotechnology, Inc.) recognizes human NCAM, whereas antibodies H-293, H-104, and M-106 (Santa Cruz Biotechnology, Inc.) recognize the α2, α5, and β1 integrins, respectively. The mAb 47A () binds to myosin. C20 is a goat polyclonal anti-Fn antibody (Santa Cruz Biotechnology, Inc.).
Cells were grown attached to tissue culture plastic surfaces (Nunc), as previously described (). In brief, CHO-derived LR-73 fibroblasts were grown in α-MEM with 10% FBS. Rat L6 myoblasts were grown in DME containing 10% FBS (GM), and were subcultured before reaching confluency to avoid selecting for nonfusing variants. Cell concentrations were determined using a particle counter (Beckman Coulter). For myoblast differentiation, 10 L6 cells/cm were seeded in 60-mm dishes. After 3 d, the media was switched to DME with 2% horse serum (DM). 4–7 d later, cultures were assessed for differentiation by hematoxylin (Sigma-Aldrich) staining and microscopic examination (), or by lysing and assessing myosin levels by Western blotting.
100-mm dishes were seeded with 2 or 4 × 10 cells/plate for LR or L6, respectively. 24 h later, cells were cotransfected by calcium phosphate coprecipitation with 5 μg of cDNA, 0.5 μg of pSV2(neo), and 10 μg of carrier DNA isolated from LR-73 cells. Double transfections were performed in the same manner, either by cotransfecting both cDNAs at once (for LR cells) with pSV2(neo) or with 0.5 μg of pBabe(puro) to supertransfect L6 transfectants; transfectants were isolated by selection with 400–600 μg/ml neomycin (G418; Invitrogen) or 1 μg/ml puromycin (Sigma-Aldrich). After 10–14 d, resistant clones were pooled and sorted for high expression by FACS using mAbs J22 or 123C3. Although pooled populations of many clones were used, two independent transfections of L6 cells were performed to ensure no clonal variation occurred, with identical assay results. Note that the L6 (NCB) cells were pooled colonies resistant to both G418 and puromycin, and that although data from L6 and LR-73 parental cells is shown, no difference between these cells and pooled G418-resistant clones transfected with the pSV2(neo) alone has been noted ().
Cells were collected with PBS-citrate containing 4 mM EDTA (PBSCE) for NCAM-expressing transfectants (because of the sensitivity of the NCAM external domain to trypsinization; ) or 0.063% trypsin in PBS-citrate for CEA transfectants. 2.5 × 10 cells were resuspended in ice-cold PBS with 2% FBS (PBSF). Cells were incubated for 30 min with mAb at a dilution of 1:50–100, washed with PBSF, and incubated with FITC-conjugated goat anti–mouse antibody (Jackson ImmunoResearch Laboratories) diluted 1:100. After an additional 30 min, cells were pelleted, resuspended in PBSF, and analyzed using a FACScan instrument (Becton Dickinson).
Cells were seeded at 10 cells/cm on day 0. On day 2, subconfluent cultures were collected with PBSCE, resuspended in GM, and incubated at 37°C for 30 min. Cells were washed with serum-free media and resuspended in serum-free media at 4 × 10 cells/ml. 100 μl/well of this suspension was added to wells from a 96-well plate coated with Fn, Vn, or collagen I (CHEMICON International, Inc.) and incubated for 1 h at 37°C. Wells were washed with PBS containing Mg, and adherent cells were stained with crystal violet. Wells were washed again with PBS, and the bound stain was solubilized with 0.05 M NaHPO, pH 4.5, plus 25% ethanol. Staining was quantified with a microplate reader (Bio-Tek Instruments) at 570 nm. Statistical significance was determined using a test ().
Proteins in cellular lysates were resolved by SDS-PAGE and transferred electrophoretically to a 0.45-μm PVDF membrane (Millipore). Immunoblotting was performed as previously described (), with antibody binding detected using ECL Plus reagent (GE Healthcare).
Triton X-100 solubility was determined as previously described (). In brief, subconfluent cell cultures were rinsed with PBS, collected with PBSCE, and rendered single-cell suspensions by passing through a 27-gauge needle. 10 cells/ml were resuspended in ice-cold lysis buffer containing 1% Triton X-100 and the protease inhibitors aprotonin (Roche), leupeptin (Roche), and PMSF (Sigma-Aldrich). Lysates were syringed with a 27-gauge needle, incubated on ice for 15 min, and centrifuged at 13,500 for 20 min at 4°C. Soluble fractions were removed, and the pellets were resuspended in 0.9 vol of lysis buffer and 0.1 vol 10% SDS. The relative amounts of soluble versus pellet protein were determined by immunoblotting. For sucrose gradient ultracentrifugation, two T175 flasks were seeded with 10 cells/cm. 2 d later, cells were collected with PBSCE, and lysed with 1 ml of 1% Brij-98 (Sigma-Aldrich) in sucrose gradient buffer (10 mM Tris, pH 8.0, and 140 mM NaCl) containing aprotonin, leupeptin, and PMSF for 30 min at 4°C. 1 ml of ice-cold 80% sucrose was added to this lysate and overlayed successively with 2 ml of 35% sucrose and 1 ml of 5% sucrose. Lysates were centrifuged with a rotor (SW55; Beckman Coulter) for 19 h at 45,000 RPM at 4°C. 400-μl fractions were collected from the top of the gradient, and equal volumes of each fraction were assessed by immunoblotting.
L6 cells from 2–4 T175 flasks were collected with PBSCE, pooled, and lysed with 500 μl of 1% Brj-98 in sucrose gradient buffer for 30 min on ice. The lysate was then added on top of 11 ml of 12.5% sucrose, and centrifuged for 1 h at 12,300 RPM (∼18,700 ) in an SW41 rotor (Beckman Coulter). 25 460-μl fractions were collected from the top, and assayed by immunoblotting. In certain cases, cells were pretreated with mβCD (Sigma-Aldrich) for 15 min at 37°C before Brj-98 lysis, to disrupt membrane rafts.
4 × 10 L6 cells were seeded in three 100-mm dishes for each transfectant. 2 d later, cells were washed with PBS and incubated, with gentle rocking at 4°C, with either 1 ml of 1 mM DTSSP (Pierce Chemical Co.) in PBS or with PBS alone. After 1 h, unconjugated DTSSP was neutralized with 100 mM Tris, pH 7.4. Cells were lysed with 400 μl/plate of 60 mM n-Octyl β-D glucopyranoside (Sigma-Aldrich) in lysis buffer containing protease inhibitors. Lysates were pooled and syringed to reduce viscosity. 1.2 ml of each lysate was precleared by rotation with 75 μl of Protein A/G Plus–Agarose beads (Santa Cruz Biotechnology, Inc.) for 3 h at 4°C. Precleared lysates were then diluted with an equal amount of lysis buffer, and divided into three aliquots, receiving no antibody, 5 μg 123C3, or 5 μg J22. Samples were rotated overnight at 4°C, and then 75 μl of Protein A/G Plus–Agarose beads were added. 3 h later, the beads were washed five times with lysis buffer and resuspended in 75 μl 1× Laemmli sample buffer for analysis by Western blotting
L6 transfectants were seeded in 8-well Lab-Tek Permanox chamber slides (Nunc) at a density of 10 cells/well. 2 d later, cells were washed with PBSF and incubated with primary antibodies 123C3 (at a dilution of 1:100 in PBSF) and rabbit polyclonal anti-CEA (1:2,000 dilution) for 30 min at RT. Cells were washed with PBSF, and then incubated at RT for 30 min, in the dark, with a 1:250 dilution of both Cy2-conjugated goat anti–rabbit and rhodamine-conjugated goat anti–mouse secondary antibodies. Cells were then washed twice with PBSF, and fixed by incubation with 4% formaldehyde for 10 min at 4°C, followed by 100% methanol for 20 min at 4°C. Samples were then mounted using fluorescent mounting medium (DakoCytomation). Localization of stained proteins was observed using a LSM 510 Axiovert 100M confocal microscope with a Plan-Achromat 63×/1.4 NA oil differential interference contrast objective (both Carl Zeiss MicroImaging, Inc.).
Adhesion assays were performed as previously described (). In brief, 10 LR cells were seeded in 80-cm tissue culture flasks (Nunc), and collected 2 d later by incubation with PBSCE. 3 × 10 cells were resuspended in 3 ml α-MEM containing 0.8% FBS and 10 μg/ml DNase I (Roche), syringed with a 27-gauge needle to obtain single-cell suspensions, and allowed to aggregate at 37°C with stirring at 100 rpm using a small magnetic stirring bar. Aliquots were removed at the indicated times, and the cells were counted with a hemocytometer to determine the percentage of single cells. For the aggregate assay, cells were prepared as for adhesion assays, but the number of cells present in each of ∼50 multicellular aggregates was scored after 1 h in suspension.
On day 0, 10 cells/well were seeded in a 96-well plate. 2 d later, cells were washed with PBSF and incubated for 30 min at 37°C in DME with 50 μl of 10 μg/ml human Fn (BD Biosciences), along with, where indicated, 5 μg/ml primary mAb and 30 μg/ml donkey anti–mouse secondary antibody (Jackson ImmunoResearch Laboratories) to further cross-link CEA or NCAM constructs. Cells were then washed three times with PBSF, and fixed with 4% formaldehyde. Bound Fn was determined by incubation with anti-Fn antibody, C20, at a dilution of 1:100 in 3% BSA (Sigma-Aldrich) in PBS (PBSB) for 90 min at RT, having blocked nonspecific binding by incubation for 1 h at RT with PBSB. Cells were washed with PBSB, and incubated with HRP-conjugated rabbit anti–goat secondary antibody (Jackson) at a dilution of 1:2,500 in PBSB for 1 h. After incubation with a hydrogen peroxide solution containing ABTS (Sigma-Aldrich), bound Fn was determined with a microplate reader at 405 nm, with a reference wavelength of 490 nm. |
Mammalian epidermis functions as a barrier to prevent both water loss to the terrestrial environment and entry of toxic and pathogenic agents into the organism (; ). Embryonic ectoderm is specified to an epidermal fate at murine embryonic day (E) 8.5, regulated by the p63 transcription factors (). At E9.5, this single layer of basal cells express cytokeratin (K) 5 and K14 (). During midembryogenesis, these basal cells continue to divide as a single-layered epithelium to increase the surface area of the developing embryo. Stratification of the basal epithelium initiates at E12.5 with asymmetric cell divisions perpendicular to the basement membrane (). By E15.5, suprabasal cells initiate the differentiation program and express K1/K10 (). Establishment of the epidermal permeability barrier initiates at E16.5 on the dorsal surface and spreads ventrally to achieve a fully competent barrier by E18 (). Barrier establishment requires cross-linking of the cells in the upper layer, which then constrains increases to the surface area of the embryo. Although the barrier must be acquired before the typical end of gestation, it is not advantageous to develop a fully competent barrier too early in development because of the need for continued growth.
Interfollicular epidermal cells retain the ability to self-renew under both homeostatic and injured conditions by maintaining mitotically active cells (; ; ; ; ). Terminal differentiation begins when basal cells concomitantly withdraw from the cell cycle and lose adhesion to the basement membrane. In the intermediate spinous layers, the cells assemble a durable cytoskeletal framework that provides mechanical strength to resist physical trauma. In the upper granular layer, a cornified envelope (CE) is assembled directly underneath the plasma membrane by sequential incorporation of precursor proteins. Lipid-containing lamellar bodies fuse with the plasma membrane and attach to the CE scaffold, sealing the now enucleated cells together to create the “bricks and mortar” barrier at the skin surface (). Recent experimental results have also demonstrated a selective role for tight junctions in establishing the epidermal barrier (). This process of differentiation from a mitotically active basal cell to a terminally differentiated squame is maintained throughout life as part of epidermal regeneration and maturation.
Mouse models with targeted ablations of genes encoding keratinocyte transcription factors have demonstrated that barrier acquisition is a coordinated and regulated process (). Our earlier experiments demonstrated that the transcription factor () is necessary to establish the epidermal barrier in utero (). To elucidate further the transcriptional networks regulating this process, we examined the specific role of , the most highly expressed member of the GATA family of transcription factors in interfollicular epidermis.
GATA-3 expression in interfollicular epidermis is first detected in the immediate suprabasal layer at E15.5 (). At E16.5 and thereafter, GATA-3 is expressed in both basal and immediate suprabasal layers (). −/− embryos die at E11 (), and pharmacologic rescue of these embryos until E17.5 revealed a role for in development of the inner root sheath of the hair follicle (). To elucidate the necessary function of GATA-3 in terminal stages of epidermal differentiation in vivo, we used the cre–loxP system.
K14- genotype, hereafter referred to as mutants (; ). Quantitative PCR of amplicons both within and outside the locus on mutant and control littermate epidermal genomic DNA demonstrated that >97% of the epidermal cells had deleted the locus (unpublished data). The residual amplification in mutants may be from melanocytes or Langerhans cells, resident in the epidermis. Deletion of GATA-3 mRNA and protein was also demonstrated by Northern and immunohistochemical analysis of mutant newborn skin ().
mutants, born at the expected Mendelian ratio, are distinguishable from their littermates at birth by a lack of prominent whiskers. In the perinatal period, mutants do not feed and can be identified by the lack of a typical “milk spot.” Equally striking is the apparent desiccation of the mutant skin during the perinatal period, which takes on a thin, erythemic, wrinkled appearance (). During the first 6 h after birth, mutant mice lose an average of 5% of their birth weight. By comparison, littermates who do survive >24 h when unfed do not appear to desiccate and lose significantly less weight during a comparable period (). Specifically, individual mutants lose weight at a rate that is >3.5- and 7.8-fold greater than the standard deviation of the control littermates. Because other barrier-deficient animal models display a similar perinatal lethality, we tested this directly (; ).
As compared with control littermates, mutant newborns exhibit a significant increase in the rate of transepidermal water loss across their skin surface (P < 0.001; ). The increase in the rate of transepidermal water loss and weight loss is of a similar order of magnitude (). mutants' impaired skin barrier is unable to retain water in the terrestrial ex utero environment, which results in dehydration and, ultimately, lethality. To complement the studies that measure water loss across the skin surface, we investigated mutants' competence to exclude percutaneous dye penetration. As previously shown, dye exclusion in control littermates (visualized as white areas) initiates on the dorsal surface at approximately E16.5 and spreads ventrally, resulting in complete dye impermeability by E17.5. A transient delay of 0.5 d is observed in both the initiation and completion of the dye exclusion of the mutants (). Because the mutants are able to exclude dye penetration before birth, this delay does not explain the increased rate of transepidermal water loss and subsequent lethality. However, the biophysical properties of the skin barrier that regulate the relative permeability of small molecules, infectious agents, water, and gases across this surface are still poorly understood.
To determine which of the three known components of the barrier is disrupted in mutant newborns, we analyzed tight junctions, CEs, and lipid composition. Although total lipid content was similar, mutants exhibit a selective defect in lipid synthesis. mutants have a decreased level of glucosylceramides and its derivative ceramide EOS (). Ceramide EOS is one of the precursors of sphingolipids, which interact with free lipids to organize the lipid lamellar structures in the stratum corneum (SC; ). Ultrastructural analysis of mutant skin, preserved to maintain lipid structures, revealed a paucity of lamellar bodies in the SC. In addition, these lamellar bodies contain only a few disorganized membrane leaflets and irregular vacuoles (). This analysis points to a specific defect in lipid content and organization underlying the selective barrier impairment. In contrast, the other two elements of the barrier appear normal. Specifically, egression of a subcutaneously injected dye halted at occludin-positive structures, indicating that the tight junctions in mutants are fully competent (; unpublished data). In addition, the CEs of the mutants appear normal: mature, plump, and rigid (unpublished data).
To investigate the etiology of mutants' barrier defect, we examined the histology, differentiation, and proliferation status of embryonic mutant skin. At E1 5.5, when GATA-3 is initially expressed, the histology of mutant skin appears normal (). At E16.5, nuclei persisted in the presumptive granular layer of –deficient epidermis, consistent with a differentiation defect and delay in barrier acquisition ( and ). At E17.5, granular cells of –deficient epidermis were properly enucleated and differentiated, again consistent with overcoming the delay in acquiring a selective barrier ( and ). mutant newborn epidermis appears thinner with a disorganized basal layer (). Immunohistochemical analysis of –deficient newborn epidermis demonstrated that the structural proteins K14 (basal) and loricrin (granular) were expressed in the proper cell layer (). Ultrastructural analysis of the mutant newborn epidermis revealed the absence of filaments that connect the keratohyalin granules (), again suggesting that the terminal differentiation program may be impaired. The rate of proliferation of mutant basal cells is similar to controls, as measured by BrdU immunohistochemistry and cell cycle FACS analysis (unpublished data).
To circumvent the perinatal lethality of mutants and investigate the role GATA-3 plays in epidermal homeostasis, we grafted E18.5 mutant and control littermate skin onto nude mice. The gross morphology of grafted mutant skin confirmed the previously reported role of in hair follicle specification (; ). Previous studies have established that hyperproliferation and acanthosis (thickened epidermis) are compensatory responses to impaired epidermal barrier (). Histological analysis of the grafted mutant skin displayed both these hallmark features (). mutant epidermis is ∼10 cell layers thick, with an increase in K1-positive suprabasal cells, whereas both control grafted and hairless nude epidermis are approximately three to four cell layers thick (). Although proliferation was increased in the mutant epidermis, it was restricted to the basal cells (unpublished data). These grafting studies suggest that mutant epidermis retains an inherent barrier defect that extends beyond the perinatal period.
To identify the pathways of gene expression that are affected by the loss of GATA-3 during development, we analyzed microarray data from mutant and control littermate dorsal skin isolated from three distinct epidermal stages of development: at E15.5, the initial defect in barrier acquisition; at E16.5, the compensatory acquisition of barrier to exclude small molecules; and at newborn, the selective barrier deficiency upon exposure to the terrestrial environment. This tripartite experiment enabled us to query the genes and pathways affected by GATA-3 that led to the delay in epidermal differentiation and the persistent lipid and barrier defect.
At all developmental stages, lipid synthesis and modification was identified as the most significant and commonly affected pathway in the mutants, consistent with the lipid defect observed in these animals (). Down-regulated at all epidermal developmental stages are prostaglandin-endoperoxide synthase 1 (; greater than threefold), 1-acylglycerol 3 phosphate -acyltransferase 5 (; greater than three- to ninefold), and sphingosine-1-phosphate phosphatase 1 (; greater than two- to fivefold). The family of Elongation of very long fatty acids–like (Elovl) genes, , , , and , encoding lipid biosynthetic proteins, is also down-regulated in mutants.
To determine if genes in the lipid biosynthetic pathway are direct targets of GATA-3, we used a genomic approach. First, to identify potential cis-acting regulatory elements in the lipid genes, we performed a multispecies alignment of the human sequences compared with mouse, rat, and dog homologues. Between mouse and human, ∼5% of the genomic sequence is under positive selection; i.e., alignable and conserved (). Only one third of these regions are predicted to encode an exon of a gene. The other regions of alignment are postulated to encode RNA genes or regulatory elements. An example of the multispecies alignment of the proximal promoter and first intron of AGPAT5 with the program MultiPipMaker is shown in (). MultiPipMaker identifies two blocks of noncoding sequence conservation: distal to the first exon and proximal to the second exon. To refine this analysis, we used TRANSFAC to query whether GATA-3 binding sites were predicted within these blocks of conserved sequence, with a consensus binding sequence of G A T A/T A/G (; ). Examination of the conservation tracks on the University California Santa Cruz genome web browser enabled us to rapidly determine whether these predicted GATA-3 sites are conserved between species. Examples of two highly conserved GATA-3 sites (GATTA and GATTG) as well as one not conserved (GATTc) and one sequence conserved only with dog (GATTA) are given in . Finally, to determine if GATA-3 binds in vivo to these sites, we immunoprecipitated chromatin with a GATA-3–specific antibody. Two overlapping amplicons (+0.9 and +1.0 from AGPAT5 transcription start site), which contain these highly conserved GATA-3 binding sites, were specifically enriched 3.8- and 3.2-fold in the GATA-3 chromatin immunoprecipitated DNA. Sequences in the proximal promoter (−0.2) and more distal in the AGPAT5 gene (+15.5 and +39.2) were not enriched in the GATA-3 chromatin immunoprecipitated DNA (). Although a similar genomic analysis of PTGS and SGPP1 were performed, we did not identify multispecies conserved GATA-3 binding sites, which might suggest that the criteria for inclusion were very stringent. In summary, GATA-3 binds in vivo to a region in the first intron of the lipid acyltransferase gene AGPAT5 that contains highly conserved GATA-3 binding sites.
In addition to the defects in lipid synthesis, mutants display a developmental delay in the expression of structural proteins (). Specifically at E15.5, genes encoding the cornification proteins hornerin and loricrin, as well as the differentiation proteins K1 and involucrin, are down-regulated in the mutants. At E16.5, expression of late CE genes (LCE 1B, 2B, 2C, 3A, and 4B) are either absent or decreased by more than fivefold in the mutants, demonstrating a continuation of the differentiation delay. Previous work has shown that late CE protein expression immediately precedes in utero dye impermeability in a patterned fashion (). Therefore, we postulate that this delay in late CE gene expression underlies the delay in barrier acquisition visualized in . Because of their temporal expression during development, the genes down-regulated at E15.5 are different than E16.5, but both expression profiles reflect a delay in differentiation. By the newborn stage, mutants express the vast majority of these genes encoding epidermal differentiation and cornification proteins at normal levels. This transcriptional profiling provides the molecular underpinnings to interpret the morphological changes in the mutant skin, observed during development.
The transcriptional profile of mutant newborn skin also reflects the pathways invoked to compensate in the ex utero terrestrial environment for an intrinsic barrier defect. Classic studies have shown that barrier deficiency results in increased DNA synthesis and acanthosis (). mutants express high levels of K6 in the suprabasal layers of the interfollicular epidermis (unpublished data), consistent with many other examples of K6/K16 in hyperproliferative conditions (). Repetin is a filaggrin-like protein that has been postulated to specifically aggregate K6/K16 filaments, explaining its expression under these conditions (). Unexpectedly, mutants express K13 protein in the spinous layer of the epidermis (). K13 is normally expressed only in stratified but not cornified epithelium, such as tongue and esophagus. K13 expression in epidermis has previously only been reported in papillomas at high risk of converting to squamous cell carcinoma (). The expression of K13 could suggest a role for GATA-3 in squamous cell carcinoma progression. Alternatively, K13 expression could reflect a similar underlying state of the skin that is common to both barrier impairment and tumor progression, such as mounting an inflammatory response.
The first line of cutaneous defense against infection by microorganisms is the proteinaceous/lipid skin barrier. Augmenting this physical barrier are both the innate and adaptive immune systems (; ; ). Antimicrobial peptides, effectors of innate immunity, are expressed by keratinocytes and have distinct but overlapping reactivity against bacteria, fungi, and enveloped viruses (). Antimicrobial peptides are induced to provide a rapid defense, which is particularly important in fetal skin before maturation of immunological memory (). Epithelial defense is a significantly affected pathway in mutant newborns, including a strong up-regulation of the antimicrobial peptides, secretory leukocyte proteinase inhibitor, adrenomedullin, S100A8, S100A9, and β-defensin 1 and 3 (; ; ).
To investigate the specificity of the transcriptional profile of mutant newborn skin, we compared these results with a similar analysis of barrier-impaired −/− skin (). First, the levels of are not altered in mutants and vice versa, suggesting that these transcription factors are not epistatic but distinct in their regulation of epidermal differentiation (unpublished data). Second, the nature of the barrier deficiencies in −/− and mutants appears completely distinct. −− mutants exhibit a specific defect in CE maturation, with normal synthesis but abnormal extrusion of lipids, and a persistent dye penetration even as newborns. Molecularly, is the only “lipid synthesis pathway” gene with decreased levels in mutants. Both mutants do up-regulate genes common to a hyperproliferative state, including K6, K16, and repetin. However, the greatest similarity between the two barrier-deficient mutant mice is an up-regulation of the epithelial defense genes. −/− and mutants show a similar up-regulation of the innate immune effectors, secretory leukocyte proteinase inhibitor and β-defensin 3 (). However, whereas −/− mutants show a strong up-regulation of β-defensin 6, mutants more strongly up-regulate adrenomedullin, S100A9, S100A8, and β-defensin 1. Thus, and mutants, genetically distinct models of barrier impairment, both activate an innate immune response, but they do so through up-regulation of distinct antimicrobial peptides.
These results demonstrate 's specific role in epidermal barrier acquisition. mutant embryos exhibit a transient delay in differentiation, demonstrated by percutaneous dye penetration. Similar delays in dye exclusion were observed in mice with targeted deletions of genes encoding CE proteins, envoplakin, and loricrin (; ). However, envoplakin- and loricrin-deficient mice survive the perinatal period, perhaps because of the compensatory up-regulation of other structural proteins. In contrast, deficiency in the epidermis results in a perinatal lethality with an inherent barrier defect that extends postnatally, as demonstrated by grafting experiments. Underlying mutant's barrier defect is a severe defect in lipid synthesis, in particular, ceramide EOS and glucosylceramides. The electron micrographs of Gata-3 mutant skin, postfixed to maintain lipid structure, are reminiscent of similar findings in infants with severely affected type 2 Gaucher disease. Mutations in β-glucocerebrosidase, the enzyme that catalyzes the hydrolysis of glucosylceramide to ceramide, underlie Gaucher disease (). Type 2 Gaucher disease and a mouse model, with a targeted deletion of β-glucocerebrosidase, manifest at birth with a primary barrier deficiency and display abnormal loosely packed lamellar body–derived sheets in the SC (). Our findings suggest that GATA-3 may act to regulate this important process of lipid biosynthesis, and genes in this pathway should be tested as a potential modifiers to explain the wide phenotypic variation observed among Gaucher patients (). Our studies identified AGPAT5, which catalyzes an essential step in the synthesis of all glycerolipids, as a direct target in vivo for GATA-3 (). Future studies will address the hierarchical transcriptional regulation of lipid synthesis in the skin.
Morphological and transcriptional analyses at distinct developmental stages revealed both GATA-3's regulation of differentiation and lipid synthesis pathways and the compensatory responses to impaired barrier. For example, although the newborn barrier-deficient mutant skin is hypocellular, the grafted mutant skin is acanthotic or hypercellular, as a compensatory response to the impaired barrier in the terrestrial environment ( and ). Analysis of mutants at only one developmental stage would have revealed the specific defect in lipid biosynthesis but would have been refractive to elucidating the transient delay in expression of genes encoding differentiation and cornification proteins. Transcriptional profiling at multiple developmental stages brings clarity to pathways affected by and responding to 's loss.
This process is remarkably well conserved, as GATA transcription factors are also essential to specify the fate and regulate differentiation of epidermal cells in . The cell biology of the epidermis closely resembles that of mammals, including intermediate filament networks and cell connections through adherens and tight junctions (). GATA transcription factor ELT-1 specifies epidermal cell fate (). Subsequently, ELT-5 and -6 (adjacent genes encoding GATA factors) are required throughout development to regulate epidermal cell differentiation ().
Mammalian lung and skin are both epithelia at the interface between the body and the environment that form proteinaceous lipid barriers. Although lung is a branched simple epithelium and the composition of the barriers is distinct, there are remarkable similarities between the systems. At the transcriptional level, corticosteroids and thyroid hormone accelerate barrier maturation in utero of both epidermis and alveoli (). Just as is necessary for the terminal stages of epidermal development, () plays an important role in the terminal stages of lung development (). GATA-6 is the only known GATA factor expressed in the distal epithelium of the developing lung. Expression of a dominant-negative form of GATA-6 in these alveolar cells resulted in a defect in terminal differentiation and proximal airway development. These GATA-6 transgenic mice die perinatally with defects in lipid (surfactant protein) synthesis and decreased expression of , a gene encoding a water channel. Similar to GATA-3's role in epidermal barrier, GATA-6 is necessary for maturation of the proteinaceous lipid barrier that regulates alveoli gas exchange ().
Extending the well-established paradigms from hematopoietic cells, it is intriguing to speculate whether GATA-3 will have similar interactions with family members of other transcription factors in the skin. GATA-1 acts upstream of EKLF (KLF1) during erythroid development, and GATA-3 acts upstream of LKLF (KLF2) during lymphocte development (; ). The expression of GATA-3 and KLF4 in basal and suprabasal cells, respectively, is consistent with GATA-3 acting upstream of KLF4. levels are unchanged in mutants, which could reflect compensatory autoregulation or parallel pathways.
Both and mutants exhibit an epidermal barrier deficiency, but each activates distinct antimicrobial peptides, effectors of innate immunity. Innate immunity is important before the adaptive immune system mounts a response and particularly during the first year of human life, as the adaptive immune system is maturing. These findings demonstrate that newborn skin can mount a robust activation of an innate immunity. Moreover, an analysis of and mutant newborns demonstrates that genetically distinct barrier impairments activate overlapping but distinct innate immune responses. The comparison of - and –deficient newborn epidermis will be very informative to unravel the complex immune response to barrier impairment.
Barrier disruption is a hallmark characteristic of common inflammatory skin disorders, such as atopic dermatitis (more commonly known as eczema) and psoriasis (). Recent work has examined the distinct innate immune responses of psoriasis and atopic dermatitis (). In particular, patients with atopic dermatitis have an increased tendency to develop both disseminated viral skin infection after smallpox vaccine inoculation and recurrent infections because of inadequate innate immune response (). In contrast, barrier-impaired keratitis-ichthyosis-deafness patients develop recurring yeast infections. Mutations in the epidermal cornification protein filaggrin were recently reported to underlie atopic dermatitis, focusing attention on the role that barrier impairment plays in this disorder (). Because of the naive state of T and B cells in newborn mice, a full investigation into this complex innate/adaptive immune response requires adult epidermal-specific targeting of and .
allele.
K14- mice were generated by crossing mice with
K14 mice. Genotyping was done as previously described (; ). The morning of the plug was 0.5 d after coitum. E18.5 dorsal skin was grafted onto nude mice in an area that the mice could not scratch, and these mice were individually housed. All animal studies were approved by the National Human Genome Research Institute animal care and use committee, and all mice were housed in our Association for Assessment of Laboratory Animal Care–accredited facility.
Dye penetration assays were performed with X-gal at pH 4.5 for 4 h at 37°C as previously described (). After staining, embryos were photographed under a dissecting scope (MZFLIII; Leica) using a digital camera (AxioCam; Carl Zeiss MicroImaging, Inc.), and images were acquired with OpenLab software (Improvision). Transepiderrmal water loss was measured using a Tewameter (Courage + Khazaka).
Routine histology and paraffin staining were performed as described previously (). For immunofluorescence, frozen sections were fixed in 10% formalin/PBS and stained with primary antibodies: rabbit polyclonal antibodies against GATA-3 (Segre 379-2b; 1:100), K14 (1:1,000; Covance,), K1 (1:1,000; Covance), Loricrin (1:500; Covance), K13 (1:500; a gift from S. Yuspa, National Cancer Institute, Bethesda, MD), and α6 integrin rat polyclonal antibodies (MAB1982; 1:100; Chemicon). Fluorescent secondary antibodies were Alexa 488 goat anti–rabbit (1:400) and Alexa 594 goat anti–rat (1:200). Slides were mounted with DAPI glycerol media, containing SlowFade Gold antifade, to counterstain nuclei (Invitrogen). Fluorescent staining was imaged with a microscope (Axioplot; Carl Zeiss MicroImaging, Inc.) and photographed with a camera (CoolSNAP; Photometrix).
RNA was isolated from the dorsal skin of newborns and embryos, incubated in RNALater (Ambion), snap frozen, homogenized in TRIzol (Invitrogen) using tissue lyser (QIAGEN), and processed according to the manufacturer's instructions. Northern blot was hybridized with probes for and . Microarrays were done on independent samples for newborns ( = 4) and E15.5 and E16.5 embryos ( = 3).
or , and mutant mice are
K14-. Complimentary RNA was labeled according to the manufacturer's recommendations and hybridized onto Affymetrix 430 2.0 A+B mouse arrays. These arrays contain 45,000 probe sets, representing 34,000 well-substantiated mouse genes. We identified ∼20,000 probes as present in mouse skin during the developmental windows analyzed in these experiments. Microarray results were analyzed by Genesifter using a test (P < 0.05) and Benjamini and Hochberg correction (VizX Labs). Confirmation of fold changes was made with quantitative PCR on cDNA from and mutants on a TaqMan light cycler (Applied Biosystems) with SYBR Green mix (Invitrogen), and primers spanning exon boundaries are listed in Table S1 (available at ).
Pipmaker and MultiPipmaker were performed with repeat masked sequences () with mouse (Chr8:18,841,481-18,861,523), human (Chr8:6,548,286-6,569,868), rat (Chr16:75,769,607-75,791,434), and dog (Chr16:61,666,596-61,692,310) sequences (). Coordinates for blocks 1 and 2 are human (Chr8:6,553,816-6,554,997 and Chr8:6,568,473-6,569,799, respectively). The overlap of amplicon +0.9 and +1.0 in which the GATA-3 conserved sequences are identified is Chr8:6,554,370-6,544,430. Mouse and human sequence coordinates are relative to February 2006 and March 2006 releases, respectively. TRANSFAC was accessed through a National Human Genome Research Institute site license ().
ChIP was performed on human MCF-7 cells, an epithelial cell line that expresses GATA-3 and the lipid biosynthetic genes, including AGPAT5. Chromatin was immunoprecipitated with a GATA-3 antibody (SC-9009; Santa Cruz Biotechnology, Inc.) binding to the endogenous protein. Other reagents were provided in the ChIP-IT kit (Active Motif), and we followed the manufacturer's instructions. DNA/GATA-3 antibody complexes were immunoprecipitated with protein G and A beads. DNA was quantified with QuantiTect SYBR Green PCR kit (QIAGEN). Primers are listed in Table S2 (available at ), and amplification was quantified on a TaqMan light cycler (Applied Biosystems). Binding of GATA-3 to chromatin immunoprecipitated DNA was measured as the change in the number of cycles required to cross a threshold, normalized to sonicated, reverse-cross-linked input DNA.
Whole backskin was removed, placed on a paper towel, and fixed in modified Karnovsky's fixative (2% paraformaldehyde, 2% glutaraldehyde, 0.1 M cacodylate buffer, pH 7.3, and 0.06% CaCl) overnight at 4°C. Samples were washed twice in 0.1 M cacodylate buffer after fixation before embedding. Lipids were extracted into chloroform: methanol mixtures and analyzed by thin-layer chromatography as previously described (). Lipid masses were used to calculate weight percentages. Ruthenium tetroxide transmission EM was performed as previously described ().
Tables S1 and S2 provide the sequences of the primers used for quantitative RT-PCR and ChIP, respectively. Online supplemental material is available at . |
My father is a chemist, and when I was a kid he would bring me to his lab. I spent a lot of time there. Then, when I was in high school, I became increasingly interested in chemistry and biochemistry. I decided to go to college with biochemistry as a major.
I went to Nankai University in Tianjin, which is my hometown and one of the largest cities in the northeastern part of China.
Yes, back in '95. I did my Ph.D. studies in biochemistry and biophysics with Dr. Yue Xiong at UNC Chapel Hill.
At that time in China, most of the universities didn't have enough resources to carry out cutting-edge biomedical research. Like lots of my classmates, I applied to graduate programs in the U.S. to get further education.
For one of my last experiments as a graduate student, I used RNAi to knock down my favorite gene and study its function. It helped me graduate!
At that time I realized how powerful the technique was. But people didn't know the mechanism involved. That got me thinking.
Back then, people knew that siRNAs could lead to the degradation of the target mRNAs, but they didn't know how. I joined Greg Hannon's lab at CSH as a postdoctoral fellow to study the mechanisms of RNAi and took advantage of my previous training in studying protein function.
The first approach was to look for proteins that directly interact with the siRNA. We found that one of the RISC components, Argonaute, binds directly to the siRNA. Then, later on, we found out that Argonaute itself functions as the slicer enzyme that cleaves the target mRNA.
Once cleaved, the mRNA is defective and will be degraded. That's how siRNA leads to the degradation of its targets.
The paper was obviously a quite exciting and important finding, but on the other hand, we were really expecting to find a new enzyme that cleaved the targets. Argonaute had been known for several years; we just didn't know its exact function. Obviously, you always hope to find a novel protein so you'll have more to work on. But instead it was kind of the end of that chapter. So in a way, it was a little disappointing. You shouldn't write that…[laughs].
We wanted to find the endogenous/physiological function of the RNAi machinery. The first endogenous small RNA (currently known as miRNA) was discovered back in 1993 in . However, the real significance of the discovery was not realized until 2000/2001 when several groups reported the presence of large numbers of miRNAs in both invertebrates and vertebrates. Such conservation suggested they were important.
Sequence specificity of RNA silencing is based on base pairing between the small RNA and the potential targets. In the case of siRNAs, the targets are perfectly matched. But for the miRNA, you don't necessarily have perfectly matched targets.
If the targets are not perfectly matched, they cannot go through the cleavage and degradation steps. However, those targets can still be silenced by the miRNA. So the question was, How does miRNA silence gene expression?
It's still the biggest unanswered question. It's related to siRNA-mediated silencing. They both use the same core protein components: the Argonaute proteins and the RNA-induced silencing complex (RISC), but the mechanisms are different.
Some studies have shown that miRNA can lead to mRNA degradation just like siRNA does, although it doesn't cleave its targets. It's possible that miRNA and Argonaute could lead to deadenylation of the target mRNAs, which would then lead to degradation of the mRNA. Other studies have shown that there's no degradation at all but that miRNA instead represses mRNA translation. Almost all the steps of translation including initiation, elongation, and termination have been suggested as the mechanism of miRNA-mediated silencing.
We showed miRNA could lead to the sequestration of target mRNAs into cytoplasmic bodies, and this prevented the mRNAs from being translated. Other studies suggested sequestration is not required, but that binding of Agonaute/RISC to the mRNAs is sufficient to silence them. Everyone seems to agree that miRNA will silence gene expression, but the exact mechanism is still highly debated.
Yes. We decided to look at where the protein components that carry out the silencing localize in the cell. We found, back in 2003, that argonaute proteins localized into very specific foci in the cytoplasm. At that stage we didn't know what the foci were. It was just a very interesting observation.
People first observed similar foci back in '97. They found that a nuclease that degrades mRNA localizes into specific foci in the cell. But again, at that time, nothing was known in terms of function. Later, Roy Parker's group showed that mRNA decapping and degradation occurs in these P-bodies.
The formation of P-bodies is RNA dependent. So, it's a complicated situation: you need mRNA to form those foci, but if the foci's purpose is to just degrade mRNA, then if they do a good job, you shouldn't see them, right?
So when you see them, either they are not efficiently degrading those mRNAs so you see the accumulation, or these foci may function to sequester and store the mRNA, not degrade it. The P-bodies could just sequester mRNAs away from the translational machinery.
I think whether mRNA will be degraded might depend on the protein complex associated with that particular mRNA. Obviously, we haven't figured out what proteins decide these fates yet.
But that could explain why there are so many different reports: some groups see mRNA degradation and some groups don't. Perhaps the P-body functions like a sorting machine for either degrading or protecting and sequestering mRNA. Of course, miRNA is just one branch that can bring mRNA into P-bodies. There are several other pathways, like nonsense-mediated or ARE-mediated RNA decay. These could provide the mRNA substrates and trigger P-body formation. Clearly, in these cases, mRNA will be degraded eventually.
What are the components of P-bodies? Are all P-bodies the same or different? How do they assemble together and what regulates their assembly? Also, how does sequestered mRNA get back into cytoplasm for translation?
From the cell biology point of view, we have a long list of P-body components. We want to study whether these colocalize in the same P-bodies, what are their particular functions, and how do they modulate the silencing activity of miRNA?
We are also trying to purify P-bodies. Right now there is no biochemical way to purify them, so it's hard to argue as to what percentage of the RISC or silenced mRNA is in the P-body as opposed to in the cytoplasm.
Also, a major question for the miRNA field is, What are the real genes that they repress? Because miRNAs are not perfectly matched to their targets, genome database searches give you hundreds of potential hits. There are very few confirmed miRNA targets at this point.
If we can biochemically purify P-bodies, and if our idea about sequestering is right, then the important targets should be there in the P-bodies.
Yeah, the P-body finding—that it could function as both the degradation sites for mRNA and the storage sites for translationally repressed mRNA—I think is exciting. It kind of opens up a new field, a new direction for research. |
Emerin is a type II integral membrane protein residing principally at the inner nuclear membrane (INM) (), where it interacts with a number of other proteins such as lamin A/C () barrier-to-autointegration factor () and β-catenin (). Emerin has also been shown to interact with proteins that are principally found at the outer nuclear membrane (ONM), namely nesprin 1α () and nesprin 2 (). Emerin was identified by positional candidate cloning as the gene responsible for the X-linked form of Emery Dreifuss muscular dystrophy (EDMD) (). The autosomal-dominant form of the disease is caused by mutations in the gene LMNA, which encodes lamins A and C (). Two hypotheses have been formulated to explain why ubiquitously expressed proteins such as emerin, lamin A, and lamin C should cause such highly tissue-specific diseases. These are referred to as the “structural” and the “gene expression” hypotheses. The gene expression hypothesis proposes that emerin and lamins are involved in tissue-specific gene expression and disease may arise from the downstream effects of mutations on chromatin structure or gene expression (). The structural hypothesis proposes that emerin and lamins contribute to the structural integrity of the cell by acting as a load-bearing center underneath the nuclear envelope (NE) (). In this hypothesis, absence of emerin or lamins in disease would render contractile cells like skeletal and cardiac muscle vulnerable to damage, leading to cell death and tissue damage ().
Supporting the structural hypothesis, there is accumulating evidence that the NE is closely linked and connected to its surrounding cytoskeleton. Work on two protein families, the nesprins and Sun proteins, reveals the existence of “bridging” complexes, referred to as the LINC complexes, which span the NE, thus connecting, the INM with the actin cytoskeleton (). A recent study showed that disorganization of the actin, vimentin, and tubulin cytoskeletons arose as a consequence of the absence of lamins A and C in mouse embryonic fibroblasts (), directly supporting the idea that the NE is a load-bearing center in animal cells. Here, we investigate how absence of emerin in human fibroblasts affects cytoskeleton organization. We show that emerin interacts with β-tubulin to anchor the centrosome at the ONM. This unexpected finding provides further support for the structural hypothesis and provides the first clue as to how the tubulin cytoskeleton is connected to the NE.
We wished to investigate whether cytoskeletal abnormalities are induced by the absence of emerin in cells. To this end, human dermal fibroblasts (HDF) from healthy individuals and from X-EDMD patients (which were null for emerin) were investigated for possible abnormalities in the actin, vimentin, and tubulin cytoskeleton (). In stark contrast to findings in lamin A/C-null mouse embryonic fibroblasts, there were little or no differences in the organization of any of the cytoskeletal elements in emerin-null HDFs. Surprisingly, however, we did observe that the centrosome was not positioned next to the nucleus in emerin-null HDFs (, arrowheads).
To confirm this finding, we used an antibody against pericentrin to specifically investigate the position of the centrosome. In normal HDF the centrosome was positioned next to or within 1.5 μm of the NE. In contrast, in four independent emerin-null HDF cell lines the centrosome was >3.0 μm distant from the NE (). To further investigate this phenomenon we looked at centrosome position in a cell line from an X-EDMD carrier, which has approximately equal numbers of emerin-positive and emerin-null cells. We found that in emerin-positive cells the centrosome was positioned next to the NE, whereas in emerin-negative cells the centrosome was >3.0 μm away from the NE. Finally, we investigated the centrosome position in a lamin A/C–null HDF line in which emerin was located entirely within the endoplasmic reticulum (). In the lamin A/C–null cell line the centrosome was also >3.0 μm away from the NE, indicating that absence of emerin from the NE was the cause of the centrosome mislocalization (). To confirm that mislocalization of the centrosome was specific to absence of emerin from the NE, we investigated centrosome positions in a fibroblast cell line from a patient with Greenberg dysplasia, which were null for the INM protein lamin B receptor (LBR) (). Like EDMD, Greenberg dysplasia and the related disorder Pelger-Huey anomaly are characterized by nuclear morphological defects (). However, in LBR-null fibroblasts, emerin was located at the NE and similarly the centrosome was positioned adjacent to the NE (), suggesting that centrosome mislocalization is specific to loss of emerin from the NE. To verify these results, we performed knockdown of emerin by siRNA in normal HDFs (). In HDFs transfected with the scrambled siRNA, centrosomes were found adjacent to the NE. In contrast, in HDFs that were transfected with siRNA specific for emerin, the centrosome was located at a distance of >3.0 μm away from the NE, similar to the distances observed in X-EDMD cells.
This very intriguing result raised the important question of how a protein that is localized in the INM could affect the position of the centrosome, an organelle that is localized at the ONM. To investigate whether a yet-unidentified emerin binding partner could help explain the observed phenomenon, we used recombinant emerin peptides in coprecipitation experiments to identify new emerin binding partners. The peptide was used as a bait to precipitate interacting partners from the egg extracts, which were in turn chosen because they store very large quantities of cytoskeleton and centriolar proteins in a soluble form (). Bands that coprecipitated with emerin were cut from the gel and identified by mass spectrometry. Interestingly, β-tubulin was identified as the most consistent emerin binding protein in this assay. To confirm that emerin is a microtubule (MT) binding protein, MT cosedimentation experiments were performed in which purified MTs were polymerized by taxol and incubated with the same emerin peptide (aa 73–180) or two different emerin peptides corresponding to its chromatin binding domain (aa 1–70) or most of the nucleoplasmic domain (aa 1–176) (). Emerin 73–180 and 1–176 efficiently cosedimented with MTs, whereas emerin 1–70 did not bind to MTs. To estimate the stoichiometry of emerin/microtubule interactions we calculated the tubulin/emerin binding ratios. Emerin 1–176 bound to tubulin at an approximate ratio of 1:8, which is close to the binding ratios of known microtubule-associated proteins (MAPs) (e.g., Enconsin; ). Emerin 73–180 bound to tubulin at an approximate ratio of 1:24, and this weaker interaction is likely due to misfolding of this peptide. Collectively, these data suggest that emerin is a novel MT-interacting protein.
The interaction of emerin with β-tubulin led us to investigate whether MTs are involved in the attachment of the centrosome to the NE. To investigate this possibility, normal and X-EDMD fibroblasts were treated with nocodazole and its effects on MT organization and centrosome position were investigated (). As expected, nocodazole treatment led to the depolymerization of the MT network. Interestingly, when normal HDFs were treated with nocodazole the centrosome was observed to be located >3.0 μm away from the NE, just as was observed in emerin-null fibroblasts. As a control, normal HDFs were treated with latrunculin B to depolymerize the actin cytoskeleton. In latrunculin B–treated HDFs the centrosome was located adjacent to the NE, implying that only disruption of the tubulin cytoskeleton leads to an emerin-null phenocopy. We confirmed this finding using biochemical fractionation to determine whether centrosomes cosedimented with the nucleus in a range of HDF lines (Fig. S1, available at ). In normal HDFs, nuclei and centrosomes cosedimented at a 1:1 ratio. In emerin-null HDFs or normal HDFs treated with nocodazole, nuclei and centrosomes cosedimented at a 1:0.4 ratio, again showing that centrosomes were detached from the NE.
All the above experiments provide strong evidence that emerin links centrosomes to the NE via a MT association. Given that emerin is a protein of the INM, this raised the question as to whether emerin acts via another protein that crosses the NE. We therefore investigated whether either SUN domain proteins or one of the nesprins is also mislocalized in emerin- null HDFs as a first step to determining whether these proteins might be involved in centrosome localization. We could not detect any change in the distribution of SUN1 or nesprin 1 (not depicted) or SUN2 or nesprin 2 (Fig. S2, available at ) in these cells, suggesting that these proteins were not involved in centrosome localization. As a result of this finding we decided to reinvestigate the localization of emerin in normal fibroblasts. Using digitonin permeabilization, we showed that a considerable fraction of emerin was concentrated at the ONM, with a further dispersed fraction in the peripheral ER, with two independent anti-emerin antibodies, whereas lamins A/C and SUN2 were undetectable under similar conditions and therefore located exclusively at the INM (). The anti-Sun2 antibody used in this assay recognizes a luminal domain, indicating that not only is the INM intact, but also the ONM, strengthening the finding that a fraction of emerin resides at the ONM. To confirm that the protein detected at the ONM was indeed emerin, we performed siRNA knockdown of emerin on control fibroblasts and again stained the cells with anti-emerin antibodies after digitonin permeabilization. In these experiments knockdown was ∼70% efficient, and whereas emerin was detected at the ONM in fibroblasts treated with scrambled siRNA, staining was eliminated in cells transfected with siRNA specific to emerin (). As a further control we also stained X-EDMD fibroblasts, (which are null for emerin) or an X-EDMD carrier in which emerin is absent from ∼50% of cells. We found that staining of the ONM was undetectable in emerin-null fibroblasts, further supporting the presence of emerin at the ONM. This finding implies that emerin residing at the ONM can interact directly with centrioles via MTs. It has recently been reported that localization of emerin at the INM is dependent upon the presence of both nesprin1α and 2β (). To investigate whether nesprins might also be involved in localization of emerin to the ONM, we transfected normal HDF with a dominant-negative Sun1 mutant that causes loss of the ONM form of nesprin 2 (). We found that emerin was still detected at the ONM in the presence of this mutant, suggesting that nesprin2 does not cause the localization of emerin to the ONM ().
Overall, our results show that emerin interacts directly with MTs and that emerin and MTs both are necessary for the association of the centrosome with the NE. Our findings are supported by previous work, which showed a colocalization of emerin with β-tubulin in mitotic cells () and an enrichment of emerin at the kinetochores, near the spindle poles, during NE reassembly (). Recent evidence has demonstrated how complexes involving lamins A/C, the SUN domain proteins, and the nesprins link the actin and intermediate filament cytoskeletons to the NE in mammalian cells (; ). Our current data now reveal how the tubulin cytoskeleton via the centriole interacts with the NE in human fibroblasts.
HDF cells and LBR-null cells (provided by K. Hoffmann, Charité Humboldt University, Berlin, Germany) were grown in DME supplemented with 10 U/ml penicillin, 50 μg/ml streptomycin, and 10% vol/vol NCS, and maintained at 37°C in a humidified atmosphere containing 5% CO until 80% confluence. Serial passage was performed in the presence of trypsin and 0.5% EDTA.
HDFs were fixed with methanol/acetone (1:1) on ice for 5 min or with 4% paraformaldehyde for 12 min at RT followed by permeabilization either with 1% Triton X-100 (for 5 min at RT) or with Digitonin (40 μg/ml for 2 min, on ice). Primary antibodies used and their dilutions are described in . Anti-Sun2 was provided by B. Burke (University of Florida, Gainesville, FL) and anti-Nesprin 2 K1 antibody was provided by I. Karakesisoglou (University of Cologne, Cologne, Germany). FITC- or TRITC-conjugated secondary antibodies were obtained from Stratech and chromatin was visualized with DAPI/Mowiol.
One-dimensional SDS-PAGE was performed according to . For immunoblotting, proteins were transferred onto nitrocellulose membranes (Schleicher & Schuell) using the Mini Trans-Blot system (Bio-Rad Laboratories). HRP-conjugated secondary antibodies were obtained from Jackson ImmunoResearch Laboratories. Bands were visualized by enhanced chemiluminescence (ECL reagents; GE Healthcare).
Emerin-specific siRNA duplexes were obtained from Ambion. The sequence of sense nucleotides was as follows: 5′-GGUGGAUGAUGACGAUCUUtt-3′. RNAi transfection procedure was performed as in . Specific silencing of emerin was confirmed by three independent experiments.
For the pull-down experiments, purified emerin peptide 73–180 was immobilized on Ni-beads and incubated with the cytosolic fraction of egg extracts for 4 h at 4°C, on a roller. Non-specific binding was removed with 250-mM NaCl washes. Emerin together with co-eluting proteins were resolved by SDS-PAGE. egg extracts were prepared and fractionated to generate membrane-free cytosol as described by .
Tubulin was purified from bovine brain according to the method of . Microtubules were polymerized in the presence of taxol as described previously (). 10 μg of purified emerin peptides 1–70 and 73–180 were mixed with 20 μg of taxol-stabilized MTs and centrifuged at 200,000 . No MTs were added to control samples. Samples were analyzed by SDS-PAGE. The amount of the emerin fragments was quantified in the gel using NIH Image software and normalized by the mean. To measure the relative binding ratio of tubulin to emerin in the microtubules, microtubule cosedimentation assays were performed as described above, then the samples were separated by SDS-PAGE and Coomassie brilliant blue R-250 was used to stain the gels. The gels were quantified using NIH Image software. The average density values were multiplied by the measured area values to get the absolute intensity, and then the background values were subtracted from the protein band values.
is molecular mass (50 kD for tubulin, 22 kD for aa 1–176 fragment, and 15 kD for aa 73–180 fragment).
Normal HDFs were transfected with 2 μg of plasmid SS-HA-Sun1L-KDEL () with the Amaxa Biosystems Nucleofection system using the Basic Nucleofector kit for primary mammalian fibroblasts (VPI-1002) as described by the manufacturer. Cells were fixed 24 h after transfection with 4% paraformaldehyde and underwent sequential permeabilization. Cells were initially permeabilized with digitonin and stained with the rabbit-emerin AP8 antibody. Cells were then fixed again with 4% paraformaldehyde incubated with 1% Triton X-100 and stained with a mouse-HA tag antibody to detect the Sun1 construct.
Cells were grown to 90% confluence and collected by centrifugation. Cell pellets were resuspended in Nuclear Isolation Buffer (NIB) (250 mM NaCl, 3 mM MgCl, 10 mM Tris-HCL, pH 7.6, 0.5% Nonidet, and protease inhibitor cocktail; 1 ml NIB/10 cells) and incubated on ice for 15 min. Nuclei were released with a Wheaton homogenizer and isolated on coverslips by centrifugation (4,000 rpm for 10 min at 4°C) through a 30% sucrose cushion.
Fig. S1 shows nuclear isolation experiments in normal and X-EDMD cells with increased salt concentration. Fig. S2 shows Sun 2 and Nesprin 2 staining in normal and X-EDMD HDFs. Online supplemental material is available at . |
C
e
n
t
r
i
o
l
e
s
h
a
v
e
d
u
a
l
r
o
l
e
s
i
n
m
a
m
m
a
l
i
a
n
c
e
l
l
s
:
a
s
t
h
e
c
o
r
e
o
f
t
h
e
c
e
n
t
r
o
s
o
m
e
,
t
h
e
y
p
a
r
t
i
c
i
p
a
t
e
i
n
f
o
r
m
i
n
g
t
h
e
m
i
t
o
t
i
c
s
p
i
n
d
l
e
,
a
n
d
,
i
n
q
u
i
e
s
c
e
n
t
c
e
l
l
s
,
t
h
e
y
m
i
g
r
a
t
e
t
o
t
h
e
c
e
l
l
c
o
r
t
e
x
t
o
f
u
n
c
t
i
o
n
a
s
b
a
s
a
l
b
o
d
i
e
s
f
o
r
p
r
i
m
a
r
y
c
i
l
i
a
f
o
r
m
a
t
i
o
n
.
L
i
k
e
c
e
n
t
r
i
o
l
e
s
,
b
a
s
a
l
b
o
d
i
e
s
c
o
n
s
i
s
t
o
f
a
c
o
r
e
s
t
r
u
c
t
u
r
e
(
w
i
t
h
p
e
r
i
p
h
e
r
a
l
c
o
m
p
o
n
e
n
t
s
)
o
f
n
i
n
e
t
r
i
p
l
e
t
m
i
c
r
o
t
u
b
u
l
e
s
o
r
g
a
n
i
z
e
d
i
n
t
o
a
c
y
l
i
n
d
e
r
.
T
h
e
c
y
l
i
n
d
e
r
c
a
n
b
e
f
u
r
t
h
e
r
d
i
v
i
d
e
d
i
n
t
o
d
i
s
t
i
n
c
t
r
e
g
i
o
n
s
f
o
r
w
h
i
c
h
f
e
w
o
f
t
h
e
m
o
l
e
c
u
l
a
r
c
o
m
p
o
n
e
n
t
s
o
r
f
u
n
c
t
i
o
n
s
a
r
e
k
n
o
w
n
.
E
a
c
h
d
o
m
a
i
n
f
u
l
f
i
l
l
s
a
r
o
l
e
f
o
r
b
a
s
a
l
b
o
d
i
e
s
a
n
d
t
h
e
i
r
a
t
t
a
c
h
e
d
c
i
l
i
a
.
T
h
e
s
e
p
o
o
r
l
y
d
e
f
i
n
e
d
f
u
n
c
t
i
o
n
s
i
n
c
l
u
d
e
b
a
s
a
l
b
o
d
y
a
s
s
e
m
b
l
y
,
b
a
s
a
l
b
o
d
y
c
o
r
t
i
c
a
l
a
t
t
a
c
h
m
e
n
t
,
p
r
o
t
e
i
n
r
e
c
r
u
i
t
m
e
n
t
f
o
r
c
i
l
i
a
r
y
a
s
s
e
m
b
l
y
,
m
i
c
r
o
t
u
b
u
l
e
n
u
c
l
e
a
t
i
o
n
,
a
n
d
c
e
l
l
u
l
a
r
p
o
l
a
r
i
t
y
.
B
a
s
a
l
b
o
d
y
d
e
f
e
c
t
s
l
e
a
d
t
o
a
r
a
n
g
e
o
f
d
i
s
e
a
s
e
s
,
i
n
c
l
u
d
i
n
g
B
a
r
d
e
t
-
B
i
e
d
l
s
y
n
d
r
o
m
e
a
n
d
P
o
l
y
c
y
s
t
i
c
K
i
d
n
e
y
D
i
s
e
a
s
e
.
A
n
u
n
d
e
r
s
t
a
n
d
i
n
g
o
f
t
h
e
b
a
s
a
l
b
o
d
y
f
u
n
c
t
i
o
n
a
n
d
d
y
s
f
u
n
c
t
i
o
n
i
n
d
i
s
e
a
s
e
r
e
q
u
i
r
e
s
a
c
o
m
p
r
e
h
e
n
s
i
v
e
i
n
v
e
n
t
o
r
y
o
f
t
h
e
p
r
o
t
e
i
n
s
t
h
a
t
c
o
m
p
r
i
s
e
t
h
e
b
a
s
a
l
b
o
d
y
a
s
w
e
l
l
a
s
t
h
e
s
t
r
u
c
t
u
r
a
l
c
o
n
t
r
i
b
u
t
i
o
n
o
f
e
a
c
h
.
T
h
i
s
i
s
t
h
e
f
i
r
s
t
s
t
u
d
y
t
o
c
o
m
b
i
n
e
a
b
a
s
a
l
b
o
d
y
p
r
o
t
e
o
m
e
w
i
t
h
a
n
u
l
t
r
a
s
t
r
u
c
t
u
r
a
l
a
n
a
l
y
s
i
s
o
f
n
e
w
l
y
i
d
e
n
t
i
f
i
e
d
b
a
s
a
l
b
o
d
y
c
o
m
p
o
n
e
n
t
s
.
W
e
r
e
v
e
a
l
2
4
n
e
w
b
a
s
a
l
b
o
d
y
p
r
o
t
e
i
n
s
a
n
d
d
e
s
c
r
i
b
e
t
h
e
s
p
e
c
i
f
i
c
d
o
m
a
i
n
s
o
f
l
o
c
a
l
i
z
a
t
i
o
n
f
o
r
1
9
.
P
r
o
t
e
i
n
s
w
e
r
e
i
d
e
n
t
i
f
i
e
d
f
r
o
m
v
a
r
i
o
u
s
d
o
m
a
i
n
s
w
i
t
h
i
n
t
h
e
b
a
s
a
l
b
o
d
y
,
a
l
l
o
w
i
n
g
u
s
t
o
p
r
e
d
i
c
t
t
h
e
i
r
r
o
l
e
s
i
n
b
a
s
a
l
b
o
d
y
f
u
n
c
t
i
o
n
.
To describe a comprehensive inventory of the molecular components that comprise the basal body, two separate isolation techniques were used in combination with the high throughput shotgun proteomic technique, multidimensional protein identification technology (MudPIT;). In one preparation, cell cortices (pellicles) with and without associated basal bodies were isolated to identify proteins dependent on γ-tubulin for basal body localization (; see Pellicle preparation section in Materials and methods). γ-Tubulin functions in microtubule nucleation and is involved in the early steps of basal body assembly (). In , γ-tubulin is required for both the assembly and maintenance of basal bodies; transcriptional repression leads to the loss of basal bodies after 20 h (). Pellicles were prepared from cells in the presence and absence of γ-tubulin (). In the second preparation, the oral apparatuses were isolated from cells (). The oral apparatus is composed of >100 ciliated and nonciliated basal bodies interconnected by a framework of microtubules and filaments to generate a feeding structure (; ). The associated basal bodies were extracted from the oral apparatus substructure (see Basal body lysate preparation section in Materials and methods). Basal body proteins extracted from the oral apparatus are reported to reassemble intermediate basal body assembly structures in vitro (). Using these two isolation techniques, we identified proteins necessary for intermediate structures in the basal body assembly pathway. Pellicle preparations allowed for the isolation of basal body components that are dependent on γ-tubulin for basal body localization, giving us insight into proteins involved in the maturation and maintenance of basal bodies. Proteins isolated from the oral apparatus allowed for the identification of components involved in the formation of the cartwheel, an early structure in basal body assembly (; ; ).
Proteins from all samples (pellicles with basal bodies, pellicles without basal bodies, and oral apparatuses) were identified using MudPIT (see Mass spectrometry section in Materials and methods). To identify proteins specifically found in the pellicle sample with basal bodies, components unique to each pellicle sample were tabulated (; ). Proteins identified from the sample lacking γ-tubulin and basal bodies were subtracted from the study. Of the 398 proteins that were removed, only one known basal body component (α-tubulin) was identified. α-Tubulin is a ubiquitous protein that is found throughout the cell cortex (), although it is greatly reduced from pellicles lacking γ-tubulin by Western blotting ().
Combining the data from the two approaches yielded a basal body proteome of 355 proteins (Table S1, available at ); 175 proteins were dependent on γ-tubulin for association with the basal body, and 220 proteins were extracted from the isolated oral apparatuses. 40 proteins were present in both samples. All 355 proteins were annotated for homologous proteins by BLAST searches in the , , , and databases; gene ontology codes; and presence in proteomics studies of related structures (Table S2; ; ; ; ; ; ; ; ; ; ). Based on these analyses, the proteins were classified into eight broad categories: basal body, candidate, cilia, other, metabolism, mitochondria, transcription/translation, and not conserved (). 22 of the 355 proteins (6%) were identified as known basal body/centriole components from previous studies (Table S1; ; ; ; ; ; ; ; ). Approximately one third of the total proteins was classified as known or candidate basal body proteins, several of which were verified through this study. One third were likely contaminating proteins; another quarter were proteins that had no obvious homology to vertebrate cells (see Mass spectrometry section in Materials and methods). Because the centriole is a highly conserved organelle, the nonconserved proteins were not studied further. However, it is worth noting that these proteins may be important for future studies in understanding basal body duplication and assembly. By analogy, Bld10p, which is found in , is involved in the early assembly steps of the basal body; however, an obvious Bld10 candidate has not been found in other species ().
We classified 79 proteins as centriole/basal body candidates because they fell into at least one or more of the following three categories. First, proteins were chosen that contained domains or structural motifs that were previously found to associate with basal bodies, centrioles, or microtubules. These included domains that interact with microtubule structures (e.g., assemblin, glyceraldehyde-3-phosphate dehydrogenase, Lis2, and EF hand) or domains found in many known centriole/centrosome proteins (e.g., WD40 repeats, HEAT repeats, DM10, 14-3-3 domains, or coiled-coil domains). Second, proteins were classified as candidates if homologues were identified by previous studies using proteomic, bioinformatic, or comparative genomic approaches in a variety of organisms, including , , , , and human cells, to identify molecular components of centrioles (), centrosomes (), or cilia (; ; ; ; ; ). Many proteins were identified in multiple studies, confirming that not only is the structure and function of microtubule-organizing centers conserved across phylogeny but the molecular components are conserved as well (Table S2). This comparative proteomics approach enabled cross validation of proteins to make them stronger candidate basal body components. Proteins identified through this approach include 17 of the POC (proteome of centriole) and BUG (basal body proteins with upregulated genes) proteins found in the first reported proteomic analysis of centrioles (). Although each of the aforementioned studies identified new potential protein components of the studied structure, localization studies are necessary to confirm the identity as a basal body/centriole component. Finally, we focused on proteins in which mutations in the human orthologues are known to cause ciliary dysfunction and disease. For example, mutations in LIS-1 cause mental retardation, neurodegeneration, and male sterility (; ), mutations in Spag-6 cause hydrocephalus (), and mutations in Parkin coregulated gene (PACRG) result in sterility (). orthologues for each of these proteins were identified.
Proteins classified as candidates through the aforementioned secondary screen were then selected for biological validation by GFP fluorescence localization in live cells. We generated GFP fusions to 40 genes, all of which were expressed (not depicted), and 24 of the proteins localized to basal bodies (summarized in ; also see Figs. S1–3; available at ). This is in contrast to the diffuse localization of GFP alone (Fig. S1). We used centrin (Cen1), a known basal body component and proteomic hit found in our study, as a positive control for proper basal body localization (). Each new basal body component localized in the canonical basal body pattern for both cortical rows and the oral apparatus (; and Fig. S1–3). In addition, have 17 microtubule-containing structures in the cell () and several proteins localized to multiple microtubular structures. For example, we identified proteins that localize to the micronuclear spindle (e.g., Bbc23) and cilia (e.g., Spag-6) in addition to basal bodies. Of the 24 newly confirmed basal body components, all but one (BBC14) are conserved in most of the eukaryotes examined.
To begin to understand the function of each newly identified basal body component, we determined the ultrastructural localization using immunoelectron microscopy (see Immunoelectron microscopy section in Materials and methods). Detailed localization data provide critical clues for function, and this is the first study describing protein localization to basal body domains in combination with a basal body proteome. Basal bodies consist of a highly conserved structure built on a core of nine symmetrically arrayed triplet microtubules with peripheral components necessary for microtubule nucleation, basal body anchoring to the plasma membrane, polarity establishment, and protein docking. Distinct regions exist for which few functions are known (). Assembly of the nascent organelle in ciliates reveals several intermediate structures in basal body formation (; ; ). Localization of a component to an intermediate structure implies not only a role at a given stage in assembly but also information about the role the protein plays in basal body function. In this study, we define the localization of 19 newly identified basal body components to intermediate structures found in the assembly pathway and to mature basal bodies.
Basal body assembly begins with the appearance of an amorphous disk structure near the base or proximal end of the existing mother basal body. Bbc14, Bbc31, and Poc1 localized to this site of assembly, as does centrin (; ; ; ). The cartwheel, which consists of a central hub and nine radial spokes, assembles on top of the amorphous disk. Three proteins (Poc1, Bbc82, and Sas6a) localize to the cartwheel of mature basal bodies, a domain also observed early in assembly (, S2, and S3). The Bld10 protein also localizes to cartwheels and is essential for basal body assembly (). In addition, Bbc29 localized to the proximal end of the cartwheel (Fig. S1). Thus, Bbc29, Poc1, Bbc82, and Sas6a may all be crucial for early basal body assembly and/or maintenance.
At the tip of each spoke of the cartwheel, the A and subsequently the B and C microtubules are nucleated perpendicular to the mother basal body. These microtubules form the triplet blades for the cylindrical structure of the probasal body. During probasal body maturation, the cylinder elongates and separates from the mother as it inserts into the membrane, forming a structure that will nucleate a cilium. The mature and functional basal body exhibits several distinct morphological domains for which we identified new proteins localizing to each ( and ).
Bbc23, Bbc30, Bbc52, Bbc57, Bbc73, and Bbc78 localized to the microtubule scaffold or walls of the core cylinder (, S1, and S2). These proteins appeared as a sheath surrounding the core structure. Similarly, ɛ-tubulin (; ) and actin () are found surrounding the basal body cylinder. Although ɛ-tubulin appears to have a conserved role in assembling or maintaining the triplet microtubule structure, actin contributes to normal ciliary structure and motility ().
In addition to the general localization of proteins along the length of the basal body cylinder, regions of discrete localization were also observed. The proximal one third of the microtubule scaffold is surrounded by an electron-dense collar structure. Six proteins (Bbc30, Bbc53, Eno1, Ftt18, Ftt49, and PACRG) localized to the collar ( and Figs. S1–3). The collar is used as a site of attachment for cortical structures that contribute to the highly organized cytoskeleton of the cell (). Finally, Cen1 and Ftt49 localized to a common site approximately equidistant from the ends of the basal body. This domain was defined as the midpoint and may be analogous to satellite structures found in centrioles.
The distal-most region of the mature basal body is the transition zone, which appears as two sheets in longitudinal sections (). Proteins localizing to the transition zone play roles in creating a foundation for the nucleation of ciliary microtubules and/or as a docking site for protein transport. Bbc20, Bbc23, Bbc52, Bbc53, Bbc73, PACRG, and Spag6 localized to the transition zone ( and Figs. S1–3). Specifically, Bbc53 and Spag6 localized to the center of the transition zone at the site of central pair assembly. This localization is consistent with the role of Spag6 in the assembly and/or maintenance of ciliary central pair microtubules (), whereas Bbc20 (a Lisencephaly-1 domain–containing protein) localized to the side of the transition zone similar to intraflagellar transport protein localization at basal bodies (). It should be noted that this localization is distinct from the transition fiber localization observed for IFT52 (), although both regions seem to function in intraflagellar transport motility. Mutations in Lis-1 domain proteins cause several pathologies in humans, including mental retardation and neurodegeneration.
The mature basal body lumen is filled with an opaque material extending from the cartwheel nearly to the terminal plate. To our knowledge, only centrin and γ-tubulin have been localized to this region and are thought to form a continuous filament scaffold throughout the lumen (; ). We have identified one additional luminal basal body component, Bbc57.
The ultrastructural localization of new basal body proteins is illustrated in a compilation model describing the molecular architecture of the organelle (). By localizing basal body components into specific domains, we can begin to consider distinct functional roles for these proteins as well as possible interacting proteins. We have identified 24 new basal body protein components and have defined a high resolution map for 19. These data provide a detailed, albeit preliminary, molecular view of the basal body structure. Similar structural, biochemical, and genetic studies have contributed to the comprehensive picture of the yeast spindle pole body (for review see ), nuclear pore complexes (), and kinetochores (). A comprehensive understanding of basal body and centriole function throughout the cell cycle requires an inventory of the components and how each contributes to the overall structure. Our proteomic and molecular approach combined with future genetic and biochemical studies will provide a comprehensive molecular map for understanding the assembly and functions of basal bodies and centrioles.
strains CU428, SB1969, B2086 ( Stock Center, Cornell University), and TTMGII (a gift from M. Gorovsky, University of Rochester, Rochester, NY) were used in this study. Unless specified, cells were grown in super proteose peptone (SPP) media (2% proteose peptone, 0.1% yeast extract, 0.2% glucose, and 0.003% Fe-EDTA) at 30°C. TTMGII cells were grown in SPP with 1% proteose peptone at 22°C. For conjugation, midlog phase cells were washed and resuspended in starvation media (10 mM Tris, pH 7.5). After 18–20 h, equal numbers of each mating type were combined and incubated at 30°C.
TTMGII () cells were grown to midlog phase in SPP with 1 mg/ml CdCl to maintain γ-tubulin expression. To deplete γ-tubulin, cells were washed three times in starvation media and diluted to 3 × 10 cells/ml without CdCl for 24 h at 22°C before pellicle preparation.
Pellicles were prepared from 1 liter of TTMGII cells grown to ∼2 × 10 cells/ml in SPP with and without 1 mg/ml CdCl. Cells were spun for 15 min at 600 and 4°C and were resuspended in 15 ml of ice-cold homogenization buffer (HB; 20 mM Hepes, pH 7.0, 40 mM NaCl, 0.3 M sucrose, and 2 mM MgCl) plus protease inhibitors (1 μg/ml leupeptin, 15 μg/ml E64, 10 μg/ml chymostatin, and 10 μg/ml antipain). Cells were spun again for 1 min, resuspended in 15 ml HB, left on ice for 10 min, transferred to a Dounce homogenizer, and lysed with 40 strokes. The lysed cells were loaded onto a sucrose step gradient (1.46 and 1 M sucrose) and spun at 2,500 rpm for 10 min at 4°C. Pellicles were collected from the interphase layer and washed with HB (the protocol was adapted from ). Electron micrographs show that much of the cytoplasm and contaminating organelles were removed through this preparation. Pellicles were solubilized in HB + 1% NP-40 and TCA precipitated.
Basal body protein lysates were prepared using modified methods that were previously published (). 1.5 liters of cells were grown to ∼3 × 10 cells/ml in SPP, harvested, and washed with cold 0.12 M sucrose. All of the following procedures were performed at 4°C. Cells were incubated in isolation buffer (IB; 1 M sucrose, 1 mM EDTA, 0.1% 2-mercaptoethanol, 10 mM Tris, pH 7.3, at room temperature, and 0.75% Triton X-100) for 4 h to isolate oral apparatuses from the cell cortex. Samples were then washed with 2% Triton X-100 in IB before douncing in a 50-ml tissue homogenizer. Samples were washed again with 2% Triton X-100 in IB, filtered through an 8.0-μm polycarbonate filter (Nucleopore Filter Corp.), and spun, and the isolated oral apparatuses were resuspended in IB without detergent (). To isolate basal body proteins (basal body lysate) from the oral apparatus framework, KCl was added to a final concentration of 1 M and incubated for 18 h with gentle stirring. Samples were then spun at 30,000 g for 30 min to pellet the oral apparatus framework, leaving the soluble basal body lysate. The lysate was then dialyzed into storage buffer (10 mM MES, pH 6.7, 150 mM KCl, 0.5 mM MgSO, 1 mM EGTA, 0.5 mM DTT, and 1 M sucrose) and TCA precipitated.
TCA precipitates from purified basal bodies were resuspended in 200 mM NaCO, pH 11, adjusted to 8 M urea, reduced, and alkylated as reported previously (). 5 μg proteinase K was added to the sample and incubated at 37°C for 5 h in a Thermomixer (Brinkmann; ). The digestion was stopped by the addition of formic acid to 5%, microcentrifuged at 18,000 and 4°C for 15 min to remove particulates, and subsequently pressure loaded on a 6-cm-long fritted () microcapillary-fused silica column (250-μm inner diameter) packed with 3 cm/5 μm C-18 resin (Aqua; Phenomenex) and 3 cm/5 μm strong cation exchange resin (Whatman). This precolumn was coupled via a true zero dead volume union (UpChurch Scientific) to a 15-cm/75-μm analytical column packed with 3 μm C-18 resin (Aqua; Phenomenex). Liquid chromatography/liquid chromatography/mass spectrometry/mass spectrometry was then performed as described previously (). Next, the collection of resulting ms2 spectra was searched against the preliminary sequence data for using the SEQUEST algorithm (). Data for scaffolds was obtained from The Institute for Genomic Research (TIGR; ; ). Peptide identifications were organized and filtered using the DTASelect and Contrast programs (). Filtering criteria for positive protein identifications for all purifications were the identification of two unique peptides with a false positive rate ≤5%. The proteins corresponding to the matching peptides are listed in Table S1 along with the number of peptides matched to each protein and spectral count, which gives an indication of the relative abundance of the protein in the sample. All proteins were analyzed with BLAST searches to human, , , , and databases (BLAST e value > 10).
To demonstrate the presence of γ-tubulin, α-tubulin, and centrin in pellicles, 0.5 × 10 pellicles (solubilized in SDS-PAGE buffer) per lane were separated by SDS-PAGE. Blots were performed with anti–γ-tubulin antibody (GTU88; 1:500), anticentrin (1:1,000), or anti–α-tubulin (DM1A; 1:500). Proteins were detected using an infrared scanner (Odyssey System; LI-COR).
Candidate gene ORFs (preliminary gene predictions TIGR Genome Database) were amplified from genomic DNA by PCR with attached cloning sites. The products were cloned into pIGF.1 (gift from D. Chalker, Washington University, St. Louis, MO) to generate an N-terminal GFP fusion under the control of the metallothionein promoter. Protein expression was induced by the addition of 0.2–1.0 μg/ml CdCl for 1–2 h, and cells were then washed into fresh media without induction for 2–4 h. Live cells were washed in 10 mM Tris, pH 7.4, and GFP was imaged using an upright microscope (DMRXA/RF4/V; Leica) with a CCD camera (SensiCam; Cooke) at 25°C. Images were collected using a PL-APO 63× NA 1.32 objective (Leica) and the Slidebook software package (version 3.0.6.6; Intelligent Imaging Innovations). Through z series, maximum projections were generated using ImageJ software (National Institutes of Health [NIH]).
cells were pelleted, high-pressure frozen in a machine (HPM-010; Bal-Tec), freeze substituted in 0.25% glutaraldehyde/0.1% uranyl acetate in acetone, and embedded in Lowicryl HM20. 60-nm serial sections were cut and put on nickel slot grids, blocked with 1% milk in PBS–Tween 20, and incubated with anti-GFP (a gift from J. Kahana and P. Silver, Dana Farber Cancer Institute, Boston, MA) at 1:100. 15 nm gold-conjugated secondary antibody was applied to the grids at a dilution of 1:20 (Ted Pella). Grids were poststained with 2% uranyl acetate and lead citrate. Images were collected using an electron microscope (CM10; Philips) equipped with a digital camera (BioScan2; Gatan) and digital micrograph software (Gatan). The domains of localization for each protein were determined by imaging the basal bodies of at least five cells and compiling the gold particles onto one schematic cartoon image (Figs. S1–3). Domain localization was defined if at least 20% of the total gold particles were seen at a specific region.
Figs. S1–3 describe the fluorescence and immunoelectron microscopy localization of basal body components found in this study. Table S1 shows a summary of the mass spectrometry data. Table S2 shows a summary of the comparison to other proteomics/genomics studies. Online supplemental material is available at . |
Mouse embryonic stem (ES) cells derived from the inner cell mass of blastocyst embryos have the ability to self-renew and are pluripotent. ES cell pluripotency is maintained via the LIF-gp130-STAT3, bone morphogenetic protein (BMP)–Smad-Id, and probably Wnt and mTOR signaling cascades (; ; ; ; ; ; ; ). Intracellular regulators of ES cell self-renewal include Oct4, Sox2, Nanog, and the recently implicated transcription factors Sall4, Esrrb, Tbx3, and Tcl1 (; ; ; ; ; ; ).
Using chromatin immunoprecipitation on chip analyses to map Oct4, Sox2, and Nanog target genes, a large group of genes was identified that is coregulated by these factors in different combinations, although the majority of genes was cooccupied by Oct4, Sox2, and Nanog (; ). Interestingly, many of these target genes are not expressed in ES cells.
Recent reports showed that in ES cells, many differentiation genes are silenced by Polycomb group (PcG) complexes, indicating that the epigenetic regulation of gene expression is essential for maintaining ES cell pluripotency (; ; ; ; ; ). Interestingly, many of the repressed Nanog, Oct4, and Sox2 target genes were cooccupied by PcG complexes, suggesting that ES cells are poised to enter differentiation programs but are held in check by PcG-mediated chromatin modifications. The suggestion that epigenetic regulation is an important instrument to control ES cell pluripotency versus their capacity to differentiate is further supported by the findings that the PcG protein Suz12 is required for ES cell differentiation () and that a functional NuRD (nucleosome remodeling and disruption) complex, which is involved in nucleosome remodeling, is required for the lineage commitment of ES cells ().
Apart from , , and , other genes are also highly and almost exclusively expressed during early embryogenesis (; ). One of these genes, the () gene, is specifically expressed in the inner cell mass and primitive ectoderm and is down-regulated at early primitive streak stages (). Expression is maintained in the primordial germ cells in developing embryos and in the gonads in adult animals (). Promoter analysis indicated that the murine gene is transcriptionally regulated by Oct4 and Sox2 (). The UTF1 protein was shown to repress transcription (), to activate reporter genes in an ATF2-dependent manner, and to interact with the basal transcription factor TFIID (; ). A recent study suggested a role for UTF1 in the proliferation rate and teratoma-forming capacity of ES cells (Nishimoto et al., 2005).
The purpose of this study was to determine the requirement of UTF1 for ES cell self-renewal and/or differentiation and to gain insight into its mechanistic properties. Using knockdown (KD) strategies, we determined that UTF1 is involved in ES cell differentiation. UTF1 KD perturbed ES and embryonic carcinoma (EC) cell differentiation, whereas their ability to self-renew was unaffected. UTF1 displays transcriptional repressor activity, and a combination of localization experiments, FRAP protocols, and subcellular fractionation assays indicated that UTF1 is stably chromatin associated with dynamics and biochemical properties similar to core histones.
To study the potential role of mouse UTF1 (mUTF1; hereafter UTF1) in ES and EC cell differentiation, we stably expressed UTF1 and Renilla luciferase (hereafter Renilla) siRNAs in P19CL6 EC cells. UTF1 expression levels were substantially decreased in all clones tested (), whereas expression levels of the pluripotency marker Oct4 were not affected (). Next, DMSO-induced differentiation of wild-type (wt), Renilla, and UTF1 KD cells was analyzed (). wt and Renilla KD cells differentiated normally, which was reflected by a drastic reduction in Oct4 levels around day 4, decreased UTF1 levels between days 4 and 6, and detectable GATA4 (not determined for Renilla) and Troma1 expression by day 8. Actin was used as a protein loading control. In UTF1 KD lines, the differentiation-induced down-regulation of Oct4 was either delayed (#1) or minor (#2), and both GATA4 and Troma1 were not detected. Residual UTF1 protein levels were not further down-regulated, most likely as a consequence of high Oct4 levels, a transcriptional activator of the gene.
As UTF1 was previously reported to be involved in ES cell proliferation (), we determined the doubling times of wt, Renilla KD, and UTF1 KD EC cells. UTF1 KD cells showed a 24% and 17% increase in doubling time (8.9 ± 0.3 h) compared with wt EC (7.2 ± 0.1 h) and Renilla KD (7.6 ± 0.3 h) cells, respectively. Next, the differentiation of wt and UTF1 KD EC cells was performed with different cell numbers to rule out potential cell density effects on differentiation (0.5 and 2 times the number of cells: 1.8 × 10 and 7.3 × 10 cells, respectively). Irrespective of the initial number of cells, the differentiation of UTF1 KD cells was always delayed or blocked, whereas wt cells differentiated normally (unpublished data). Summarizing, these data indicate that in EC cells, UTF1 KD results in an abrogated differentiation capacity and persistent Oct4 expression under differentiation-inducing conditions.
To extend these findings to a nontransformed mouse cell line, we tested the effect of UTF1 KD on IB10 ES cell differentiation. Renilla KD clones expressed normal levels of UTF1 and Oct4, whereas in UTF1 KD cell lines, UTF1 levels were reduced, but Oct4 expression was not affected (). In addition, UTF1 and Renilla KD ES cells are positive for AP, confirming their ES cell phenotype (). To determine whether UTF1 down-regulation also affected the differentiation potential of these cells, embryoid bodies (EBs) were generated. Where wt and Renilla KD cells formed normal EBs with high efficiency, UTF1 KD–derived EBs were irregularly shaped, much smaller in size, formed with low efficiency, and compaction was not observed ().
In agreement with observations by , UTF1 KD affected (although less dramatically) the doubling time of ES cells: UTF1 KD ES cells have a doubling time of 11.8 ± 0.7 h compared with 9.6 ± 0.7 h and 10.2 ± 0.1 h for wt (23% increase) and Renilla (16% increase) ES cells, respectively. Because UTF1 KD abrogated the ability of EC cells to differentiate, we tested whether EBs from UTF1 KD ES cells also failed to differentiate. AP staining showed that day 8 wt and Renilla EBs are largely AP negative, whereas UTF1 KD EBs still displayed substantial AP activity, suggesting that UTF1 is involved in ES cell differentiation (). To further validate this observation, the expression pattern of several germ layer–specific marker genes during EB development was determined by RT-PCR (). Both wt and Renilla ES cells show a clear up-regulation of various lineage markers. At days 3–5, Brachyury (early mesoderm) was detected, and at day 10, BMP5 (dorsal mesoderm) was detected. Endoderm markers GATA4 and GATA6 were detected at day 10, and ectoderm markers GAP43 and FGF5 were both detected at days 3–10. In contrast, both UTF1 KD cell lines showed either an absence (GATA6), minor (GATA4 and BMP5), or delayed (Brachyury) expression of these markers. However, ectoderm markers FGF5 and GAP43 were detected from day 3. Pluripotency markers like Oct4, REX1, and Nanog were detected at all time points in the various EBs, most likely as a result of the incomplete differentiation of a subset of cells.
To understand the mechanistic properties of UTF1 that underlie its involvement in EC and ES cell differentiation, a series of experiments were performed to molecularly characterize the protein. First, we determined the subcellular localization of UTF1 in EC cells (). UTF1 was clearly localized to the nucleus and excluded from the nucleoli. In addition, we found UTF1 to localize to the chromosomes at different stages during cell division. To further characterize this potential interaction between UTF1 and the DNA/chromatin, we performed subnuclear fractionations of EC cells separating free-diffusing proteins (cytosolic and nuclear), weak/strong DNA-associated proteins, and nuclear matrix (associated) proteins. UTF1 is observed exclusively in the ammonium sulfate fraction known to contain strongly DNA-associated proteins, like core histone H2A (). In contrast, Oct4 primarily localized to the free-diffusing fraction and, to some extent, to the nuclear matrix fraction, indicating that Oct4 and UTF1 have distinct chromatin-binding characteristics. To compare the observed behavior of UTF1 to that of chromatin-modifying proteins, we determined the fractions containing histone deacetylase 1 (HDAC1; ). Unlike UTF1, HDAC1 is found in the fractions containing free-diffusing and weak DNA-associated proteins. Collectively, these data suggest that UTF1 is a protein with a high affinity for chromatin, similar to that of core histones and different from chromatin-modifying proteins like HDAC1.
As a chromatin-associated protein, UTF1 is likely to be involved in gene expression regulation. To determine the effect of UTF1 on promoter activity, reporter assays were performed using constructs containing multiple copies of either the Smad-binding element (SBE) or BMP-responsive element (BRE). These reporters were used because we previously identified UTF1 as an SBE-interacting protein in a yeast 1 hybrid screen. However, more detailed analysis showed that UTF1 is not specifically involved in Smad signaling. The reporters are activated by the cotransfection of either Smad3 and 4 (SBE) or Smad1 and 4 (BRE; ). Cotransfection of UTF1 reduced the activity of Smad-stimulated SBE and BRE reporters by approximately twofold and fourfold, respectively. These data indicate that UTF1 is a transcriptional repressor.
UTF1 contains two conserved domains (CDs): CD1 (aa 55–124), which shares high homology with Myb/SANT DNA-binding domains, and CD2 (aa 271–334), which contains a putative leucine zipper. To identify its repressor domains, the effect of a series of GAL4-UTF1 (deletion) constructs was tested on a thymidine kinase (TK) luciferase reporter containing five copies of the GAL4 target sequence (UAS-TK-Luc; ). As expected, UTF1 repressed UAS-TK-Luc reporter activity (eightfold reduction compared with GAL4). Deletion of the very C-terminal 39 aa resulted in an almost 2.8-fold reduction in repressor activity. Further C-terminal deletions only marginally affected repressor activity, but when the Myb/SANT domain (aa 55–124) was deleted, an additional drop in repressor activity compared with the 1–167 and 1–134 constructs was observed. The finding that both the C terminus and Myb/SANT domain are involved in transcriptional repression was confirmed using a series of progressive N-terminal deletions. Deletion of aa 1–65 resulted in a 3.1-fold reduction of repressor activity (compare 1–339 with 66–339), and further N-terminal deletions did not affect UTF1 repressor activity except for the deletion construct (297–339) that misses part of the CD2 domain, which completely lacked repressor activity. To address the importance of the Myb/SANT domain, we generated a mutant lacking this region, which reduced repressor activity by 1.7-fold, indicating that it is important for transcriptional repression by UTF1. In addition, the Myb/SANT domain alone (33–134) also displayed considerable repressor activity (3.9-fold repression). Collectively, these experiments indicate that both the Myb/SANT domain and the extreme C terminus of UTF1 are important for transcriptional repression by UTF1.
To study its localization in living cells, UTF1 was fused to enhanced GFP (eGFP), creating eGFP-HA-UTF1 (hereafter GFP-UTF1), and was stably expressed in EC cells. To prevent localization artifacts, we used a clone that underexpressed GFP-UTF1 compared with the endogenous protein (). Subnuclear fractionation showed that GFP-UTF1, like endogenous UTF1, is almost exclusively found in the strongly DNA-associated fraction (). Reporter (UAS-TK-Luc) assays in HepG2 cells showed that GFP-UTF1 acted as a transcriptional repressor as well (see ). These data indicate that fusing GFP to the N terminus of UTF1 does not interfere with the function of the protein.
Confocal microscopy of living cells showed that GFP-UTF1 localized to the nucleus with an inhomogeneous distribution in a similar fashion as the endogenous protein ( and ). GFP-UTF1 is excluded from the nucleoli (, arrows). The punctate localization is more intense around the nucleoli and in the nuclear periphery. Time-lapse imaging of a cell counterstained with Hoechst showed the chromosomal localization of GFP-UTF1 during metaphase, anaphase, and telophase ().
In ES cells, a similar GFP-UTF1 distribution was observed: localized to the nucleus, excluded from the nucleoli (, arrows), and chromosome associated during mitosis (, arrowhead). GFP-UTF1 ES cells were AP positive () and expressed Oct4 (not depicted). Fractionation of GFP-UTF1 ES cells showed that both endogenous UTF1 (αmUTF1) and GFP-UTF1 (αHA) localized to the fraction containing strongly DNA-associated proteins ().
To study the observation that UTF1 is a stably chromatin-associated protein in a more physiological context, we analyzed the dynamic properties of UTF1 in living cells using a FRAP protocol. In EC cells, GFP-UTF1 molecules were bleached in a small strip spanning the nucleus, and subsequent fluorescent recovery in the strip was measured at 20-ms intervals (; ). The mean fluorescence intensity in the strip of several cells was plotted against time relative to the prebleach level. GFP-expressing cells showed a fast recovery of fluorescence in the strip (, green line), indicating a highly mobile protein. Fluorescence in the strip did not recover to prebleach levels as a result of the permanent bleaching of a fraction of the molecules. In contrast to GFP, GFP-UTF1 (, blue line) showed only little recovery after bleaching, indicating that the vast majority is long-term immobilized, at least for the duration of the FRAP experiments. Because mobility measurements of GFP and GFP-UTF1 in ES cells produced identical results (, green line and blue line, respectively), EC cells were used for all subsequent mobility measurements.
In terms of localization ( and ) and subnuclear fractionation behavior (), UTF1 greatly resembles core histones (). To further substantiate this observation, the mobilities of UTF1 and core histone H2B () were compared (). FRAP curves for GFP-UTF1 (, blue line) and H2B-GFP (, red line) were virtually identical, indicating the similar molecular kinetics of these proteins. Computer simulations of the FRAP procedure were used to fit the experimental data, yielding diffusion constants, immobile fractions, and residence times of all proteins tested ( and Fig. S1, available at ). Both the population of GFP-UTF1 and H2B-GFP molecules displayed an immobile fraction of ∼90% ( and Fig. S1, A and B). The duration of immobilization was much longer than the 20-s time scale of our experiments and, therefore, could only be determined with limited accuracy. For both GFP-UTF1 and H2B-GFP, a residence time in the order of minutes to hours was determined, which is in agreement with the findings of .
Subsequently, we compared the dynamic behavior of GFP-UTF1 and Oct4-GFP (). In contrast to GFP-UTF1 (, blue line), Oct4-GFP (, red line) is largely mobile. Note that Oct4-GFP fluorescence recovery is much slower than that of GFP (, green line). Computer simulations indicated that 10% of the Oct4-GFP molecules are immobile with a residence time in the order of 0.1 s ( and Fig. S1 C). In addition, the diffusion rate of Oct4-GFP (3 μm/s) suggested that the protein resides in a high molecular weight complex. The highly dynamic behavior of Oct4-GFP molecules is similar to what is found for several other DNA transacting factors like the transcription/repair factor TFIIH, the homologous recombination protein Rad54, and TFIIB during interphase (; ; ; ; ; ; ; ). These data indicate that the dynamic behavior of UTF1 is similar to that of core histones but not to that of transcription factors like Oct4.
Using GAL4-UTF1 fusions, we identified the putative Myb/SANT domain and the C terminus of UTF1 as repressor domains (). To investigate the requirement of these domains in UTF1 localization and mobility, a series of GFP-UTF1 mutants was generated. First, the repressor activity of wt and mutant GFP-UTF1 proteins was determined in reporter assays. Mutation of aa 63 (W→G) and 67 (E→K; GFP-UTF1 W63G E67K; ), two amino acids highly conserved in Myb/SANT domains, and/or deletion of the C-terminal 39 aa (GFP-UTF1 W63G E67K 1–300; GFP-UTF1 1–300) resulted in a complete loss of UTF1 repressor activity ().
In terms of localization, GFP-UTF1 and GFP-UTF1 W63G E67K display a similar distribution. Deletion of the entire Myb/SANT domain resulted in an almost completely cytoplasmic localized fusion protein (unpublished data). GFP-UTF1 1–300 also interacted with mitotic chromosomes, whereas during interphase, the protein seemed to be more dispersed (). GFP-UTF1 W63G E67K 1–300 showed a completely homogenous nuclear distribution in combination with nucleolar exclusion. Furthermore, association with mitotic chromosomes was never observed (). These data indicate that both the Myb/SANT domain and C terminus of UTF1 are required for proper localization of the protein during interphase as well as mitosis.
To determine the role of the Myb/SANT domain and C terminus in UTF1 mobility, FRAP analyses were performed. GFP-UTF1 W63G E67K–expressing cells showed an increased recovery of fluorescence in the strip (, red line) compared with GFP-UTF1 (, blue line), indicating a reduced binding efficiency. The rate of fluorescence recovery after the initial influx resembled that of GFP-UTF1, implying that the residence time of individual molecules was not affected. Computer simulations showed that the residence time of GFP-UTF1 W63G E67K molecules is similar to GFP-UTF1 molecules (in the order of minutes to hours) but that the mean immobile fraction was smaller (∼60%; and Fig. S1 D).
GFP-UTF1 1–300–expressing cells showed a complete recovery after bleaching (, red line), demonstrating that the C terminus is required for the long-term immobilization of GFP-UTF1. However, the initial fluorescence recovery in the strip was substantially slower than that of GFP (), suggesting that UTF1 resides in a high molecular weight complex and/or is still capable of transiently interacting with sites of affinity. Simulations showed that 85% of the GFP-UTF1 1–300 molecules are immobilized with a residence time in the order of 0.25 s ( and Fig. S1 E). These data indicate that the C terminus of UTF1 is required for the long-term stabilization of interactions with sites of affinity, most likely chromatin.
Remarkably, GFP-UTF W63G E67K 1–300 showed a much faster recovery of fluorescence than GFP-UTF1 1–300 and an only slightly slower recovery than GFP (), indicating that this mutant is freely mobile. This was further supported by computer simulations that predicted that 25% of the GFP-UTF1 W63G E67K 1–300 molecules was immobilized with a short residence time of 0.25 s ( and Fig. S1 F). The UTF1 mutants lacking their C-terminal 39 aa (GFP-UTF1 1–300 and GFP-UTF1 W63G E67K 1–300) displayed a marked increase in their diffusion constants compared with GFP-UTF1 and GFP-UTF1 W63G E67K (14 vs. 0.6 μm/s; ). However, because the model used for fitting the data only included one pair of binding constants (immobile fraction and residence time) and the stable binding of GFP-UTF1 and GFP-UTF1 W63G E67K is dominant in the FRAP curve, the observed low mobility of 0.6 μm/s is most likely the result of additional transient interactions similar to those of the C-terminal mutants.
To investigate whether the differential mobilities of the mutant proteins are reflected by altered distribution over subnuclear fractions, cell lines stably expressing mutant GFP-UTF1 proteins were analyzed (). As shown before (), GFP-UTF1 localized to the strongly DNA-associated protein fraction. The majority of the GFP-UTF1 W63G E67K proteins was also strongly DNA associated but was detected in the free-diffusing protein fraction as well, indicating the presence of an increased portion of mobile molecules, which is in agreement with the FRAP data. Both GFP-UTF1 1–300 and GFP-UTF1 W63G E67K 1–300 were found in the free-diffusing protein fraction, indicating that both mutants are fully mobile. Note that GFP-UTF1 1–300 was still capable of binding to mitotic chromosomes and had a punctate nuclear localization, suggesting that this protein is capable of transient interactions with sites of affinity. Throughout these experiments, endogenous UTF1 was always detected in the fraction containing strong DNA-associated proteins (unpublished data).
These results show that UTF1 is required for the proper differentiation of EC and ES cells. KD of UTF1 expression in EC and ES cells resulted in blocked or delayed differentiation but did not affect the self-renewal capacity of these cells, indicating that UTF1 is not required for ES cell self-renewal. In addition, reporter assays, subnuclear fractionations, and FRAP analyses showed that UTF1 is a stably chromatin-associated transcriptional repressor with histone-like properties like long-term DNA association and a majority of immobilized molecules. The UTF1–chromatin interaction is dependent on two separate interaction domains: the Myb/SANT domain and the extreme C terminus. The concerted action of both interaction domains causes ∼90% of the molecules to bind to sites of affinity for times similar to those of H2B (; ). Summarizing, these data indicate that UTF1 is strongly associated with chromatin in EC and ES cells and most likely also during the early stages of embryogenesis. UTF1 may establish a chromatin state that renders an ES cell susceptible to the activation of differentiation programs in response to appropriate stimuli. Despite being prone to differentiation, ES cells are kept in a self-renewing state by the combined action of self-renewal regulators like Nanog, Oct4, Sox2, and Sall4 and the recently identified Esrrb, Tbx3, and Tcl1 proteins that interfere with differentiation to epiblast-derived lineages (; ; ; ; ; ). Although the expression patterns of and these genes are identical (; ), the function of UTF1 seems opposite; where UTF1 is not required for ES cell self-renewal, it is involved in ES cell differentiation. The fact that UTF1 expression is down-regulated during ES cell differentiation is probably a consequence of the inactivation of the Oct4 gene, a transcriptional activator of the gene ().
Molecularly, UTF1 may be necessary for signaling to these self-renewal factors to allow differentiation to commence, explaining why UTF1 is dispensable for ES cell self-renewal and that its expression is down-regulated upon the initiation of differentiation. More likely, in view of the histone-like properties of UTF1, its function could be the maintenance of a specific epigenetic profile required for differentiation either by attracting chromatin-modifying proteins or by chromatin compaction. This hypothesis is supported by the observation that UTF1 has transcriptional repressor activity, an observation also made by , who showed that UTF1 in the absence of ATF2 could repress the activity of various reporter genes.
Recent reports have emphasized the role of the epigenetic regulation of gene expression in ES cell self-renewal and differentiation (; ; Boyer et al., 2006; ; ). The PcG proteins were found to silence a large set of developmental differentiation genes in ES cells. Many of these PcG targets in ES cells are cooccupied by Oct4, Sox2, and Nanog (; ), suggesting that to maintain the self- renewing state of ES cells, stem cell self-renewal regulators may directly regulate the targeting and/or activity of chromatin remodeling complexes.
The phenotype of UTF1 KD ES cells is similar to that of ES cells lacking the PcG protein Suz12 (). show that Suz12 ES cells fail to differentiate, underlining the important role of PcG-mediated silencing in lineage specification. Similarly, UTF1 may either directly or indirectly influence the epigenetic state of ES cells, thereby allowing the initiation of lineage-specific differentiation. Down-regulation of UTF1 may allow ES cells to establish a new, more somatic type of chromatin, which is analogous to observations made by , who showed that architectural chromatin proteins bind loosely to chromatin in ES cells and become immobilized upon differentiation.
In this study, we show for the first time that the ES cell protein UTF1 is a stably chromatin-associated protein that is involved in initiation of the differentiation program of ES cells. We propose that with UTF1, we have identified a principal component of the complex regulatory gene network underlying initiation of the lineage-committed differentiation of ES cells.
The mUTF1 cDNA was provided by H. Stunnenberg (Nijmegen Centre for Molecular Life Sciences, Nijmegen, Netherlands). BamHI (5′) and EcoRI (3′) sites were added by PCR, and this fragment was BamHI–EcoRI ligated into pcDNA3-HA, resulting in pcDNA3-HA-mUTF1. For PCR, the primers mUTF1 forward (5′-ATATGATATCGGATCCATGCTGCTTCGTCCCCGGAG-3′) and mUTF1 reverse (5′-ATATGAATTCTTATTGGCGCAAGTCCCCAAG-3′) were used.
P19CL6 EC cells () were grown in α-MEM (Invitrogen) supplemented with antibiotics and 10% FBS (Hyclone) at 37°C and 5% CO. HepG2 cells were maintained in DME with antibiotics and 10% FBS. For differentiation of P19CL6 cells, 365,000 cells were seeded in 6-cm ø plates in culture medium supplemented with 1% DMSO (Sigma-Aldrich). Embryonic day 14 ES cells (subclone IB10) were grown on gelatin-coated dishes in buffalo rat liver cell–conditioned medium supplemented with 1,000 U/ml leukemia inhibitory factor (Chemicon), nonessential amino acids, and 0.1 mM 2-mercaptoethanol. For confocal laser-scanning imaging, IB10 cells were seeded on a layer of STO feeder cells on gelatinized glass coverslips. To generate stably transfected cell lines, 10 cells were electroporated with 13.5 μg plasmid DNA and 1.5 μg pPGK-Hyg plasmid. Selection was performed using 200 μg/ml hygromycin, clones were picked, and cell lysates were analyzed. For analyses of AP activity, an AP detection kit (Chemicon) was used. To generate stably transfected P19CL6 cell lines, cells were transfected with FuGENE 6 (Roche) and selected with 600 μl/ml G418 or 600 μg/ml hygromycin, and clones were picked. For DNA staining, cells were cultured for 2 h in the presence of 10 μg/ml Hoechst 33258. For transient transfections, 250,000 HepG2 cells were seeded per 3.5-cm ø well. Transfections were performed using calcium phosphate coprecipitation. After 48 h, cells were harvested (reporter lysis buffer; Promega), and luciferase activity was measured (LucLite; Packard). To normalize luciferase activities, a β-galactosidase expression plasmid (pDM2LacZ) was cotransfected. β-Galactosidase activity was determined in 100 mM NaHPO/NaHPO, 1 mM MgCl, 100 mM 2-mercaptoethanol, and 0.67 mg/ml -nitrophenylgalactopyranoside.
For EB formation, ES cells were suspended from the lids of 10-cm ø dishes in 20-μl drops (5 × 10 cells/ml). After 48 h, EBs were transferred to bacterial grade Petri dishes. On day 7, EBs were transferred to gelatinized 3.5-cm ø six-well plates. On days 3, 5, and 7, pictures were taken, and total RNA was isolated on days 3, 5, and 10.
Total RNA was extracted with TRIzol (Invitrogen), treated with DNaseI (Fermentas), and reverse transcribed (RevertAid M-MuLV Reverse Transcriptase; Fermentas). Details of primer sets, cycle numbers, and annealing temperatures used in subsequent PCR reactions can be found in Table S1 (available at ). PCR products were analyzed on 2% agarose gels.
Cells were washed with cold PBS and incubated in lysis buffer (400 mM NaCl, 20 mM Tris-HCl, pH 7.8, 1% NP-40, 0.5% sodium deoxycholate, 2 mM EDTA, 2 mM DTT, and protease inhibitors) for 30 min on ice. Next, cell lysates were collected by scraping, subsequently sonicated, and cleared by centrifugation at 4°C and 14,000 rpm for 10 min. For western analysis, the following primary antibodies were used: mUTF1 (rabbit polyclonal raised by Eurogentec), Oct-4 (H-134; Santa Cruz Biotechnology, Inc.), HDAC1 (H-51; Santa Cruz Biotechnology, Inc.), histone H2A (acidic patch; Upstate Biotechnology), GATA4 (C20; Santa Cruz Biotechnology, Inc.), actin (C4; MP Biomedicals), and HA (3F10; Roche). Secondary immunodetection was performed using donkey anti–rabbit IgG-HRP (GE Healthcare), rabbit anti–rat IgG-HRP (DakoCytomation), goat anti–mouse IgG-HRP (Santa Cruz Biotechnology, Inc.), and donkey anti–goat IgG-HRP (Santa Cruz Biotechnology, Inc.). Subnuclear fractionation was performed as previously described ().
For immunofluorescence analysis, P19CL6 cells were cultured on poly-- lysine–coated glass coverslips and fixed in 2% PFA in PBS for 10 min at RT. After fixing, cells were permeabilized with 0.1% Triton X-100 in PBS. Endogenous UTF1 was detected using our UTF1 antibody followed by a goat anti–rabbit tetramethylrhodamine IgG (H+L)-conjugated secondary antibody (Invitrogen). Fluorescent images were made using a microscope (Axiophot; Carl Zeiss MicroImaging, Inc.) with a plan-NEOFLUAR 40× NA 0.70 lens. Confocal laser-scanning microscopy images of live cells were recorded with a microscope (LSM 510; Carl Zeiss MicroImaging, Inc.). GFP signal was detected using a 488-nm argon laser line and a bandpass 500–550-nm filter. Hoechst signal was monitored by excitation with a Titanium Sapphire 810-nm dual-photon laser and a bandpass 390–465-nm filter.
For FRAP experiments, a confocal laser-scanning microscope (LSM 510; Carl Zeiss MicroImaging, Inc.) was used. To measure FRAP, a 10-μm-wide strip spanning the nucleus was bleached for 120 ms at the highest intensity of the 488-nm line of a 30-mW argon laser focused by a plan Apochromat 63× NA 1.4 oil differential interference contrast lens (Carl Zeiss MicroImaging, Inc.). Recovery of fluorescence in the strip was monitored at 20-ms intervals at 0.5% of the laser intensity used for bleaching. For emission detection, a bandpass 500–550-nm filter was used.
For analysis of FRAP data, FRAP curves were normalized to prebleach values, and the best fitting curve (least squares) was picked from a large set of computer-simulated FRAP curves in which three parameters representing mobility properties were varied: diffusion rate (ranging from 0.04 to 25 μm/s), immobile fraction (0, 10, 20, 30, 40, and 50%), and time spent in the immobile state (2, 4, 8, 16, 32, 64, 128, and ∞ s). Monte Carlo computer simulations used to generate FRAP curves were based on a model of random diffusion in an ellipsoid volume representing the cell nucleus and simple binding kinetics representing binding to immobile elements in the cell nucleus. Simulations were performed at unit time steps corresponding to the experimental sample rate of 21 ms.
Diffusion was simulated by each step, deriving novel positions M(x + dx, y + dy, and z + dz) for all mobile molecules M(x, y, and z), where dx = G(r), dy = G(r), dz = G(r), r is a random number (0 ≤ r ≤ 1) chosen from a uniform distribution, and G(r) is an inversed cumulative Gaussian distribution with μ = 0 and σ = 6Dt, where D is the diffusion coefficient and t is time measured in unit time steps.
Immobilization was based on simple binding kinetics described by k/k = F/(1 − F), where F is the relative number of immobile molecules. The chance for each particle to become immobilized (representing chromatin binding) was defined as P = k = k F/(1 − F), where k = 1/t and t is the mean time spent in immobile complexes measured in unit time steps; the chance to release was P = k = 1/t. In simulations of two immobile fractions with different kinetics, two immobilization/mobilization chances were evaluated for each unit time step.
For each set of parameters, a FRAP curve was generated based on 10 molecules per nucleus (which yields similar results as averaging 10 cells containing 10 molecules or 100 cells containing 10 molecules). This number was empirically determined to produce a curve with a limited fluctuation of fluorescence (as a result of diffusion) after complete recovery.
Table S1 provides the primer sequences used for the PCR reactions displayed in G and their product sizes, annealing temperatures, and number of PCR cycles. Fig. S1 shows experimental FRAP curves and computer-simulated curves for the constructs GFP-UTF1, H2B-GFP, Oct4-GFP, GFP-UTF1 W63D E67K, GFP-UTF1 1–300, and GFP-UTF1 1–300 W63D E67K. Online supplemental material is available at . |
Telomeres are nucleoprotein structures that protect the ends of chromosomes (). A proper protective function of telomeres is accomplished by maintaining a minimum length of TTAGGG repeats as well as by the association of telomere-binding factors (for review see ). Maintenance of telomere length homeostasis is primarily achieved by telomerase, a reverse transcriptase that adds telomeric repeats de novo after each cell duplication cycle, counteracting the end replication problem (). Alternative ways to maintain telomere length have also been described, such as the so-called alternative lengthening of telomeres (ALT) mechanism, which is considered to rely on recombination among telomeric sequences ().
Recent evidence indicates that epigenetic modifications of the chromatin are also important to maintain telomere length homeostasis (). In particular, alterations in histone methylation or DNA methylation leading to the loss of heterochromatic features at telomeric and subtelomeric chromatin have been shown to result in telomere length deregulation (, ). Two of the main histone marks at compacted heterochromatin domains are the trimethylation of H3K9 and H4K20 as well as binding of the heterochromatin protein 1 isoforms (; ; ). Cells lacking the Suv39h1 and Suv39h2 histone methyltransferases (HMTases) show decreased H3K9 trimethylation at telomeres concomitant with aberrant telomere elongation (). Similarly, mouse embryonic stem (ES) cells deficient for the DNA methyltransferases Dnmt1 or Dnmt3a,3b present a marked reduction in DNA methylation at subtelomeric domains, which is accompanied by dramatically elongated telomeres (). Furthermore, mouse embryonic fibroblasts (MEFs) lacking all three members of the Retinoblastoma family of tumor suppressors (Rb, p107, and p130) show decreased levels of H4K20 trimethylation at telomeres as well as a global reduction in DNA methylation, which again are concomitant with aberrant telomere elongation (; ). These findings lead to the notion that a compacted or heterochromatic state at telomeres is required for a proper telomere length control (, ). However, more studies are needed to identify the different activities that participate in the assembly and regulation of telomeric chromatin as well as the mechanisms by which epigenetic alterations lead to telomere length deregulation.
In particular, the recently discovered Suv4-20h1 and Suv4-20h2 HMTases are prime candidates to carry the trimethylation of H4K20 at telomeres, as suggested by their ability to trimethylate H4K20 as well as to interact with heterochromatin protein 1 (; ). In addition, the fact that all Rb family members can bind in vitro to Suv4-20h HMTases () also suggests that these enzymes are responsible for H4K20 trimethylation both at pericentric and telomeric heterochromatin domains. In fact, evidence of their involvement on pericentric heterochromatin assembly has been already provided (; ). However, direct evidence for a role of these enzymes on telomere chromatin architecture and regulation has been unknown to date.
With the goal of unraveling the enzymatic activities that participate in the assembly of telomeric chromatin structure and their mechanism of action, we assessed whether Suv4-20h1 and Suv4-20h2 HMTases are directly responsible for trimethylating H4K20 at telomeres and subtelomeres. For this, we used MEFs and ES cells deficient in either Suv4-20h1 (Suv4-20h1 MEF) or Suv4-20h2 (Suv4-20h2 MEF) or simultaneously deficient for both enzymes (Suv4-20h double-null [dn] MEFs and ES cells; see Cell culture section in Materials and methods). The results presented here show that abrogation of these enzymes results in decreased H4K20me3 at telomeric and subtelomeric domains concomitant with an increase in telomere length and in sister chromatid recombination both globally (sister chromatid exchanges [SCEs]) as well as at telomeric regions (telomere SCEs [T-SCEs]). In contrast, other heterochromatic marks at telomeres such as H3K9me3 and heterochromatin protein 1 binding were unaltered by the loss of these enzymes. In summary, we demonstrate that Suv4-20h HMTases actively participate in the correct assembly of telomeric chromatin and that their abrogation impacts telomere length homeostasis as well as telomere integrity.
The two main histone marks characteristic of telomeric and subtelomeric heterochromatin are the trimethylation of H3K9 and the trimethylation of H4K20 (, ; , ). The mammalian enzymes responsible for the trimethylation of H3K9 at telomeres are the Suv39h1 and Suv39h2 HMTases (). However, the enzymes responsible for the trimethylation of H4K20 at telomeres have not been identified yet. To test whether the Suv4-20h1 and Suv4-20h2 HMTases are responsible for the trimethylation of H4K20 at telomeres, we studied telomeric chromatin in MEFs deficient in Suv4-20h1 or Suv4-20h2 or simultaneously deficient for both enzymes (Suv4-20h dn cells).
First, we measured the length of TTAGGG repeats using Southern blot terminal restriction fragment (TRF) analysis in Suv4-20h1, Suv4-20h2, and Suv4-20h dn MEFs derived from two different embryos of each genotype (see TRF analysis section in Materials and methods). As shown in , subtelomeric digestion with MboI released TRF fragments of higher molecular weight in Suv4-20h2 and Suv4-20h dn cells than in wild-type cells, indicating telomeres of greater length. In contrast, Suv4-20h1 cells did not show increased telomere length compared with the wild-type controls (). In agreement with the fact that these TRF fragments corresponded to telomeric sequences, they were digested with the Bal31 exonuclease (), which degrades DNA from the chromosome ends. We also confirmed elongated telomeres in Suv4-20h dn ES cells compared with the corresponding wild-type controls (). Of note, telomeres were slightly longer in ES cells compared with MEFs of the same genotype (), which is in agreement with the fact that they have higher telomerase activity and correspond to a more primitive developmental stage.
The long-telomere phenotype associated with Suv4-20h deficiency was confirmed by quantitative FISH (Q-FISH) of metaphase nuclei using a telomere-specific peptide nucleic acid (PNA) probe, which specifically measures the length of TTAGGG repeats at all individual chromosome ends (see Q-FISH analysis section in Materials and methods). Again, Suv4-20h2 and Suv4-20h dn MEFs presented significantly elongated telomeres of 36.5 ± 17.0 kb and 37.4 ± 17.7 kb, respectively, compared with 33.6 ± 17.2 kb in the wild-type controls (P < 0.001; Wilcoxon rank-sum test; ). This was in contrast to Suv4-20h1 MEFs that showed similar length or slightly shorter telomeres compared with the wild-type controls (Wilcoxon rank-sum test; P < 0.008; ). In agreement with elongated telomeres in Suv4-20h2 and Suv4-20h dn MEFs, these cells also showed a statistically significant higher percentage of telomeres longer than 50 kb and a lower percentage of telomeres shorter than 20 kb compared with the wild-type controls (χ test; P < 0.001 for all comparisons with wild-type MEFs; ).
Interestingly, telomere length deregulation in Suv4-20h2 and Suv4-20h dn MEFs did not result in increased chromosomal aberrations involving telomeric repeats as detected by Q-FISH (see for representative examples of chromosomal aberrations and for quantification), suggesting that these abnormally elongated telomeres retain a normal ability to protect chromosome ends from end to end fusion events. Altogether, these results indicate that simultaneous loss of the HMTase Suv4-20h1 and Suv4-20h2 activities results in a considerable telomere elongation in the absence of the loss of telomere capping.
To test whether Suv4-20h HMTases are responsible for H4K20me3 at telomeres, we performed chromatin immunoprecipitation (ChIP) with antibodies recognizing this modification in cells that lack either Suv4-20h1, Suv4-20h2, or both h1 and h2 HMTases. Immunoprecipitated DNA was detected by Southern blotting with a telomeric probe (see ChIP section in Materials and methods). As shown in , the loss of Suv4-20h1 HMTase does not lead to decreased H4K20me3 at telomeres. In contrast, Suv4-20h2 cells and, to a larger extent, Suv4-20h dn cells showed a marked decrease in H4K20me3 at telomeric chromatin compared with wild-type controls ( test; P < 0.005 for both comparisons; ). These results were also confirmed using ES cells doubly deficient for Suv4-20h1 and h2 enzymes (Suv4-20h dn ES cells; ). As a control, we also observed the loss of H4K20me3 at pericentric heterochromatin upon the abrogation of Suv4-20h2 or both Suv4-20h1 and h2 HMTases compared with wild-type controls ( test; P < 0.05 for both comparisons; ; ). Trimethylation of H3K9 was not substantially altered at telomeric and pericentric heterochromatin by the loss of Suv4-20h HMTases (), which is in agreement with a previous study showing a normal trimethylation of H3K9 in the presence of H4K20 trimethylation defects (). Indeed, it has been previously proposed that the assembly of telomeric and pericentric heterochromatin occurs sequentially and that trimethylation of H4K20 requires previous H3K9me3 by the Suv39h HMTases and not vice versa (; ).
We next evaluated whether Suv4-20h deficiency affected the binding density of telomere repeat–binding factors TRF1 and TRF2 to telomeres, which were previously shown to have an important role in telomere length regulation (for review see ). ChIP did not reveal any significant differences in the binding density of TRF1 or TRF2 to telomeric chromatin in the different Suv4-20 genotypes when using either MEF or ES cells (P ≥ 0.04 for all comparisons; test; ). These results suggest that Suv4-20h deficiency does not dramatically alter the binding density of TRF1 and TRF2 to telomeric chromatin, although we cannot rule out that altered expression of other telomere-binding factors (for review see ) as the consequence of Suv4-20h ablation could contribute to the long-telomere phenotype of these cells.
We next evaluated whether the adjacent subtelomeric regions, which were recently characterized as chromatin domains enriched in H4K20me3 and H3K9me3 heterochromatic marks (), were affected by abrogation of the Suv4-20h HMTases. To this end, we performed ChIP assays followed by real-time PCR to detect these modifications at two independent subtelomeric regions in chromosomes 1 and 2 (; see Real-time quantitative PCR section in Materials and methods; ). After normalization of the ChIP signals to the different inputs, we found no decrease in the levels of H3K9 trimethylation at these subtelomeric regions in the different Suv4-20h–deficient MEFs; instead, this modification was slightly increased in Suv4-20h dn MEFs compared with wild-type controls (P < 0.05 for both subtelomeric regions; ). In contrast, H4K20 trimethylation was significantly decreased at chromosome 1 and 2 subtelomeric regions in Suv4-20h2–deficient cells ( test; P ≤ 0.05; ), and this reduction was further aggravated in cells deficient for both Suv4-20h1 and h2 HMTases ( test; P < 0.01; ). Similar to that previously shown for telomeric chromatin, Suv4-20h1 cells did not show decreased H4K20me3 at subtelomeric chromatin domains (), indicating that the Suv4-20h2 HMTase is the one responsible for this histone modification at both telomeric and subtelomeric chromatin. As a control, we confirmed the lack of these epigenetic modifications in a region of the p16 promoter known to be transcriptionally active and devoid of these heterochromatic marks (; unpublished data). Finally, these results were confirmed using Suv4-20h dn ES cells, which also showed a significant decrease (P < 0.01) in H4K20 trimethylation at both chromosome 1 and 2 subtelomeric regions in the absence of changes in H3K9 trimethylation (Fig. S1, available at ).
We have recently found that subtelomeric regions are also enriched in methylated CpG residues, which are maintained by the Dnmt1 and Dnmt3a,3b DNA methyltransferases (), further supporting the idea that subtelomeric regions are heterochromatic. To asses whether the loss of H4K20 trimethylation at subtelomeres in Suv4-20h–deficient cells resulted in decreased DNA methylation at these regions, we determined the percentage of CpG residues that are methylated at the subtelomeric regions in the chromosomes 1 and 2 described above in this same section () and that were previously shown by us to be heavily methylated (). To this end, we used bisulfite sequencing of CpG residues (see Analysis of genomic subtelomeric DNA methylation section in Materials and methods). As expected, we found that both chromosome 1 and 2 subtelomeric regions were heavily methylated in wild-type MEFs (). However, abrogation of either the Suv4-20h1 or h2 enzyme did not result in the significantly decreased DNA methylation of these regions () except for a slight decrease in the frequency of methylated CpG residues in Suv4-20h dn MEFs at the subtelomeric region of chromosome 2 (χ test; P < 0.005; ), which was not observed at the subtelomeric region of chromosome 1 (χ test; P = 0.09; ). To confirm these results, we also studied subtelomeric DNA methylation in ES cells deficient for the Suv39h1 and h2 HMTases (Suv39h dn cells), which were previously shown to be upstream of the Suv420h HMTases and, therefore, are required for H4K20 trimethylation at heterochromatic domains (). As shown in Fig. S2 (available at ), we observed a normal or even increased DNA methylation at chromosome 1 and 2 subtelomeric regions in Suv39h dn ES cells compared with the wild-type controls. Altogether, these results demonstrate that Suv4-20h2 is largely responsible for the trimethylation of H4K20 at both telomeric and subtelomeric heterochromatin domains but that this modification does not seem to impinge on subtelomeric DNA methylation.
We have previously shown that loss of DNA methylation results in increased recombination between telomeric sequences, as indicated by an elevated number of T-SCEs in cells lacking the Dnmt1 and Dnmt3a,3b DNA methyltransferase activities (). This increased recombination among telomeric sequences has been described to be correlative with activation of the ALT pathway for telomere length maintenance (; ), although evidence for a causal relationship between both processes is still pending.
With this idea in mind, we hypothesized that other epigenetic alterations that may affect the structure of the telomeric or subtelomeric chromatin, such as defects in histone methylation, could also have an impact on the rate of recombination among telomeric sequences. To test this hypothesis, we determined the frequency of T-SCE events in cells that lack the Suv39h and Suv4-20h HMTase activities, which have been shown to have defects in telomere chromatin assembly and telomere length maintenance (; this study). To this end, we used the two-color chromosome orientation FISH or chromosome orientation FISH technique, which allows us to detect SCE events specifically at telomere repeats (see Chromosome orientation FISH section in Materials and methods; ; ; ). Notably, we considered it a positive T-SCE event only when it was simultaneously detected at both the leading and lagging strand telomere as an unequal exchange of telomeric signal (see representative examples in ; yellow arrows). We first analyzed MEFs lacking the Suv4-20h1 and h2 HMTases (Suv4-20h1, Suv4-20h2, and Suv4-20h dn MEFs). Suv4-20h2 and Suv4-20h dn MEFs showed a significant increase in the frequency of T-SCE events compared with the wild-type controls (χ test; P < 0.05 for both comparisons; ). In contrast, Suv4-20h1 MEFs showed a slightly decreased frequency of T-SCE events than wild-type cells, although this difference did not reach statistical significance (χ test; P = 0.3271; ). Notably, both Suv4-20h2 and Suv4-20h dn MEFs showed an increased telomere length compared with wild-type controls, whereas Suv4-20h1 showed slightly shorter telomeres (), suggesting a correlation between telomere elongation and an increase in T-SCE in these genotypes.
Next, we evaluated whether ES cells lacking either the Suv4-20h or Suv39h HMTases also presented deregulated telomeric recombination. We detected a significant increase in the frequency of T-SCE events per chromosome in both Suv4-20h dn and Suv39h dn ES cells (χ test; P < 0.0001 for both cases; ), although the phenotype was more severe in the Suv39h dn ES cells. Notably, the frequencies of T-SCE in ES cells were higher than in MEFs (compare T-SCE frequencies in wild-type and Suv4-20h dn MEFs with ES cells in ), which is in agreement with a previous study's finding that telomeres in ES cells are more recombinogenic than those of other cell types (). Importantly, these results suggest that not only DNA methyltransferases but also other chromatin-modifying activities such as the Suv4-20h and Suv39h dn HMTases are necessary to prevent sister chromatid recombination events between the highly repetitive telomeric sequences. Indeed, Suv39h dn ES cells showed similarly elevated T-SCE frequencies to those previously described for Dnmt-deficient ES cells (). To address whether Suv4-20h and Suv39h deficiencies also resulted in increased sister chromatid recombination throughout the genome, we determined the frequency of global SCE events in both MEFs and ES cells dn for either the Suv4-20h (Suv4-20h dn) or Suv39h enzymes (Suv39h dn). As shown in Fig. S3 (A–C; available at ), both MEF and ES cells deficient for either the Suv4-20h or Suv39h enzymes showed a significantly elevated frequency of SCE compared with the corresponding wild-type controls (χ test; P < 0.005 for all cases), suggesting that the global loss of H3K9me3 and H4K20me3 results in increased frequencies of recombination between sister chromatids.
To further address whether Suv4-20h dn and Suv39h dn ES cells showed an increased activation of ALT pathways, we determined the presence of the so-called ALT-associated promyelocytic leukaemia (PML) bodies or ALT-associated PML bodies (APBs), which, together with increased T-SCE, also correlate with activation of the ALT pathway (; ). This is characterized by the colocalization of telomeres and the PML protein (; arrowheads). Suv4-20h dn ES cells showed a significant increase in the frequency of cells showing APBs, which represented a 1.4-fold increase compared with wild-type controls (χ test; P < 0.0014; ). Suv39h dn ES cells showed a further increase in the frequency of cells showing APBs, which represented a 2.3-fold increase compared with wild-type controls (χ test; P < 0.0001; ). These findings agree with the increased T-SCE frequencies of both Suv4-20h dn and Suv39h dn ES cells and suggest the activation of ALT pathways associated with defective histone methylation at telomeres and subtelomeres.
Recent studies have stressed the importance of the acquisition of a specific chromatin structure at telomeres to maintain telomere length homeostasis in mammals (; for review see ). In addition, the architecture of the adjacent subtelomeric chromatin has also been proven to be critical for telomere length control (). In this study, we have identified the Suv4-20h HMTases as active regulators of the assembly of telomeric and subtelomeric chromatin by undertaking the trimethylation of H4K20 at these domains. Importantly, the loss of Suv4-20h2 activity or simultaneous abrogation of Suv4-20h1/h2 activities and, consequently, of the H4K20me3 mark lead to telomere length deregulation. These results support the notion that epigenetic alterations consisting of the loss of repressive chromatin-modifying activities lead to telomere elongation. It is interesting to note that Suv4-20h1 deficiency did not result in decreased H4K20 trimethylation at either telomeric or pericentric chromatin, suggesting that Suv4-20h2 is the main enzyme responsible for this histone modification at heterochromatic domains. Similarly, Suv4-20h1 deficiency did not result in increased telomere length or increased T-SCE frequencies, suggesting a direct correlation between the loss of H4K20 trimethylation at telomeres and the occurrence of these telomere phenotypes. It is also important to highlight that Suv4-20h deficiency is likely to have global effects in sister chromatid recombination that are not just restricted to the telomeric regions, as indicated by elevated SCE events in Suv4-20d dn MEF and ES cells compared with the wild-type controls.
Altogether, these results suggest a model in which the opening of telomeric chromatin by the loss of repressive chromatin marks could facilitate the accessibility of telomerase or other telomere-elongating activities to the telomere, leading to telomere elongation. Two main mechanisms have been described for the maintenance of mammalian telomeres: the addition of telomeric repeats by telomerase and an alternative mechanism (ALT) that relies on the recombination between telomeric sequences to maintain telomere length. So far, the accessibility of telomerase to telomeres in epigenetically altered mouse cells has not been evaluated. Regarding the derepression of ALT, we have previously demonstrated that DNA hypomethylation results in a marked increase in recombination among telomeric sequences concomitant with aberrant telomere elongation (). In the present study, we demonstrate that abrogation of either Suv39h, Suv4-20h2, or both Suv4-20h1/h2 HMTase activities also leads to elongated telomeres and increased recombination between telomeric repeats in the absence of major DNA methylation defects at subtelomeric regions. This is in contrast to previous results showing that cells lacking Suv39h HMTases also present defects in DNA methylation at pericentric repeats (). Therefore, our results suggest that the increased recombination among telomeric sequences described here for both Suv4-20h and Suv39h-deficient cells is associated with the loss of H4K20me3 and H3K9me3 histone marks and is independent of subtelomeric DNA methylation (see model in Fig. S4 and see Table S1 for a summary of the data; available at ). In agreement with this, Suv39h dn cells, which have decreased levels of both H4K20me3 and H3K9me3 marks, show more severe phenotypes in T-SCE and APB frequencies than Suv4-20h dn cells.
Finally, the results described here support the previously proposed sequential model of chromatin assembly at mouse pericentric heterochromatin () in which the Suv4-20h HMTases act downstream of the Suv39h HMTases (see model in Fig. S4). Altogether, our results demonstrate that the proper action of HMTases inhibits random recombination events and also ensures telomere length homeostasis. These findings are potentially important for cancer research because studies have shown that tumor cells undergo a series of epigenetic modifications, such as global DNA hypomethylation, and a global decrease in H4K20 trimethylation and H4K16 acetylation (; ). One can envision that in tumor cells that present these epigenetic defects, telomere length maintenance may be favored, with the consequent growth advantage that this will provide to the tumor cells.
MEFs were prepared from wild-type, Suv4-20h1, Suv4-20h2, and Suv4-20h1/h2 dn (Suv4-20h dn) embryos. Wild-type and Suv4-20h1/h2 dn ES cells were also generated. MEF cells and ES cells from wild type and Suv39h dn were previously described ().
Cells were included in agarose plugs, and TRF analysis was performed as previously described ().
Before digestion with MboI (), agarose plugs were digested with 3 U Bal31 exonuclease (New England Biolabs, Inc.) at 30°C during 90 min after 1-h preincubation in Bal31 buffer.
We prepared metaphases and performed Q-FISH hybridization as previously described (; ). To correct for lamp intensity and alignment, we analyzed images from fluorescent beads (Invitrogen) using the TFL-Telo program (gift from P. Lansdorp, Terry Fox Laboratory, British Columbia Cancer Research Centre, Vancouver, Canada). Telomere fluorescence values were extrapolated from the telomere fluorescence of lymphoma cell lines LY-R (R cells) and LY-S (S cells) with known telomere lengths of 80 kb and 10 kb, respectively. There was a linear correlation (2 = 0.999) between the fluorescence intensity of the R and S telomeres. We recorded the images using a CCD camera (FK7512; COHU) on a fluorescence microscope (DMRB; Leica). A mercury vapor lamp (CS 100 W-2; Philips) was used as a source. We captured the images using Q-FISH software (Leica) in a linear acquisition mode to prevent the oversaturation of fluorescence intensity. We used the TFL-Telo software () to quantify the fluorescence intensity of telomeres from at least five metaphases for each data point.
ChIP assays were performed as previously described (). In brief, after cross-linking and sonication, 3 × 10 cells (MEFs or ES cells) were used per immunoprecipitation with anti-H3K9me3 or anti-H4K20me3 antibodies (Upstate Biotechnology) or anti-TRF1 raised in our laboratory against full-length mouse TRF1 and anti-TRF2 (provided by E. Gilson, Ecole Normale Superieure de Lyon, Lyon, France). The immunoprecipitated DNA was transferred to a nitrocellulose membrane using a dot blot apparatus. The membrane was then hybridized with either a telomeric probe or a probe recognizing major satellite sequences, which is characteristic of pericentric heterochromatin. Quantification of the signal was performed with ImageQuant software (Molecular Dynamics). The amount of telomeric and pericentric DNA after ChIP was normalized to the total input signal for each genotype.
The presence of subtelomeric sequences of chromosomes 1 and 2 in ChIP samples was detected using primers Chr1-bis-end-5′ (TTAGGACTTCTGGCTTCGGTAG) and Chr1-bis-end-3′ (AGCTGTGGCAGGCATCGTGGC) for chromosome 1 and Chr2-tris-end-5′ (GAATCCTCCCTGTAGCAGGG) and Chr2-tris-end-3′ (GTACATAACCGATCCAGGTGTG) for chromosome 2. We performed real-time quantitative PCR and calculated the ΔΔC value, which represents the cycle threshold difference between the subtelomeric primer pair and the input primer pair. In all cases, PCR was repeated at least five times. ChIP values were represented relative to those of wild-type cells.
Confluent MEF and ES cells were subcultured in the presence of BrdU (Sigma-Aldrich) at a final concentration of 1 × 10 M and were then allowed to replicate their DNA once at 37°C for 24 h and 12 h, respectively. Colcemid was added in a concentration of 0.2 μg/ml for MEFs and 1 μg/ml for ES cells for the last 4 h and 1 h, respectively. Cells were then recovered, and metaphases were prepared as described previously (). Chromosome orientation FISH was performed as previously described (, ) using first a (CCCTAA) PNA probe labeled with Cy3 and then a second (TTAGGG) PNA labeled with fluorescein (Applied Biosystems). Metaphase spreads were captured on a fluorescence microscope (DMRB; Leica).
Genomic SCEs were visualized using an adapted Fluorescence-Plus-Giemsa protocol (). In brief, confluent MEFs and ES cells were subcultured in the presence of BrdU (Sigma-Aldrich) at a final concentration of 1 × 10 M and were then allowed to replicate their DNA twice at 37°C for 48 h and 24 h, respectively. Colcemid was added in a concentration of 0.2 μg/ml for MEFs and 1 μg/ml for ES cells for the last 4 h and 1 h, respectively. Cells were then recovered, and metaphases were prepared as described for the Q-FISH (). Slides were stained in 200 μg/ml Hoechst 33258 in water for 30 min at room temperature, rinsed with distilled HO, and allowed to air dry. Slides were then mounted with 2× SSC and exposed to UV light for 30 min. After being rinsed with distilled HO, slides were air dried. Finally, they were stained with Leishman's stain (Sigma-Aldrich) for 4 min. Metaphases were captured using brightfield microscopy (AX70; Olympus) and analyzed for harlequin staining. Each color switch was scored as one SCE.
For confocal immunostaining experiments, 50,000 cells (MEFs or ES cells) were seeded in multiwell immunofluorescence slides (LABTEK n°154534; Nunc). After 24 h in culture, cells were washed twice with 1× PBS and fixed for 20 min at 4°C in 4% PFA in 1× PBS. Cells were then washed twice with 1× PBS and treated with 0.1% Triton X-100 in 1× PBS at RT for 7 min followed by two washes with PBS. After blocking with 10% BSA (in PBS) at 37°C for 20 min, cells were incubated with rabbit anti–mouse PML polyclonal antibody (gift from P. Freemont, Cancer Research UK, London, UK) diluted 1:1,000 in blocking solution overnight at 4°C. After washing twice with 0.05% Triton X-100 in PBS for 5 min, cells were incubated with a secondary goat anti–rabbit IgG conjugated with AlexaFluor488 and diluted 1:300 in blocking solution for 1 h at RT. After immunostaining, telomeric FISH was performed as described previously () with minor modifications. In brief, slides were first washed twice with PBS and fixed for 2 min in 4% formaldehyde. After washing three times with PBS, slides were dehydrated in ethanol series (70, 90, and 100%) and air dried. Hybridization with Cy3-labeled telomeric PNA probe was performed as described for Q-FISH analysis. After hybridization, slides were washed twice with 50% formamide/10 mM Tris, pH 7.2/0.1% BSA for 15 min followed by three washes with 0.1 M Tris/0.15 M NaCl, pH 7.5/0.08% Tween 20 for 5 min. Slides were then dehydrated with ethanol series and air dried. Finally, slides were counterstained with 0.2 μg/ml DAPI in Vectashield (Vector Laboratories). Images were captured at room temperature in a confocal microscope (TCS SP2-A-OBS-UV; Leica) and analyzed by confocal software (Leica).
The methylation status of the different subtelomeric genomic DNA regions was established by PCR analysis after bisulfite modification. Bisulfite genomic sequencing was performed as described previously (). Automatic sequencing of 7–14 colonies for each sequence was performed to obtain statistical data on the methylation status of every single subtelomeric CpG island. Bisulfite genomic sequencing primers were designed against subtelomeric regions in chromosome 1. The primer sequences used were CACCTCTAACCACTTAAACCTAACAA and GGGGTAGATATTTAGGGAAGG, which flank positions 197042227–197042569 of chromosome 1 in the mouse National Center for Biotechnology Information (NCBI) 36 genome assembly, and TTACCAATACCACCATTCCTCCA and GAGAGTAGTTAATTAGATGAGGAATA, which flank positions 181837807–181838281 at chromosome 2 in the mouse NCBI 36 genome assembly. Results were analyzed with the BiQ Analyzer program.
A Wilcoxon rank-sum test was used to calculate statistical significance differences for telomere length differences. A two-tailed test was used to assess the statistical significance of the observed differences in ChIP and RT-PCR analyses. Prism (version 4; GraphPad) and Excell software (version 2003; Microsoft) were used for statistical calculations.
To calculate the statistical significance of differences in subtelomeric DNA methylation, SCE, T-SCE, and APB frequencies, we used the χ test. The two-sided p-values presented were obtained from a 2 × 2 contingency table analyzed by the χ test (including Yates' continuity correction). Instat (version 3.05; GraphPad) was used for the calculations. In all cases, differences are significant for P < 0.05, very significant for P < 0.01, highly significant for P < 0.001, and extremely significant for P < 0.0001.
Fig. S1 shows the defective assembly of subtelomeric chromatin in Suv4-20h–deficient ES cells. Fig. S2 shows normal DNA methylation of subtelomeric domains in Suv39h-deficient ES cells. Fig. S3 shows increased global SCE in cells deficient for HMTase Suv39h and Suv4-20h. Fig. S4 depicts a model for chromatin assembly at mammalian telomeres and subtelomeres. Table S1 provides a summary of the effects of Suv39h and Suv420h deletion on telomere length, telomere recombination, and telomere structure. Online supplemental material is available at . |
In eukaryotes, the removal of introns from pre-mRNAs requires the five phylogenetically conserved small nuclear RNP (snRNP) particles (U1, U2, U4, U5, and U6 snRNPs; for reviews see ; ). The formation of functional spliceosomal snRNPs is a complex event (for reviews see ; ; ), and several discrete nuclear domains, such as Cajal bodies (CBs), interchromatin granule clusters (IGCs), and nucleoli have been implicated in their maturation and/or storage (). The snRNPs, along with >100 other splicing factors, assemble onto pre-mRNA to form the spliceosome, and it is this dynamic macromolecular machine that orchestrates the excision of introns and the ligation of exons through two successive trans-esterification reactions (for review see ). Spliceosomal assembly and splicing itself, which are key events in the maturation of pre-mRNAs, are tightly coupled to RNA transcription (for reviews see ; ). Accordingly, nascent RNA polymerase (RNAP) II transcripts were previously shown to recruit splicing factors, such as the snRNPs and SR (serine-arginine rich) proteins (; ; ; ) and, more recently, the exon junction complexes (EJCs), which mark the ultimate products of splicing, exon–exon junctions (for review see ).
Although data on the spatial and temporal recruitment of splicing factors onto a template pre-mRNA abound, very little is still known about the essential characteristics of a spliceosomal snRNP that contribute in vivo to its association with nascent transcripts. Previous work on U1 and U2 snRNPs highlighted the importance of the base pairing of their RNA moieties to cis-acting sequences on pre-mRNAs, the intronic 5′ splice site (SS) and the branch point sequence (BPS), respectively (; ; ; ). In the case of the U1 snRNP, however, it was shown that the base pairing of its 5′ end with the 5′ SS is only one of several interactions that contribute to the formation of a U1 snRNP–pre-mRNA complex () and occurs after an initial recruitment of the U1 snRNP (). Interestingly, cleavage of the 5′ end of the U1 small nuclear RNA (snRNA) has no effect on the rate of association of the U1 snRNP with a consensus 5′ SS RNA oligonucleotide in vitro (). Rather, recognition of the 5′ SS by the U1 snRNP appears to be driven by its overall protein complement. Which of the several U1 snRNP proteins and which sequence elements of the U1 snRNA are critical for its targeting to nascent transcripts is still unclear, however. The same question also remains unanswered for the other spliceosomal snRNPs, and, in light of their complex intranuclear trafficking before engaging pre-mRNA splicing (for review see ), it cannot be addressed directly using in vitro systems.
The lampbrush chromosomes (LBCs) of amphibian oocytes exhibit unique structural characteristics that make it possible to study the recruitment of snRNPs to nascent transcripts in vivo. In particular, these extended diplotene bivalent chromosomes display numerous lateral loops of chromatin that correspond to regions of intense transcriptional activity by RNAPII (for review see ). The chromosomal loops are composed of two distinct domains: the first domain corresponds to a decondensed euchromatin axis that can be demonstrated using antibodies against the RNAPII transcriptional machinery or various chromatin components (). The second domain corresponds to nascent RNP fibrils, which are formed from nascent pre-mRNAs associated with a cortege of factors involved in their maturation. These RNP fibrils create a dense RNP matrix around the loop axis that is readily observable by phase contrast or differential interference contrast (DIC) microscopy. Indeed, the elongation of transcripts along the axis is reflected in a characteristic thin to thick morphology of the loops (; ; for review see ). We show here for the first time that newly assembled snRNPs, which are formed upon cytoplasmic injection of fluorescein-labeled snRNAs, associate rapidly with nascent transcripts, and we used this novel cytological assay to begin dissecting the molecular mechanisms that regulate the association of splicing snRNPs with active transcriptional units. In particular, we demonstrate that U1 and U2 snRNPs need not be functional for their association with elongating transcripts. We also characterize the first stem loop domain of the U1 snRNA as a structure that is both necessary and sufficient for targeting the U1 snRNP to nascent pre-mRNAs. Finally, we present evidence that pre-mRNA splicing occurs on the LBC loops and that recruitment of the splicing snRNPs to active transcriptional units is independent of their integration into a spliceosome.
It was previously established that in vitro–synthesized spliceosomal snRNAs injected into the cytoplasm of amphibian oocytes assemble into functional snRNP particles () that are competent to splice pre-mRNA. Curiously, in these experiments, the subnuclear localization of the newly formed snRNPs does not correspond to the steady-state distribution of the endogenous snRNPs, which are associated with CBs, IGCs (B-snurposomes in the oocyte), and chromosomal loops. Notably, although the newly formed snRNPs accumulate within CBs, their association with IGCs is very weak, and their targeting to chromosomal loops was never documented ().
Chromosomal loops are likely sites of pre-mRNA processing, however, and because injected synthetic spliceosomal RNAs can rescue splicing in oocytes depleted of the corresponding endogenous snRNA (; ), our hypothesis was that injected fluorescent snRNAs do associate with chromosomal loops but at a concentration too low to be detected without amplification. To test that idea, fluorescein-conjugated U1, U2, U4, and U5 snRNAs were synthesized and injected into the cytoplasm of oocytes, and their was fate monitored over time on fixed nuclear spreads. A two-antibody detection system was used to enhance the fluorescent signals, and, as expected, all four snRNAs entered the nucleus and associated with CBs. shows the targeting of the U1 snRNP to both CBs and IGCs only 1 h after cytoplasmic injection. The same result was obtained with U2, U4, and U5 snRNAs. In contrast, the nonspliceosomal U7 snRNP (discussed in the next paragraph), which was used here as a negative control, did not associate with IGCs but strongly targeted CBs.
Also, for the first time, we were able to demonstrate their association with the active transcriptional units (). Unlike a previous study (), we found that fluorescent snRNPs target the chromosomal loops rapidly after injection because a weak but specific signal was also detected in these nuclear domains as soon as 1 h after injection. Detailed analyses of the loop staining using laser-scanning confocal microscopy revealed an association of the fluorescent snRNPs with the nascent RNPs rather than with the axial chromatin (, inset). This loop distribution is identical to that of the endogenous snRNPs as previously determined by in situ hybridization (). Importantly, labeling of the loops cannot be attributed to an incorporation of free fluorescent UTP (possibly produced by degradation of the injected snRNAs) into nascent transcripts because the injection of 200 pmol of fluorescent UTP fails to generate any detectable signal (unpublished data). Instead, staining of the loops is most likely caused by association of the snRNAs in their snRNP conformation.
To test whether the presence of the fluorescent snRNPs on chromosomes was the result of a genuine recruitment rather than that of random binding, we analyzed the subnuclear distribution of a synthetic fluorescent U7 snRNA after cytoplasmic injections. Just like the spliceosomal snRNAs, the U7 snRNA assembles into snRNP that is subsequently recruited to the nucleus (; ; ; ). The nuclear U7 snRNP comprises part of the processing machinery responsible for the maturation of histone pre-mRNAs (; ; ; ; ), and it was previously shown by in situ hybridization that >90% of the nuclear U7 snRNA associates with CBs and is absent from chromosomes (; ). Thus, the U7 snRNP is not expected to interact with chromosomal loops. and show that the newly made fluorescent U7 snRNP was efficiently targeted to CBs but not to chromosomes.
To further test whether snRNPs are recruited to active sites of transcription, oocytes were treated with the transcription inhibitor actinomycin D before nuclear spread preparation. Such treatment results in a complete loss of chromosomal signal, as shown in for the U1 snRNP. These data further support the conclusion that association of the fluorescent snRNPs with chromosomes depends on the presence of nascent transcripts. Although there is no RNAPI activity on LBCs, both RNAPII and RNAPIII are actively engaged in transcription. The sites of RNAPIII transcription have been mapped to ∼90 distinct chromosomal loci (). These sites are not visible by light microscopy because they lack the density of an RNP matrix but are readily detected by immunofluorescence using anti-RNAPIII antibodies (). An antibody directed against one of the specific subunits of RNAPIII, RPC53, was used in to show that a newly assembled fluorescent U1 snRNP does not associate with RNAPIII transcriptional units. This result is in agreement with the fact that RNAPIII transcripts are not substrates of the spliceosome. Identical results were obtained with the U2 snRNP (unpublished data).
U1 and U2 snRNPs are thought to be involved early in the stepwise formation of the spliceosome onto a target pre-mRNA, and they both display a short sequence that hybridizes to the 5′ SS or BPS of an intron, respectively (for review see ). In the case of the U1 snRNA, the 5′ SS recognition sequence lies within its first 20 nucleotides. To test whether the recruitment of the U1 snRNP to transcriptional units requires its hybridization with pre-mRNAs, a fluorescent U1 snRNA truncated from its first 20 residues, U1(ΔSS) snRNA, was synthesized and injected into the cytoplasm of oocytes. It was established that removal of these residues of the U1 snRNA does not prevent the assembly of a U1(ΔSS) snRNP with its full protein complement (; ), and, as expected, the newly made U1(ΔSS) snRNP was rapidly recruited to the nucleus. Interestingly, in addition to accumulating within CBs and IGCs, the U1(ΔSS) snRNP targeted the chromosomal loops just as well as the full-length U1 snRNP (). A similar deletion analysis was performed for the U2 snRNP in which the BPS recognition sequence was removed. Such a U2(ΔBPS) snRNA can no longer engage splicing by hybridizing with an intronic BPS, yet the newly formed U2(ΔBPS) snRNP associates with chromosomal loops as well as with CBs and IGCs, identical to wild-type U2 snRNP (). Together, these data demonstrate that the recruitment of U1 and U2 snRNPs to nascent transcripts is not directed by hybridization of their snRNA moieties to cis-acting signals on pre-mRNAs. Importantly, they also highlight the fact that the association of U1 and U2 snRNPs with elongating transcripts can be uncoupled from their function in splicing.
In the case of the U2 snRNP, its splicing activity depends greatly on modification of the U2 snRNA by 2′--methylation and pseudouridylation (; ). In particular, the modification of several residues within the first 29 nucleotides of the U2 snRNA is critical for the formation of a mature 17S snRNP particle (). Thus, we produced a fluorescently labeled U2 snRNA deleted of these residues (U2(Δ29) snRNA), injected it into the cytoplasm of stage V oocytes, and analyzed its nuclear distribution 18 h later on nuclear spreads. shows that the newly assembled U2(Δ29) snRNP associates well with the chromosomal loops, further supporting the idea that the association of snRNPs with active RNAPII transcriptional units is independent of their ability to engage splicing. In addition, IGCs are brightly labeled, but, surprisingly, CBs appear to be only weakly stained (, arrow), especially when the fluorescent signal is compared with that of the full-length U2 snRNP (). Because CBs are implicated in the internal modification of the spliceosomal snRNAs, one possible explanation is that lack of the first 29 residues, among which many are modified, renders the U2(Δ29) snRNA a poor substrate for the modification machinery and, as a result, reduces its overall residence time within CBs.
Three proteins are known to be specific for the U1 snRNP: U1A, U1C, and U1-70K. Both U1-70K and U1A bind directly to the U1 snRNA through stem loop I and stem loop II, respectively, whereas U1C interacts with U1-70K (; ). Because U1C was previously implicated in association of the U1 snRNP with pre-mRNAs in vitro (), we tested whether the deletion of stem loop I would impact the subnuclear distribution of the U1 snRNP. The first 47 residues of U1 were deleted, and the resulting mutant U1(Δ47) snRNA was injected into the cytoplasm of stage V oocytes. Nuclear spreads were prepared 18 h later. shows that U1(Δ47) snRNP accumulates in CBs but fails to associate with both IGCs and the nascent transcripts on chromosomal loops. Because the removal of the first 20 residues of the U1 snRNA does not disrupt its chromosomal targeting (), we concluded that stem loop I is the structure present within the first 47 nucleotides that is critical for the association of the U1 snRNP with nascent transcripts. To test that idea, we constructed a chimeric RNA by fusing stem loop I to the 3′ end of the U7 snRNA, which is exclusively found associated with CBs ( and ). The resulting U7/U1(I) snRNA was injected into stage V oocytes, and its subnuclear distribution was analyzed on nuclear spreads (). Remarkably, stem loop I alone is sufficient to promote targeting of the U7 snRNP to chromosomal loops and IGCs. Surprisingly, CBs are only weakly labeled (; arrows). This result was unexpected, as the U7 snRNP is known to accumulate in CBs at very high concentrations (; ), and suggests that stem loop I is important to regulate the kinetics of U1 snRNP exchange between CBs and the nucleoplasm.
The observation that several mutant U1 and U2 snRNPs, which cannot participate in the assembly of the spliceosome, still target chromosomal loops prompted us to ask whether the association of snRNPs with active transcriptional units could be uncoupled from the splicing reaction itself. An efficient way to inhibit pre-mRNA splicing is to deplete the oocyte of U2 snRNAs using an antisense oligonucleotide–RNase H degradation strategy (; ). In the absence of U2 snRNP, formation of the A complex (a spliceosomal intermediate containing both U1 and U2 snRNPs) and, thus, splicing itself is inhibited (; ; ). Importantly, splicing can be rescued by a cytoplasmic injection of in vitro–made U2 snRNAs (; ).
We first showed that depletion of the U2 snRNA results in the loss of the U2 snRNP–specific protein U2B″ from the chromosomal loops (). Interestingly, U2B″ was found to relocalize from chromosomes, IGCs, and CBs to nucleoli. The significance of U2B″ relocalization is not known, but we subsequently used it in all of our experiments as a cytological indicator of successful U2 snRNA depletions. Because microinjected DNA oligonucleotides are short lived, we were able to show that newly injected U2 snRNA could reestablish the normal distribution pattern of U2B″ in U2-depleted oocytes (). This result further validates our previous conclusion that association of fluorescent snRNAs with the chromosomal loops reflects the targeting of fully mature snRNPs.
We then asked whether pre-mRNA splicing occurs on the chromosomal loops and whether it is prevented by depletion of the U2 snRNA. During pre-mRNA splicing, the spliceosome stably deposits a large proteinaceous complex named the EJC ∼20 nucleotides upstream of exon–exon junctions (for review see ). Such EJCs influence the cellular fate of spliced mRNAs, with which they remain associated during nuclear export and until they are displaced by translating ribosomes. One of the EJC core subunits, Y14 (; ), is deposited after exon–exon ligation (). Importantly, then, deposition of Y14 on nascent transcripts is a reliable indication of splicing activity. To test whether EJCs are present on chromosomal loops, Y14 was expressed in fusion with an HA tag, and its subcellular distribution was analyzed using the anti-HA antibody mAb 3F10. shows that upon injection of HA-Y14 transcripts into the cytoplasm of stage V oocytes, a protein with the expected molecular mass of ∼24 kD is synthesized and efficiently recruited to the nucleus. There, it associates with CBs, IGCs as previously reported in somatic nuclei (), and, to a lesser extent, with nucleoli. In addition and in agreement with the fact that pre-mRNA splicing occurs cotranscriptionally, Y14 also associates with nascent transcripts. Remarkably, the depletion of U2 snRNA results in a complete loss of Y14 from chromosomal loops (), indicating a lack of spliceosomal activity on nascent RNP fibrils. Finally, a cytoplasmic injection of fluorescently labeled U2 snRNAs restores the presence of the U2 snRNP and Y14 on chromosomal loops ( and ). Together, these data show that pre-mRNA splicing occurs on the chromosomal loops in the presence but not in the absence of U2 snRNP.
Finally, we tested whether U1, U4, and U5 snRNPs could still be recruited to transcriptional sites in the absence of any splicing activity. It was shown previously that the presence of a fully functional U1 snRNP is critical to transcription and, thus, to the maintenance of chromosomal loops in amphibian oocytes (). As expected, we found that the U1 snRNP still associates with the nascent transcripts of the chromosomal loops in U2 snRNA–depleted oocytes (). This result is also consistent with an early recruitment of the U1 snRNP to the pre-mRNA template, as it would be in the canonical model of splicing, which proposes a stepwise assembly of the spliceosome. In such a model, the U4/U6.U5 tri-snRNP is recruited only after the formation of the A complex. Although the U5 snRNP is commonly used as a representative member of the U4/U6.U5 tri-snRNP, it is present in both the major (U2 type) and minor (U12 type) spliceosomes. Thus, the U4 snRNP, a specific member of the U2-type spliceosome, was also used here as a marker of the U4/U6.U5 tri-snRNP. Surprisingly, in the absence of U2 snRNP, the U4/U6.U5 tri-snRNP is still recruited to nascent transcripts (), indicating that the A complex is not required. Together, these data demonstrate that the splicing activity present on the chromosomal loops does not direct the association of snRNPs with nascent RNP fibrils.
The removal of most introns requires a conserved 5′ SS, a BPS followed by a polypyrimidine tract, and a 3′ SS. Although current models propose that the spliceosome assembles onto the target pre-mRNA in an ordered process, it is still unclear which early intermediate complexes form in vivo and in which order. The establishment of one of these intermediates, the A complex, involves base pairing of the U1 and U2 snRNAs to the 5′ SS and BPS, respectively. Whether removal of the 5′ SS recognition sequence on U1 snRNA results in a nonfunctional U1 snRNP is difficult to assess, as the requirement of U1 snRNA itself for intron removal in vitro depends on the pre-mRNA template as well as the concentration of SR proteins in the chosen splicing extract (, ). In addition, although one study indicates a strict requirement of the 5′ end of the U1 snRNA for intron removal (), others present the hybridization of U1 snRNA 5′ end to pre-mRNA as a nonessential stabilizing force () that might influence the transition between spliceosomal intermediate complexes (). In the case of the U2 snRNA, however, the requirement of the BPS recognition sequence for efficient pre-mRNA splicing has been well established (; ; ).
Interestingly, we have shown here that U1(ΔSS) and U2(ΔBPS) snRNAs, which cannot hybridize to introns, are assembled into snRNPs and target the nascent transcripts on chromosomal loops. One interpretation is that the respective base pairing of U1 and U2 snRNAs with the 5′ SS and BPS is not essential for their association with nascent transcripts in vivo. This is in agreement with previous work showing that initial recruitment of the U1 snRNP to pre-mRNAs appears to be mediated by U1 snRNP proteins in a 5′ SS–independent manner (; ; , ; ). In addition, an in vitro study showed that hybridization of the U1 snRNA to target pre-mRNAs is dispensable for early intermediate formation and intron removal (). In particular, the U1 snRNP was recently shown to be cotranscriptionally recruited to pre-mRNAs with mutations in the 5′ SS that abolish hybridization with the U1 snRNA (), suggesting that the 5′ SS/U1 snRNA base pairing occurs after an initial recruitment phase (; ).
Another possibility stems from the structural organization of the chromosomal loops. In amphibian oocytes, the RNAPII loops are readily visible by light microscopy because of their dense RNP matrix, which is composed of the nascent RNAPII transcripts and associated maturation factors. Surprisingly, some of these factors, such as the 3′ end processing factor CstF77, are only involved in the late steps of pre-mRNA maturation (). Therefore, the presence of CstF77 over the entire length of the loops () suggests that some pre-mRNA processing factors might associate with nascent RNP particles but remain inactive until the occurrence of their corresponding cis-acting RNA elements. Thus, the efficient recruitment of U1(ΔSS) and U2(ΔBPS) snRNPs to nascent transcripts could be the result of a staging event in which snRNPs would first be recruited to the nascent RNP fibrils and be maintained there until spliceosomal assembly could occur. In this model, the initial recruitment of snRNPs would rely, in part, on already associated heterogeneous nuclear RNPs, such as the SR proteins, whose presence was previously shown to require intronic sequences on the pre-mRNA ().
We show here that deletion of the first 47 nucleotides of U1 snRNA, which contain both the 5′ SS recognition sequence and stem loop I, has a dramatic effect on its subnuclear distribution. The resulting U1(Δ47) snRNP still accumulates strongly within CBs, but it fails to target IGCs and the chromosomal loops. Although these data demonstrate that a discrete region of U1 snRNA is critical for its intranuclear trafficking, it also raises the question of how stem loop I regulates the association of the U1 snRNP with two subnuclear domains that are distinct in structure and functions. The lack of association of the U1(Δ47) snRNP with nascent RNP fibrils could to be caused, in part, by the fact that the U1C protein cannot associate with the U1(Δ47) snRNA in the absence of stem loop I (). U1C was previously implicated in the binding of pre-mRNAs by the U1 snRNP (; ; ), and, thus, its absence from a U1(Δ47) snRNP could result in the loss of chromosomal targeting. There is no pre-mRNA splicing activity occurring in IGCs, however. Instead, one demonstrated function of these nuclear bodies is to serve as reservoirs for RNAPII maturation factors, which are subsequently recruited to active transcriptional sites (; for review see ). In light of the current model in which newly assembled snRNPs transit through CBs for modification and assembly before their association with IGCs (; ; for review see ), an attractive possibility is that stem loop I is essential to regulate kinetic exchanges of the U1 snRNP between CBs and the nucleoplasm. In particular, stem loop I might be essential for U1 snRNP to exit CBs. Interestingly, we showed that stem loop I is not only sufficient to direct the association of the nonspliceosomal U7 snRNP to nascent transcripts and IGCs, but it also modifies the association of the U7 snRNP with CBs. Indeed, although the normal fluorescent U7 snRNP accumulates greatly in CBs, this association is dramatically reduced by its fusion with stem loop I. Importantly, the chromosomal association of chimeric U7/U1(I) snRNP demonstrates that a snRNP, which cannot participate in splicing, can be targeted to nascent transcripts. In agreement with this idea, we find that both U2(ΔBPS) and U2(Δ29) snRNPs, which are nonfunctional (; ; ; ), are recruited efficiently to nascent transcripts.
However, one cannot exclude two other interesting possibilities. The first one is that the U4/U6.U5 tri-snRNP was previously shown to recognize the 5′ SS in the absence of the U2 snRNP in vitro (). In addition, the U5 snRNP was demonstrated to interact with the 5′ SS before the start of splicing (), and, more recently, the U4/U6.U5 tri-snRNP together with the U1 snRNP was proposed to comprise part of a very early intermediate that presumably plays an important role in defining the 5′ SS (). Thus, the observed recruitment of the U1 snRNP and tri-snRNP to chromosomal loops in the absence of the U2 snRNP might reflect the formation and stalling of this early intermediate form on nascent transcripts.
The second possibility comes from the development over the last decade of a different model for spliceosome assembly. A large RNP complex named the penta-snRNP containing all five splicing snRNPs in equal stoichiometric abundance and at least 13 other proteins was purified in yeast (), and a similar complex was found in mammals (). The penta-snRNP forms in the absence of a pre-mRNA template and, thus, challenges the canonical view of stepwise assembly of the spliceosome. Importantly, when supplemented with an snRNP-depleted extract, the penta-snRNP was competent to splice synthetic substrates as a unitary particle, providing evidence for a preassembly model of splicing wherein all five snRNPs engage the pre-mRNA in one step as a single complex (). Therefore, another interpretation of the U1 snRNP and tri-snRNP association with chromosomal loops in the absence of splicing is that snRNPs are recruited to the nascent transcripts as part of a preassembled complex. In that case, however, one would have to assume that such a complex could be formed and recruited to the transcriptional units without the U2 snRNP.
In the course of our study, we used the deposition of EJCs onto nascent transcripts as an indication of splicing, as it allows the simultaneous monitoring on nuclear spreads of all RNAPII transcriptional units in the same oocyte. EJCs are recruited cotranscriptionally by the spliceosome to mark exon–exon junctions after intron removal (for review see ), and, accordingly, we demonstrate here that Y14, a subunit of the EJC, targets the numerous LBC lateral loops. In the absence of the U2 snRNA, spliceosomal assembly and, thus, pre-mRNA splicing is inhibited (; ; ), which is illustrated on nuclear spreads by the loss of Y14 from LBCs. Interestingly, we recently obtained evidence that Magoh, another EJC subunit, distributes similarly to Y14 in the oocyte. This result was expected, as Y14 and Magoh were shown previously to interact (; ). We are now currently using the advantageous spatial resolution offered by LBCs together with the fact that these chromosomes can now be visualized in in vivo–like conditions (unpublished data) to characterize the kinetics of the association of Y14, Magoh, and splicing factors with the active transcriptional units.
The DNA templates U1, U2, U4, U5, U7/U1(I), U1(Δ47), U1(ΔSS), U2(Δ29), and U2(ΔBPS) used for the transcription of fluorescein-labeled RNA were amplified by PCR using the high fidelity Deep Vent DNA Polymerase (New England Biolabs, Inc.) from corresponding human U1 (), amphibian U2 and U5 (), amphibian U7 (), or chicken U4B (; ) clones.
The templates for producing P-labeled riboprobes (anti-U2 and -U5) for Northern blotting were amplified with GoTaq DNA Polymerase (Promega). In all cases, the 5′ primer used for amplification contains the T3 or T7 promoter for direct transcription with the respective polymerase. Amplified DNA fragments were gel purified using 0.45-mm cellulose acetate spin-X plastic centrifuge tube filters (Corning Inc.). The U2 DNA template was produced by a two-step PCR in which the first step deletes the BPS recognition sequence and two residues on each side (residues 31–40 of the U2 snRNA) and the second step introduces the T3 promoter. The DNA templates for the transcription of fluorescein–U7 snRNA and HA-Y14 mRNA were prepared by linearizing the pUC9/T7-X.l.U7 snRNA vector with PvuII (Invitrogen) and of pcDNA3.1/HA-tagged human Y14 vector with XbaI (Invitrogen). All DNA templates were phenol extracted, ethanol precipitated, and washed with 70% ethanol before transcription.
The DNA primers (Integrated DNA Technologies) used were as follows (T3 and T7 promoters are underlined): U1 (5′-GATACTTACCTGGCAGGGGAG and 3′-CAGGGGAAAGC- GCGAACGCAGTCCCCCAC), U1(ΔSS) (5′-GCATACCATGATCATGAAG and 3′-CAGGGGAAAGCGCGAACGCAGTCCCCCAC), U1(Δ47) (5′-CGCAGGGCCAGGCTCAGCC and 3′-CAGGGGAAAGCGCGAACGCAGTCCCCCAC), U2 (5′-GCATCCTTTCGCCTTTGC and 3′-AAGTGCACCGGTCCTGGAGGT ACTGC), U2(ΔBPS) first step (5′-ATCGCTTCTCGGCCTTTTGGCTAAGATCAATGTTCTTATCAGTTTAATATCTG and 3′-AAGTGCACCGGTCCTGGAGGTACTGC), U2(ΔBPS) second step (5′-GC
ATCCTTTCGCCTTTGC and 3′-AAGTGCACCGGT- CCTGGAGGTACTGC), U2(Δ29) (5′-CG AGTGTAGTATCTGTTCTTATC and 3′-AAGTGCACCGGTCCTGGAG- GTACTGC), U4 (5′-AGCTTTGCGCAGT- GGCAGTATC and 3′-CAGTCTCCGTAGAGACTGTCA), U5 (5′-CGG- AATTC and 3′-ATACCTGGTGTGAACC- AGGCTTC), U7/U1(I) (5′-ATACCATGATCATGAAGGTGGTTCTCCAAG- TGTTACAGCTC and 3′-TGTGGCTCCTACAGAG), anti-U2 (5′-CG
AAGTGCACCGGTCCTGGAG and 3′-ATCGCTTCTCGGCCTTTTGG), and anti-U5 (5′-CGTACCTGGTGTGAACCAGGCTT and 3′ ATACTCTGGTTTCTCTTCAAATTC and 3′-TGTGGCTCCTACAGAG).
Fluorescently labeled snRNA and mutants were transcribed with T3 or T7 RNA polymerases (Stratagene) in the presence of 125 nM ATP, 62.5 nM GTP, 125 nM CTP, 62.5 nM UTP, 25 nM ChromaTide fluorescein-12–UTP (Invitrogen), and 125 nM mG(5′)ppp(5′)G cap analogue (GE Healthcare). P-labeled riboprobes were transcribed with T7 RNA polymerase (see previous two paragraphs) except that the cap analogue was omitted in the reaction and the fluorescein-coupled UTP was replaced by 50 μCi α-[P]UTP (GE Healthcare). The HA-Y14 mRNA was transcribed with T7 RNA polymerase similarly except that UTP was present at 125 nM and fluorescein-12–UTP was omitted. Recombinant RNasin Ribonuclease Inhibitor (Promega) was present in all transcription reactions. After 2 h of incubation at 37°C, transcription reactions were treated with RQ1 RNase-free DNase (Promega) for 15 min at 37°C. All labeled RNAs were purified using NucAway Spin columns (Ambion) equilibrated with water. The HA-Y14 mRNA was phenol extracted, ethanol precipitated, washed, and resuspended in water.
Female adult was anesthetized in 0.15% tricaine methane sulfonate (MS222; Sigma-Aldrich), and fragments of ovary were surgically removed. Oocytes were defolliculated for 2 h at room temperature in saline buffer OR2 () containing 0.2% collagenase (type II; Sigma-Aldrich). Stage IV–V oocytes were isolated and maintained in OR2 at 18°C. Oocytes were always injected into the cytoplasm with volumes of 20–30 nL. Glass needles were prepared using a horizontal pipette puller (P-97; Sutter Instrument Co.). All injections were performed under a dissecting microscope (S; Leica) using an injector (nanojet II; Drummond). For snRNP targeting assays, ∼10–20 fmol of the respective in vitro–made fluorescent snRNAs were injected per oocyte. For the U2 depletion experiments, 50 ng of the following DNA oligonucleotides (Integrated DNA Technologies) were used per oocyte: U2b oligonucleotide (complimentary to residues 28–42 of the U2 snRNA) CAGATACTACACTTG and C oligonucleotide (sequence unrelated to any RNA) TCCGGTACCACGACG.
It is noteworthy to mention that the injection of any DNA oligonucleotide into amphibian oocytes has several reversible nonspecific effects, including a transient inhibition of transcription as visualized by a loss of the lateral chromosomal loops and their reformation over time (). In all experiments, oocytes were thus incubated for a minimum of 18 h after DNA oligonucleotide injection before preparing nuclear spreads to allow for the recovery of their transcriptional activity. When actinomycin D treatment was required to inhibit transcription, oocytes were incubated in OR2 medium containing 10 μg/ml actinomycin D (Sigma-Aldrich).
Single nuclei were isolated and homogenized in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 0.2% SDS. Total nuclear RNAs were phenol extracted, ethanol precipitated, and fractionated on an 8 M urea, 1× Tris borate EDTA, and 8% polyacrylamide gel in 1× Tris borate EDTA electrophoresis buffer using an electrophoresis system (Mini-PROTEAN 3; Bio-Rad Laboratories). The RNA was electrophoretically transferred to a Zeta probe membrane (Bio-Rad Laboratories) in 1× Tris acetate EDTA transfer buffer using the Mini Trans-Blot cell (Bio-Rad Laboratories). The RNA was UV cross-linked (12 kJ/cm) to the membrane using a cross-linker (Spectrolinker; Spectronics Corp.). The membrane was blocked for 10 min with hybridization buffer (171 mM NaHPO, 79 mM NaHPO, 1 mM EDTA, pH 8.0, and 7% SDS), probed overnight with P-labeled antisense U2 and U5 snRNA probes at 10 cpm/ml in hybridization buffer, and washed twice for 30 min with wash buffer (13.7 mM NaHPO, 6.3 mM NaHPO, 1 mM EDTA, pH 8.0, and 1% SDS). Blocking, hybridization, and washing were performed at 65°C with rotation in an incubator (Isotemp Hybridization; Fisher Scientific). A phosphorscreen was exposed for 1 h and scanned with the Cyclone Storage Phosphor system (PerkinElmer).
To express HA-tagged Y14, 25 nL (0.5 ng/nL) HA-Y14 mRNA was injected into the cytoplasm of stage V oocytes. After a 50-h incubation, 10 oocytes, cytoplasms, or nuclei were hand isolated using jeweler's forceps and homogenized in 10 mM Tris-HCl, pH 8.0, 1 mM EDTA, and 0.2% SDS. The crude extract was centrifuged at 22,000 at 4°C for 10 min. The clarified extract was collected and fractionated on a 12% polyacrylimide gel using an electrophoresis system (Mini-PROTEAN 3; Bio-Rad Laboratories). Immunoblotting was then performed as described previously () with the anti-HA antibody mAb 3F10 (Hoffmann-La Roche, Inc.) used at 50 ng/ml.
Nuclear spreads were prepared as described previously () and fixed with 2% PFA in PBS + 1 mM MgCl for 1 h at room temperature. After fixation, nuclear spreads were rinsed in PBS and blocked with 0.5% BSA (Sigma-Aldrich) + 0.5% gelatin (from cold water fish) in PBS (blocking buffer) for 10 min. Spreads were incubated with primary antibody for 1 h at room temperature, washed for 30 min with two changes of PBS, incubated with secondary antibody for 1 h at room temperature, and washed again for 30 min with two changes of PBS before they were mounted in 50% glycerol/PBS containing 1 mg/ml phenylenediamine and 10 pg/ml DAPI. When a red fluorescent DNA counterstain was desired, spread preparations were incubated with 1 μM Syto61 (Invitrogen) in PBS for 20 min at room temperature and briefly rinsed in PBS before mounting. In these preparations, DAPI was omitted from the mounting medium. |
Nuclear mRNA transport is often thought of in terms of translocation through the nuclear pore, but mRNA export also requires intranuclear progression of transcripts from the gene to the nuclear pore. In some genetic diseases, failed export of a mutant mRNA is critical to the phenotype, yet typically it is not well understood how nuclear export is impeded or whether mutant mRNA accumulates at a specific point within the nuclear structure. In fact, the examination of mRNA blocked at a specific step in export may help illuminate the path whereby mRNA normally transits from the gene to the nuclear pore. The analysis of human disease gene mutations that impact nuclear metabolism of the mRNA provides an avenue to study both disease pathogenesis and the interrelationship between nuclear structure and steps in mRNA biogenesis. In addition, the study of naturally occurring disease alleles in patient cells provides the advantage that the mutant mRNA is expressed in a normal structural and physiological context.
This study examines the intranuclear fate of normal and triplet repeat–expanded transcripts in myotonic dystrophy type 1 (DM1) to define the point in the nuclear structure where the progression of normal and mutant transcripts diverge, which, in turn, provides insight into the step at which mutant mRNA transport is blocked. In DM1, expansion of a CTG triplet repeat occurs in the 3′ untranslated region of the gene encoding dystrophia myotonica protein kinase (DMPK), a serine/threonine protein kinase (; ; ; ). Normal alleles have 5–35 repeats, whereas DM1 alleles have 50 to >1,000. It has been shown that the mutant genes encode mRNAs containing expanded repeat sequences and that these RNAs are sequestered in the nucleus, where they accumulate in discrete nuclear foci (; ; ). However, it is not known whether normal mRNA associates with any defined intranuclear nuclear structure after its transcription and whether this differs for the mutant RNA.
In addition to the defect in mRNA transport in DM1, it is hypothesized that the expanded repeat RNAs alter the nuclear distribution or activity of specific CUG-binding proteins, which, in turn, affects the alternative splicing of other pre-mRNAs (; ). Multiple proteins such as CUG-binding protein (, ; ), heterogeneous nuclear RNP H (hnRNP H; ), double-stranded RNA–binding proteins (), or transcription factors Sp1 and retinoic acid receptor γ () have been identified as interacting with CUG repeat RNA. MBNL1 (muscleblind-like protein 1), a known regulator of alternative splicing (), is an especially strong candidate. In DM1 cells, MBNL1 is sequestered in mutant RNA foci (; ; ), and the loss of MBNL1 results in a phenotype that has similarities to DM1 ().
Understanding how mutant RNA transport is blocked requires an appreciation of the relationship of the normal RNA to nuclear structure. The mammalian nucleus is comprised of several nonmembrane bound compartments, each containing specific subsets of macromolecules. Of particular relevance here, a host of pre-mRNA metabolic factors concentrate markedly in 10–30 irregularly bordered domains (; ; ; ; ), which are defined here by staining for the spliceosome assembly factor SC-35 (). These SC-35 domains (also known as speckles or splicing factor compartments) are also highly enriched in poly(A) RNA (; ). SC-35 domains are associated with a specific subset of active genes (; ; ), with many genes and gene-rich R-band DNA clustering around their periphery (). Importantly, the genes position at the immediate edge of an SC-35 domain but not within them, whereas several pre-mRNAs have been shown to detach from the gene and enter the domains (; ), which contain multiple distinct mRNAs ().
Several observations suggest that the egress of specific gene transcripts from a peripheral gene into the domain may be linked to postsplicing steps rather than just splicing itself. For ( α) RNA, it was shown that most introns are removed at the gene just outside the domain, which is consistent with most splicing being cotranscriptional (). Some genes lacking introns become associated with domains upon activation (). More recently, several factors linked to postsplicing complexes required for mRNA export have also been shown to be concentrated within these domains (; ; ).
In osteogenesis imperfecta type I, a splicing defect in intron 26 of the collagen 1 α1 gene results in an abnormal accumulation of the mutant collagen RNA within SC-35 domains, indicating an impediment to the egress of the RNA from the domain (). This study suggested that SC-35 domains are a screening point for properly processed mRNA, a concept supported by a study that described the movement of poly(A) RNA through SC-35 domains (). We have proposed that nuclear export of some specific mRNAs begins by passage through the SC-35 domain neighboring the gene. This process may consist of two distinct steps: entry of the newly formed mRNA into the domain and exit of export-ready mRNA from the domain. The study of osteogenesis imperfecta type I provides evidence for the second step, as it was shown that the exit of mutant mRNA from domains was blocked or impeded. However, it is not clear whether entry into the domain from the peripheral gene is also a discrete step.
In this study, we examine the status of normal and mutant RNA relative to intranuclear structure in both normal myoblasts and myoblasts from individuals with DM1. Although both normal and mutant transcripts detach from the gene, they show distinct distributions relative to SC-35 domains. Normal transcripts consistently accumulate within the SC-35 domain, emanating from a gene precisely positioned at the outer edge of the domain. In contrast, mutant transcripts detach from the gene but no longer progress within the SC-35 domain and accumulate in multiple round foci that accumulate at the outer boundary of SC-35 domains. We provide direct visual evidence that mutant RNA in foci is spliced despite its exclusion from the splicing factor–rich domains. We also examine the role of the MBNL1 protein in the sequestration of RNA carrying the repeat expansion and demonstrate that loss of MBNL1 restores the ability of the mutant RNA to accumulate in domains. Our results suggest that egress of transcripts from the gene into an SC-35 domain is a normal step in the nuclear progression of this mRNA, and transcripts from the mutant allele are blocked from normal entry into SC-35 domains. This block correlates with the failed cytoplasmic export of mutant RNA, suggesting that export is not blocked at the nuclear pore but at a very early step in nucleoplasmic transport.
Although several active genes and their mRNAs associate closely with SC-35 domains, this association is locus specific, as has been observed for slightly more than half of the 25 active endogenous genes studied (). To determine whether the normal gene and its RNA associate with SC-35 domains, a sequential hybridization technique (; ) was used to detect the mRNA in one color using a cDNA probe (2.3 kb) and the gene in a second color using a cosmid probe (∼40 kb). As shown in , after hybridization to normal differentiated myotube nuclei, RNA is usually seen in two irregular tracklike accumulations that are larger than and juxtaposed to the locus signal. RNA hybridization with the cDNA probe and costaining with an antibody to the spliceosome assembly factor SC-35 demonstrate that these RNA accumulations are found within SC-35 defined domains (), with >84% ( = 316) of signals examined (see ) clearly overlapping part or most of the domain. 3D analysis ( and Video 1, available at ) confirms that the RNA lies largely within the domain.
DNA hybridization of the gene locus also shows that 70–90% associate with an SC-35 domain (n = 136 signals; ). We initially defined association to indicate that two entities appear to be in direct contact as viewed by light microscopy, with no visible separation (). By this strict criterion, 70% of loci associate with a domain, but an additional 20% show just a minute hairline separation (, left inset); thus, up to 90% of genes localize to the region of active genes that surrounds an SC-35 domain. This contrasts with the much lower theoretically and empirically derived random association rates; based on the volume occupied by SC-35 domains (∼3–10%; ; ) and on association frequencies of many nonassociated loci, the random association frequencies are ∼5–25% (; ; ). Most importantly, although the loci are clearly associated with the domain edge, they are not within it like the RNA signal. Thus, although this may not represent all of the mRNA, an accumulation of transcripts emanates from one side of the gene, which is positioned at the domain boundary, into the SC-35 domain. Although is not as complex and highly expressed, this structural arrangement is very similar to that shown for mRNAs such as and myosin heavy chain, which were shown to represent posttranscriptional mRNA that has detached from the gene and accumulates within an adjacent SC-35 domain (; ).
Although SC-35 domains are enriched for poly(A) RNA, much of this RNA remains in the domain long after transcription inhibition (>6 h), leading to suggestions that poly(A) RNA in the domains is comprised of nuclear RNA that is not exported to the cytoplasm (see Discussion). Thus, because RNA is consistently within splicing factor–rich domains, a question may remain as to whether this focus of RNA is trapped in the SC-35 domain or whether it can be chased from domains after transcription inhibition, as may be expected for an exported mRNA. To approach this question, we visualized RNA in normal myotube nuclei after transcription inhibition using dichlorobenzimidazole ribofuranoside. Loss of discernible mRNA foci in myotube nuclei was scored (), and a half-life of these RNA accumulations was calculated to be ∼43 min. This half-life is similar to the export rate found for a fluorescently labeled mRNA that also associated with nuclear speckles () and is far shorter than the >15-h half-life reported for the expanded CUG repeat–containing RNA foci that are retained in the nucleus in DM1 cells ().
The collective aforementioned results define the nuclear distribution of normal mRNA, in which recently synthesized transcripts associate with a specific nuclear compartment, laying a foundation for investigating any potential reorganization of the mutant mRNA. Our results demonstrate that accumulation within SC-35 domains is a normal part of the mRNA biogenesis and suggest that it is linked to the export of some (if not all) DMPK mRNA.
We next determined the relationship of nuclear-retained RNA containing triplet repeat expansion to SC-35 domains. A previous study briefly examined the relationship of mutant DMPK mRNA foci to SC-35 domains and concluded that there was no substantial association (). Hybridization to mRNA in DM1 myoblasts from a homozygous DM1 patient () shows that the mutant mRNA containing expanded CUG repeats is typically seen within nuclei in numerous round, discrete spots, which we refer to as mutant RNA granules (MRGs) as opposed to the two more irregular tracks seen for the mRNA in normal muscle cell nuclei (). The numerous MRGs in DM1 myoblasts are not the result of several copies of the gene as confirmed by DNA hybridization (). In DM1 nuclei, 92% of the numerous MRGs (on average more than four per nucleus in homozygous muscle cells) associate with SC-35–defined domains (). However, this association is markedly different than the RNA foci in normal muscle nuclei, which are reproducibly found within the SC-35 domain (). As quantified in , the bulk of the associations of MRGs with SC-35–defined domains shifts from internal (complete overlap) in normal muscle cells to an association with the edge of the domains in DM1. This edge association is apparent in most cases by simple 2D analysis; in cases in which the RNA signal initially would be scored to overlap the SC-35, further 3D analysis found that it was most typically either in a hole of lower density SC-35 signal or actually above or below the SC-35 domain (). 3D visualization of DM1 nuclei (Video 2, available at ) supports the localization of these RNA foci predominantly at the outer edge of SC-35 domains.
It has previously been demonstrated that MRGs do not colocalize with other nuclear structures, such as PML bodies, exosomes, and the perinucleolar compartment (). However, the presence of multiple MRGs in close association with SC-35 domains suggested the possibility that the RNA could be in paraspeckles, which are novel nuclear domains that abut SC-35–defined speckles (). Simultaneous detection of the mutant RNA foci and the paraspeckle protein PSP1 clearly demonstrated that the signals were not overlapping (unpublished data).
We then combined simultaneous DNA, RNA, and SC-35 detection in heterozygous DM1 myotube nuclei. As in normal muscle nuclei, genes in DM1 myoblasts show similar positioning at the edge of SC-35 domains. Thus, the change in RNA association is not caused by a shift in gene localization relative to SC-35 domains. shows that the two alleles are positioned at the edge of splicing factor domains. One allele has an irregular dispersed track of RNA extending from it into the adjacent domain, which is clearly characteristic of transcripts released from the normal allele into the SC-35 domains in normal muscle cells (). The other allele is distinct from the normal allele in that it has only a minute RNA signal (nascent transcripts) that does not extend into the domain, similar to DNA and RNA signals seen in homozygous mutant DM1 nuclei (). This relationship of the genes, RNA, and SC-35 is apparent in 3D as seen in Video 3 (available at ).
In , we also see that MRGs associate with domains other than the ones adjacent to the DNA loci. This suggests that MRGs have drifted from the transcription site but clearly still show an affinity for the SC-35 domains. However, unlike the normal RNA, these MRGs are restricted to the periphery and do not enter the domains. This shift in localization relative to the SC-35 compartment in DM1 demonstrates that the normal path of RNA from the gene into the domain is blocked for RNA carrying the triplet repeat expansion. In contrast to mutant RNA in osteogenesis imperfecta type I, in which mutant transcripts entered the domain normally but failed to exit properly (), some block to export in DM1 occurs at the entry of the repeat-expanded transcripts into the domain. Although we cannot exclude that some normal transcripts are exported without entering SC-35 domains, these results show that the block in export is correlated with a failure of mutant RNA entry into the domain.
The failure of the mutant RNA to enter SC-35 domains raises the question as to whether splicing occurs normally despite the repeat expansion in the 3′ untranslated region, which prevents their accumulations within these splicing factor–rich structures. Using Northern analysis of extracted cellular mRNA, an earlier study () was unable to detect intronic sequences in RNA derived from DM1 or normal cells, suggesting that splicing was unaffected. We directly addressed whether the mutant RNA foci still contain introns by visualizing RNA in intact DM1 cells using a PCR-generated probe encompassing the intron 9 simultaneously detected with a cDNA probe. In normal muscle nuclei that have just two high concentration sites of largely unspliced pre-mRNA emanating from each normal allele, intron 9 signals intermingle with much of the localized cDNA signal (). In contrast, the numerous bright foci of mutant RNA in DM1 cells (visualized by their enrichment of MBNL1 protein; ) lack the intron 9 signal (). Similar results are seen when using a cDNA probe to detect RNA foci (unpublished data). This result illustrates that the mutant RNA in the MRGs of DM1 cells is spliced. Thus, although blocked entry into the SC-35 domain could still relate to failed export of the mRNA, it does not appear to affect splicing. This is consistent with other evidence that collagen pre-mRNA splicing occurs primarily at the edge of the SC-35 domain, whereas more mature mRNA is within the domain ().
In normal myotube nuclei, MBNL1 is seen in a broadly distributed, somewhat punctuate pattern by immunostaining (). Interestingly, MBNL1 does not appear particularly concentrated in splicing factor–rich SC-35 domains but does show sites of punctuate concentration, some of which abut domains in normal cells. In DM1 muscle cells, the distribution of MBNL1 changes markedly. As shown in , we find that the MBNL1 nucleoplasmic signal is dramatically reduced in DM1 myotube nuclei and that MBNL1 is concentrated with mutant RNA in the very bright MRGs. These results corroborate evidence from and that MBNL1 becomes sequestered with the repeat RNA accumulations coincident with MBNL1-dependent changes in alternative splicing patterns in muscle.
Myotonic dystrophy type 2 (DM2) has phenotypic similarities to DM1 (), and, in DM2 cells, we also see that MBNL1 protein becomes sequestered with nuclear foci of mutant repeat RNA (unpublished data), which is consistent with the idea that overlapping clinical features in DM1 and DM2 may be caused by the common sequestration of MBNL1 (). However, in DM2, the repeat expansion is in a quadruplet repeat (CCTG) in the first intron of a zinc finger protein, znf9, on chromosome 3 ( ). Although DM2 cells contain repeat RNA foci, unlike DM1, the foci contain only intron sequences, and the intronic DM2 repeat does not impede znf9 mRNA export or translation (; ). Thus, because the repeat RNA foci in DM2 do not contain mRNA blocked from export, their relationship to SC-35 domains was of less interest. We did note that there was some association of the MBNL1/ intronic RNA foci around SC-35 domains, although to a much lesser extent (51%; = 604; unpublished data) than the DM1 MRGs (92%). Whether the aggregation with MBNL1 is partially responsible for the collection of MRGs around SC-35 domains is not known, but the important point is that MBNL1 sequestration is common to both DM1 and DM2, whereas blocked export of the mutant mRNA only occurs in DM1, possibly contributing to the unique clinical features of DM1.
Recent publications have suggested that MBNL1 and hnRNP H actually have roles in the formation of repeat RNA foci and RNA sequestration in DM1. have demonstrated that reduction of MBNL1 levels by RNAi disrupts repeat RNA foci in proliferative DM1 myoblasts. Similarly, have found that siRNA-mediated knockdown of hnRNP H restores the export of repeat-containing RNA out of the nucleus. These two proteins interact, as MBNL1 is responsible for hnRNP H association with repeat foci (). Thus, MBNL1 would appear to play a key role in the retention of mutant RNA in the nucleus by forming the MRG aggregates.
If the normal biogenesis and export of RNA are accomplished via SC-35 domains, these reports raise the possibility that knockdown of MBNL1 might alter the localization of repeat-containing RNA relative to SC-35 domains so that it is within domains, as seen for the normal RNA. To address this possibility, we performed siRNA-mediated knockdown of MBNL1 in differentiated DM1 muscle cell cultures. To determine the effectiveness of the RNAi in individual cells, we performed RNA in situ hybridization on siRNA and control coverslips using CAG repeat oligonucleotides and stained for MBNL1. Coverslips were scored for differentiated nuclei that had repeat RNA foci with and without MBNL1 (). As seen in the graph in , 27% of differentiated muscle nuclei treated with siRNA ( = 148) had CUG repeat RNA foci (MRGs) without substantial levels of MBNL1, whereas CUG RNA colocalized with MBNL1 foci in all of the control nuclei ( = 142). This in situ approach demonstrates effective transfection and siRNA depletion of MBNL1 expression in a substantial subset of differentiated cells.
Although reported that MBNL1 was necessary for mutant RNA focus integrity, we found intact RNA foci in cells in which MBNL1 expression was down-regulated by RNAi 5 d after transfection. We believe there is a difference in results because our analysis was performed on nonproliferative cultures under low serum differentiation conditions, whereas the cultures used by were dividing during the 5 d after transfection. We suggest that in their experiments, mutant RNA foci may have broken down during mitosis and not reformed afterward as a result of dilution and depletion of previously made MBNL1. Our evidence is in agreement with the statement by that the absence of MBNL1 is sufficient to prevent nucleation of the mutant RNA foci.
A triple label experiment in which CUG RNA, MBNL1, and SC-35 were detected was used to investigate changes in CUG RNA localization upon MBNL1 knockdown (). In nuclei in which MBNL1 expression was suppressed, we noted that there was commonly a single SC-35 domain that contained an irregular, rather diffuse accumulation of repeat RNA that was distinct from the round MRGs (; and see 3D visualization in Video 4, available at ). Based on scoring by two independent investigators, these signals were far more common in the cells in which MBNL1 was knocked down (63% of nuclei; = 39) than in control cells or in untransfected cells from the same experiment (8% of nuclei; = 37; ). These signals were reminiscent of diffuse RNA accumulations from the normal allele detected with a cDNA probe (; and ). Thus, we suspected that these signals did not arise from the preexisting RNA/protein aggregates (MRGs) but are new repeat-containing transcripts arising from the affected allele, which would not have any newly synthesized MBNL1 with which to complex as a result of the siRNA-mediated inhibition.
To address the possibility that the diffuse CUG RNA signals were new transcripts emanating from a gene locus, we performed another triple label experiment on MBNL1 knockdown cells. In this experiment, we detected SC-35 followed by hybridization for CUG RNA with an oligonucleotide probe and a DNA hybridization with a genomic probe. Analysis of this experiment provided numerous clear examples of gene loci abutting more diffuse CUG RNA (mutant RNA) signals within SC-35 domains (). In nuclei in which the diffuse RNA signal and two gene loci were visible, the dispersed RNA signal was always adjacent to one of the genes. These results demonstrate that depletion of MBNL1 by RNAi allows newly transcribed RNA containing expanded CUG repeats to enter the SC-35 domain abutting the gene locus.
Findings presented here further our understanding of nuclear structure and mRNA transport and, at the same time, provide insight into the cellular pathogenesis of DM1. We show that newly synthesized RNA accumulates within the interior of an SC-35–defined domain in normal muscle nuclei, having emanated from a gene positioned at the domain's edge. In contrast, the mutant transcripts in DMI detach from the gene but are not within the domain; rather, they accumulate in multiple granules that gather at the edge of domains. This change in domain association occurs in concert with mutant DMPK RNA retention in the nucleus. Although the transcripts do not appear to enter the SC-35 domain, they are nonetheless spliced. Down-regulation of MBNL1 changes the distribution of newly transcribed mutant RNA such that it is now found within the SC-35 domain.
Our studies examine the detailed relative distributions of gene/RNA/protein associated with normal versus mutant alleles but in static images that capture a window of time. A major advantage of this approach is that it provides information about the real endogenous gene and RNA, which are expressed in a native structural as well as physiological context. Although this does not provide direct visualization of molecules in live cells, our findings collectively provide evidence of the route occupied by at least a substantial fraction of these native transcripts. Because transcripts clearly emanate from the gene, for the sake of this Discussion, we consider the gene point A and will consider any accumulation of transcripts at a resolvable distance from the corresponding gene to have moved with respect to it.
Collectively, these findings suggest several important points regarding the relationship of the SC-35 domains to mRNA metabolism and transport. The consistent positioning of both homologous genes at the outer edge of an SC-35 domain adds to the body of evidence that specific active genes are organized relative to SC-35 domains, providing an example of an associated gene that is not particularly complex or highly expressed. We interpret the accumulation of transcripts within the SC-35 domain (to one side of the gene at the domain edge) to indicate a normal step in the path of mRNA for at least a substantial fraction of transcripts. The fact that mutant transcripts do not enter these domains further supports the idea that entry into the domains is a step that can be blocked by mutation. Because MRGs contain spliced mRNA, passage into the splicing factor–rich domain does not appear to be required for splicing, which is consistent with other evidence that most splicing occurs at the domain periphery (). These findings are consistent with the idea that postsplicing steps linked to export may occur within the domain. Blockage of mutant transcripts before entering SC-35 domains provides further evidence for the structural integrity of the SC-35 domain (with which normal and mutant alleles remain associated), which exists independently of the presence of mRNA within it. Consistent with this, the normal mRNA does not occupy the whole domain of splicing factors, as might be expected if the domain was merely factors bound to this pre-mRNA. Loss of RNA signals upon transcription inhibition demonstrates that the normal RNA accumulation in SC-35 domains does not represent RNA just trapped within the domain but is chased as would be expected for a transported mRNA. Remarkably, of the two mutant pre- mRNAs studied thus far, both have shown abnormal accumulations at or within the SC-35 domains and not at other sites such as the nuclear envelope or nucleolus. Whereas mutant COL1A1 RNA in osteogenesis imperfecta accumulates to abnormal levels within the domain, mutant transcripts accumulate outside the domain. This supports a model in which passage into and release from these domains are distinct steps in the normal path of some pre-mRNAs.
Although we demonstrate that many transcripts are not randomly dispersed upon initial release from the gene, these results should not be misinterpreted as excluding the possibility that export of these mRNAs may also involve multidirectional diffusion, which is consistent with other models of RNA trafficking (; ). These collective findings support a model of early steps in mRNA export and maturation as presented in .
This study provides evidence of a relationship between mRNA and SC-35–defined speckles that is disrupted when a CUG expansion in DM1 prevents normal mRNA export. Despite numerous studies describing the association of specific mRNAs with SC-35–defined speckles (; ; ; ), the idea that SC-35 domains contain mRNAs and/or play a role in mRNA export remains somewhat controversial (). Therefore, we will briefly address the two major pieces of evidence that generate this uncertainty. First, the belief that SC-35 speckles do not contain appreciable short-lived pre-mRNAs is based on early findings that tritiated uridine incorporation labels the speckles (interchromatin granule clusters) very little relative to the surrounding nucleoplasm (; ; for review see ). However, a study did find newly synthesized BrUTP-labeled RNA concentrated within these domains (). Most importantly, labeling methods such as [H]uridine or bromo-UTP are wholly nonspecific and do not necessarily reflect the distribution of pre-mRNA. These methods primarily label unknown heterogeneous nuclear RNA () and introns (most of pre-mRNA mass) that are rapidly removed and disperse during the labeling period (; for review see ). Evidence in support of this concern has been strengthened in recent years by findings that transcription of nongenic DNA is far more widespread throughout the genome than previously anticipated (; ). Furthermore, copious amounts of RNA throughout the nucleoplasm are detected by RNA hybridization using Cot-1 DNA, composed largely of repetitive elements such as Alu and long interspersed nuclear elements (), which recent findings suggest is largely nongenic transcription (; ; ). In short, uridine incorporation detects a great deal of RNA that is not mRNA and, as such, does not accurately represent the distribution of pre-mRNA in the nucleus.
Second, the demonstration that the poly(A) RNA signal often remains in SC-35 domains upon transcription inhibition (; ) is often interpreted to indicate that this poly(A) RNA is not mRNA but a putative long-lived, polyadenylated, structural RNA. However, transcriptional inhibition has complex effects on nuclear RNA distributions and transport that varies with the RNA. In a dramatic example of this, we recently identified an abundant noncoding polyadenylated nuclear RNA, NEAT1 (), that actually enters the SC-35 domains upon treatment with certain transcription inhibitors (unpublished data). This illustrates the difficulty of interpreting the nature of the poly(A) RNA in domains from transcription inhibition experiments. Our study demonstrates that at least some specific mRNAs, such as RNA, do leave the SC-35 domain upon transcription inhibition. In addition, recent studies of labeled RNAs introduced into live cells indicate that poly(A) RNA () and specific mRNAs () passage through domains and that export-ready mRNA is present in speckles (). Thus, although our study takes a different approach to study naturally occurring mutations of an endogenous RNA, our findings complement and substantially extend approaches that seek to understand the behavior of endogenous molecules by examining labeled, microinjected RNAs in live cells.
In addition to its relevance for fundamental nuclear structure, this study contributes insight into the cellular pathogenesis of DM1. We have identified a point in nuclear structure at which the paths of normal and mutant mRNA diverge. As this difference correlates with cytoplasmic mRNA export, it defines the intranuclear step at which the block in transport of many or all mutant transcripts likely occurs. We provide evidence that the formation of MRGs mediated by MBNL1 is responsible for blocking the mutant RNA from entry into the domain. Our analysis indicates that depletion of newly synthesized MBNL1 allows newly transcribed RNA carrying CUG repeat expansions to accumulate within SC-35 domains in a manner similar to that seen with normal RNA. MBNL1 may act in concert with other factors, such as hnRNP H, to bind CUG repeat RNA and form the RNA/protein aggregates. Thus, both the nuclear retention of mutant DMPK RNA as well as sequestration of specific nuclear factors like MBNL1 and hnRNP H are likely involved in the complex DM1 phenotype.
The normal muscle cultures used were human skeletal muscle myoblasts (HSMMs; Clonetics) and 50Mb. Myoblast strain 50Mb (normal myoblast preparations flow sorted to substantially remove contaminating fibroblasts) was provided by H. Blau (Stanford University, Palo Alto, CA; ). It was obtained from the vastus lateralis muscle of a 10-yr-old male with no known muscle pathologies. For propagation, cells were cultured at subconfluent density in serum-rich medium with medium changes every other day. To induce muscle differentiation, cultures were grown to near confluence and maintained in low serum medium without further medium changes until the appearance of myotubes. Propagation medium was Ham's F-10 supplemented with 20% FCS, 1% vol/vol chick embryo extract (60 Å ultrafiltrate; Invitrogen), 1 mM insulin, and 1 mM dexamethasone (Sigma-Aldrich), whereas differentiation medium contained DME low and 2% horse serum. Both contained 200 U/ml penicillin and 200 mg/ml streptomycin. HSMMs were grown in SkGM BulletKit media (Cambrex) with an additional 1.5% FBS. The differentiation media was DME–F-12 with 2% horse serum added when the cells reached ∼60% confluence.
Homozygous DM1 myoblasts (DM(SW)) were derived from a severely affected homozygote (provided by L. Timchenko, Baylor College of Medicine, Houston, TX; )). In our hands, this cell line ceased to differentiate after a few passages. Heterozygous DM1 myoblast cultures F1D and myoblast C were obtained from C. Thornton (Wellstone Muscular Dystrophy Cooperative Research Center, Rochester, NY). These myoblasts were grown in Ham's F-10 with 0.12% NaHCO, 15% FBS, 5% defined supplemental calf serum, and 2 mM -glutamine and penicillin/streptomycin. Fusion medium was described in the previous paragraph. DM2 fibroblasts were provided by L. Ranum (University of Minnesota, Minneapolis, MN). DM2 fibroblasts were grown in DME with 10% FBS, penicillin/streptomycin, and -glutamine.
RNA was detected using a 2.3-kb cDNA clone pRMK (). A genomic clone, cosmid F18894 (L08835) overlapping the gene, which was obtained from the Human Genome Center (Lawrence Livermore National Laboratory, Livermore, CA), was used for DNA detection. The 2-kb intron 9 probe was generated by PCR amplification from the F18894 clone using primers GAGTTGCAGGATCAGTCTTGGA and TGCTGATTCTCTGGTGGAGAAC. gene and RNA detection were performed using PCR-generated probes encompassing exons 2–5 of the znf9 gene. Probes were generated by PCR amplification of a bacterial artificial chromosome (Rp11-814L21; BACPAC Resource Center, Children's Hospital Oakland Research Institute, Oakland, CA) using primers AAACCTTTGCCATCACCATC and GCCTTGCAAATTGTCTGGAT. Biotin or fluorescein end-labeled oligonucleotide probes (CAGCAGCAGCAGCAGCAG and CAGGCAGGCAGGCAGGCAGGCAGGCAGG) were used to detect DM1 (CUG) and DM2 (CCUG) repeat RNAs, respectively. All probes except direct-labeled oligonucleotides were labeled using nick translation as described previously (; ).
SC-35 domains were detected with an antibody to the spliceosome assembly factor SC-35 (Sigma-Aldrich; ) or to an antibody to the splicing coactivator SRm300 (provided by B. Blencowe, University of Toronto, Toronto, Canada; ). MBNL1 was detected with monoclonal antibodies 3B10 and A2764, which were obtained from C. Thornton. An antibody to hnRNP H was obtained from D. Black (Howard Hughes Medical Institute/University of California, Los Angeles, Los Angeles, CA).
The methods used in this study, including procedures for nonisotopic probe preparation and FISH, have been published in detail previously (; ). Two-color detection of RNA and genes simultaneously was performed by sequential hybridization and detection of RNA followed by fixation, NaOH treatment, and DNA hybridization (). In some experiments, before or after hybridization, samples were stained with antibodies as described previously (, ).
RNAi was performed by methods adapted from . Double-stranded siRNA to MBNL1 (5′-CACUGGAAGUAUGUAGAGAdTdT-3′) was obtained from Dharmacon. Fresh cultures of myoblasts from DM1 patients (myoblast C) were cultured and differentiated as described in the Cell culture section. 5-d differentiated myoblast cultures on coverslips in six-well plates were rinsed twice with Opti-MEM (Invitrogen). For each coverslip, 5 μl Oligofectamine (Invitrogen) and 375 pmol MBNL1 siRNA in a 1-ml total volume of Opti-MEM were preincubated for 20 min at room temperature and added to the cells. Controls had no siRNA. Transfection was incubated at 37°C overnight and replaced with differentiation media (DME + 5% horse serum). Cells were extracted and fixed at 5 d after transfection.
Images presented in this study are either single plane or extended focus images from z stacks (step size = 100 nm) and were acquired using an Axiovert 200 microscope or an Axiophot microscope (Carl Zeiss MicroImaging, Inc.) equipped with a 100× NA 1.4 planApo objective using Axiovision (Carl Zeiss MicroImaging, Inc.) or MetaMorph (Universal Imaging Corp.) imaging software. The multiband pass dichroic and emission filter sets (model 83000; Chroma Technology Corp.) were used with excitation filters set up in a wheel to prevent optical shift. Images were captured on an Orca-ER camera (Hamamatsu) or a CCD camera (200 series; Photometrics). Z stacks were further processed using constrained iterative deconvolution in Axiovision 4.1 (Carl Zeiss MicroImaging, Inc.) and displayed as extended focus projections in some cases. Rendered images are maximum value projections. Results shown are from multiple experiments and were scored by several investigators.
Video 1 demonstrates the colocalization of normal DMPK RNA and an SC-35 domain in 3D. Video 2 is a 3D representation showing that mutant DMPK RNA forms foci that are predominantly at the edge of SC-35 domains in DM1 muscle cell nuclei. Video 3 demonstrates in 3D that both DMPK gene loci associate with SC-35 domains in heterozygous DM1 muscle but only one locus has visible RNA accumulating in the adjacent domain. Video 4 is a 3D representation showing that upon siRNA-mediated knockdown of MBNL1, CUG RNA (representing mutant DMPK RNA) localizes within an SC-35 domain adjacent to a DMPK gene locus. Online supplemental material is available at . |
Subcellular localization of mRNAs can locally control the protein composition of distinct regions within the cell. Neurons provide an ideal system for understanding how subcellular mRNA localization is regulated. The widely separated cytoplasmic extents of neuronal dendrites, axons, and cell body allow one to ask how local extracellular stimuli may alter populations of localized mRNAs and ultimately modulate the local protein composition of that subcellular domain. Much recent effort has focused on how neuronal RNA trafficking and localized translation are regulated (; ; ). Transport of mRNAs and translational machinery into axons along with subsequent local protein synthesis is needed to initiate growth responses and for growing neurons to respond to environmental stimuli (, ; ; ; , ; ; ; ; ).
Despite increasing knowledge of stimuli that can trigger axonal protein synthesis, knowledge of the specificity of these autonomous responses has been quite limited (). Injury of peripheral axons triggers localized translation of importin β and vimentin mRNAs, and these nascent protein products generate a retrograde signaling complex (; ). In cultures of developing neurons, the guidance cue semaphorin 3A (Sema3A) activates the localized translation of RhoA mRNA (), and neurotrophins increase the localized synthesis of axonal β-actin (). A study aimed at determining the scope of locally synthesized proteins argues that axons have the potential to synthesize many different proteins (), raising the questions of if and how the expression of these proteins may be regulated in the axonal compartment.
In addition to translational control, regulating the delivery of mRNAs to subcellular regions can modulate localized protein synthesis by altering which mRNAs are locally available for translation. Evidence for this is seen in cultures of developing cortical neurons in which bath application of neurotrophins can increase the delivery of β-actin mRNA to the axonal growth cone (). In the present study, we show that the levels of individual axonal mRNAs are differentially regulated by the local stimulation of axons with growth-promoting and growth-inhibiting stimuli. Quantitative analyses of axonal mRNAs showed that nerve growth factor (NGF), brain-derived neurotrophic factor (BDNF), neurotrophin-3 (NT3), myelin-associated glycoprotein (MAG), and Sema3A can specifically increase or decrease levels of individual transcripts. These alterations in axonal mRNA levels were accompanied by opposite changes in the cell body mRNA levels, suggesting that ligand-dependent alterations in anterograde transport rates exist. With in situ hybridization and heterologous expression of a chimeric mRNA containing the rat β-actin mRNA localization element, the alterations in axonal mRNA levels seen by quantitative analyses correspond to the relative enrichment or depletion of individual mRNAs from axonal regions directly adjacent to ligand sources. These findings argue that diverse extracellular signals bidirectionally regulate the transport of numerous mRNAs within axons to influence local protein synthesis.
We previously used a proteomics approach to identify locally synthesized proteins from cultures of adult rat dorsal root ganglion (DRG) neurons (). This approach was limited to the most abundant axonally synthesized proteins. Because we were able to view only a fraction of the axonal mRNAs using this method, we reasoned that a more global assessment of axonal mRNA content would be needed to test for the specific regulation of axonal mRNA localization. For this, axonal RNA was isolated from dissociated cultures of DRG neurons after 20–22 h in vitro as previously described (). The L4-5 DRGs were conditioned by in vivo sciatic nerve crush 7 d before culture. These injury-conditioned sensory neurons show rapid transcription-independent, translation-dependent process outgrowth over 24 h in culture (; ). The purity of axonal preparations was verified by the absence of γ-actin and microtubule-associated protein 2 mRNAs (Fig. S1, available at ; ; ). Amplified cDNAs prepared from axonal RNAs were used to hybridize to Atlas cDNA arrays containing ∼4,000 rat cDNAs. Localized mRNAs detected from hybridizations of four separate axonal preparations are summarized in Table S1. According to these data, the injury-conditioned DRG axons have the capacity to synthesize >200 different proteins, including transmembrane proteins (e.g., Kv3.1a and HCN4) and components of the translational machinery (e.g., ribosomal proteins) that were not detected in our previous proteomics screen ().
Guidance cues that invoke axon turning or collapse have been shown to regulate axonal protein synthesis (, ; ; ). To determine whether axonal mRNA localization is specifically regulated in the DRG neurons, we asked whether the local application of growth-promoting or growth-inhibiting stimuli to DRG axons can alter the localization of individual mRNAs. A panel of 50 axonal mRNAs from the aforementioned array data and a previous proteomics study (), which broadly represents axonal mRNAs encoding different protein types, was used for these analyses. Neurotrophins were chosen for the growth-promoting ligands because their TrkA, TrkB, and TrkC receptors are expressed by all of the DRG neurons (). Sema3A and MAG were used as growth-inhibitory stimuli because these ligands induce axonal retraction or repulsion in rat DRG neurons (; ). The axonal compartment was exposed to ligands immobilized on microparticles for 4 h in the presence of the RNA polymerase II inhibitor 5,6-dichlorobenzimidazole riboside (DRB). Microparticles with immobilized BSA, AP, or human IgG Fc domain were used as controls for the neurotrophin, Sema3A, and MAG, respectively. Axonal mRNA levels for each of the 50 transcripts were determined using axonal RNA isolates for reverse transcription followed by quantitative PCR (qPCR). The qPCR results are detailed in . These data show both ligand and transcript specificity for the regulation of axonal mRNA levels.
summarizes changes in axonal levels for a few mRNAs to illustrate the specificity of these ligand-dependent responses. Note that each stimulus can increase or decrease axonal levels of an individual mRNA. For the neurotrophins, NGF and BDNF showed an overall similar regulation of axonal mRNA levels except for 43-kD growth-associated protein (GAP-43) mRNA, which was uniquely affected by NGF (). NT3 modulation of axonal mRNA levels appeared distinct from NGF and BDNF. For example, axonal levels of the mRNA encoding the Kv3.1a potassium channel was decreased by NGF and BDNF, but NT3 increased axonal levels of Kv3.1a mRNA (). Similar to the neurotrophins, MAG and Sema3A increased and decreased the localization of individual axonal mRNAs (). With the exception of vimentin mRNA, MAG and Sema3A altered axonal mRNA levels distinctly from the neurotrophins ( and ). For example, Sema3A and MAG decreased axonal β-actin mRNA levels, whereas the neurotrophins consistently increased axonal β-actin mRNA levels. Several transcripts that were not responsive to the neurotrophins showed altered axonal levels with MAG and/or Sema3A (). Interestingly, many of the mRNAs tested showed no significant change (P > 0.01) in response to the 4-h ligand stimulation (). This result, combined with the differences in axonal mRNA levels seen among the individual neurotrophins, indicates that the regulation of mRNA localization is highly specific at the level of individual transcripts and for ligands that activate similar intracellular signaling pathways.
Over the 4-h period used to stimulate the aforementioned axons, DRB decreased new RNA synthesis >90% based on the incorporation of α-[P]UTP into RNA isolated from cultures of injury-conditioned DRGs (Fig. S2 A, available at ). Although we cannot exclude the possibility that alterations in axonal mRNA stability contributed to the changes in axonal mRNA levels, these metabolic labeling experiments suggest that chemotropic agents can alter the localization of existing mRNA populations. If this is the case, any transport-dependent changes in axonal mRNA levels should be accompanied by an opposite change in the levels of that transcript in the cell body. Although our previous analysis of cytoskeletal mRNAs did not show depletion of cell body levels after NGF-induced increases in mRNA levels in the axons, the high expression of cytoskeletal mRNAs in the DRG cultures could complicate the detection of cell body depletion during this and previous short-term experiments (). To more rigorously test for ligand-dependent alterations in axonal mRNA content, we extended treatment duration from 4 to 12 h and evaluated axonal levels of a subset of mRNAs that showed increased or decreased axonal levels with NGF (). qPCR analyses of cell body and axonal RNA content showed that the decrease in axonal levels of Ca1.2 and Kv3.1a mRNAs seen after NGF exposure resulted in a statistically significant increase (P ≤ 0.01) in the cell body content of these mRNAs. In contrast, increased axonal levels of peripherin, RP-L22, RP-S17, and thymosin β4 mRNAs resulted in decreased cell body mRNA content (). These data suggest that the sensory neurons draw on a pool of preexisting mRNAs in the cell body to alter the delivery of individual mRNAs into the axons in response to ligand stimulation. In addition, these data indicate that extracellular signals can lead to a new steady state in the distribution of specific mRNAs between the cell body and axon that is likely independent of transcription.
RNAs are transported on microfilaments in fibroblasts, whereas most other cellular systems, including neurons, have been shown to use microtubules for long-range transport of mRNAs (; ; ; ; ; ). To determine whether the neurotrophin-dependent changes in axonal mRNA levels in require intact cytoskeleton, dissociated cultures were pretreated with cytochalasin D or colchicine to disrupt microfilaments or microtubules, respectively. For this analysis, we examined a subset of the transcripts shown in that was comprised of five mRNAs with increased transport with NGF, five mRNAs with decreased transport with NGF, and five mRNAs with no response to NGF. The disruption of microfilaments decreased the NGF-dependent alterations in axonal mRNA levels but not to the extent that was seen with the microtubule-depolymerizing agent (). Although colchicine treatment modestly reduced overall axonal mRNA levels, the NGF-induced changes in axonal levels were almost completely abolished by colchicine (). Specificity for the effect of colchicine on NGF-dependent RNA localization is indicated by the lack of any effect of colchicine or cytochalasin D on the nonresponding mRNAs (). These observations further suggest that neurotrophins regulate axonal mRNA levels by altering the rates of transport of mRNAs from the cell body.
A change in delivery of mRNAs into axons with peripheral stimulation could require that instructive signals from the axons be retrogradely transmitted to the neuronal cell body (or nucleus). To determine whether the aforementioned alterations for mRNA levels require stimulus localized to the axonal compartment, we compared axonal mRNA levels after the application of immobilized NGF to axons versus bath-applied NGF (i.e., soluble), which would simultaneously stimulate axonal and cell body compartments. Approximately equivalent levels of soluble versus immobilized ligand were applied based on our previous assessments of TrkA phosphorylation by the NGF microparticles (). Roughly twofold more β-actin mRNA accumulated in the axons with local stimulation than with bath-applied NGF (). Peripherin and Kv3.1a mRNAs showed no response to bath-applied neurotrophins, whereas axonal HCN4 mRNA appeared more sensitive to bath-applied NGF than localized ligand sources. With bath application of NGF, calreticulin, and HSP70 mRNAs, two transcripts that did not respond to localized NGF showed alterations in axonal mRNA levels, indicating that these transcripts uniquely respond to the soluble ligand (). Collectively, these data indicate that the stimulus derived from localized neurotrophin sources is qualitatively and quantitatively different from the nonlocalized stimulus of the soluble ligand.
To determine the role of Trk receptors in neurotrophin-dependent axonal mRNA localization, cultures were pretreated with K252A at levels that specifically inhibit Trk tyrosine kinase activity (). For the mRNAs that increased with axonal NGF stimulation, K252A pretreated cultures showed axonal levels that were nearly indistinguishable from neurons exposed to control microparticles (). Axonal levels of the five nonresponding transcripts were not affected by K252A treatment (). Surprisingly, mRNAs with NGF-dependent decreases in axonal content showed a complete reversal, exhibiting increased axonal levels with Trk inhibition and peripheral NGF stimuli (). Both Kv3.1a and HCN4 mRNAs, for which levels were decreased by ∼1.5-fold in the axons treated with NGF, showed a one- to twofold increase in axonal levels in cultures treated with K252A (). This indicates that local sources of NGF can signal through Trk receptors to bidirectionally modulate the axonal localization of individual mRNAs.
The neurotrophin-dependent activation of phosphatidyl inositol-3 kinase (PI3K) and Ras–MAPK pathways contribute to the local trophic and tropic effects of NGF and other neurotrophins (). We used pretreatment with MEK1 and PI3K inhibitors (PD98059 and LY29004, respectively) to test whether these signaling pathways play a role in the neurotrophin-dependent regulation of axonal mRNA localization. The five mRNAs that previously did not respond to NGF remained unaffected overall by PD98059 and LY29004, indicating that basal activity of PI3K and MEK1 did not contribute to their axonal localization (). For most of the regulated mRNAs, NGF's effects on their axonal levels were attenuated by inhibition of the MAPK pathway with PD98059 (). However, two transcripts behaved differently. The NGF-dependent attenuation of Kv3.1a mRNA's axonal localization required PI3K activity but was unaffected by the MEK1 inhibitor (). The increased axonal localization of β-actin mRNA seen with NGF was attenuated by the inhibition of either PI3K or MEK1 (). All other mRNAs that localized in response to NGF (e.g., vimentin and peripherin) required MEK1 but not PI3K (). Thus, a single ligand can uniquely regulate the axonal localization of individual mRNAs using different downstream signaling pathways.
provides no information on where mRNAs are localized within the axon. Because axonal β-actin mRNA showed divergent regulation with growth-promoting versus growth-inhibiting stimuli and appeared more sensitive to localized ligand sources, we used the well-characterized localization elements of β-actin mRNA to drive the axonal localization of heterologous mRNAs encoding a reporter protein. For this, the 3′ untranslated region (UTR) of enhanced GFP (eGFP; ) was replaced with 3′ UTRs from the rat β-actin or γ-actin mRNAs (eGFPβ-actin and eGFPγ-actin, respectively). β-actin mRNA 3′ UTR contains a zipcode element that directs the transport of this mRNA in fibroblasts, myocytes, and neurons; γ-actin mRNA does not contain any similar element, and the transcript is retained in the perinuclear region (; ; ). To facilitate expression in the adult DRG neurons, we generated adenoviruses (AVs) that express these reporter cDNAs (AV-eGFPγ-actin and AV-eGFPβ-actin). In injury-conditioned DRG cultures infected with AV-eGFPγ-actin, reporter fluorescence accumulated in the cell body and did not extend into the axonal compartment (Fig. S3, available at ). In contrast, GFP signal was seen in the cell body and at foci along the axonal processes, including the growth cones in cultures infected with AV-eGFP β-actin (Fig. S3). The myr domain of this eGFP construct likely restricts diffusion of the eGFP product in the axonal compartment, providing a measure of localized protein synthesis as previously reported (). RT-PCR from axonal RNA also confirmed the differential localization of eGFP β-actin versus eGFPγ-actin mRNAs in the DRG cultures (unpublished data). Thus, similar to axons of developing cortical neurons (; ), the 3′ UTR of β-actin mRNA is sufficient for axonal localization in adult rat sensory neurons.
Because the β-actin 3′ UTR appeared to direct axonal localization of eGFP mRNA in the DRG cultures, we next considered whether axonal localization of eGFPβ-actin is modulated by growth-promoting and growth-inhibiting stimuli. Analyses of axons exposed to NGF microparticles showed that GFP signals under the control of β-actin 3′ UTR accumulated directly adjacent to the ligand source during a 50-min exposure ( and Video 1, available at ). Axons typically showed a sprouting or turning response upon contact with NGF microparticles (Fig. S4). BSA microparticles did not affect the intensity of GFP signals or the directionality of axonal growth (Fig. S4). At the completion of a 50-min exposure, axonal GFP fluorescence was significantly greater adjacent to NGF when compared with BSA microparticles (5.29 ± 0.07-fold for NGF vs. BSA; P ≤ 0.001). Although there is an inherent experimental delay in when video sequences can be initiated with this approach (i.e., as the particles settle onto the coverslip), there was also a consistent increase in GFP fluorescence adjacent to NGF microparticles over the course of the imaging sequences (1.6 ± 0.09-fold for t = 50 vs. t = 0 min; P ≤ 0.001). Similar to the effects of kinase inhibitors upon the endogenous β-actin mRNA shown in , the inhibition of Trk or downstream MEK1 or PI3K activity prevented any GFP accumulation adjacent to the NGF source ( and S4 and Video 2). Thus, the dynamic redistribution of mRNA likely directly impacts the translation and accumulation of protein.
Microtubule-depolymerizing agents were used to determine whether an increase in GFPβ-actin signals was the result of RNA accumulation at NGF sources. No NGF-dependent accumulation of GFP was seen in cultures exposed to colchicine ( and Video 3, available at ). The continued increase in GFP signals in the growth cone (distal to NGF source) indicates that the colchicine treatment did not affect translation of the eGFPβ-actin mRNA that had already accumulated in the growth cone. Because colchicine completely blocked NGF-dependent increases in axonal β-actin mRNA levels in the qPCR experiments (), the majority of the increased GFP signals shown in can be attributed to the subcellular localization of eGFPβ-actin mRNA rather than to the translational activation of any eGFPβ-actin mRNA already residing within the axon.
Because microtubule depolymerization would alter both retrograde and anterograde transport, we tested whether the NGF effect on eGFPβ-actin required retrograde signaling. For this, DRG cultures were transfected with dynein heavy chain (Dync1h1) siRNAs (). The transfected cultures showed a decrease of Dync1h1 protein, and transfected neurons showed a selective depletion of retrograde but not anterograde transport (Fig. S2 B and Video 4, available at ). These siRNA-transfected neurons also showed no significant alteration (P > 0.05) in eGFPβ-actin signals over 50 min of exposure to NGF microparticles (0.91 ± 0.11-fold for NGF vs. BSA; ). DRG cultures treated with erythro-9-(2-hydroxy-3-nonyl)adenine hydrochloride (EHNA), which has been shown to inhibit dynein ATPase activity (; ), had similar depletion of retrograde transport (Video 5) and showed no significant change (P > 0.05) in axonal GFP fluorescence in response to NGF (0.95 ± 0.10-fold for NGF vs. BSA). Thus, the NGF-dependent increase in localization of the axonal reporter mRNA appears to require retrograde transport.
To determine how growth-inhibiting stimuli can deplete axonal β-actin mRNA levels, we examined the effect of immobilized MAG on axonal GFP signals in AV-eGFPβ-actin–infected DRG cultures. GFP signals were relatively excluded from axonal regions adjacent to MAG sources ( and Video 6, available at ) and often caused the axon to turn away from the MAG source (Fig. S4). After 50-min exposure to microparticles, GFP signals were significantly decreased with MAG-Fc microparticles when compared with the control IgG Fc microparticles (0.68 ± 0.1-fold for MAG-Fc vs. IgG Fc; P ≤ 0.001). The eGFPβ-actin translation product was relatively depleted from the vicinity of MAG-Fc (Fig. S4); this focal exclusion of GFP adjacent to MAG sources is likely why MAG caused only a small reduction in β-actin mRNA by qPCR. In contrast to NGF's effects on eGFPβ-actin signals, a statistically significant attenuation of axonal GFP signals adjacent to MAG stimuli was still seen when retrograde transport was inhibited with Dync1h1 siRNA (0.75 ± 0.07-fold for MAG vs. IgG Fc; P ≤ 0.01; ) or EHNA (0.73 ± 0.12-fold for MAG vs. IgG Fc; P ≤ 0.01; not depicted). Thus, the MAG-dependent depletion of GFP signals appeared to be a local effect adjacent to axonal stimuli.
The growth-inhibitory effects of MAG can be overcome by elevating neuronal cAMP levels (, ; ). To determine whether the MAG-dependent changes in the localized production of eGFPβ-actin could be altered by cAMP, cultures were treated with a cell-permeable nonhydrolyzable cAMP analogue (dibutyral cAMP [db-cAMP]) before exposure of axons to immobilized MAG. In db-cAMP–treated cultures, eGFPβ-actin signals accumulated directly adjacent to the MAG microparticles ( and Video 7, available at ). The GFP signals showed a significant increase adjacent to MAG-Fc microparticles after cAMP treatment when compared with IgG Fc control (4.47 ± 0.06-fold for MAG-Fc + db-cAMP vs. IgG Fc + db-cAMP; P ≤ 0.001), and, similar to the NGF response, axons turned acutely after contact with the MAG microparticles (Fig. S4). Pretreatment with db-cAMP did not alter the response seen by contact with NGF or control microparticles (unpublished data).
FISH was used to determine whether focal stimulation of axons with growth-promoting versus growth-inhibiting ligands could similarly alter the local accumulation of endogenous mRNAs in these axons. For this, injury-conditioned DRGs were plated onto laminin-coated coverslips with adherent neurotrophin, MAG, and Sema3A microparticles. After 18 h in culture, cells were fixed and analyzed by FISH for mRNA and immunofluorescence for neurofilament. Differences in mRNA signal intensity adjacent to the immobilized agent were specific both at the level of the transcript and ligand. Consistent with the transfected eGFPβ-actin, endogenous β-actin mRNA was enriched adjacent to neurotrophin sources and decreased adjacent to MAG and Sema3A sources (). The disparity between the responses elicited by NGF and NT3 seen in the qPCR experiments for peripherin and Kv3.1a mRNAs was also evident in the localization of these transcripts along axons exposed to immobilized NGF and NT3. Peripherin mRNA increased at the site of NGF stimuli but decreased adjacent to NT3 stimuli (). The opposite pattern was seen with Kv3.1a mRNA: FISH signals for Kv3.1a were decreased adjacent to NGF stimuli but increased adjacent to NT3 stimuli (). Together, these findings show that the exquisite ligand specificity for mRNA localization seen in the qPCR experiments corresponded to localized accumulation or depletion of individual transcripts directly at the site of ligand exposure.
Targeting mRNAs and translational machinery to subcellular loci is being increasingly recognized as a means to locally control the protein composition of cellular domains. Modulating the levels of individual mRNAs in different subcellular regions could alter the populations of proteins generated in these regions by locally altering the availability of templates for the local translational machinery. Because of the distances separating neuronal processes from their cell body, neurons are an appealing cellular model for testing how local stimuli alter the trafficking of mRNAs into subcellular regions. Both the dendritic and axonal compartments of neurons have been used to study the localization of single transcripts, but the specificity of such changes has not been addressed for a broad population of mRNAs. Localized protein synthesis has been shown to provide a means for axons to autonomously respond to guidance cues and injury (). Studies showing altered translation in axons have either not provided any analyses of which proteins are locally generated or have focused on single proteins (; ; ; ; ; ). In the present study, we show that both growth-promoting and growth-inhibiting stimuli can differentially localize mRNAs to points of ligand stimulation.
The specificity in these mRNA localization responses is exhibited at multiple levels, including differential responses to growth-promoting versus growth-inhibiting ligands, differential responses to individual growth-promoting ligands, and even differences within downstream signaling pathways for responses to an individual ligand. The microtubule-dependent changes in axonal mRNA levels were reflected by a reciprocal decrease or increase in cell body mRNA content, suggesting a ligand-dependent alteration in delivery of mRNAs from the cell body. However, our data do not completely exclude the possibility that localized mRNA stability may also contribute to axonal mRNA content. Likewise, it is possible that both mRNA transport and stability are affected by distinct signaling pathways. Regardless of the specific mechanism, our findings indicate that extracellular signals can change the local concentrations of axonal mRNAs, which likely provides unique specificity to the localized protein synthetic responses.
NGF (Harlan), BDNF, NT3 (Alomone Labs), MAG-Fc (R&D Systems), and Sema3A-AP () were covalently coupled to 15-μm-diameter polystyrene microparticles according to the manufacturer's instructions (Polysciences). The following control proteins were also immobilized onto polystyrene microparticles or particles: BSA (Sigma-Aldrich) for NGF, BDNF, and NT3; human IgG-Fc (R&D Systems) for MAG-Fc; and AP for Sema3A-AP (provided by Y. Goshima, Yokohama University, Yokohama, Japan; ). Efficiency of absorption was determined by Bradford assay for unbound protein. To determine the mechanisms involved in modulation of axonal mRNA transport, DRG cultures were treated with the following pharmacological agents 30 min before the addition of immobilized ligands: 200 nM K252A (Calbiochem), 50 μM PD98059 (Biomol), 50 μM LY294002 (Biomol), 10 μM Lavendustin (Biomol), 10 μM Olomucine (Biomol), 1 mM db-cAMP (Sigma-Aldrich), 10 μg/ml cycloheximide (Sigma-Aldrich), 10 μM cytochalasin D (Sigma-Aldrich), and 1 μg/ml colchicine (Sigma-Aldrich). 50 nM EHNA (Sigma-Aldrich) was added to cultures 3 h before use.
All animal surgeries and euthanasia were performed according to institutional Animal Care and Use Committee guidelines under approved protocols. Primary DRG cultures were prepared from Sprague Dawley rats that had been injury conditioned 7 d before by sciatic nerve crush at midthigh level (). Dissociated cultures were prepared from L4-L5 DRGs as previously described (). Cultures were plated at moderate density on membrane inserts (for axonal isolation, see next paragraph) or at low density on coverslips (for live cell imaging and FISH analyses, see respective sections below).
The culture method for isolating DRG axons from cell bodies and nonneuronal cells has been previously described (; ). In brief, dissociated DRGs were plated into tissue culture inserts containing porous membranes (8-μm-diameter pores; BD Falcon), which were coated with poly--lysine (Sigma-Aldrich) and laminin (Upstate Biotechnology). Axons were isolated after 16–20 h in culture by scraping away the cellular content from the upper or lower membrane surfaces (yielding axonal or cell body preparations, respectively). The purity of the axonal preparations was tested by RT-PCR for microtubule-associated protein 2, γ-actin, and β-actin mRNAs.
Axonal RNA was isolated as described in the previous section, and the purity was confirmed by RT-PCR for β-actin, γ-actin, and microtubule-associated protein 2 (Fig. S1 A). 200 ng was used as a template for RT-PCR using the SMART PCR cDNA Synthesis kit (CLONTECH Laboratories, Inc.) to generate full-length double-stranded cDNA. Aliquots of the amplification were removed from the PCR every third cycle from 12–30 cycles and used for Southern blotting. Southern blots were probed with P-labeled β-actin cDNA to test for linearity (Fig. S1 B). The aliquot below where linearity was lost was used to generate a P-labeled probe cDNA using the Advantage 2 PCR System (CLONTECH Laboratories, Inc.) and hybridized to Atlas Plastic Rat 4K Microarrays (CLONTECH Laboratories, Inc.). Hybridization signals were detected by phosphorimaging, and results were analyzed using Atlas Image 2.7 software (CLONTECH Laboratories, Inc.). For comparison between arrays, individual hits were normalized across the individual array and assigned a relative intensity.
The culture method for treatment of intact DRG axons has been previously described (). In brief, DRGs were cultured on a porous membrane in the presence of 80 μM of the RNA synthesis inhibitor DRB throughout the culture period (Sigma-Aldrich). After 16–20 h in culture, the axonal compartments were selectively exposed to the protein-coupled microparticles by placing the inserts into dishes with shallower wells in which the bottom surface of the insert directly contacted coated microparticles along the bottom of the wells. Microparticles of 15 μm diameter were used for these stimulations to restrict any passage through the 8-μm pores of the membrane, which limited stimulation to the axonal compartment. After 4 h of treatment, membranes were rinsed in PBS, and the axonal compartment was isolated as described in Cell culture and axonal isolations. RNA was extracted from the fractionated cultures using the RNAqueous Micro kit (Ambion) and quantified using fluorometry with the RiboGreen reagent (Invitrogen). To normalize the axonal mass between samples, flow through from the affinity-based RNA isolation was used to measure the protein content of the axonal samples by fluorometry with the NanoOrange reagent (Invitrogen). All axonal RNA samples were normalized to protein content before RT-PCR analyses ().
To visualize the localized effects of immobilized ligand exposure to axons, 4.5-μm microparticles were used. For live cell imaging (see Live cell imaging section below), fluorescent microparticles were added to cultures growing on poly--lysine/laminin chambered coverslips. For the FISH studies (see FISH section below), 7-d injury-conditioned DRGs were cultured overnight in the presence of carboxylated polystyrene microparticles with immobilized neurotrophins, MAG, or Sema3A. In these experiments, the ligand-immobilized microparticles were coupled directly to the laminin surface, which prevented them from being washed away in subsequent FISH/immunofluorescence steps. For this, ligands were coupled to microparticles overnight according to the manufacturer's protocol (Polysciences). Microparticles were then added to the coated coverslips for 4 h. Unreacted sites were blocked with ethanolamine, and coverslips were washed and used for standard DRG culture.
For analyses of axonal transcripts, normalized axonal RNAs (∼50 ng each) were used as a template for reverse transcription using the iScript RT kit (Bio-Rad Laboratories). The reverse transcription reactions were diluted 10-fold, and the purity of each axonal preparation was assessed by RT-PCR for γ-actin and microtubule-associated protein 2 mRNAs, which are expressed at high levels in rat DRG cultures but are excluded from the axonal compartment as previously described (). Validated axonal RNA preparations were then used for reverse transcription qPCR. In brief, reverse-transcribed axonal RNA was amplified using the Prism 7900HT sequence detection system (Applied Biosystems) with 2× SybrGreen Master Mix (QIAGEN) according to the manufacturer's standard cycling parameters. Rat brain RNA was used as a template to generate standard curves for all primer pairs. A robotic system (Biomek 2000; Beckman Coulter) was used to standardize the pipetting of samples and reagents into 384-well plates for qPCR. All samples were assayed in quadruplicate from at least three independent experiments each. In addition to controlling for axonal number based on protein content, the relative levels of each transcript were normalized to the 12S mitochondrial ribosomal RNA control by the comparative threshold method (C) to provide an internal control for reverse transcription efficiency and axonal content. RNA values are expressed relative to control (BSA for NGF, BDNF, and NT3 treatments; Fc for MAG-Fc treatment; and AP for Sema3A treatment).
Injury-conditioned DRGs were cultured overnight ± 80 μM DRB. Culture medium was then supplemented with 125 mCi/ml α-[P]UTP (GE Healthcare). After a 4-h labeling period, total RNA was extracted and quantified by fluorometry as described in Localized treatment of axons. The specific activity was determined by liquid scintillation counting. These labeled RNA samples (2 μg each) were electrophoresed in a 6% acrylamide gel. After electrophoresis, the gel was stained with ethidium bromide, imaged under UV to verify RNA loading and integrity, dried, and used for autoradiography.
Chimeric reporter cDNA constructs were generated by replacing the 3′ UTR of the αCamKII-eGFP constructs (provided by E. Schuman, California Institute of Technology, Pasadena, CA; ) with that of the rat γ-actin and β-actin mRNAs. cDNA encoding 3′ UTRs of these rat mRNAs were isolated by RT-PCR from rat brain RNA template using the following primers engineered with Not1 and Xh1 restriction sites (actin components are underlined): sense β-actin (5′-AAGGAAAAAAGCGGCCGC-3′), antisense β-actin (5′-TTTAACTCGAG3′), sense γ-actin (5′-AAGGAAAAAAGCGGCCGC-3′), and antisense γ-actin (5′-TTTAACTCGAG3′). PCR products were cloned into pTOPO vector (Invitrogen) and sequenced. Sequences were compared with γ-actin and β-actin 3′ UTRs published in GenBank; verified cDNA inserts were subcloned into the eGFP construct to generate peGFPγ-actin and peGFPβ-actin. These plasmids were tested for expression and subcellular localization of the encoded eGFP by transfecting naive DRG cultures using LipofectAMINE 2000 (Invitrogen). Once validated, the eGFP plus 3′ UTR cassettes were digested with Nru1 and Xho1 and subcloned into Pme1 and Xho1 sites of pVQ-CMV-kNpA shuttle plasmid (Viraquest) for the generation of AV. The in vitro recombination and generation as well as packaging and titering of AVeGFPγ-actin and AVeGFPβ-actin were provided as a fee for service (Viraquest).
For adenoviral-based expression, dissociated DRG neurons were exposed to 150 MOI AVeGFPβ-actin or AVeGFPγ-actin for 15 min at 37°C after the last trituration step in dissociating DRGs for culture. Cultures were plated onto chambered coverglass (Nalgene) coated with poly--lysine and laminin. Cultures were grown for 16–20 h when ≥70% of cells showed GFP expression.
Dynein-based transport was diminished using ON-TARGET Plus SMARTpool siRNA targeting Dync1h1 (GenBank/EMBL/DDBJ accession no. ; Dharmacon). siGLO-Red reagent (Dharmacon) was used for identifying transfected neurons in the live cell imaging experiments (see Live cell imaging section below). After 12 h in vitro, DRG cultures were transfected with siRNAs using DharmaFECT3 as per the manufacturer's instructions (Dharmacon). The cultures were exposed to 50 nM of total siRNA (25 nM Dync1h1 and 25 nM siGLO-Red) plus 1.0 μl DharmaFECT3 in a total culture volume of 1 ml. After 24 h, the culture medium was replaced, and the cells were allowed to grow for 72 h.
For this, control and siRNA-treated cultures were lysed in radioimmunoprecipitation assay buffer (0.1% SDS, 50 mM Tris-Cl, pH 6.8, 150 mM NaCl, 0.5% NP-40, and 2 mM EDTA), cleared by centrifugation, and normalized for protein content by Bradford assay (Bio-Rad Laboratories). Lysate were denatured, fractionated by SDS/PAGE, and transferred to polyvinylidene difluoride membrane (). Blots were blocked in Tris-buffered saline with 0.1% Tween 20 (TBST) plus 5% nonfat dry milk for 1 h and then incubated overnight at 4°C in rabbit anti–dynein HC antibody (1:200; Santa Cruz Biotechnology, Inc.). Blots were rinsed several times in TBST and incubated with HRP-conjugated anti–rabbit IgG (1:5,000; Cell Signaling) for 1 h at room temperature. Blots were washed for 30 min in TBST and developed with ECL (GE Healthcare).
NGF-coated and/or MAG-coated Fluoresbrite YO Carboxylate Microspheres (4.5 or 6.0 μm; Polysciences) were added to cultures of infected DRGs at a low density and allowed to settle for 20 min before imaging. GFP-expressing neurons that contacted a fluorescent protein-bound microparticle were imaged by confocal microscopy using an inverted laser-scanning system (TCS/SP2 LSM; Leica) on an inverted microscope (DMIRE2; Leica) fitted with an environmental chamber to maintain a humidified temperature of 37°C with 5% CO. A 63× NA 1.4 oil immersion objective (Leica) was used for all imaging. The pinhole was set for 5 airy U to allow the acquisition of emission over the full thickness of the axon. The 488-nm laser line was used to excite eGFP and fluorescent microparticles; eGFP emission was collected at 498–530 nm, and microparticle fluorescence was collected at 575–600 nm. For time-lapse sequences, images were collected every minute over 50 min using the LCS confocal software package (Leica); the resultant avi file was converted to mov using the QuickTime media player (Apple Computer) and Sorenson compression (Sorenson Media). All image sequences were subjected to identical postprocessing for γ correction.
ImageJ (National Institutes of Health [NIH]) was used to quantify GFP signal intensity in these video sequences using original, unprocessed gray-scale images matched for laser intensity, photo multiplier tube voltage, and offset. For this, pixels/micrometer were quantified in a 10-μm axon segment spanning 5 μm proximally and distally from the center of the particle. The mean signal intensity was determined from five or more axons per condition for at least three separate experiments. Because mRNAs could accumulate before initiation of the imaging sequence, we concentrated our analyses on the t = 50 min images. Unless otherwise indicated, the ratio of mean pixels/micrometer of t = 50 min for NGF- and MAG Fc–treated cultures versus BSA- and IgG Fc–treated cultures, respectively, ± SD is presented. For the db-cAMP–, EHNA-, and siRNA-treated cultures, NGF and MAG signals were compared with similarly treated control cultures (BSA for NGF and IgG Fc for MAG-Fc) for the t = 50 min image.
Effects of siRNAs and EHNA on axonal transport were visualized by live cell imaging of LysoTracker dye (Invitrogen). LysoTracker was added at a final concentration of 50 nM and incubated for 20 min at 37°C. Medium was then changed, and vesicular movement was imaged over 2.5 min, with image acquisition every 5 s. The presence of siGLO signal in the neuronal cell body was used to identify siRNA-transfected neurons. To visualize axonal protein synthesis, infection with the AV-eGFPβ-actin was concurrent with the siRNA transfection. The NGF- or MAG-coated microparticles were added to cultures, and effects on local synthesis were imaged as described in the beginning of this section.
FISH was performed as previously described with minor modifications for the DRG cultures (). Oligonucleotide probes complementary to β-actin mRNA (at positions 3,187–3,138 and 3,446–3,495), peripherin (at positions 868–917, 1,263–1,312, and 1,382–1,431), and Kv3.1a (at positions 3,341–3,390 and 3,045–3,094) were designed using Oligo6 software (Molecular Biology Insights) and checked for homology to other mRNAs by BLAST. Probes were synthesized with amino group modifications at four positions each and labeled with digoxigenin succinamide ester as per manufacturer's instructions (Roche). The DRG cultures were fixed in buffered 4% PFA, equilibrated in 1× SSC with 40% formamide, and incubated at 37°C for 12 h in hybridization buffer (40% formamide, 0.4% BSA, 20 mM ribonucleoside vanadyl complex, 10 mg/ml salmon testes DNA, 10 mg/ml tRNA, and 10 mM sodium phosphate in 1× SSC) containing 20 ng of probe. Hybridization was detected by immunofluorescence using Cy3-conjugated mouse antidigoxigenin (1:1,000; Jackson ImmunoResearch Laboratories); neurofilament protein was detected by colabeling with chicken anti–neurofilament heavy subunit (1:1,000; Chemicon) followed by FITC-conjugated anti–chicken antibody (1:500; Jackson ImmunoResearch Laboratories).
Quantitative imaging of FISH signals was performed on an upright epifluorescent microscope (DM RXA2; Leica) with a CCD camera (ORCA-ER; Hamamatsu). All images were acquired using OpenLab 5.0 software (Improvision) and matched exposure time, gain, and offset with no postprocessing. All quantitative measurements were performed on the original 16-bit images using ImageJ (NIH). The point of the axon corresponding to the center of the particle was used as a reference point for measuring pixel intensity of the Cy3 emission. A 5-μm segment of axon corresponding to 2.5 μm proximal and distal to the particle center was used as the zero point. From this, the pixels/micrometer was quantified in three bins proximal to and distal to the particle center (5-μm length each plus 5-μm bin at particle center). Background was subtracted from the intensity values, and subtracted signal intensity in each 5-μm bin was normalized to the mean intensity over the entire 35-μm axon region that was measured.
Table S1 shows array-based identification of axonal mRNAs with the mean intensity of four axonal RNA preparations. Fig. S1 shows the purity of the axonal RNA used to generate the cDNA to probe the arrays (A) and confirms the linear amplification of the cDNA by virtual Northern blotting (B). Fig. S2 shows P[UTP] labeling of DRG cultures ± DRB to confirm the effectiveness of RNA polymerase II inhibition (A) and presents a Western blot confirming the reduced levels of dynein heavy chain in DRGs treated with the Dync1h1 siRNA for 72 h (B). Fig. S3 shows cultured DRGs infected with either eGFPβ-actin 3′ UTR or eGFPγ-actin 3′ UTR reporter constructs. Fig. S4 shows the cultured DRG response to immobilized sources of NGF (A), BSA (B), MAG-Fc (D), or IgG-Fc (E). The figure also shows that pretreatment with K252A before the addition of immobilized NGF shows a requirement for TrkA signaling (C) and that pretreatment with db-cAMP before the addition of immobilized MAG-Fc shows a reversal of MAG's effects by elevation of cAMP signaling (F). Video 1 shows live cell imaging of a neuron exposed to immobilized NGF (video of cell shown in A). Video 2 shows a neuron pretreated with PD98059 before NGF exposure (video of cell shown in B). Video 3 shows a DRG pretreated with colchicine before NGF exposure (video of cell shown in C). Video 4 shows a neuron transfected with Dync1h1 siRNA and stained with LysoTracker to show reduced retrograde but not anterograde transport. Video 5 shows a neuron treated with the dynein inhibitor EHNA and stained with LysoTracker to show reduced retrograde movement. Video 6 shows a neuron exposed to immobilized MAG-Fc (video of cell shown in A). Video 7 shows a neuron treated with db-cAMP before exposure to immobilized MAG-Fc (video of cell shown in C). Online supplemental material is available at . |
Signal transduction of TGF-β cytokines is mediated by an evolutionarily conserved mechanism that depends on the Smad proteins to transduce extracellular stimulus into the nucleus (; ). At unstimulated state, Smads spontaneously shuttle across the nuclear envelope and distribute throughout the cells (; ; ; ). Upon TGF-β stimulation, the receptor-activated Smads (i.e., Smad2/3 downstream of TGF-β, and Smad1/5/8 downstream of bone morphogenetic proteins [BMPs]) are phosphorylated, assemble into complexes with Smad4, and become mostly localized in the nucleus. Such signal-induced nuclear translocation of activated Smads is essential for the TGF-β–dependent gene regulations that are critical for embryonic development and homeostasis. The molecular machinery responsible for this process, especially how the activated Smads are imported as complexes, is not entirely clear ().
Previous studies on this subject used mostly in vitro methods, including reconstituted nuclear import assay which suggested either an importin-independent or importin β–mediated mechanism for nuclear import of Smads (; , ; ). The question is whether such conclusions apply to phosphorylated Smads in intact cells. Another broader issue is if additional factors, other than those mediating nuclear translocation by themselves, may be important for either activating Smads or targeting activated Smads into the nucleus. One example is recently demonstrated requirement of kinesin in guiding intra-cytoplasmic movement of Smads toward the cell surface receptor ().
Forward genetic screens in have been instrumental in identifying core components of the TGF-β pathway (). Recently, the RNAi technology offers a complementary cell-based approach to functionally identify molecules that mediate TGF-β signaling. Several critical elements of Decapentaplegic (Dpp; BMP) signaling including phosphorylation of Mothers against decapentaplegic (Mad), nuclear accumulation of phospho-Mad and Medea, and transcriptional up-regulation of (), have been characterized in tissue culture cells (; ). This, together with a collection of dsRNAs targeting the entire annotated genome, allowed us to genetically dissect the Dpp pathway and investigate molecular requirements for nuclear targeting of Smads upon stimulation ().
In this study, we describe a genome-wide RNAi screening that uncovered moleskin (Msk) as a required component in nuclear import of Dpp-activated Mad. Both genetic and biochemical studies further validated this finding. Msk belongs to a family of proteins that were originally discovered for their ability to bind the small GTPase Ran, hence the name RanBP (Ran-binding protein) (). Many RanBPs have been demonstrated to mediate nuclear import or export of various molecules and are since referred to as karyopherins (importins or exportins) (; ). We show that the mammalian Msk orthologues, Imp7 and Imp8 (also known as RanBP7 and 8), are responsible for nuclear import of both TGF-β and BMP-activated Smads in mammalian cells. Furthermore, we provide evidence that Smads are direct nuclear import cargoes of Msk/Imp7/8. Our data also revealed that in contrast to activated Smads, unphosphorylated Smads may enter the nucleus via Msk/Imp7/8-independent pathways, suggesting multiple routes for nucleocytoplasmic shuttling of Smads at basal state.
We used nuclear translocation of Mad as the readout in our RNAi screening because this is an early event in Dpp signaling. When Flag-Mad was conditionally expressed in S2R+ cells, it was detected diffusively throughout the cell (). In contrast, when the Dpp receptor kinases Punt and Thickvein (Tkv) were coexpressed, which caused Mad phosphorylation, the bulk of Flag-Mad gradually became predominantly localized to the nucleus (). With this cell line (Mad+R), we performed an RNAi screening in which the cells were treated with a library of ∼21,300 dsRNAs individually targeting over 95% of the annotated genome (). dsRNAs against the GFP and the / combination were used as negative and positive controls, respectively. After 3 d of incubation with dsRNAs, the Mad+R cell line was induced to express Flag-Mad, Punt, and Tkv, and the subcellular location of Flag-Mad was visualized with anti-Flag immunofluorescence staining followed by high throughput automated microscopy. Upon visual inspection of images obtained from duplicate screenings, we identified 346 dsRNAs that caused diffused distribution of Flag-Mad throughout the cell compared with the negative control dsRNA. Many of the genes corresponding to the 346 dsRNAs contain domains suggestive of their functions, and can be broadly categorized as in . The complete list of strong and weak hits in the primary screening can be accessed at the RNAi Screening Center (DRSC) website ().
The candidate hits were selected in an anonymous manner (see Materials and methods). Indeed, among the hits that gave strong phenotype were and , which confirmed that Mad phosphorylation is prerequisite for its nuclear accumulation and that the screening was robust (). Of particular note among the primary hits was , a karyopherin that was previously suggested to be required for nuclear import of activated ERK (dERK) (). The RNAi library used here contains dsRNA targeting many molecules known to be involved in nuclear transport, including importins, exportins, and nucleoporins. But besides , none was among the 346 hits identified in the primary screening (). In this study we focus on the analysis of ; the validation and bioinformatics analyses of the other hits will be presented elsewhere.
To verify the effect of RNAi in the primary screening, we designed and tested a second non-overlapping dsRNA against . Indeed, depletion of Msk by a different dsRNA also led to severely impaired nuclear concentration of Mad, and the effect was as potent as the positive control RNAi targeting Punt and Tkv (). This result strongly suggests that the block of Mad nuclear translocation we observed in the screening was not due to off-target effects of the dsRNA. In contrast to RNAi against Punt and Tkv, RNAi of did not affect C-terminal phosphorylation of Mad (), suggesting that Msk functions downstream of Mad phosphorylation, perhaps in transporting Mad into the nucleus.
Because the screening was performed using a S2R+ cell line overexpressing exogenous Flag-Mad, we wanted to determine if endogenous Mad is under the same regulation by in a different cell line. Dpp treatment of S2 cells resulted in predominant nuclear distribution of phosphorylated Mad, as revealed by immunofluorescence staining using a phospho-Mad–specific antibody, PS1 () (). Depletion of Msk by RNAi clearly resulted in more diffusive distribution of phospho-Mad (changing the nucleus/cytoplasm ratio from 2.9 to 1.1), while not affecting the level of Mad phosphorylation at its C terminus ( and Fig. S1 A, available at ).
Msk has been suggested to cooperate with the importin β homologue Ketel in nuclear import of dERK, because mutations in either or inhibited nuclear accumulation of dERK (). Moreover, the mammalian orthologues of Msk have been shown to function in conjunction with importin β (; ). Thus, we tested if might also be involved in nuclear translocation of Mad. Knockdown of by RNAi resulted in reduced phosphorylation of endogenous Mad (Fig. S1 A), but nevertheless phospho-Mad was still detected predominantly within the nucleus (). Quantitation of phospho-Mad staining intensity in the nucleus and cytoplasm confirmed that RNAi against did not affect nuclear accumulation of phospho-Mad (; changing the nucleus/cytoplasm ratio from 2.9 to 3.1, > 50). Similar observations were also made in S2R+ cells (unpublished data). Western blot analysis of cytoplasmic and nuclear fractions of S2 cells further validated that Msk, but not Ketel, is required for nuclear accumulation of phospho-Mad (). As shown in , although depletion of Ketel resulted in reduced amount of phospho-Mad through an unknown mechanism, the majority of phospho-Mad was still present in the nucleus (). Classic NLS–mediated nuclear import is dependent on importin β (). Indeed, RNAi against clearly impaired nuclear accumulation of classic NLS–fused GFP, while depletion of Msk had no effect (Fig. S1 B). This result verified that RNAi against was effective, and nuclear transport of Dpp-activated Mad is independent of the importin β homologue Ketel.
In both S2R+ and S2 cells, treatment with Dpp results in transcriptional activation of , a known Smad target gene in mammalian cells as well (; ). When Msk was depleted by RNAi, the Dpp- induced increase in expression was completely abolished (). The blocking effect of RNAi on expression was as strong as that caused by / RNAi (). Thus, consistent with being an essential factor for nuclear import of Mad, Msk is critical for the transcriptional output of Dpp.
To address the question if Msk is directly involved in transporting phospho-Mad into the nucleus, we tested protein–protein interaction between endogenous Msk and Flag-tagged Mad. Indeed, endogenous Msk coimmunoprecipitated with Flag-Mad from S2 cell extract (). Under our experimental conditions, both basal state and phosphorylated Mad displayed comparable interaction with Msk (). This suggests that binding of Msk is not unique to phospho-Mad, and Msk alone may not account for why only phospho-Mad accumulates in the nucleus. Therefore, although Msk is crucial for phospho-Mad to enter the nucleus, additional factors are involved to retain only phospho-Mad in the nucleus.
The above observations in cells identify as a new regulator in the Dpp pathway whose function is critical for nuclear accumulation of Dpp-activated Mad.
Mutations in the gene resulted in embryonic lethality (; ). Thus, to evaluate the functions of Msk in vivo, we used : to generate null (
) clones in the developing eye imaginal disc (; ). The clones were marked as GFP negative. In eye discs of third instar larvae, the PS1 antibody detected two stripes of phospho-Mad–containing cells around the morphogenetic furrow, consistent with the established role of Mad in eye development () (). The phospho-Mad signal in the anterior stripe is weaker and more diffused compared with that in the posterior stripe (). In the posterior stripe, phospho-Mad–positive cells span 5–6 cells wide, and quantitation of cell staining showed that 20.8% of the cells ( = 1,401) had phospho-Mad concentrated in the nucleus (). The rest of the cells within the posterior stripe have either undetectable or diffusive phospho-Mad staining ().
clones are small in size and number compared with the wild-type clones, consistent with previous reports that
mutation led to growth disadvantages (; ). In
clones that straddle the posterior stripe, we detected a significantly smaller number of cells (4%, = 142) with strong phospho-Mad staining concentrated in the nucleus (). In comparison, in wild-type clones generated by , the number of cells with high phospho-Mad signal distinctively in the nucleus is as high (22%, = 491) as in genetically unmodified wild-type cells (20.8%, = 1,401; ).
The observed phenotype is consistent with our RNAi results in cell culture, which suggests a defect in nuclear import of phospho-Mad. Because we did not observe a considerable accumulation of cytoplasmic phospho-Mad, it is possible that in vivo the un-imported phospho-Mad is rapidly degraded or dephosphorylated. Although we cannot rule out other possibilities attributing to the observations in , it is clear that in vivo, cells with loss-of-function mutation in would have defects in phospho-Mad–mediated signaling.
Msk has two homologues in mammals, Imp7 and Imp8, each sharing over 50% identity in amino acid sequence with Msk. Imp7 and Imp8 themselves are ∼60% identical. Based on in vitro assays, Imp7 has been suggested to import ribosomal proteins, histone H1, HIV reverse transcription complexes, and glucocorticoid receptor into the nucleus (; ; ; ). Imp8 was recently shown to support nuclear import of the signal recognition particle 19 (SRP19) in vitro ().
To investigate the roles of Imp7 and 8 in nuclear transport of Smads in mammalian cells, we designed siRNA duplexes that are effective in knocking down Imp7 or 8 individually (). Although Imp7 and 8 siRNA duplexes had no effects on BMP2-induced phosphorylation of Smad1 (), immunofluorescent staining with phospho-Smad1–specific antibody showed that knockdown of either Imp7 or 8 resulted in a more diffusive distribution of phospho-Smad1 after BMP2 stimulation, while in control siRNA transfected cells phospho-Smad1 was mostly present in the nucleus (). Corresponding to this defect in nuclear accumulation of Smad1, the BMP2-induced transcriptional activation of Smad6 was also suppressed in cells transfected with siRNA against either Imp7 or 8 (). Therefore, similar to their counterpart, Imp7 and 8 are critical for nuclear translocation of BMP-activated Smad1 in mammalian cells.
TGF-β and BMP pathways are similar in the general signaling mechanism, but differ in the receptor kinases and Smads that are used for signaling (). We thus investigated if Imp7 and 8 are shared by TGF-β and BMP pathways in transporting different receptor-activated Smads into the nucleus. Again, knockdown of either Imp7 or 8 severely inhibited nuclear accumulation of Smad2 and 3 in response to TGF-β stimulation (). Such observation was made in both HeLa and HaCaT cells, and Smad2/3 phosphorylation in response to TGF-β was not affected by the same siRNA against Imp7 or 8 (). The block of Smad2/3 nuclear accumulation was also manifested in substantially reduced transcriptional activation of the TGF-β target gene Smad7 (). Therefore, Smads downstream of TGF-β also depend on Imp7 or 8 for nuclear translocation.
Because both Imp7 and Imp8 are required for nuclear transport of Smads, we next examined the relative contributions from these two. The efficiency of Smads nuclear translocation could be quantitated by counting the number of cells exhibiting “nucleus only” versus “cytoplasmic” distribution of Smads. We found that indeed when transfected at the same final concentration, combining siRNAs targeting Imp7 and 8 was more effective in inhibiting nuclear accumulation of Smad2/3 than individual siRNA against either Imp7 or 8 alone (). This suggests that Imp7 and 8 are likely to act in parallel in mediating nuclear translocation of TGF-β–activated Smad2/3.
Because siRNAs against Imp7 and 8 are highly effective in blocking nuclear accumulation of Smad2/3, we examined if reintroducing Imp7 and 8 cDNAs would rescue the RNAi phenotype. To this end, we generated silent mutations in Imp7 and 8 sequences and verified that the mutants are no longer targeted by the Imp7 or Imp8 siRNA, respectively (not depicted; see Materials and methods). Such mutant cDNAs were transfected into HeLa cells 2 d after the siRNA transfection. Indeed, upon TGF-β stimulation, Smad2/3 accumulation in the nucleus was restored only in cells that expressed the rescuing HA-tagged Imp7 or 8 plasmids (). This result validated that the defects in Smads nuclear translocation observed in and were specifically due to depletion of endogenous Imp7 and 8.
When overexpressed in HeLa cells, Imp7 was diffusively distributed throughout the cell while more Imp8 was detected in the nucleus than in the cytoplasm (). Such patterns of Imp7 and 8 localization remained the same upon TGF-β stimulation (). Overexpression of Imp7 or 8 had no detectable effects on the distribution of endogenous Smad2/3 at both basal and TGF-β–stimulated states (). This suggests that in HeLa cells, endogenous Imp7 and 8 are not limited in quantity to support nuclear translocation of Smad2/3.
We next investigated Smads interactions with Imp7 and 8 by coimmunoprecipitation experiments. Flag-tagged Smad1 or Smad2 were overexpressed in 293T cells and immunoprecipitated with anti-Flag antibody. In both cases, HA-tagged Imp7 or Imp8 coimmunoprecipitated with either Smad1 or Smad2 (). Constitutively active BMP receptor (ALK3-QD) or TGF-β receptor kinase (ALK5-TD) was cotransfected to induce C-terminal phosphorylation of Smad1 and Smad2, respectively. Such phosphorylation of Smads did not affect their interaction with Imp7 or 8 (). Because phosphorylated Smad1, 2, and 3 readily assemble into complexes, our results suggest that monomeric and multimeric forms of Smads have similar interactions with Imp7 or 8 (; ). These observations are consistent with our finding in cells ().
For detailed analysis of Smad-Imp7/8 interaction, we focused on Smad3. We produced GST-fusions of the MH1, MH2, and linker plus MH2 domains of Smad3 in and tested their ability to pull down endogenous Imp7 and 8 in Hela cells. When comparable amount of GST fusion proteins were used, both the MH1 (aa 1–155) and the linker plus MH2 (aa 146–425) domains were able to bind endogenous Imp7/8, with the MH1 domain exhibiting stronger interaction (). The same assay barely detected any interaction between the Smad3 MH2 (aa 231–425) domain and Imp7/8 (). Therefore, interaction with Imp7/8 appears to involve multiple interfaces in the MH1 and linker regions of Smad3. The MH1 domains of Smad2 (aa 1–185) and Smad3 are highly similar except for two insertions in Smad2 that prevent Smad2 from binding to DNA (). But apparently such differences did not affect Smad2 binding to Imp7/8 through the MH1 domain (). Bacterially produced GST-Imp8 was able to pull down purified recombinant Smad1 or Smad3, suggesting that Imp8 could directly interact with Smad1 or Smad3 ( and Fig. S2, available at ).
One characteristic among importins is that the interaction with their cargoes is regulated by Ran in its GTP-bound form (; ). To test if this is true between Smad3 and Imp8, we first pulled down HA-Imp8 using GST fusion of full-length Smad3. After washing off the unbound proteins, RanQ69L-GTP (the Q69L mutation locks Ran in its GTP-bound state) or BSA was added to the GST beads for further incubation (). Indeed we found that comparing to the BSA control, RanQ69L-GTP caused more release of Imp8 into the supernatant and correspondingly resulted in a decrease of Imp8 remaining bound to GST-Smad3 on the beads (). This suggested that association of Smad3 with Imp8 was disrupted upon binding of Ran-GTP, supporting the notion that Smad3 is a nuclear import cargo of Imp8.
Without TGF-β stimulation, Smads undergo spontaneous nucleocytoplasmic shuttling and are distributed evenly in both the nucleus and cytoplasm in many types of cells (; ; ). Prompted by the observation that unphosphorylated Smads interact with Imp7 and 8, we examined if Imp7 and 8 are also required for basal state Smads import into the nucleus. In S2R+ cells, Flag-Mad was detected throughout the cells without exogenous Dpp (). The presence of Flag-Mad in the nucleus is not likely due to autocrine Dpp secreted by the cells, because further blocking any residual Mad phosphorylation by RNAi against and did not eliminate the presence of Mad in the nucleus (). Treatment with dsRNA targeting also had no effect on the presence of Mad in the nucleus at basal state (). Because RNAi against / and have been validated to be highly potent (), we concluded that nuclear import of basal state Mad does not rely on Msk, in contrast to Dpp-activated Mad.
Similar to cells, knockdown of Imp7 and Imp8 individually or in combination did not reduce the amount of Smad2/3 in the nucleus of unstimulated HeLa cells (). Therefore, in both and mammalian cells, although Msk and Imp7/8 interact with basal state Smads, the presence of Smads in the nucleus without TGF-β stimulation is not critically dependent on Msk or Imp7/8.
Overexpressing Fox H1 in HeLa cells is another way to drive endogenous Smad2/3 into the nucleus without TGF-β stimulation (). Such nuclear accumulation of Smad2/3 could be due to nuclear sequestration of the shuttling Smad by the nucleus-bound Fox H1, and has been observed with other Smad-interacting transcription factors such as ATF3 (). Transfection of siRNA targeting Imp7 or Imp8, individually or combined, did not alter the “nucleus only” pattern of Fox H1, and did not affect Fox H1–induced nuclear concentration of Smad2/3 (). In control siRNA-transfected cells, 82.3% ( = 34) of Fox H1–positive cells contained endogenous Smad2/3 predominantly in the nuclei, whereas in cells with double-knockdown of Imp7 and Imp8, 84.3% ( = 32) of Fox H1–expressing cells have Smad2/3 in the nucleus. Thus, the above observations, from both and mammalian cells, led us to conclude that unphosphorylated Smads may be able to enter the nucleus through additional mechanisms, such as those described in previous studies (; ; ).
Genome-wide RNAi screening in this study offers a genetic approach to uncover new elements in TGF-β signal transduction. Here we identify and validate with in vivo evidence that Msk and its mammalian orthologues Imp7 and 8 are critical components in transporting TGF-β–activated Smads into the nucleus. Biochemical evidence further suggests that Msk/Imp7/8 directly import phospho-Smads as cargoes.
Although there appears to be some discrepancy between these new findings and our previous reports that importins are dispensable for the nuclear import of Smads, these observations can be reconciled (, ). Our present and previous studies, based on different approaches, may have revealed different nuclear import mechanisms used by basal and activated Smads to enter the nucleus. There are important differences comparing Smads import with or without TGF-β stimulation. Unphosphorylated Smads are monomers, but phosphorylated Smads are assembled into complexes with Smad4 and are thus much larger in size (; ). Moreover, as phospho-Smads accumulate in the nucleus they have to move across the nuclear pore against an ascending concentration gradient of Smads already in the nucleus, whereas unphosphorylated Smads never reach a higher concentration in the nucleus than in the cytoplasm. Thus, importing phospho-Smad complexes and unphosphorylated Smad monomers may entail different mechanisms, with or without the participation of importins. Indeed, our RNAi data in both and mammalian cells suggest that nuclear import of the two forms of Smads is very different regarding the requirement of Msk/Imp7/8. This type of differential requirement for import factors is not unique to Smads. In fact, STATs (signal transducers and activators of transcription) in the interferon pathway are another example in which the latent STATs are imported by an importin-independent mechanism, whereas the phosphorylated STATs depend on importins to accumulate in the nucleus (; ; ). It is also interesting to note that phospho-Smads were still detected in the nucleus upon RNAi-mediated knockdown of Msk/Imp7/8. Although we cannot rule out the trivial explanation that this may be due to incomplete depletion of the targeted proteins, this observation may also suggest additional import mechanisms for activated Smads. We recognize that our previous finding of importin-independent nuclear import of Smads was largely based on an in vitro reconstituted nuclear import assay (, ). Although this in vitro system is widely accepted, it may not fully recapitulate nuclear import of activated Smads in cells (). Based on our RNAi data, regarding the requirement of importins, the conclusion drawn from the in vitro import assay may not apply to phospho-Smads in intact cells. However, the current study does not necessarily contradict the previous suggestions that direct Smad–nucleoporin interaction is critical for nuclear import of Smads (; ).
Our data showed that Msk/Imp7/8 interacted with Smads regardless of their phosphorylation status; thus, additional factors must be involved to explain why only TGF-β/BMP– activated Smads can accumulate in the nucleus. Because basal- state Smads are actively exported out of the nucleus (; ; ), it is possible that retaining only phospho-Smads in the nucleus requires blocking Smads nuclear export, a scenario that has been demonstrated for Smad4 (). This hypothesis would be consistent with findings in live cells, in which TGF-β signaling led to reduced mobility of Smad2 in the nucleus ().
Because Msk, Imp7, and Imp8 are shown to be critical for targeting phospho-Smads into the nucleus, it is conceivable that regulatory inputs to this nuclear import factor would impact TGF-β signaling. Although we did not notice any changes in subcellular localization of Msk or Imp7/8 in response to TGF-β in cultured cells ( and unpublished data), during embryonic development, Msk distribution changed between cytoplasm and nucleus in a dynamic fashion (). Moreover, Msk is phosphorylated on tyrosine residues with yet-unknown functional consequences (). If and how Msk localization is regulated and by what signals are completely open questions at present.
A number of mitogen-induced phosphorylation events in the linker region of Smad have been suggested to inhibit TGF-β–induced nuclear translocation of Smads in and mammalian cells (, ; ; ). Because part of the Imp7/8 binding was mapped to the linker region of Smad3, it will be interesting to determine if linker phosphorylation would affect the interaction between Smads and Imp7/8 and hence the rate of nuclear import. It is also worth noting that Msk has been genetically implicated in the nuclear import of activated ERK in . Such convergence on the same molecule for nuclear import raises the possibility of cross-talk between MAP kinase and TGF-β pathways at the level of nuclear translocation of key signal transducers.
The dsRNA library targeting the whole genome and the format for screening have been described previously (). dsRNAs were deposited in 384-well plates, and in each plate one well is reserved for dsRNA (negative control) and one for combined and dsRNA (positive control). 10 S2R+ cells in 10 μl serum-free media were seeded in each well and incubated for 1 h, after which 30 μl of serum-containing media was added. After incubation for 3 d, the cells were induced with 0.5 mM CuSO for 3 h followed by fixation with 4% paraformaldehyde in PBS (10 min) and immunofluorescence staining with anti-Flag antibody (Sigma-Aldrich). Automated microscopy was performed using the Discovery1 system (Molecular Devices). The screening was repeated once. The wells containing selected hits were reported to DRSC, which in turn revealed the identities of the dsRNAs contained in those wells. The complete dataset including strong and weak hits in the two screenings is available at .
S2 and S2R+ cells were cultured in Schneider media with 10% fetal bovine serum (Invitrogen) and transfected with Effectene (QIAGEN). HeLa and 293T cells were cultured in DME with 10% fetal bovine serum and transfected with Lipofectamine 2000 (Invitrogen). Cells with or without treatment (1 nM Dpp, 100 pM TGF-β, or 100 ng/ml BMP2 as indicated; all from R&D Systems) were processed for immunofluorescence staining as described previously (). PS1 was a gift from P. ten Dijke (Leids University, Netherlands). Alexa 488– or Alexa 633–conjugated anti–rabbit secondary antibodies (Invitrogen) were used as indicated, and cells were mounted in Vectorshield (Vector Laboratories). Immunofluorescence microscopy and image acquisition were done with an inverted microscope (20×/0.45, 40×/0.6, 60×/1.40; Eclipse TE2000-S; Nikon) and a digital camera (SPOT RT-KE; Diagnostic Instruments, Inc.) using vendor-provided software. For confocal microscopy, a DMIRE2 inverted microscope and the TCS scanning system from Leica were used. The images were captured wither lasers at: UV (DAPI), 488 nM (Alexa 488 or GFP), and 633 nM (Alexa 633) wavelengths at room temperature using vendor-provided software. 20×/0.70, 40×/(1.25–0.75), and 63×/1.40 oil immersion objectives were used for low and high magnification images. Adobe Photoshop was used to adjust the brightness and contrast of the entire images if necessary. Final figures were assembled using Adobe Photoshop. For quantitation purposes, confocal sections with the strongest signal were selected, and the staining intensity in the nucleus and cytoplasm was measured using NIH ImageJ. Only cells with unsaturated signals were chosen for such analysis.
P{
mC=} P{FRT}80B and P{ FRT}80B (gifts from A. Vrailas, Emory University, Atlanta, GA) were used to generate clones in the developing eye imaginal disc as described previously (; ). Third instar stage larvae were dissected and fixed in 4% paraformaldehyde in PBS for 30 min at room temperature. The imaginal discs were permeabilized with 0.3% Triton X-100/PBS for 30 min at room temperature and blocked with 5% normal horse serum/0.3% Triton X-100/2%BSA/PBS for 5 h at room temperature. PS1 was used at 1:2,000 dilution in the blocking buffer and the incubation lasted overnight at 4°C. After washing, the discs were stained with Alexa 633–conjugated anti–rabbit secondary antibody (2 μg/ml; Invitrogen) for 2 h at room temperature. The imaginal discs were mounted in Vectorshield/PBS (1:1; Vector Laboratories) and examined with confocal microscopy as described above.
General procedures for generating dsRNA and RNAi in S2 and S2R+ cells were as described previously (). Two amplicons corresponding to different coding regions of , DRSC11340 and DRSC23929, were used to generate nonoverlapping dsRNA targeting . Sequence information for the two amplicons can be found at the DRSC website (). For HeLa and HaCaT cells, siRNA was transfected using HiPerFect at a final concentration of 40 nM (QIAGEN). siRNA targeting Imp7: GAUGGAGCCCUG CAUAUGA dTdT (a gift from M. Stevenson, University of Massachusetts Medical School, Worcester, MA); siRNA targeting Imp8: GAGAUCTTCCGAACUAUUAdGdT. To generate the rescue constructs, the coding sequences targeted by the siRNAs were mutated into Imp7: GATGGAGC TCAATGA; Imp8: GAGATCTTGACATA. The residues changed by mutagenesis are underlined.
S2 cells were suspended in 20 mM Hepes, pH 7.6, 5 mM MgCl, 10 mM KCl, 1 mM EGTA, 1 mM EDTA, 250 mM sucrose, and 0.025% digitonin (EMD Biosciences). Cells were passed through 18 G syringes three times and incubated on ice for 5 min. The homogenate was centrifuged at 800 for 10 min. The supernatant was collected as the cytoplasmic fraction. The pellet was further extracted with 20 mM TrisCl, pH 7.5, 250 mM NaCl, and 0.5% NP-40 to yield the nuclear fraction.
For coimmunoprecipitation experiments, or 293T cells were lysed in 20 mM Tris Cl, pH 7.4, 200 mM NaCl, 5 mM MgCl, 20 mM NaF, 20 mM NaPO, 20 mM β-glycerolphosphate, 0.5% NP-40, and 2 mM DTT supplemented with protease inhibitors. Cell extracts were incubated with anti-Flag conjugated to agarose beads (Sigma-Aldrich) at 4°C for 4–16 h, followed by 3× wash in the lysis buffer before immunoblotting. Anti-Msk was a gift from L. Perkins (Harvard Medical School, Boston, MA).
Full-length human Imp8, Smad1, and Smad3 were produced in and purified as GST fusions. GST-Smad1 (a gift from F. Liu, Rutgers University, NJ) and GST-Smad3 were digested with thrombin to remove the GST moiety (Novagen). 10 μg of GST-Imp8 on beads was incubated with Smad1 or Smad3 (0.5–1 μg/μl) at 4°C with rotation for 4 h. The buffer contained 20 mM TrisCl, pH 7.5, 200 mM NaCl, 0.05% NP-40, 5% glycerol, and PMSF. The beads were washed 3× in the same buffer and the bound proteins were analyzed by immunoblotting.
Fig. S1 shows the controls for the RNAi experiments in Fig. S2 shows the SDS-PAGE of purified recombinant proteins used in D. Online supplemental material is available at . |
A hallmark of cancer is evasion of apoptosis (), which links cancer genetics and cytotoxic chemotherapies inextricably together (). Apoptosis induced by chemotherapeutic agents has been attributed to the induction of DNA damage. One of the key molecules involved in response to DNA damage is the tumor suppressor protein p53 (; ). The loss of p53 response is thought to promote genomic instability () that can lead to increased resistance to chemotherapeutic agents. In normal unstressed cells, the p53 protein is present at very low levels because of continuous degradation mediated by Mdm2, a protein that is also transcriptionally activated by p53 (). Thus, p53 and Mdm2 are linked to each other through an autoregulatory negative feedback loop (). Disruption of the p53–Mdm2 complex is the pivotal event in p53 activation after DNA damage (; ; ). In addition, recent papers have suggested that enhanced translation of p53 mRNA is also an important step in the induction of p53 in stressed cells (; ; ), although the mechanisms remain largely unknown.
Translation of eukaryotic mRNAs is predominantly regulated at the level of initiation (; ; ), when the ribosome is recruited to the mRNA. The eukaryotic translation initiation factor (eIF) complex eIF4F is required for this multistep process and is composed of the cap-binding protein eIF4E; the RNA helicase eIF4A; and the scaffold protein eIF4G, which provides binding sites for eIF4E, eIF4A, and the poly(A)-binding protein (PABP; ; ; ). eIF4A is required to unwind the second structure in the 5′ untranslated region (UTR). The helicase activity of eIF4F should be proportional to the amount of the secondary structure in the 5′ UTR, which would otherwise affect translational efficiency (; ). The efficiency of translation initiation is tightly coupled with cell cycle progression and cell growth, with translational induction occurring in response to mitogenic stimulation (; ). Such changes in translation are normally mediated by alterations in the expression or phosphorylation status of the various translation initiation factors involved (; ; ). Hypophosphorylated eIF4E–binding protein 1 (BP1) competes with eIF4G for binding to eIF4E and prevents formation of the eIF4F complex (; ; ). In addition, the interaction of eIF4E with its partners can be regulated by the availability of free eIF4G, which may be regulated at the levels of synthesis and turnover (). Despite suggestions that the control of translation may be regulated by growth-factor signaling (; ), the relative contribution of translational effects of these signaling pathways in their corresponding cellular activities and the mechanisms involved have remained unclear.
Insulin-like growth factor 1 receptor (IGF-1R) is a membrane-associated tyrosine kinase receptor that plays an important role in cell growth, transformation, and protection of cells from a variety of apoptotic stimuli (; ; ). IGF-1R signaling protects cells from apoptosis mainly through the phosphoinositide-3 kinase (PI-3K)–Akt and Ras–Raf–MAPK pathways (; ; ). Inhibition of IGF-1R has been shown to block tumor growth and sensitize cells to antitumor treatments (), indicating that IGF-1R is a promising target for cancer therapeutics (). In other situations, however, IGF-1R signaling contributes to cell death (). Overexpression of the C terminus of the IGF-1R β subunit induced apoptosis in culture cells (; ), suggesting that IGF-1R has intrinsic proapoptotic features. Remarkably, a potential role of IGF-1R in mediating cell death in vivo was suggested by the findings that mice exhibited enhanced resistance to oxidative damage compared with wild-type mice (). Furthermore, it has been reported that IGF-1R signaling may be able to potentiate p53 induction (; ), which can induce apoptosis. Although the antiapoptotic functions of IGF-1R have been well established (; ), the mechanism by which IGF-1R sends a proapoptotic signal is not well known. A better understanding of the proapoptotic function of IGF-1R may reveal more rational approaches for cancer therapies targeting IGF-1R signaling.
In this paper, we have used mouse embryonic fibroblasts (MEFs) and a specific IGF-1R inhibitor, AG1024, to decipher the role of IGF-1R in regulating cellular apoptosis induced by the chemotherapeutic agent etoposide and the mechanisms involved. We found that inhibition of IGF-1R reduces DNA-damage–induced apoptosis through translational inhibition of p53 and Mdm2 expression. Our results not only provide insights into the role for IGF-1R in the p53-induced apoptotic response but also reveal a critical role for translational regulation of the p53–Mdm2 feedback loop by IGF-1R signaling.
We observed that R MEFs, in which the gene has been knocked out (), were insensitive to apoptosis induced by the DNA-damage agent etoposide compared with R MEFs (; see Materials and methods). Detection of the cleavage of apoptotic markers caspase-3 and poly (ADP-ribose) polymerase (PARP) supported this observation (). Additionally, treatment with the IGF-1R kinase inhibitor AG1024, which suppressed both the autophosphorylation activity of IGF-1R and its downstream signaling (), reduced apoptosis in response to etoposide in R MEFs (). Furthermore, transient expression of plasmids encoding the wild-type IGF-1R (IGF-1R-WT) but not the kinase-inactive IGF-1R (IGF-1R-YF) in R MEFs resulted in an increased apoptotic response to etoposide (). Collectively, these results suggest that functional IGF-1R renders MEFs more susceptible to etoposide-induced apoptosis.
Because p53 is a key mediator of apoptosis induced by DNA damage (), we examined whether apoptosis of MEFs induced by etoposide depended on functional p53. Both R and R MEFs transfected with dominant-negative p53 (GFP-p53DD) exhibited a reduced apoptotic response to etoposide (), indicating that p53 is required for the apoptotic response of MEFs to etoposide. Given that p53 transcriptional activity is required for p53-dependent apoptosis after DNA damage (), we next investigated whether IGF-1R inhibition could impair p53 activation. To this end, we performed luciferase assays using p53-reponsive elements (p53bs-luc) and unstimulated elements (p53ms-luc). The p53bs-luc reporter had higher relative luciferase activity in R than in R MEFs after DNA damage (), implying that DNA-damage–induced p53 activation is impaired in R MEFs.
Because p53 activation after DNA damage is associated at least in part with p53 accumulation (), we next analyzed the induction of p53 protein levels in R and R MEFs. Titration experiments revealed a substantial increase in the amount of p53 protein as well as its downstream targets p21 and Mdm2 in response to etoposide in R compared with R MEFs (). Furthermore, AG1024 attenuated p53 induction followed by etoposide treatment in R but not in R MEFs (), suggesting that IGF-1R–mediated sensitization of MEFs to p53 accumulation was dependent on IGF-1R kinase activity. In agreement with p53 expression, p21 and Mdm2 induction in response to etoposide treatment was also impaired in R but not in R MEFs after AG1024 treatment (). To test the generality of our observations, we next examined whether the lack of IGF-1R could reduce p53 induction in response to other anticancer agents, such as doxorubicin and Taxol. We found that in R MEFs, the induction of p53 and p21 in response to doxorubicin or Taxol was impaired (Fig. S1 A, available at ). However, despite impaired p53 induction, R MEFs exhibited enhanced apoptotic responses to doxorubicin and Taxol (Fig. S1 B), suggesting that impaired p53 induction in R MEFs may not always translate into reduced apoptosis. Because p53 induction may also result in G1 cell cycle arrest in response to DNA damage (), we next examined the cell cycle profiles of R and R MEFs after DNA damage. Treatment with etoposide induced cell cycle arrest at the G1/S and G2/M checkpoints in R MEFs, whereas R MEFs exhibited a reduced G1 arrest (Fig. S1 C), which is consistent with the impaired p53 induction observed in R MEFs.
To determine whether IGF-1R inhibition could impair p53 accumulation and apoptosis in human tumor cells, we treated human hepatocellular carcinoma SK-hep1 and human colon cancer HCT116 cells with AG1024. Treatment of these cells with AG1024 impaired p53 accumulation as well as apoptosis in response to etoposide (Fig. S2, A and B, available at ). Notably, AG1024-treated and -untreated cells showed a similar level of apoptosis in response to etoposide (Fig. S2 B), indicating that IGF-1R inactivation cannot protect these cells against DNA-damage–induced apoptosis in the absence of p53. In addition, we tested whether IGF-1R inhibition could protect cells from p53-independent apoptotic stimuli such as ionomycin, which causes calcium flux. demonstrates that the ability of ionomycin to induce apoptosis was unaffected in R MEFs. Similarly, ionomycin induced comparable levels of p53-independent cell death in both untreated and AG1024-treated HCT116 cells (Fig. S2 C). Thus inactivation of IGF-1R antagonizes the ability of etoposide to increase p53 abundance and activity and thereby impairs p53-dependent functions including apoptosis and cell cycle arrest.
To define the mechanisms that underlie attenuated p53 response to etoposide in R MEFs, we next investigated the integrity of DNA-damage checkpoint pathways in R MEFs. Phosphorylation of p53 on ser18 (corresponding to serine 15 in human p53) contributes to p53 activation after DNA damage through increased binding to the p300 coactivator protein (). We found that etoposide treatment induced similar levels of ser18 phosphorylation of p53 in both R and R MEFs (unpublished data). In addition, the experiment to detect p53 localization revealed that the etoposide-induced p53 protein in both R and R MEFs was localized in the nuclei (unpublished data), again indicating that inactivation of IGF-1R impairs p53 induction without affecting the DNA-damage signaling to p53.
Because DNA damage increases p53 protein levels mainly by up-regulating p53 protein stability (; ), we reasoned that IGF-1R inhibition might regulate p53 accumulation in response to DNA damage by influencing p53 protein stability. Indeed, treatment of R MEFs with etoposide increased p53 stability (). Importantly, there was no measurable difference in p53 stability in etoposide-treated R and R MEFs (). Likewise, p53 protein was stable in untreated and AG1024-treated SK-hep1 cells after etoposide treatment (Fig. S3, A and B, available at ). Surprisingly, we detected a higher stability of p53 protein in untreated R MEFs than in untreated R MEFs (). Similarly, the IGF-1R inhibitor also stabilized p53 protein in SK-hep1 cells (Fig. S3 B). To confirm that a lack of IGF-1R activity can stabilize p53 protein, R and R MEFs were pulse labeled with [S]methionine/cysteine followed by a 4-h chase. The results showed that the half-life of p53 protein was ∼15 and 60 min in R and R MEFs, respectively (), again demonstrating an increased half-life of p53 protein upon IGF-1R inhibition.
Because the degradation of p53 is mediated by the ubiquitin–proteasome pathway, we next examined the amount of ubiquitin that is conjugated to p53 for degradation. The results showed a remarkable decrease in p53–ubiquitin complexes in R MEFs and AG1024-treated SK-hep1 cells ( and Fig. S3 C), implying that IGF-1R inhibition may increase p53 stability by reducing p53 ubiquitination.
Because the ubiquitin ligase Mdm2 is a key regulator of p53 protein turnover (), we tested whether Mdm2 was involved in the regulation of p53 stability by IGF-1R inhibition. R MEFs as well as AG1024-treated Sk-hep1 cells expressed lower levels of Mdm2 protein compared with R MEFs and untreated Sk-hep1 cells, respectively ( and Fig. S3 D). Furthermore, AG1024 treatment led to the down-regulation of Mdm2 protein in wild-type HCT116 cells and HCT116 cells (), implying that Mdm2 expression is down-regulated in a p53-independent manner in response to IGF-1R inhibition. RT-PCR analysis revealed no detectable difference in mdm2 mRNA levels in HCT116 and cells upon IGF-1R inhibition (), suggesting a translational or posttranslational role of IGF-1R signaling in regulating Mdm2 expression. We therefore examined Mdm2 protein synthesis by metabolic labeling assay. The S-labeling experiments revealed a reduced synthesis of S-labeled Mdm2 in either or HCT116 cells upon AG1024 treatment (). The reduction in S incorporation was not caused by the reduced stabilization of Mdm2 because treatment of HCT116 cells with AG1024 did not alter the half-life of Mdm2 protein (). In fact, using a S-pulse label analysis, we demonstrated that the half-life of Mdm2 protein in untreated and AG1024-treated HCT116 cells was ∼55 and 60 min, respectively (). Thus, these results suggest that inhibition of IGF-1R activity decreases the translational rate of mdm2 transcripts and consequently the expression levels of Mdm2 protein, therefore increasing p53 protein stability.
It should be noted that IGF-1R inhibition did not up-regulate the steady-state levels of p53 protein in either of the examined MEFs or tumor cells ( and , and see ), although degradation of p53 protein had been severely attenuated. It is therefore conceivable that, despite decreased p53 turnover, IGF-1R inhibition might maintain low levels of p53 protein by reducing p53 synthesis. Northern blot analysis revealed similar levels of p53 mRNA in R and R MEFs (); therefore, we reasoned that IGF-1R inhibition might counterbalance the effects of the enhancement of p53 protein stability by reducing p53 synthesis at the translational level. We did observe a reduction in [S]methionine/cysteine–labeled p53 in R MEFs (). Similarly, treatment of SK-hep1 cells with IGF-1R inhibitor also decreased synthesis of S-labeled p53 (Fig. S4 B, available at ), whereas p53 mRNA levels remained constant (Fig. S4 A). Collectively, these results suggest that decreased p53 mRNA translation may neutralize reduced p53 degradation in response to IGF-1R inhibition. Thus, our analyses indicate that R MEFs and AG1024-treated cells are refractory to p53 induction after DNA damage because of the prolonged half-life of p53 and reduced p53 synthesis.
The observation that protein synthesis of Mdm2 and p53 proteins is reduced after IGF-1R inhibition suggests a possible role for IGF-1R in translational regulation of gene expression. It has been reported that growth-factor signaling could regulate mRNA translation by modulating the general translation initiation factors (; ). We therefore tested whether the lack of IGF-1R activity altered overall protein synthesis and activity of the eIF4F complex. Compared with R MEFs, R MEFs had a slower rate of incorporating amino acids into protein (). Similar levels of inhibition were obtained in SK-hep1 cells with the administration of AG1024 (Fig. S4 C). Furthermore, IGF-1R inhibition had no measurable effect on the levels of eIF4A, PABP, and eIF4E proteins, but resulted in a reduction in eIF4G abundance (, bottom). In addition, the hyperphosphorylated form of eIF4E–BP1 was also reduced upon IGF-1R loss (, bottom). To determine whether these modulations could disrupt the eIF4F complex, we next examined the association of eIF4E with other translation initiation factors by pull-down on mGDP–sepharose resin. The precipitation assay showed reduced association of eIF4G, eIF4A, and PABP with eIF4E, whereas the amount of eIF4E–BP1 in the precipitate was increased in R MEFs (, top). The IGF-1R inhibitor induced similar alterations of translation initiation factors and impaired the formation of the translation initiation complex in SK-hep1 cells (Fig. S4 D). Together, these results suggest an important role for IGF-1R signaling in the regulation of translation initiation processes.
Cellular mRNAs differ hugely in the amount of eIF4F required for efficient translation (; ; ). Alterations of the general translational apparatus may preferentially affect the translation of weak mRNAs with extensive secondary structure in their 5′ UTR (; ; ; ; ). We next investigated whether the modulations of the basal translational machinery by IGF-1R inhibition could evoke a selective translational effect. To this end, we examined the translation levels of several proteins with short half-lives upon IGF-1R inhibition because the levels of short-lived proteins are believed to be more sensitive to translational inhibition (). We observed no change in the translation levels of short-lived proteins after IGF-1R inhibition, including p27 and c-fos (, bottom panels). These results indicate that the translational depression in response to IGF-1R inhibition might be caused by an mRNA-specific mechanism.
Although it is likely that the attenuated translation initiation induced by the impaired eIF4F system contributes to decreased p53 and mdm2 mRNA translation in response to IGF-1R inhibition, there might be additional mechanisms, including the regulation of translation elongation or termination on mRNA, for the observed effects of IGF-1R inhibition on p53 and mdm2 mRNA translation. We examined the impact of eIF4F complex disruption on translation using a dicistronic mRNA construct that contains the FLAG-tagged p53, Mdm2, or c-fos coding region flanked by the corresponding 5′ and 3′ UTRs and a GFP coding sequence (, left). The respective coding region was translated in a cap-dependent manner, whereas the translation of the gfp sequence is driven by the cricket paralysis virus (CrPV) internal ribosome entry site (IRES), which is independent of translation initiation factors (). We found that the expression of p53 and Mdm2 was down-regulated by AG1024 treatment, whereas the levels of c-fos were unaltered (, right). The cytomegalovirus (CMV) promoter in the constructs drove similar levels of gfp mRNA expression under all conditions (, right), thus excluding the possibility that there are differences in the promoter activity or transfection efficiency in AG1024-treated and -untreated cells. Importantly, GFP protein levels were unaltered after IGF-1R inhibition (, right), indicating that the initiation factor–independent translation is not inhibited. Interestingly GFP expression driven by the control vector (pIRES-GFP) was higher than that driven by other constructs, presumably because of the interference of the insert sequence (, right). Together, these findings suggest that translational control of p53 and Mdm2 expression by IGF-1R signaling is regulated at the level of initiation.
Weak mRNAs are subjected to gene-specific regulation under conditions that reduce the efficiency of translation initiation owing to the presence of long, highly structured 5′ UTRs (; ; ). We therefore predicted the secondary structures of the 5′ UTRs of p53, mdm2, and c-fos mRNA using the program MFOLD (; ). Consistent with the idea that the weak mRNA has a highly structured 5′ UTR, the sequences of the p53 and mdm2 5′ UTR but not the c-fos 5′ UTR were predicted to form several highly structured stem loops (unpublished data).
To determine whether the UTRs of p53 or mdm2 mRNA are sufficient on their own to mediate IGF-1R signaling– dependent translational regulation, we generated a series of constructs that contain a reporter sequence encoding firefly luciferase flanked by the UTRs of p53, mdm2, or c-fos mRNA () and then transfected the constructs into SK-hep1 cells. We found that in the absence of the flanking UTRs or the presence of c-fos UTRs, AG1024 does not inhibit the translation of the reporter mRNA (). In contrast, the translatability of reporter mRNA containing p53 or mdm2 UTRs was decreased by AG1024 treatment ().
Because the mechanisms by which the 5′ and 3′ UTRs confer translational control of specific mRNAs may be different (), we examined the impact of the 5′ and 3′ UTRs of p53 and mdm2 mRNA on translational efficiency using chimeric luciferase reporter constructs (). We found that the three reporter mRNAs (p53–CUTR–luc, Mdm2–CUTR–luc, and c-fos–CUTR–luc) lacking their respective 5′ UTRs were less translated (). Nonetheless, AG1024 inhibited the reporter mRNA translation in the presence of p53 or mdm2 5′ UTR but not in the presence of their respective 3′ UTRs (). In contrast, IGF-1R inhibition did not influence the luciferase activity of the reporter construct c-fos–NUTR–luc and c-fos–CUTR–luc (). Moreover, the IGF-1R inhibitor attenuated the translatability of hybrid reporter mRNA containing p53 or mdm2 5′ UTR and c-fos 3′ UTR (), further demonstrating that the translational control of p53 and mdm2 by IGF-1R inhibition is mediated by the respective 5′ UTR. Collectively, these data indicate that the 5′ UTR of p53 or mdm2 mRNA is sufficient to enable the IGF-1R signaling-dependent control of protein translation.
The PI-3K–Akt–mTOR (molecular target of rapamycin) pathway has been demonstrated to regulate general protein synthesis and translation of selected mRNAs (; ). We found that inhibition of PI-3K by LY294002, or mTOR by rapamycin, had no effect on p53 and Mdm2 expression (Fig. S5 A, available at ), which suggests an mTOR-independent mechanism for IGF-1R–mediated mRNA- specific translational regulation. Extracellular signal-regulated kinase (ERK) signaling has also been shown to promote translation by facilitating assembly of the translation initiation complex (). PD98059, a specific inhibitor of MAPK and ERK kinase, did not alter the amount of p53 and Mdm2 (Fig. S5 A). Furthermore, treatment of cells with LY294002 (rapamycin) or PD98059 did not affect luciferase activity driven by p53–UTR–luc or Mdm2–UTR–luc (Fig. S5 C). It therefore appeared that the PI-3K–Akt–mTOR and ERK pathway, although inactivated after IGF-1R inhibition, may not be involved in reducing p53 and mdm2 translation. It has been suggested that active glycogen synthase kinase (GSK)-3β phosphorylates and inhibits the translation initiation factor eIF2B (). Because IGF-1 signaling inactivates GSK-3β and promotes protein synthesis (), we examined whether inhibition of IGF-1R activity could reduce p53 and mdm2 translation through activation of GSK-3β. The reduction of Mdm2 levels in AG1024-treated SK-hep1 cells was not inhibited by GSK-3β inhibitors SB216763 or SB415286, which blocked β-catenin degradation (Fig. S5 B). Likewise, GSK-3β inhibitors had no effect on the luciferase activity of the chimeric reporter constructs (Fig. S5 D), further indicating that GSK-3β plays no part in the translational inhibition of p53 and Mdm2 by IGF-1R inactivation.
Although p53 is frequently mutated in >50% of human cancers (), a large fraction of cancers express wild-type p53, which may be regulated by other mechanisms such as amplification of Mdm2 () or deregulation of growth-factor signaling (; ). In this study, we demonstrate that inactivation of IGF-1R signaling impairs p53 accumulation after DNA damage through translational modulation of the p53–Mdm2 feedback loop. On the one hand, the translation of both p53 and mdm2 mRNA is attenuated upon IGF-1R inhibition. On the other hand, p53 protein becomes stabilized in response to IGF-1R inhibition because of reduced Mdm2 protein levels and is thus insensitive to further up-regulation of protein stability. IGF-1R inhibition therefore acts on p53 through two competing pathways (decreasing p53 protein synthesis and increasing p53 protein stability).
It is conceivable that p53 protein levels are determined by a balance between the opposing effects of IGF-1R signaling. In different cell types, the balance of the two competing pathways is likely to be different. Consistent with this idea, a lack of IGF-1R activity led to reduced p53 protein levels in MEFs (), whereas in HCT116 and SK-hep1 cells there was no detectable difference in p53 expression levels upon IGF-1R inhibition ( and ). Moreover, in MCF-7 cells the IGF-1R inhibitor up-regulated p53 protein levels with reduced p53 and mdm2 mRNA translation (unpublished data), further supporting the notion that the opposing effects of IGF-1R signaling on p53 are dependent on cell type.
Previous papers showing that activation of IGF-1R signaling decreases p53 expression in many systems are not contradictory to our findings of translational regulation of p53 by IGF-1R signaling, as these papers do not reveal whether IGF-1R signaling could regulate p53 mRNA translation (; ; ). In fact, our results indicate that a reduction in p53 mRNA translation by itself induced by IGF-1R inhibition may not always reflect and/or translate into a decline in p53 expression. Furthermore, IGF-1 signaling has been reported to be able to up-regulate p53 expression (). Thus, it is possible that the down-regulation of p53 expression upon IGF-1R activation that was observed in previous studies is cell-context dependent and additionally might be associated with an increase in p53 translation. Our results also provide a possible explanation for previous observations that Mdm2 expression is up-regulated by IGF-1 signaling (; ).
There are two general forms of translational control: mRNA-specific regulation and global control of protein synthesis (; ; ). Importantly, these two forms of regulation are not mutually exclusive (). We found that despite a reduction in global translation, the effect of IGF-1R inhibition on p53 and mdm2 mRNA translation is mRNA specific because the 5′ UTR of p53 and mdm2 mRNA rather than the 5′ UTR of c-fos mRNA imposed the translational regulation by IGF-1R signaling (), nor did we observe a change in c-fos and p27 mRNA translation after IGF-1R inhibition (). mRNA-specific regulation is either acquired by alterations of the general translational machinery or conferred by specialized mRNA binding factors (; ; ). Previous papers have documented a translational regulation of p53 and Mdm2 expression through the interactions of mRNA binding factors with the corresponding mRNAs (; ; ). Our findings from this study suggest a different mechanism by which IGF-1R signaling regulates p53 and mdm2 mRNA translation. We showed that IGF-1R inhibition led to reduced eIF4G expression and decreased eIF4E–BP1 phosphorylation ( and S4 D), both of which in turn attenuated the formation of the eIF4F complex and may impair cap-dependent translation initiation. Consistently, repression of p53 and mdm2 mRNA translation by IGF-1R inhibition was at the level of initiation, not elongation or termination, because there was no decrease in CrPV IRES–driven EGFP translation (). However, although it is likely that these observed inhibition effects are at least in part mediated by impairing the activity of the eIF4F complex, there could be additional mechanisms for the attenuated translation of p53 and mdm2 mRNA upon IGF-1R inhibition.
Many growth regulators are encoded by weak mRNAs, translation of which is highly eIF4F dependent and more sensitive to small perturbations in eIF4F complex formation (; ; ; ; ). The mechanisms of gene-specific translational regulation by IGF-1R signaling presented in this paper may therefore not be limited to regulation of p53 and Mdm2 but may rather be of general significance in translational regulation of gene expression. It will be interesting to determine how many genes could be regulated at the translational level by IGF-1R signaling and how many physiological effects of IGF-1R signaling could occur through translational effects. Although in our studies we show that IGF-1R signaling regulates p53 and mdm2 translation independent of Ras and the PI-3K–Akt–mTOR pathway, we cannot exclude the possibility that these pathways may be involved in IGF-1R–dependent translational regulation of other weak mRNAs.
Two well-documented hallmarks of cancer are deregulation of cell proliferation and evasion of apoptosis (). IGF-1R not only transmits mitogenic growth signals but also governs survival pathways, both of which are conducive to increased tumor growth (; ). However, IGF-1R signaling has also been proposed to be involved in inducing contradictory signals, including proapoptotic signaling (), on malignancy in different environments (; ), though how IGF-1R functions as a proapoptotic factor is unclear. The findings presented in this paper implicate IGF-1R as a proapoptotic factor by modulating the response of p53 to DNA damage.
Because p53 is involved in cellular responses to oxidative damage (), our findings provide an explanation for the increased resistance observed in mice when challenged with oxidants (). Our data is also consistent with the notion that growth signals have the potential to sensitize cells to apoptosis (). IGF-1R has been shown to be involved in TNF-α–induced apoptosis () and in a nonapoptotic form of cell death (), both of which seem not to depend on p53 function. Thus IGF-1R signaling can participate in both p53-dependent and -independent cell death. Together, these results provide an interesting contrast to other papers that showed that inactivation of IGF-1R sensitizes cells to apoptosis induced by chemotherapeutic drugs (). Yet as shown in our studies, IGF-1R inhibition not only impairs p53-dependent apoptosis but also inactivates the PI-3K–Akt and ERK pathways, which have been shown to be important for the antiapoptotic activity of IGF-1R signaling (; ; ). Therefore, upon IGF-1R inhibition, it is the balance between attenuated p53-dependent apoptosis and inactivated survival pathways that determines whether a cell survives or dies in response to stress. One might expect that the inclination of the balance would be dependent on cell type and the nature of apoptotic stimuli. Consistent with this idea, the loss of IGF-1R sensitized cells to doxorubicin- and Taxol-induced apoptosis (Fig. S2 B), although p53 induction was attenuated (Fig. S2 A).
Our findings may have important implications for the design of therapeutic protocols that involve the targeting of IGF-1R signaling. In tumors with functional p53, where p53 is critical for chemotherapeutic response (), small molecular therapy targeting IGF-1R, when used together with chemotherapy, may lead to the attenuation of cytotoxicity of chemotherapeutic drugs. However, because IGF-1R is important for cancer cell growth and survival, such therapy between courses of chemotherapy may well be useful ().
R MEFs lacking have been described previously (). R MEFs were obtained from R MEFs stably transfected with a plasmid containing human IGF-1R cDNA. Both cell lines were provided by R. Baserga (Thomas Jefferson University, Philadelphia, PA) and cultured in DME medium supplemented with 10% FBS (Invitrogen). SK-hep1, HCT116 , and cells (provided by B. Vogelstein, Johns Hopkins University, Baltimore, MD) were maintained in standard medium.
AG1024, LY294002, PD98059, rapamycin, etoposide, doxorubicin, Taxol, and cycloheximide (CHX) were obtained from Calbiochem. SB216761, SB415286, and ionomycin were obtained from Sigma-Aldrich. 7-methyl GTP (mGTP) and mGTP-Sepharose 4B were obtained from GE Healthcare.
For determination of the sub-G1 population, 10 MEFs were transfected with 8 μg of the indicated plasmids and combined with or without 1 μg cDNA coding for GFP. Transfections were performed using a transfection system (Nucleofector; Amaxa) according to the manufacturer's instructions. 70% transfection efficiency of cells was obtained using solutions and programs recommended by the manufacturer. For reporter assay, cells were transfected with the indicated reporter plasmids by jetPEI transfection reagent (Polyplus). The empty pCMV-Tag2B vector was added to adjust total DNA amount to 1 μg per well.
After electrophoresis and transfer of samples onto Immobilon membrane (Millipore), the blots were probed with the following antibodies: anti-Mdm2 (SMP14; Santa Cruz Biotechnology, Inc.; 2A10; Oncogene Research Products); anti–caspase-3, anti-PARP, anti–α-catenin, and anti-eIF4E (BD Biosciences); anti–eIF4E–BP1, anti-PABP, anti-pERK (T202/Y204), anti– p-p70 S6k (T389), anti-p70 S6k, anti-akt, anti–p–IGF-1R (Y1131), and anti-p27 (Cell Signaling Technology); anti-p53 (FL-393), anti–c-fos (H-125), anti–IGF-1R (C-20), anti–p-akt (Ser473), anti-ubiquitin (FL-76), anti-GFP (FL), and anti-actin (I-19; Santa Cruz Biotechnology, Inc.); anti-FLAG (Sigma-Aldrich); anti-eIF4A (provided by H. Trachsel, University of Bern, Bern, Switzerland); anti-eIF4G (provided by S. Morley, University of Sussex, Brighton, UK). The membranes were exposed to x-ray film (Kodak), which was scanned (Scanjet 3570c; Hewlett-Packard) using software (Photo and Imaging 2.0; Hewlitt-Packard). The analysis of the images was performed with imaging software (Photoshop 8.0; Adobe).
Cell death was determined according to the percentage of sub-G1 DNA content by flow cytometry. For untransfected cells, cells were collected and fixed with 70% cold ethanol overnight at −20°C. In transfected cells, after drug treatments for the indicated times, transfectants were collected and resuspended in 1% paraformaldehyde at room temperature for 10 min, centrifuged, and fixed in 70% cold ethanol at −20°C overnight. Fixed cells were then incubated in PBS containing 50 μg ml RNase A (Sigma-Aldrich) for 1 h at 37°C, followed by 30 μM propidium iodide (Sigma-Aldrich) staining. In each assay, either 10,000 (untransfected) or 50,000 (transfected) cells were collected by FACScan (BD Biosciences) and analyzed with software (WinMDI version 2.8; provided by J. Trotter, Scripps Research Institute, La Jolla, CA).
Cells cotransfected with the indicated constructs and the PRL-SV40 vectors were harvested in lysis buffer and analyzed using a luciferase assay reagent according to the manufacturer's instructions (Dual-Luciferase reporter assay system; Promega). The reporter activity was expressed as arbitrary luciferase units (firefly/renilla).
For [S]methionine/cysteine label analysis, cells were incubated in methionine/cysteine–free DME (Invitrogen) and supplemented with 10% dialyzed FBS (Invitrogen) for 1 h. 0.3 mCi [S]methionine/cysteine (GE Healthcare) was next added in 0.5 ml of free medium for 30 min. Lysates were prepared for the immunoprecipitation assay. Immunoprecipitated proteins were resolved by SDS-PAGE and visualized by PhosphorImaging. For quantitative analysis of total protein synthesis, lysates were prepared by standard protocols and the incorporation of S incorporation for cells was measured by using a liquid scintillation counter (LS 6500; Beckman Coulter).
For [S]methionine/cysteine pulse-label analysis cells were starved for methionine/cysteine for 1 h, and then pulse labeled with 0.75 mCi [S]methionine/cysteine for 1 h and chased for the indicated times with unlabeled methionine/cysteine (1 mg ml) added. The S-labeled p53 or Mdm2 in the immunoprecipitates from each time point was resolved and quantified by PhosphorImaging and normalized to that of the zero time point.
The mGTP pull-down assay was performed as described previously (). In brief, after washing with cold PBS, 2 × 10 cells were lysed in 1 ml NLB (50 mM Hepes, pH 7.4, 150 mM NaCl, 2 mM EDTA, 2 mM NaVO, protease inhibitor cocktail, and 0.5% NP-40) and then extracts were clarified by centrifugation at 10,000 (4°C for 10 min). Supernatants were then incubated with 1 ml NLB including mGTP-Sepharose 4B (60 μl of 50/50 slurry) at 4°C for 1 h. The beads were centrifuged at 2500 , and then washed with NLB. The mGTP-agarose was resuspended in 100 μl NLB containing 100 μM mGTP at 4°C for 30 min. The elute was collected and diluted with an equal volume of 2× SDS sample buffer and boiled. The eIF4E-bound proteins were analyzed by SDS-PAGE and immunoblotting.
Fig. S1 presents the expression analysis of p53 and p21 and the apoptosis analysis of R and R MEFs upon doxorubicin or Taxol treatment and the cell cycle distribution of R and R MEFs after etoposide treatment. Fig. S2 shows that inhibition of IGF-1R attenuates etoposide-induced p53 accumulation and apoptosis in tumor cells and has no effect on ionomycin-induced p53-independent apoptosis. Fig. S3 shows that inactivation of IGF-1R leads to enhanced p53 protein stability in tumor cells. Fig. S4 shows that inhibition of IGF-1R activity results in a reduced p53 translation. Fig. S5 shows that IGF-1R inhibition impairs p53 and Mdm2 translation through an ERK- and GSK-3β–independent and probably PI-3K–Akt–mTOR–independent mechanism. Online supplemental material is available at . |
Carcinogenesis, caused by physical, chemical, or viral mechanisms, is a multistage process of coordinated acquisition of favorable genetic lesions and complex interactions between tumor and host tissues that ultimately leads to an aggressive metastatic phenotype (). A mouse skin carcinogenesis model was well-established by application of 7,12-dimethylbenz(a)-anthracene (DMBA) and 12--tetradecanoylphorbol-13-acetate (TPA). DMBA is an initiator, whereas TPA is a classic tumor promoter. In this model, activator protein-1 (AP-1) activation appears to play a causal role in skin tumor promotion (). The AP-1 transcription factor family consists of 18 different combinations of Jun-Jun or Jun-Fos proteins as well as the closely related ATF and MAF transcription factors. The Jun family of proteins includes c-Jun, JunB, and JunD, whereas the Fos family of proteins includes c-Fos, FosB, Fra-1, and Fra-2 (). The and genes are inducible by a broad range of extracellular stimuli. Upon activation, AP-1 binds to TPA-response elements 5′-TGAG/CTCA-3′ to transactivate many effector genes, thus regulating cell proliferation, tumor promotion, cell cycle progression, growth arrest, and apoptosis ().
AP-1 was first considered as a mediator of tumor promotion because of its ability to alter gene expression in response to tumor promoters such as TPA and UV irradiation (). Indeed, TPA and UV, as well as reactive oxygen species, activate AP-1 (). Both TAM67- and c-fos–deficient mice have been used to establish the role of AP-1 in skin carcinogenesis induced by UV and DMBA/TPA (; ; ). TAM67 is a transactivation domain deletion mutant of c-Jun that acts to sequester Jun and Fos family proteins in low activity complexes (). Overexpression of TAM67, driven by a K14 promoter, reduces AP-1 activity and dramatically inhibits the formation of tumors induced by DMBA/TPA (), as well as of squamous cell carcinoma induced by UV (). In contrast, c-Fos–deficient mice are resistant to the malignant progression of skin tumors (). Suppression of DMBA/TPA-induced tumor formation in manganese superoxide dismutase–overexpressing transgenic mice is also associated with modulation of AP-1 signaling (). Thus, AP-1 activation is required for both DMBA/TPA- and UV-induced skin carcinogenesis.
Sensitive to apoptosis gene (SAG) was initially cloned as a redox-inducible gene that encodes an evolutionarily conserved really interesting new gene (RING) finger protein () and was later found to be the second family member of regulator of cullins-1 (ROC1)/RING box protein 1 (; ; ; ; ), the RING component of the Skp1-cullin1–F-box protein (SCF) E3 ubiquitin ligases that promotes ubiquitination and degradation of a variety of protein substrates (). SAG/ROC/Rbx and cullins form the core ubiquitin ligase, whereas the F-box proteins determine its specificity by recognizing the substrates (). We have recently established an autofeedback loop in which SAG is a direct transcriptional target of AP-1. Upon induction by AP-1, SAG promotes ubiquitination and degradation of c-Jun to inhibit AP-1 activity and AP-1–induced neoplastic transformation in a mouse epidermal cell model (). We extended this work to an in vivo transgenic model in which SAG expression is driven by a K14 promoter, and we report that SAG, upon targeted expression in the epidermis where an F-box protein (Fbw-7) is expressed, reduces TPA-induced c-Jun levels, inhibits AP-1 activity, cell proliferation, and eventually DMBA/TPA-induced skin carcinogenesis. However, SAG expression in tumor tissues, where another F-box protein, β-transducin repeat-containing protein 1 (β-TrCP1), is overexpressed, reduces inhibitor of κBα (IκBα) levels and activates nuclear factor κB (NF-κB), resulting in apoptosis inhibition and enlarged tumor size. Thus, it appears that SAG inhibits tumor formation at the early stage by targeting c-Jun/AP-1 and promotes tumor growth at the later stage by targeting IκBα/NF-κB in a manner dependent on the availability of F-box proteins.
We have recently shown that SAG is a novel AP-1 target and that, upon induction, SAG inhibits AP-1 activity and AP-1–induced neoplastic transformation by promoting c-Jun ubiquitination and degradation in cultured cells (). Because AP-1/c-Jun is actively involved in promoting skin carcinogenesis induced by DMBA/TPA (; ), we tested our hypothesis that SAG would act as an inhibitor of skin carcinogenesis. A SAG transgenic construct, which, driven by the K14 promoter, targets gene expression mainly to the epidermis (; ), was made () and used to generate the K14-SAG transgenic lines in the mouse FVB/N strain that is sensitive to DMBA/TPA carcinogenesis (). Out of a total of 106 mice produced, after three consecutive microinjections of K14-SAG transgenic construct into FVB/N one-cell embryos, we identified 13 K14-SAG transgenic mice via PCR genotyping (not depicted). These positive mice were followed up for the potential expression of the SAG transgene by RT-PCR. Expression of trans-SAG mRNA was detected and confirmed with sequencing (not depicted). To determine the gene copy number, Southern blot analysis was performed in each transgenic line. Two lines, 345 and 710, which contained intact transgene in a single insertion site with transgene copy numbers of ∼20 and 30, respectively, were identified (not depicted) and chosen for Northern analysis for mRNA expression. As shown in , compared with nontransgenic lines, which expressed a low endogenous level of SAG mRNA, both lines had increased levels of SAG mRNA, with a higher level seen in the 710 line, reflecting a good correlation between transgene copy numbers and SAG mRNA levels. Western blot analysis of epidermal skin tissues also showed a copy number–dependent increase of FLAG-tagged transgenic SAG protein, with a level two- to threefold higher than that of the endogenous SAG (, FLAG-SAG and Endo-SAG). These two transgenic lines, 345 and 710, with different levels of SAG transgenic expression, were chosen for a two-stage DMBA/TPA carcinogenesis study.
Our previous work in cultured JB6 epidermal cells has shown that SAG inhibits TPA-mediated c-Jun induction by promoting c-Jun ubiquitination and degradation (). We extended this observation to primary keratinocytes isolated from several 1- to 2-d-old postnatal pups from SAG transgenic (SAG-Tg) lines 345 and 710, as well as control nontransgenic littermates. Two pairs of primary keratinocytes were either treated with DMSO vehicle control or TPA for 4 h. As shown in , TPA treatment induced c-Jun expression up to 8- to 10-fold in keratinocytes from nontransgenic littermates, whereas the same treatment caused only a three- to fourfold c-Jun induction in keratinocytes from SAG transgenic lines (, top). Expression of transgenic SAG was readily detectable in cells from transgenic lines, but not from the controls (, middle). The results demonstrated that SAG transgenic expression inhibits TPA-induced c-Jun accumulation. We further confirmed this by in situ immunohistochemical (IHC) analysis. As shown in , c-Jun was not detectable in the epidermal layer of acetone-treated mouse skin (top left). TPA application induced strong c-Jun nuclear staining in the majority of epidermal cells in SAG-Tg(−) mice (, middle, inset for higher magnification). In SAG-Tg(+) mice, both the number of cells with c-Jun nuclear staining and the staining intensity in c-Jun–positive cells were reduced (, bottom, arrows in inset for c-Jun–negative staining cells). No difference in epidermal staining of two control proteins, c-Fos and p53, was observed between the two lines. The c-Fos (, middle column) is a strong TPA-inducible protein, known to be degraded by UBR1 E3 ligase (), whereas p53 (right column), a weak TPA responder, is mainly degraded by Mdm2 E3 ubiquitin ligase (). To further demonstrate the effect of SAG transgenic expression on c-Jun levels induced by TPA in situ, two consecutive sections in the same areas of mouse skin tissues from SAG-Tg(+) and SAG-Tg(−) mice were prepared, respectively, with one section stained for FLAG-SAG transgenic expression (, top) and the other for c-Jun expression (bottom). An inverse relationship between SAG transgenic expression (detected by anti-FLAG antibody) and c-Jun induction by TPA was clearly shown. Thus, SAG, upon ectopic expression, reduced TPA-induced c-Jun levels both in primary cultures and in mouse skin in vivo. It is noteworthy that the epidermal layers were thicker in SAG-Tg(−) mice than that in SAG-Tg(+) mice (see below).
We next determined whether expression of the SAG transgene inhibited AP-1 binding and transactivation as a result of c-Jun ubiquitination and degradation in in vivo mouse skin tissues collected from SAG-Tg(+) mice and their negative littermates, SAG-Tg(−). As shown in , application of acetone solvent control did not induce any AP-1 DNA binding activity, regardless of SAG transgenic expression (lanes 1–4). However, application of TPA twice a week for 4 wk produced substantial induction of AP-1 binding in three out of three SAG-Tg(−) lines tested (, lanes 5–10). Significantly, SAG transgenic expression caused a remarkable reduction of TPA-induced AP-1 binding in three out of three SAG-Tg(+) mice (, lanes 12–18). This dramatic difference in AP-1 binding was not due to load discrepancy, as demonstrated by similar levels of nuclear histone H2A among all eight samples (). Furthermore, because AP-1 binding can be supershifted by c-Jun antibody (, lanes 6, 8, 10, 13, 15, and 17) and completely blocked by 50× cold oligonuclotide (lanes 11 and 18), it was concluded that the AP-1 complex contained c-Jun and the binding was specific. Thus, short term TPA application caused substantial AP-1 activation, which was remarkably inhibited by SAG transgenic expression.
To further confirm whether reduced AP-1 binding can be correlated with reduced AP-1 transactivation in vivo, we crossed AP-1 luciferase reporter mice () with SAG-Tg(+) mice to generate mice that are heterozygous for AP-1–Luc reporter in SAG-Tg(+) or SAG-Tg(−) background, respectively. The AP-1 activity, as reflected by luciferase light units, was significantly induced in ear-punched tissues after TPA application. SAG transgenic expression significantly reduced this AP-1 activity (, P < 0.01). Collectively, it is clearly demonstrated in both in vitro and in vivo settings that SAG, upon transgenic expression, significantly reduced TPA-induced AP-1 activities (both DNA binding and transactivation) by reducing TPA-induced c-Jun accumulation.
It is well established that, at the early stage of DMBA/TPA carcinogenesis, TPA induces hyper-proliferation of the epidermis (), likely through AP-1 activation. We therefore determined that SAG transgenic expression, which inhibits AP-1 activity, would block TPA-induced hyperproliferation with two independent assays. The first assay was hematoxylin and eosin (H&E) staining, followed by measurement of the thickness of the epidermal layer. Representative figures, shown in as well as , clearly demonstrate that SAG transgenic expression remarkably reduces the thickness of the epidermis. The difference is statistically significant (P < 0.01). The second assay was the BrdU incorporation assay, which measures cells with active DNA synthesis. As shown in , BrdU-labeled cells were mainly localized in the basal cell layer. The number of these positive cells was significantly reduced upon SAG transgenic expression (P < 0.01). To further demonstrate the effect of SAG transgenic expression on cell proliferation in situ, two consecutive sections in the same areas of skin tissues from SAG-Tg(+) and SAG-Tg(−) mice were prepared, respectively, with one section stained for FLAG-SAG transgenic expression (, top) and the other for BrdU staining (bottom). An inverse relationship between SAG transgenic expression and the number of BrdU-positive cells was clearly shown. Thus, it appears that SAG transgenic expression reduces the thickness of the epidermis by inhibiting the proliferation of epidermal cells, likely through SAG-induced c-Jun reduction and AP-1 inactivation at the early stage of DMBA/TPA carcinogenesis.
Finally, we determined if SAG transgenic expression alters the differentiation pattern of the epidermal cells in response to TPA treatment by IHC staining of keratin-5, -6, and -10, three commonly used skin differentiation markers (; ). Whereas there is no difference in keratin-5 and -6 staining in the skin epidermal layers between TPA-treated SAG-Tg(−) and SAG-Tg(+) mice (unpublished data), keratin-10 staining does reveal some differences. Several layers (≥2) of keratin-10–negative basal cells were readily detected in some areas of the skin from SAG-Tg(−) mice, whereas in SAG-Tg(+) mice, keratin-10–negative cells were essentially restricted to the single basal cell layer (, brackets). Moreover, a small subset of basal cells in SAG-Tg(+) mice appeared to express keratin-10 precociously. These results suggest that SAG transgenic expression alters the inhibitory effects of TPA on expression of spinous cell differentiation markers ().
Two independent SAG transgenic lines, 345 and 710, with different levels of SAG transgenic expression, along with SAG-Tg(−) littermates, were used to determine the effect of SAG expression and SAG dosage on tumor formation during DMBA/TPA two-stage carcinogenesis. A single application of DMBA (100 nmol/0.2 ml acetone) as initiator was given, followed by TPA (5 nmol/0.2 ml acetone, twice a week) promotion for 20 wk. As shown in , FVB/N mice are quite sensitive to the DMBA/TPA two-stage carcinogenesis protocol, and papillomas started to appear after 4–6 wk of TPA promotion in both groups (top). It should be noted that almost 95% of nontransgenic 345 mice developed papillomas after 9 wk of TPA application, whereas only 40% of nontransgenic 710 mice developed papillomas (, top). This apparent discrepancy of incidence in control mice is likely caused by the use of TPA from two different vendors (TPA from Sigma-Aldrich was used for the 345 line, whereas TPA from Qbiogene was applied to the 710 line). Nevertheless, the percentage of mice with one or more papillomas was lower in SAG-Tg(+) mice than that in SAG-Tg(−) littermates throughout the promotion period (, top). Statistical analysis revealed a significant difference between the SAG-Tg(+) and SAG-Tg(−) mice in the greater SAG-expressing 710 line (P < 0.05), but not in the lower SAG-expressing 345 line (P > 0.05), suggesting a SAG dosage effect. Remarkably, the mean number of papillomas per mouse was significantly lower in both SAG-Tg(+) lines than that in SAG-Tg(−) littermates (P < 0.0001; , middle). These data clearly demonstrate that SAG transgenic expression inhibits skin papilloma development induced by DMBA/TPA. Interestingly, when mean tumor size was measured, the papillomas in both SAG-Tg(+) lines were statistically bigger than those of their transgenic negative littermates (, bottom; P = 0.0026 and P = 0.0043 for the 710 and 345 lines, respectively). The observed decrease of mean tumor size in the 710 line between weeks 13.5 and 14.5, and weeks 18.5 and 19.5 (, bottom right), was caused by the death of two mice bearing large tumors. These results demonstrate that SAG transgenic expression promoted tumor growth after tumor formation.
We also measured the rate of conversion from papilloma to carcinoma in the 345 line after 20 wk of TPA promotion and in the 710 line after 29 wk of TPA promotion (9 wk after termination of 20 wk of TPA application). The data are summarized in . In the 345 line, the conversion rate was very low (<10%), with no statistical difference between SAG-Tg(+) and SAG-Tg(−) mice. In the 710 line, the conversion rate was much higher, reaching 25%, as expected, caused by the additional 9 wk of tumor growth after TPA promotion. But again, no statistical difference (P > 0.05) was observed between the SAG transgenic lines and their negative littermates, although invasive squamous carcinomas were more frequently seen in SAG-Tg(+) mice. Thus, SAG transgenic expression has no significant role in promoting papilloma-to-carcinoma conversion.
The increased tumor size seen in SAG-Tg(+) lines could result from either increased proliferation or decreased apoptosis in the papillomas. To determine which, we first assayed BrdU incorporation in papillomas derived from SAG-Tg(−) and SAG-Tg(+) mice. As shown in , taken from two representative areas of two independent tumors, many BrdU-positive cells were seen across the tumor tissues, indicating that the rate of BrdU incorporation was much higher in tumors than in skin tissues after just a few TPA exposures (). However, no difference was found between the two groups, indicating that SAG transgenic expression had no effect on DNA synthesis, thus excluding the potential difference in the rate of cell proliferation. We next determined the rate of apoptosis using the TUNEL assay, because we have previously shown that SAG, upon overexpression, protects cells or tissues from apoptosis (; ; ). In two representative areas of two independent tumors shown in , TUNEL-positive cells were either localized at the junction of living and keratinized cells () or scattered across the tumor tissues (). In either case, the number of TUNEL-positive cells was remarkably reduced in papillomas derived from the SAG-Tg(+) mice (), compared with those from the SAG-Tg(−) mice (). The difference was statistically significant (; P < 0.01). Similar results were observed using IHC staining for cleaved active caspase-3 (). Cells with active caspase-3 staining, indicative of apoptosis, were either clustered in one region () or scattered across tumor tissues (). More active caspase-3 staining was observed in tumors derived from SAG-Tg(−) mice than in those derived from SAG-Tg(+) mice (; P < 0.01). Thus, increased tumor size seen in SAG-Tg(+) mice was not caused by accelerated proliferation, but rather by reduced apoptosis.
To determine the potential mechanism by which SAG expression inhibited apoptosis, we measured p65 nuclear translocation and NF-κB activation in papillomas derived from SAG-Tg(−) and SAG-Tg(+) mice. NF-κB is a well-known cellular survival factor that, upon activation, protects cancer cells from apoptosis (). NF-κB (a heterodimer of p50/p65) is activated after translocation from cytoplasm to the nucleus after the degradation of IκB, a substrate of SCF–β-TrCP1 E3 ubiquitin ligase (). It has been previously shown that constitutive expression of a SAG phosphoactive form leads to reduction of IκBα levels (), suggesting that SAG could activate NF-κB by promoting IκBα degradation. As shown in , IHC staining revealed that, although the cell boundary is less evident because of the nature of p65 immunostaining, p65 was clearly localized outside of the nucleus but inside the cytoplasm of the majority of tumor cells derived from SAG-Tg(−) mice (, arrows in inset for tumor cells without nuclear staining). In contrast, p65 was translocated to the nucleus, with strong nuclear staining in the majority of tumor cells from SAG-Tg(+) mice (, arrows in inset for positive staining cells). Thus, SAG transgenic expression promotes p65 translocation from the cytoplasm to the nucleus.
We next applied a DNA-binding gel retardation assay to determine if NF-κB is activated. HeLa cells, treated with TNF-α, were used as positive controls (, lanes 1–4). Consistent with the data concerning p65 localization, although mainly localized in the cytoplasm, NF-κB was inactive and failed to bind to its consensus DNA sequence in five out of five papillomas harvested from five individual SAG-Tg(−) mice (, lanes 5–10). In contrast, nuclear localized NF-κB was active and bound to its consensus DNA sequence in four out of four papillomas harvested from four individual SAG-Tg(+) mice (, lanes 11–18). Because the NF-κB DNA binding can be supershifted by the p65 antibody (, lanes 3, 14, and 17) and blocked by specific cold oligonucleotides (, lanes 4, 15, and 18), we concluded that it is specific. To determine the potential changes in AP-1 activity in these tumors, the same samples were subjected to an AP-1 binding assay (). Compared with the skin tissue exposed to TPA (, lanes 1–3), the AP-1 binding activity in these tumor samples was very low (, lanes 4–16) and varied between tumors. However, no difference was observed between SAG-Tg(−) (, lanes 4–10) and SAG-Tg(+) (lanes 11–16) mice, indicating that, in tumor tissues, AP-1 is largely inactive, regardless of SAG transgenic expression. Collectively, we concluded that the increased tumor size seen in SAG-Tg(+) mice was caused by reduced apoptosis, resulting at least in part from SAG-mediated p65 nuclear translocation, followed by NF-κB activation.
SAG is a RING component of the SCF E3 ubiquitin ligases, whose substrate-degrading activity requires F-box proteins that recognize substrates (). It is well established that the F-box proteins that recognize c-Jun and IκBα are Fbw7/Cdc4 (; ) and β-TrCP1 (), respectively. We therefore determined the levels of c-Jun, IκBα, Fbw7, and β-TrCP1 during DMBA/TPA-induced skin carcinogenesis. Consistent with results in primary keratinocytes (), the c-Jun levels were very low or undetectable in mouse skin tissues treated with acetone solvent (, top, lanes 1 and 2), and dramatically induced after TPA treatment in SAG-Tg(−) mice (, top, lanes 3 and 4). TPA-mediated c-Jun induction was substantially reduced in SAG-Tg(+) mice (, lanes 5 and 6). The c-Jun levels were, however, very low in tumor tissues regardless of SAG transgenic expression (, lanes 7–9). In contrast, the levels of IκBα (, second panel) were quite high in skin tissues as well as tumor tissues from SAG-Tg(−) mice (lanes 1–7), but remarkably reduced in tumor tissues from SAG-Tg(+) mice (lanes 8 and 9). The levels of β-TrCP1 (, third panel) were very low or undetectable in skin tissues (lanes 1–6), but remarkably increased in tumor tissues, regardless of SAG transgenic expression (lanes 7–9). Because the Fbw7 levels were found to be very low in the cells, and could only be detected by immunoprecipitation-coupled Western blotting analysis (), we performed the analysis using the same Fbw7 antibody (). Fbw7 was only detected in skin tissues (, lanes 1–4), but not in tumor tissues (, lanes 5–8), regardless of SAG transgenic expression.
We further measured the expression of c-Jun, IκBα, β-TrCP1, and transgenic SAG in four independent tumors from four individual SAG-Tg(−) or SAG-Tg(+) mice, respectively (). Again, very low levels of c-Jun (, top) and relatively high levels of β-TrCP1 (third panel) were detected, regardless of SAG transgenic expression, which was only shown in SAG-Tg(+) tumors (fourth panel). In contrast, the level of IκBα (, second panel) was quite high in tumors from SAG-Tg(−) mice, but dramatically reduced in tumors from SAG-Tg(+) mice, strongly suggesting that SAG transgenic expression promotes IκBα degradation when β-TrCP1 levels are high enough. Finally, we measured the expression of other components of SCF E3 ubiquitin ligase, including Cul-1, Skp1, and ROC1, a SAG family member. They were all expressed constitutively, with no difference between mouse skin tissues and tumors (), indicating that they are unlikely to be critical in regulation of skin carcinogenesis. Collectively, these results showed that SAG regulation of AP-1 and NF-κB activity is dependent of the levels of F-box proteins Fbw7 and β-TrCP1 with SAG as a rate-limiting factor.
The SCF E3 ubiquitin ligases comprise the largest family of E3 ubiquitin ligases that ubiquitinate a variety of regulatory proteins for 26S proteasome degradation (). The core SCF E3 ubiquitin ligase is a complex of ROC1–cullin or SAG–cullin that recruits E2, whereas the substrate specificity of the SCF complex is determined by the F-box proteins that bind to Skp1 and cullin through their F-box domain, and to substrates through their WD40 or LRR domains (). The SCF E3 ligase promotes ubiquitination and degradation of c-Jun through the F-box protein Fbw7/Cdc4 (; ; ), whereas the SCF E3 ligase promotes ubiquitination and degradation of IκB through the F-box protein β-TrCP1 (), which leads to NF-κB activation and apoptosis protection. Thus, SCF E3 ubiquitin ligases regulate cell cycle progression and apoptosis through timely ubiquitination and degradation of their regulators ().
SAG is a stress-responsive protein and is inducible by a variety of stimuli, including hypoxia, metal ions, and the tumor promoter TPA (; ; ). Upon induction or overexpression, SAG acts as a survival protein to protect cells and normal tissues from apoptosis in several in vitro and in vivo models (; ; ). We have recently found that, as a RING component of SCF E3 ubiquitin ligases, SAG promotes ubiquitination and degradation of c-Jun, an essential member of the AP-1 transcription factor. As a consequence, TPA-induced and AP-1-mediated neoplastic transformation in the JB6 epidermal cell model is inhibited upon SAG expression or enhanced upon SAG siRNA silencing (). We demonstrated, using an in vivo transgenic mouse model, that SAG, upon targeted expression in mouse skin epidermal cells driven by K14 promoter, inhibited DMBA/TPA-induced skin carcinogenesis, as indicated by a prolonged latent period and a reduced tumor frequency. This is very likely attributable to inhibition of TPA-induced c-Jun accumulation, followed by inactivation of AP-1 activity at the early stage of carcinogenesis. The fact that targeted SAG transgenic expression is two- to threefold higher than that of endogenous SAG suggested that SAG is a rate-limiting factor for the modulation of activity of SCF E3 ubiquitin ligases in mouse skin or tumors induced by DMBA/TPA. Indeed, our data showed that the two- to threefold increase in SAG caused a two- to threefold reduction in TPA-induced c-Jun levels and TPA-induced AP-1 binding and transactivation, which also correlated with an approximately twofold reduction in papilloma formation. Two previous studies using transgenic mice expressing TAM67, a dominant-negative form of c-Jun, also showed a good correlation between inhibition of AP-1 and inhibition of skin carcinogenesis. It was observed that a twofold decrease in UVB-induced AP-1 activation in the epidermis of TAM67 transgenic SKH-1 mice correlated with a twofold reduction in the number of skin tumors induced by UVB (), whereas a 4.5-fold reduction of TPA-induced AP-1 activation correlated well with a fivefold inhibition of papilloma formation (). Collectively, it appears that blockage of c-Jun/AP-1 activity has a dose-dependent effect on skin carcinogenesis.
It appeared paradoxical that the same SAG protein, upon targeted expression, would inhibit tumor formation at the early stage, but promote tumor growth at the later stage of carcinogenesis. This may be explained by differential expression of two F-box proteins, Fbw7/Cdc4 and β-TrCP1, during carcinogenesis. Fbw7/Cdc4 is a p53-dependent, haploinsufficient tumor suppressor gene (). Mutational inactivation and loss of heterozygosity of Fbw7 is seen in several human cancers (; ). Fbw7 binds to phosphorylated c-Jun and promotes its ubiquitination and subsequent degradation. As a result, SCF shortened the c-Jun protein's half-life and inhibited AP-1 activity and apoptosis. Consistent with this, knock down of Fbw7 expression via RNAi increased the level of phosphorylated c-Jun, followed by enhanced AP-1 activity and apoptosis induction (; ; ). We show in this paper that at the early stage of DMBA/TPA carcinogenesis, Fbw7 is expressed and cooperates with rate-limiting SAG to promote c-Jun degradation, leading to reduction of AP-1 activity ( and ) and inhibition of proliferation and tumorigenesis ( and ). In the late stage of carcinogenesis, when tumors are formed, the expression of Fbw7 is down-regulated and undetectable (). Although the mechanism is still elusive, our observation of loss of Fbw7 expression during skin carcinogenesis suggests that it is a change favored and selected for in tumor formation. It appears that the lack of Fbw7 prevents SAG's ability to promote c-Jun degradation, leading to a similar AP-1 activity (although very low, due to the lack of TPA induction) in tumors from SAG-Tg(−) and SAG-Tg(+) mice (). Furthermore, because AP-1 is not activated in the tumors () and c-Jun levels are very low in these tumors, regardless of SAG transgenic expression (), our model strongly suggests that AP-1 is unlikely to play a considerable role in the growth of DMBA/TPA-induced tumors, although c-Jun/AP-1 has been implicated in apoptosis regulation that could influence the tumor growth in many other systems ().
In contrast to Fbw7, β-TrCP1 was found to be overexpressed in colon cancers with associated activation of both β-catenin and NF-κB and inhibition of apoptosis (). Increased levels of β-TrCP1 were also found in pancreatic carcinoma cells, which correlated with constitutive NF-κB activation and chemoresistance (). Likewise, targeting β-TrCP1 via siRNA silencing or overexpression of a dominant-negative mutant suppressed growth and survival of human breast cancer cells (). Furthermore, mouse β-TrCP2 was overexpressed in skin tumors generated by DMBA-TPA two-stage carcinogenesis, causing a constitutive activation of NF-κB through reduction of IκBα (). Consistent with this, we found that, regardless of SAG transgenic expression, the levels of β-TrCP1 were very low or not detectable in mouse skin, but significantly elevated in tumors induced by DMBA/TPA, likely caused by the activation of Ras-Raf pathways as a result of DMBA-initiated Ras mutation (; ). However, only SAG-overexpressing tumors had reduced IκBα levels and activated NF-κB activity, suggesting that SAG is a rate-limiting factor in SCF–β-TrCP1–mediated IκBα degradation ( and ). It is worth noting that NF-κB has been previously shown to regulate spontaneous skin carcinogenesis with inconsistent results. Upon targeted overexpression by a K5 promoter of a superrepressor form of IκBα (leading to continued inactivation of NF-κB) in mouse skin, an increased basal rate of apoptosis and the spontaneous development of squamous cell carcinomas were observed (). In contrast, targeted ablation of α-catenin, an adherens junction protein, resulted in NF-κB activation and formation of squamous cell carcinoma in the skin (). Nevertheless, our results showed that NF-κB was marginally activated at the early stage (unpublished data), but significantly activated in tumors from SAG-Tg(+) mice, resulting in apoptosis inhibition in the DMBA/TPA two-stage carcinogenesis model.
Based on these observations, we propose a model to elucidate the role of SAG in skin carcinogenesis (). At the early stage of carcinogenesis, SAG is induced by carcinogens/tumor promoters via AP-1 transactivation as a cellular protective mechanism (). Upon induction, SAG cooperates with the tumor suppressor Fbw7 to promote ubiquitination and degradation of c-Jun, thus inactivating AP-1 and inhibiting carcinogenesis. At this stage, because the β-TrCP1 level is very low, SAG had little, if any, effect on NF-κB. However, at the later stage, when tumors are formed, SAG, in the presence of overexpressed β-TrCP1, promotes IκBα degradation, leading to the activation of NF-κB, inhibition of apoptosis, and enlargement of tumor size. At this stage, SAG has little, if any, effect on AP-1 because TPA is no longer present to induce AP-1 and Fbw7 is no longer detectable. Indeed, SAG overexpression was detected in a subset of human colon cancers and nonsmall cell lung carcinomas and was associated with worse patient prognoses (; ). Thus, SAG targeting for induction or inhibition appears to be stage dependent. It might be beneficial to induce SAG in normal tissues or at the early stage of carcinogenesis for cancer prevention and to inhibit SAG at the late stage, when tumors have formed, for cancer therapy via apoptosis induction ().
A 375-bp cDNA fragment encoding FLAG-tagged human SAG protein with a BamHI site was subcloned into the single BamHI site of the K14/human growth hormone (hGH) expression vector (; ). The construct was digested with HindIII–KpnI; the fragment containing K14-promoter–FLAG-SAG PolyA was excised and injected into a FVB/N mouse egg by the Transgenic Core at the University of Michigan. The University of Michigan Animal Care and Use Committee approved the procedures for the use of laboratory animals. Transgenic mice were genotyped by direct PCR using tail biopsy DNA with primers FLAG-SAG–Tg01 (5′-CGGGATCCGCCACCATGGACTACAAGGACGACGATGACAAGGCCGACGTGGAAGACGGAG-3′) and hGH-02 (5′-GGGAATGGTTGGGAAGGCACTG-3′; ). Primary murine skin keratinocytes were isolated from the epidermis of 1- to 2-d-old mice ().
Analyses were performed as described previously (). The antibodies against Fbw7, c-Jun, IκBα, Skp1, cullin-1, and histone H2A were obtained from Santa Cruz Biotechnology, Inc. β-TrCP1 from Zymed Laboratories, β-actin from Sigma-Aldrich, and SAG and ROC1 as described previously ().
The skin or tumor specimens were excised and fixed in 4% PFA-PBS for 3 d at 4°C and embedded in paraffin. 4-μm-thick sections were cut for immunostaining with ABC kits (Vector Laboratories). The primary antibodies used were against c-Jun, c-Fos, p53 (Santa Cruz Biotechnology, Inc.), FLAG-tag (Sigma-Aldrich), cleaved caspase-3 (Cell Signaling Technology), mouse keratins-5 and -6 (Covance), and mouse keratin-10 (Lab Vision). Normal goat serum was used as a negative control. The papilloma-to-carcinoma conversion was determined through histopathological observation after H&E staining. Keratin-14 (Vector Laboratories) was stained to show the origin of the tumor cells. A microscope (BX51; Olympus) with objective lenses (UplanFI 40×; Olympus) was used. The experiments were performed at a temperature of 25°C. A mounting medium was used for the imaging medium (Cytoseal 60; Richard-Allan Scientific). The fluorochromes used were green fluorescence and propidium iodide (For TUNEL assay). A camera (DP70; Olympus) and acquisition software (DP70-BSW, version 01.02; Olympus) were also used. All images, pictures, and figures were prepared with Illustrator CS2 and Photoshop CS2 (Adobe).
Nuclear extracts were prepared from mouse skin and tumors as described previously (; ) and subjected to gel retardation assay (), using oligonucleotides containing a typical AP-1 binding site, AP-1-01 (5′-CGCTTGATGAGTCAGCCGGAA-3′), or NF-κB site, NFκB-01 (5′-AGTTGAGGGGACTTTCCCAG-3′).
SAG-Tg(+)/710 line mice were crossed with AP-1–luc SKH mice (). The SAG–AP-1–Luc mice were typed by PCR analysis of tail DNA. The primers used were Luc-01 (5′-GCGGAATACTTCGAAATGTCCGTTCGGTTGG-3′) and Luc-02 (5′-CCTTAGGTAACCCAGTAGATCCAGAGGAATTC-3′). The SAG primers were FLAG-SAG–TG-01 and hGH-02. AP-1–Luc activity was measured in a total of 10 AP-1–Luc(+)/SAG(−) mice and seven AP-1–Luc(+)/SAG(+) mice. The mice were treated with 50 μl of acetone on the left ear and 50 μl TPA (5 nmol) on the right ear. After 24 h, three ear punches (per ear) of 3 mm were collected. Tissues were ground with a microtube pestle and lysed in 1× passive lysis buffer for 1 h, followed by luciferase activity measurement ().
After a single application of DMBA followed by eight treatments with TPA, three mice from the SAG-Tg(+) line and three of their SAG-Tg(−) littermates were killed 24 h after the last TPA treatment, along with two mice in each group treated with acetone controls. Samples were collected and 4-μm sections were stained with H&E. A total of 16 measurements were made for each mouse in a systematic manner under a microscope, starting from the top of the basement membrane to the bottom of the strateum corneum. The skin thickness from each mouse was then averaged within the group, expressed as milimeters after 400× (mm/400), and subjected to Student's test for statistical difference.
SAG-Tg(+) and SAG-Tg(−) mice (for 345 line: K14-SAG, = 15, and control, = 13; for 710 line: K14-SAG, = 17, and control, = 17) at 7–8 wk of age were used for the DMBA/TPA two-stage carcinogenesis protocol. A single dose (100 nmol) of DMBA (Sigma-Aldrich) in 0.2 ml of acetone was applied topically to the shaved backs of mice. 2 wk after initiation, TPA (5 nmol; Sigma-Aldrich for 345 line and Qbiogene for 710 line) in 0.2 ml of acetone was applied twice weekly to the skin for 20 wk. The presence of the tumor and the size of these tumors were measured twice a week using a digital caliper (Fisher Scientific).
SAG-Tg(+) and SAG-Tg(−) mice were injected i.p. with 2 ml/100g (body weight) BrdU-labeling reagent. Mice were killed 2 h later and skin tissues or tumors with diameters of 5–7 mm were collected and fixed in 4% PFA-PBS. The assay was performed with a 5-bromo-2′-deoxy-uridine labeling and detection kit II (Roche). Slides were developed with NBT/BCIP and counterstained with Eosin (Richard-Allan Scientific). For quantification of BrdU-positive cells in mouse skin, two random areas were selected from each mouse (three mice per group), and the number of positive cells out of a total of 1,000 epidermal cells was counted. For quantification of BrdU-positive cells in tumors, two random areas were selected from each tumor. The number of positive cells out of a total of 1,000 cells was counted in each area. Student's test was performed to determine the statistical difference.
Tumors with a size of ∼5–7 mm in diameter (10 of each from the SAG-Tg(+) and SAG-Tg(−) mice) were excised and fixed in 4% PFA-PBS and embedded in paraffin. 4-μm-thick sections were cut, and deparaffinized slides were used for the TUNEL assay by means of the in situ cell death detection kit (Roche). The slides were stained with the TUNEL reaction mixture, along with the label solution for the negative control, and counterstained with propidium iodide counterstaining solution. The slides were mounted and analyzed under a fluorescence microscope. For quantification of TUNEL-positive cells, two random areas were selected from each tumor. The number of positive labeling cells (in green) out of a total of 1,000 cells was counted in each area. Student's test was performed to determine the statistical differences.
Data were analyzed using SAS v9.1 (SAS Institute). All hypotheses were tested at the 0.05 significance level. The volume of each individual tumor was recorded from each SAG-Tg(+) or SAG-Tg(−) mouse for a period of 20 wk of TPA promotion to reflect a growth rate of each tumor. Tumor volume was log transformed, and a random slopes and intercepts model (), implemented in SAS PROC MIXED, was used to estimate the growth rate of each tumor and to analyze the effect of SAG-Tg(+/−). This analysis appropriately considers the clustering of multiple tumors with animals. Counts of tumors were analyzed using log-linear models () in SAS PROC CATMOD. |
A prominent feature common to most neurodegenerative diseases is the accumulation of misfolded proteins in cytoplasmic inclusions, such as Lewy bodies in Parkinson's disease (PD) and neurofibrillary tangles in Alzheimer's disease (AD). The accumulation of misfolded proteins in these diseases most likely occurs due to a chronic imbalance in the generation and clearance of misfolded proteins. Misfolded proteins can be generated by genetic mutations () or oxidative modifications (). Misfolded proteins are prone to aggregation and have the potential to impair cellular functions (). Cells combat the buildup of misfolded proteins either by chaperone-mediated refolding or by proteasomal degradation (). Alternatively, when the proteasome becomes overwhelmed or impaired, misfolded proteins are transported by the retrograde dynein motor complex to pericentriolar inclusions called aggresomes (; ; ). It has been proposed that aggresome formation is a specific and active cellular response serving to sequester potentially toxic misfolded proteins (). Recent evidence suggests that there are similarities between Lewy bodies and aggresomes (). However, the molecular mechanisms involved in recognition and targeting of misfolded proteins to aggresomes remain unclear.
Ubiquitination is a dynamic post-translational modification that serves diverse cellular roles (; ). Ubiquitin is covalently attached to a substrate protein through the formation of an isopeptide bond between the C-terminal glycine residue of ubiquitin and the ɛ-amino group of a lysine residue on the substrate by a cascade of enzymatic reactions involving an E1 ubiquitin-activating enzyme, an E2 ubiquitin-conjugating enzyme, and an E3 ubiquitin protein ligase. Successive conjugation of ubiquitin molecules to one of the seven internal lysine residues (K6, K11, K27, K29, K33, K48, and K63) within the preceding ubiquitin molecule results in formation of a polyubiquitin chain (). K48-linked polyubiquitination acts as the canonical signal for targeting the substrate to the proteasome for degradation. In contrast, K63-linked polyubiquitination has recently been shown to have a proteasome-independent role in the regulation of several cellular processes, including endocytosis, signal transduction, and DNA damage (; ). Although K48-linked polyubiquitination has been extensively studied, relatively little is known about the molecular mechanisms underlying polyubiquitination via K63 or other lysine linkages.
Mutations in the E3 ligase parkin account for ∼50% of all recessively transmitted early-onset PD cases (; ; ). Interestingly, parkin-associated PD is devoid of Lewy bodies (; ; ), suggesting that parkin function may be required for the formation of these inclusion bodies (). Parkin is present in Lewy bodies in sporadic PD () and is recruited to aggresomes in cells treated with proteasome inhibitors (; ; ; ). However, the precise role of parkin in the biogenesis of Lewy bodies or aggresomes is unclear. Two studies suggest that parkin is capable of mediating K63-linked polyubiquitination (; ). However, the cellular role of parkin-mediated K63-linked polyubiquitination and the molecular mechanism by which parkin mediates K63-linked polyubiquitination remain undefined.
DJ-1 is a ubiquitously expressed protein that is mutated in an autosomal recessive, early-onset form of PD (; ). We and others have previously shown that the PD-linked L166P mutation disrupts DJ-1 protein folding, resulting in a misfolded protein that is prone to aggregation (; ). The folding state of other parkin-interacting proteins has not been investigated. Therefore, in this study we investigated the role of parkin in ubiquitinating misfolded proteins, using the L166P mutant DJ-1 as a model. Our findings reveal a novel proteasome-independent role for parkin in the regulation of aggresome formation and have important implications for understanding the biogenesis of Lewy bodies in PD.
We and others have previously shown that the L166P mutation-induced misfolded DJ-1 is selectively polyubiquitinated and degraded by the proteasome (; ). The molecular machinery that recognizes and ubiquitinates L166P mutant DJ-1 is unknown. We have proposed that L166P mutant DJ-1 might be a substrate for the E3 ligase parkin (). As shown in , immunoprecipitation of hemagglutinin (HA)-tagged L166P mutant DJ-1, but not wild-type DJ-1, was able to coprecipitate Myc-tagged parkin, providing evidence for an interaction of these two proteins in vivo (). Given the reported interaction of Parkin and L166P mutant DJ-1 with Hsp70 (; ; ), we considered the possibility that these three proteins might be present in a complex. However, coimmunoprecipitation experiments indicated that, even with high levels of Hsp70 overexpression, Hsp70 did not associate with L166P mutant DJ-1 under the experimental conditions used (Fig. S1, available at ). Furthermore, we found that GST-parkin bound Myc-tagged L166P mutant DJ-1 expressed in SH-SY5Y cells, but not Myc-tagged wild-type DJ-1 or endogenous DJ-1 (Fig. S2, available at ). In addition, GST-parkin was able to efficiently pull down in vitro translated [S]-labeled L166P mutant DJ-1, but not wild-type DJ-1 (). Together, our data indicate that parkin directly binds L166P mutant DJ-1, but not wild-type DJ-1, and that this interaction does not require Hsp70 as a recruiting factor.
Next, we performed in vivo ubiquitination assays to examine whether the interaction with parkin targets misfolded DJ-1 for parkin-mediated ubiquitination. In agreement with previous reports (; ), we found that L166P mutant DJ-1, but not wild-type DJ-1, was selectively polyubiquitinated in cells (). The polyubiquitination of L166P mutant DJ-1 was significantly increased by coexpression of parkin in the presence of the proteasome inhibitor MG132 (). Under the same conditions, parkin did not promote the ubiquitination of wild-type DJ-1 (). Together, these results demonstrate that parkin selectively recognizes and ubiquitinates misfolded DJ-1 in vivo.
To determine whether parkin-mediated ubiquitination targets L166P mutant DJ-1 for degradation by the proteasome, we assessed the effect of parkin overexpression on the steady-state levels of wild-type and L166P mutant DJ-1 proteins (). As we and others reported previously (; ; ; ), the steady-state level of L166P mutant DJ-1 was substantially lower compared with wild-type DJ-1. Inhibition of proteasome function by MG132 selectively increased the steady-state level of the L166P mutant, but not wild-type DJ-1 (). Contrary to our expectation, we found that overexpression of parkin did not alter the steady-state levels of L166P mutant or wild-type DJ-1 (). Given the recent report that parkin promotes degradation of synphilin-1 only when the expression of parkin reaches very high levels (), we examined the effects of increased levels of parkin expression on the degradation of L166P mutant DJ-1 (Fig. S3, available at ). Even at extremely high levels of overexpression, parkin was unable to promote degradation of L166P mutant DJ-1 (Fig. S3). Consistent with our previous findings (), pulse chase experiments indicated that L166P mutant DJ-1 was highly unstable compared with the wild-type DJ-1 (). Co-expression of parkin had no effect on the stability of L166P mutant DJ-1 (). Together these results suggest that parkin-mediated polyubiquitination of mutant DJ-1 may have a nondegradative role.
One possibility that may account for the inability of parkin to promote the degradation of L166P mutant DJ-1 is that parkin may mediate an alternative form of polyubiquitination that is not associated with proteasomal degradation, such as K63-linked polyubiquitination. The best-characterized E2 enzyme in mediating K63-linked polyubiquitination is the UbcH13/Uev1a heterodimer, in which UbcH13 is the catalytic subunit (; ). A recent in vitro study has shown that the GST-fused second RING domain of parkin can bind UbcH13 and this interaction recruits the UbcH13/Uev1a heterodimer to parkin (). However, it is unknown whether full-length parkin is able to bind UbcH13 in vivo. As shown (), UbcH13 readily coprecipitated with parkin, providing evidence for an interaction between these two proteins in vivo. In addition, we observed the interaction of parkin with UbcH7 and UbcH8, but not with UbcH5, consistent with previous studies (; ). These data indicate that parkin binds multiple E2 enzymes in vivo and suggest that the ability of parkin to facilitate K48- or K63-linked polyubiquitination may be dependent on the selective recruitment of distinct E2 enzymes.
We next performed in vitro ubiquitination assays to determine if the UbcH13/Uev1a heterodimer is the cognate E2 enzyme for parkin-mediated ubiquitination of L166P mutant DJ-1. We found that parkin ubiquitinated L166P mutant DJ-1 in an UbcH13/Uev1a-dependent manner (). To investigate the linkage of parkin-mediated polyubiquitination of L166P mutant DJ-1, we used ubiquitin mutants Ub-K29, Ub-K48, and Ub-K63, which contain arginine substitutions of all of its lysine residues except the one at position 29, 48, and 63, respectively. The Ub-K29, Ub-K48, and Ub-K63 mutants are thus expected to only allow the formation of K29-, K48-, and K63-linked polyubiquitin chains, respectively. In vitro ubiquitination analyses using these ubiquitin mutants revealed that parkin was able to facilitate the polyubiquitination of L166P mutant DJ-1 in the presence of wild-type ubiquitin (Ub-WT) and Ub-K63 (). In contrast, in vitro ubiquitination reactions using Ub-K29 or Ub-K48 resulted in the accumulation of monoubiquitinated L166P mutant DJ-1 (). In addition, low levels of higher molecular weight bands were also observed at sizes corresponding to L166P mutant DJ-1 conjugated with three or four ubiquitin molecules (), and these bands could potentially represent monoubiquitination of L166P mutant DJ-1 at multiple lysine residues. Together these results demonstrate that parkin-mediated polyubiquitination of L166P mutant DJ-1 occurs in vitro via K63-linked ubiquitin chains.
We then used Ub-K48 and Ub-K63 mutants in in vivo ubiquitination assays to determine the linkage of parkin-mediated polyubiquitination of L166P mutant DJ-1 in cells (). In agreement with the results of the in vitro ubiquitination analysis (), parkin promoted robust polyubiquitination of L166P mutant DJ-1 in cells expressing Ub-WT or Ub-K63, but not in cells expressing Ub-K48 (). To provide further support for the identified K63 linkage of parkin-mediated polyubiquitination of L166P mutant DJ-1, we used another set of ubiquitin mutants, Ub-K29R, Ub-K48R, Ub-K63R, which contain a single lysine-to-arginine mutation at position 29, 48, and 63, respectively. The Ub-K29R, Ub-K48R, and Ub-K63R mutants are expected to solely disrupt the assembly of K29-, K48-, and K63-linked polyubiquitin chains, respectively. As a control, we also included a polymerization-defective mutant of ubiquitin (Ub-K0), in which all lysine residues of ubiquitin were changed to arginines, and therefore is unable to form polyubiquitin chains. We found that parkin-mediated polyubiquitination of L166P mutant DJ-1 was greatly reduced by replacement of Ub-WT with Ub-K63R, but not by the replacement with Ub-K29R (). As expected, parkin-mediated polyubiquitination of L166P mutant DJ-1 was virtually abolished by replacement of Ub-WT with Ub-K0 (). The low levels of L166P mutant DJ-1 ubiquitination observed with Ub-K63R could be due to non-K63-linked ubiquitination of mutant DJ-1 by another E3 ubiquitin-protein ligase(s) in SH-SY5Y cells. Consistent with this possibility, we found that the polyubiquitination of L166P mutant DJ-1 was reduced by replacement of Ub-WT with Ub-K48R mutant (). Furthermore, we and others have previously shown that L166P mutant DJ-1 associates with the E3 ligase CHIP (; ). Collectively, our in vitro and in vivo data provide strong evidence that parkin facilitates K63-linked polyubiquitination of L166P mutant DJ-1 in cooperation with UbcH13/Uev1a.
Our findings that parkin specifically binds and ubiquitinates L166P mutant DJ-1 prompted us to investigate the potential colocalization of these two proteins in SH-SY5Y cells by immunofluorescence confocal microscopy. HA-tagged L166P mutant DJ-1 exhibited a cytoplasmic staining pattern that displayed substantial overlap with that of Myc-tagged parkin (, top). In addition, treatment with the proteasome inhibitor MG132 caused a dramatic relocation of L166P mutant DJ-1 and parkin to a single prominent, perinuclear inclusion (, bottom).
The morphology and localization of the L166P mutant DJ-1–containing inclusions appeared similar to that of aggresomes (; ; ). To test whether these inclusions were aggresomes, we performed additional immunofluorescence confocal microscopy experiments to further characterize these inclusions (). We found that the L166P mutant DJ-1-containing inclusions were highly enriched with ubiquitin and the chaperone protein Hsp70, and surrounded by a compacted cage of the intermediate filament protein vimentin. Lamp2-immunoreactive lysosomes did not colocalize with the inclusions, but were observed in close apposition and often tightly ringed the inclusions. This finding is consistent with recent studies indicating that aggresomes may represent substrates for autophagic degradation, a process involving fusion with lysosomes (). Moreover, serial 1-μm Z-sections revealed that the misfolded DJ-1 protein was present throughout the center of the aggresome, extensively colocalized with ubiquitin, and completely surrounded by vimentin (Fig. S4, available at ).
Given the reported role of microtubule-dependent retrograde transport in aggresome formation (), we examined the effects of the microtubule-depolymerizing drug nocodazole on parkin-induced recruitment of L166P mutant DJ-1 to aggresomes (). Depolymerization of microtubules by nocodazole in MG132-treated cells disrupted delivery of L166P mutant DJ-1 to the aggresome, leading to accumulation of misfolded DJ-1 in widely dispersed preaggresome particles (). Parkin colocalized with L166P mutant DJ-1 in these preaggresome particles (), demonstrating an association of these two proteins before their transport to the aggresome. Together, these data indicate that recruitment of L166P mutant DJ-1 to the aggresome is dependent on intact microtubules and that the inclusions containing L166P mutant DJ-1 are bona fide aggresomes.
Accumulating evidence indicates that aggresomes found in cultured cells are substrates for eventual autophagic degradation. To determine if L166P mutant DJ-1-containing aggresomes are sites of autophagy we used the widely used marker of autophagic vacuoles monodansyl cadaverine (MDC) (; ) (Fig. S5, available at ). The specificity of MDC staining is further confirmed by our results, which show that MDC staining does not overlap with lamp2 (Fig. S5), a protein present on late endosomes and lysosomes. Our findings indicate that under control conditions MDC displays a widely distributed, punctate distribution with relatively low fluorescence levels (Fig. S5). However, MG132-induced proteasomal impairment results in an increase in MDC fluorescence levels and a redistribution of MDC immunoreactivity to a perinuclear region that colocalizes with the L166P mutant DJ-1 aggresomes (Fig. S5). Together these findings indicate that L166P mutant DJ-1 aggresomes may be sites of autophagy.
To determine whether parkin selectively targets L166P mutant DJ-1 to aggresomes, we assessed the effects of parkin overexpression on the accumulation of HA-tagged wild-type and L166P mutant DJ-1 in aggresomes formed in response to proteasome inhibition by MG132 (). In untreated cells the coexpression of parkin had no effect on the cytoplasmic distribution of wild-type or L166P mutant DJ-1 (). Treatment of the cells with MG132 did not alter the distribution of wild-type DJ-1, but caused the accumulation of L166P mutant DJ-1 in aggresomes in a small percentage of cells (). Interestingly, coexpression of parkin resulted in a dramatic increase in the percentage of cells containing L166P mutant DJ-1–positive aggresomes, but had no effect on the localization of wild-type DJ-1 ().
Aggresomes contain misfolded and aggregated proteins, and are often insoluble in detergents (). To determine if recruitment of L166P mutant DJ-1 to these aggresomes alters its solubility, cell lysates were separated into detergent soluble and insoluble fractions, and analyzed by Western blotting. Wild-type and L166P mutant DJ-1 were predominantly found in the soluble fraction in untreated cells (). However, MG132 treatment resulted in the selective accumulation of L166P mutant DJ-1 in the detergent insoluble fraction, and this accumulation was increased by coexpression of parkin (). These results demonstrate that parkin selectively promotes the redistribution of misfolded DJ-1 into a detergent-insoluble aggresome.
To investigate the role of parkin-mediated K63-linked polyubiquitination in aggresome formation, we analyzed and compared the formation of MG132-induced aggresomes in SH-SY5Y cells coexpressing L166P mutant DJ-1 and Myc-tagged parkin in the presence of either HA-tagged Ub-WT, Ub-K48, Ub-K63, Ub-K48R, or Ub-K63R. We observed the presence of a clearly visible aggresome containing immunoreactivity to HA-tagged ubiquitin and L166P mutant DJ-1 in most of the cells expressing Ub-WT (46%) after 16-h MG132 treatment (). Replacement of Ub-WT with Ub-K63R dramatically reduced aggresome formation and resulted in the accumulation of cytoplasmic preaggresomal particles, whereas replacement with Ub-K48R had no apparent effect (). Moreover, expression of Ub-K63 significantly enhanced aggresome formation compared with cells expressing Ub-K48, which were more likely to contain cytoplasmic preaggresomal particles (). Together, our results suggest that K63-linked polyubiquitination of misfolded DJ-1 by parkin promotes aggresome formation.
The observed microtubule-dependent recruitment of L166P mutant DJ-1 to the aggresome () suggests that the retrograde dynein motor complex may be involved in the transport of misfolded DJ-1. Recent evidence indicates that histone deacetylase 6 (HDAC6) is critically involved in aggresome formation by acting as an adaptor between polyubiquitinated proteins and the dynein motor complex (; ). We performed coimmunoprecipitation analyses to examine a potential interaction between Myc-tagged L166P mutant DJ-1 and endogenous HDAC6. We found that HDAC6 does not bind nonubiquitinated L166P mutant DJ-1 in untreated cells (). In contrast, HDAC6 strongly interacts with polyubiquitinated L166P mutant DJ-1 induced by parkin coexpression in MG132-treated cells (). Moreover, immunofluorescence confocal microscopic studies revealed that, in response to proteasome inhibition, both HDAC6 and dynein are recruited with L166P mutant DJ-1 to aggresomes ().
Although HDAC6 has been shown to bind polyubiquitinated proteins (; ), it remains unknown whether HDAC6 specifically binds K48-linked and/or K63-linked polyubiquitin chains. Coimmunoprecipitation studies in SH-SY5Y cells expressing HA-tagged Ub-K48 or Ub-K63 indicate that HDAC6 interacts with both K48- and K63-linked polyubiquitin chains, but binds more strongly to K63-linked chains (), providing evidence that HDAC6 preferentially interacts with K63-linked polyubiquitinated proteins in vivo. Together, these data suggest that parkin-mediated K63-linked polyubiquitination of L166P mutant DJ-1 serves as a signal for binding HDAC6, and thus couples the misfolded DJ-1 to the dynein motor complex for transport to the aggresome.
Given our findings that parkin-mediated K63-linked polyubiquitination mediates the recognition of misfolded DJ-1 by HDAC6 and the dynein motor complex (, , and ), we next assessed the role of parkin in the transport of misfolded proteins to the aggresome using mouse embryonic fibroblasts (MEFs) cultured from knockout (
) mice (; ).
MEFs (). In addition, pulse chase analysis indicated that the absence of parkin had no effect on the stability of L166P mutant DJ-1 ().
and
MEFs (unpublished data). In
MEFs treated with MG132, L166P mutant DJ-1 accumulated in aggresomes ().
MEFs, Myc-tagged L166P mutant DJ-1 was mostly observed in small preaggresomal particles dispersed throughout the cytoplasm ().
MEFs to properly target the misfolded DJ-1 to aggresomes (). Together these results indicate that parkin is critically involved in the transport of misfolded DJ-1 to aggresomes.
Although it is clear that cells sequester misfolded proteins into centrosomal aggresomes, how the cell recognizes these proteins for transport to the aggresome is poorly understood. Our findings support the model depicted in , and suggest that misfolded proteins are normally polyubiquitinated and efficiently degraded by the 26S proteasome. However, under pathogenic conditions in which proteasome function is impaired, parkin cooperates with UbcH13/Uev1a to mediate K63-linked polyubiquitination of misfolded proteins, promotes binding of HDAC6 and couples the misfolded protein to the dynein motor complex for transport to aggresomes.
Our results demonstrate that parkin directly binds the misfolded L166P mutant, but not wild-type DJ-1. We find that parkin selectively promotes K63-linked polyubiquitination of L166P mutant DJ-1 in an UbcH13/Uev1a-dependent manner. K63-linked polyubiquitination has been proposed to play a role in the formation of inclusion bodies (), but the underlying molecular mechanisms remain unknown. Our data show that parkin promotes the accumulation of misfolded DJ-1 into aggresomes. Furthermore, the ability of parkin to promote the sequestration of misfolded DJ-1 into aggresomes is dependent on K63-linked polyubiquitination, and expression of Ub-K48 or Ub-K63R results in the accumulation of small preaggresomal particles dispersed throughout the cytoplasm. Previous studies suggest that misfolded proteins form small aggregates within the cytoplasm, which are then transported to the forming aggresome (; ). Disruption of transport by incubation with nocodazole (), overexpression of dynamitin (; ), siRNA-mediated depletion (), or chemical inhibition () of HDAC6 results in the accumulation of small cytoplasmic aggregates. Thus the preaggresomal particles that accumulate upon expression of Ub-K48 or Ub-K63R may represent L166P mutant DJ-1-containing small aggregates. In addition to L166P mutant DJ-1, these preaggresomal particles also contained parkin, which suggests that K63-linked polyubiquitination may be important for the transport of parkin to the aggresome. Hampe et al. reported that parkin catalyzes multi-monoubiquitination of itself (auto-ubiquitination) in vitro and a fraction of parkin is multi-monoubiquitinated in vivo under basal conditions (). However, under conditions of proteasomal impairment, they found that parkin is mostly polyubiquitinated in vivo (), which might be mediated by another E3 ligase or by parkin in cooperation with additional factor(s) such as E4 (; ). Furthermore, Lim et al. reported that parkin is modified by K63-linked polyubiquitination in vivo, which was suggested to be mediated by parkin itself (). K63-linked polyubiquitination of parkin under conditions of proteasomal impairment would promote the transport of parkin to aggresomes, and this could explain the reduced formation of parkin-positive aggresomes when K63-linked polyubiquitination was inhibited.
One possibility is that K63-linked polyubiquitination may block K48-linked polyubiquitination, resulting in decreased proteasomal degradation and increased substrate accumulation. However, this does not address how these proteins are recognized and targeted to the aggresome. Our results indicate that K63-linked polyubiquitination plays an active signaling role, and that under conditions of proteasomal impairment facilitates an interaction between L166P mutant DJ-1 and the dynein adaptor protein HDAC6. HDAC6 links polyubiquitinated proteins, to the molecular motor dynein by simultaneously binding ubiquitin via a zinc-finger ubiquitin binding domain and dynein via a distinct dynein motor binding domain (). In addition, HDAC6 has been found to localize to Lewy bodies in sporadic PD patients, suggesting that HDAC6 may also play a similar role in the biogenesis of these hallmark PD inclusions. Our results further demonstrate that HDAC6 preferentially binds K63-linked polyubiquitinated proteins in vivo, suggesting that K63-linked polyubiquitination may act as a novel signal for the dynein-mediated transport and sequestration of misfolded proteins in the aggresome.
Based on the localization of parkin in aggresomes/inclusions, it has been proposed that parkin is first recruited to aggresomes in response to proteasome inhibition and then ubiquitinates misfolded proteins in these inclusion bodies ().
MEFs would be unaffected, but that there would be a decrease in the ubiquitination of the inclusions.
MEFs. These findings support a role for parkin in the transport of misfolded DJ-1 to aggresomes rather than in the ubiquitination of the misfolded proteins in aggresomes. In addition, when transport of misfolded proteins to the aggresome was disrupted with nocodazole we observed that parkin colocalized with misfolded DJ-1 in ubiquitin-positive, preaggresomal particles, suggesting that parkin associates with and polyubiquitinates misfolded proteins before their transport to aggresomes.
Whether protein inclusions are cytoprotective or cytotoxic is highly controversial. Studies indicate that several intermediates are generated during the process of protein aggregation, and suggest that the small-intermediate oligomers may be the principle toxic species (). Accumulation of aggregated proteins can cause global impairment of the ubiquitin-proteasome system (; ) and aggregated proteins also directly interfere with proteasomal function, possibly due to an inability to properly enter or exit the proteasome (; ; ). Sequestration into inclusions like aggresomes may render these oligomers inert and protect an impaired proteasomal system from further damage. Interestingly, recent evidence suggests that aggresomes might also act as a staging area for disposal by autophagy (). In support of this possibility, we find that L166P mutant DJ-1 aggresomes stained with MDC, a marker of autophagic vacuoles. Consistent with the findings of Iwata et al. (), we observed that MG132-mediated inhibition of the proteasome resulted in a redistribution of lysosomes, and often lysosomes tightly encircled the aggresome. A protective role for aggresomes has been further supported by the finding that disruption of aggresome formation using small molecule inhibitors of HDAC6 () or by knockdown of HDAC6 protein levels by RNA interference () increases the susceptibility to apoptosis induced by proteasome inhibition or misfolded protein stress.
It is widely believed that loss of parkin function would lead to decreased proteasomal degradation of a cytotoxic substrate protein, resulting in selective dopaminergic neurodegeneration (). Studies have found that Parkin promotes the proteasomal degradation of several putative substrate proteins (). However, the accumulation of these substrates in parkin knockout mouse brain and human PD brains has been conflicting (; , ; ). Thus the role these substrates play in the pathogenesis of PD remains to be determined. Recent reports indicate that parkin is able to mediate monoubiquitination (; ) and K63-linked polyubiquitination (; ; and ), two forms of ubiquitination that are not associated with proteasomal degradation. In addition, parkin has been implicated in nonproteasomal processes, including endocytosis and signal transduction (). Our findings further indicate that parkin plays a regulatory role in trafficking of misfolded DJ-1 to aggresomes. Our data reveal a critical role for parkin-mediated K63 polyubiquitination in earmarking misfolded proteins for dynein-mediated transport to the aggresome and suggest that loss of parkin function may impair sequestration of toxic misfolded proteins, thereby resulting in an increased susceptibility to aggregated protein-induced cellular dysfunction.
The expression vectors encoding HA- and Myc-tagged wild-type and L166P mutant DJ-1 were described previously (). Myc-tagged parkin and HA-tagged UbcH13 were provided by T. Suzuki (Tokyo Metropolitan Institute of Medical Science, Tokyo, Japan) and Z. Chen (University of Texas Southwestern, Dallas, TX), respectively. HA-tagged Ub-WT, Ub-K48, Ub-K63, and Ub-K0 were provided by T. Dawson (Johns Hopkins University, Baltimore, MD); and Ub-K29R, Ub-K48R, and Ub-K63R were provided by M. Wooten (Auburn University, Auburn, AL). Anti–DJ-1 antibodies P7F and P7C were described previously (, ). Other antibodies used in this study include the following: anti-HA (12CA5), anti-Myc (9E10.3, Neomarkers), anti-parkin (Cell Signaling), anti-Hsp70 (Stressgen), anti-ubiquitin (P4G7, Covance), anti-vimentin (Sigma-Aldrich), anti–β-tubulin (Boehringer Mannheim), anti-HDAC6 (Santa Cruz Biotechnology, Inc.), anti-lamp2 (Iowa Developmental Studies Hybridoma Bank), and anti-dynein (Sigma-Aldrich); all secondary antibodies were purchased from Jackson ImmunoResearch Laboratories, Inc.
SH-SY5Y cells and mouse embryonic fibroblasts (MEFs) were transfected with the indicated plasmids using Lipofectamine 2000 (Invitrogen) or TransIt-Neural (Mirus), respectively, according to the manufacturer's instructions. Cell lysates were prepared from transfected cells and immunoprecipitations performed as described previously ().
GST-tagged parkin fusion proteins were expressed in BL21 cells and affinity purified as described previously (; ). Immobilized GST or GST-Parkin (∼5 μg) were incubated with [S]Methionine-labeled Myc-tagged DJ-1 or L166P mutant DJ-1, generated using the TNT Quick Coupled Transcription/Translation System (Promega), in phosphate buffered saline for 2 h at 4°C. After extensive washes, bound proteins were separated by SDS-PAGE and visualized by autoradiography.
Transfected SH-SY5Y cells were incubated for 8 h at 37°C with the proteasome inhibitor MG132 (20 μM, Calbiochem) or vehicle dimethyl sulfoxide (DMSO, final concentration 0.1%) and protein levels analyzed as described previously ().
As described previously (), transfected SH-SY5Y cells or MEFs were labeled by incubation with Met/Cys-free DME (Invitrogen) containing 100 μCi of Met/Cys TranS-label (MP Biomedicals) for 1 h. Cells were washed and then incubated with DME containing 5× the normal concentrations of methionine and cysteine. At the indicated chase times, equal amount of proteins were immunoprecipitated with anti-Myc antibody, separated by SDS-PAGE, and analyzed by autoradiography.
In vitro ubiquitination assays were performed using a well-established reconstitution system (, ; ). In brief, L166P mutant DJ-1 was purified as described previously (). 1.5 μg of purified L166P mutant DJ-1 was incubated at 37°C in 60 μl of reaction buffer (50 mM Tris-HCl, pH 7.5, 2.5 mM MgCl, 2 mM DTT, and 2 mM ATP) containing 200 ng of E1 (Boston Biochem), 400 ng of E2 ubiquitin-conjugating enzyme (UbcH13/Uev1a) (Boston Biochem), 1.5 μg of purified GST-tagged parkin, and 10 μg ubiquitin or ubiquitin mutants (Boston Biochem). After incubation for 2 h the reaction was stopped by addition of loading buffer. Reaction products were analyzed by immunoblotting with anti-DJ-1 antibody to detect ubiquitin-conjugated DJ-1 proteins. In vivo ubiquitination assays were performed as described previously (; ).
Transfected SH-SY5Y cells were lysed in a buffer containing 50 mM Tris-HCl pH 7.6, 150 mM NaCl, 1.0% Triton X-100, and a cocktail of protease inhibitors. Lysates were centrifuged at 100,000 for 30 min at 4°C and separated into detergent-soluble and insoluble fractions as described (). Fractions were analyzed by Western blotting, and protein levels quantified using Scion Image. An ANOVA with a Tukey's posthoc test was used for statistical comparison.
SH-SY5Y cells or MEFs were processed and stained as previously described (). Hoechst 33258 (Molecular Probes) was used to stain nuclei. MDC (Sigma-Aldrich) staining and costaining was performed as described (). Analysis and acquisition was performed using a confocal laser-scanning microscope (Axiovert 100 M; Carl Zeiss MicroImaging, Inc.) with a 63×, 1.4 NA, oil-immersion objective (Carl Zeiss MicroImaging, Inc.) at room temperature. Images were exported in TIFF format using LSM-510 software (Carl Zeiss MicroImaging, Inc.), and Adobe Photoshop Version 7.0 software was used to adjust the contrast and brightness.
knockout mice, containing a targeted deletion of exon 2, were generated on a coisogenic background (129S4/SvJaeSor) and characterized as described previously (; ; ). MEF cultures were prepared using well-established methods (; ). Early passage MEFs were used for all experiments.
Transfected SH-SY5Y cells or MEFs were incubated in the presence and absence of 20 μM MG132 for 16 h, and then processed for immunofluorescence microscopic analysis of aggresome formation. An aggresome was defined as a single, perinuclear inclusion containing L166P mutant DJ-1. For each experiment, 50–100 transfected cells were randomly selected and scored for the presence of an aggresome in a blinded manner. Experiments were repeated three times, and the data were subjected to statistical analysis by ANOVA with a Tukey's posthoc test.
Figure S1 shows how L166P mutant DJ-1 and parkin are not in a complex with Hsp70. Figure S2 shows that Parkin selectively pulls down L166P mutant DJ-1, but not wild-type DJ-1. Figure S3 shows how high levels of parkin overexpression have no effect on the degradation of L166P mutant DJ-1. Figure S4 shoes that the misfolded L166P mutant DJ-1 colocalizes with ubiquitin and is concentrated in the center of the aggresome. Figure S5 shows how L166P mutant DJ-1-containing aggresomes stain with the autophagosomal-marker monodansyl cadaverine. Online supplemental material is available at . |
Duchenne muscular dystrophy (DMD) is one of the most common X-linked lethal diseases, affecting 1 in 3,500 newborn males. DMD results from mutations in the gene coding for the protein dystrophin, which localizes at the innerface of the sarcolemma. Dystrophin associates with a large complex of membrane proteins, called the dystrophin glycoprotein complex, which is important for cell membrane integrity (; ). Without the dystrophin complex to tether the actin cytoskeleton inside the muscle cell to the extracellular matrix, forces generated by the muscle fiber result in tears of sarcolemma, leading to muscle damage (for review see ). The mouse strain is the most widely used animal model for DMD, having a nonsense mutation in exon 23, which eliminates dystrophin expression (; ). Human patients with DMD and mice suffer from progressive muscle cell degeneration and regeneration episodes. Ultimately, however, the dystrophic muscle damage cannot be repaired any longer, and the dystrophic myofibers become gradually replaced, initially by fibrotic infiltrates and subsequently by fat tissue ().
DMD remains an incurable and devastating disease. Therapies based on the restoration of dystrophin expression or the administration of dystrophin stem cells are promising but are still in the preclinical phase (; ; ; ; ; ). Intense research efforts have identified muscle-specific factors regulating muscle progenitor cell (satellite cell [SC]) functions (i.e., proliferation and differentiation), which also play a key role in muscle regeneration (e.g., Pax7, MyoD family members, etc.; ; ; ; ; ; ). However, these intrinsic factors will be difficult to target throughout the musculature when developing alternative therapies to treat DMD disease.
Mounting evidence indicates a critical involvement of extrinsic factors in DMD disease progression and the recovery of injured muscles. Indeed, infiltrated inflammatory cells release several cytokines and growth factors that modulate muscle degeneration, inflammation, and regeneration (e.g., TNFα, VEGF, and nitric oxide synthase; ; ; ; ; , ; ; ). We previously reported a critical role of the protease urokinase plasminogen activator (uPA) in the recovery of experimentally injured muscle (). Among the several enzymatic functions of uPA, the most classic one is the ability to convert the zymogen plasminogen into active plasmin, whose classic role is degradation of the fibrinogen end product fibrin (from here on, we refer to both by the term fibrin/ogen). By binding of uPA to its uPA receptor (uPAR), uPAR localizes the conversion of plasminogen to plasmin to the cell surface, thereby increasing pericellular proteolysis. In addition, uPAR also allows uPA to induce intracellular signaling, thereby promoting cell proliferation and migration (; ). Importantly, uPA and plasmin promote inflammatory cell infiltration and repair of injured muscle, whereas the role of uPAR herein remains unclear (; ). As the role of uPA and uPAR in dystrophy remains unknown, we therefore intercrossed mice with mice lacking either uPA (uPA) or uPAR (uPAR) and examined disease progression and its pathological features.
We previously showed that uPA mediates the recovery of experimentally injured muscle (), but its role in dystrophy remains unknown. Therefore, we first analyzed by zymography uPA activity in muscle extracts before and after the onset of muscle degeneration. At 14 d of age (i.e., before disease onset), the activity levels of uPA were undetectable in wild-type (WT) mice and in mice (unpublished data). In contrast, after disease onset (i.e., 30 d of age), the activity levels of uPA were increased in muscle but not in WT muscle (). These changes were specific for uPA, as no lytic band corresponding to tPA (at 72 kD) was detected by zymography (). Thus, uPA activity is specifically increased in dystrophic muscle during disease.
To evaluate whether uPA would affect the disease course in mice, we intercrossed mice with uPA mice and phenotyped uPA
(from here on referred to as ) and uPA
littermates. Both genotypes were healthy at birth and did not show any signs of muscle injury or differences in muscle size before disease onset (14 d of age; ; and Table S1, available at ). Beyond 3–4 wk of age, obvious signs of muscle dystrophy were detectable in and uPA
mice. However, compared with mice, uPA
mice suffered from a much more severe dystrophinopathy, at least up to 4 mo of age, as characterized by a more widespread and extensive myofiber degeneration and necrosis (). Indeed, uPA
muscles contained larger areas of muscle damage and significantly more clusters of degenerated myofibers (P < 0.05; ). Furthermore, von Kossa–stained calcium deposits, which are typically found in necrotic myopathies (), were almost exclusively detected in uPA
but minimally in muscle (). Moreover, the number of centrally nucleated fibers (indicator of muscle regeneration) was lower in uPA
than in muscle (). Consistent with this, the mean muscle cross-sectional area and myofiber size were smaller in uPA
mice as compared with mice ( and Table S1).
To ascertain worsening in the pathology of the whole skeletal musculature, we measured the serum levels of creatine kinase (CK), a biomarker of sarcolemmal damage (). Consistent with the more severe muscle degeneration, uPA
mice showed approximately twofold higher serum CK levels as compared with mice at 2.5 mo of age (). To determine the functional status of the diseased muscle, we used grip-strength and treadmill assays. Compared with mice, muscle strength at 2.5 mo of age was substantially decreased in uPA
mice in both assays (). Altogether, these findings provide histological, biochemical, and functional evidence that uPA deficiency aggravates muscle degeneration and attenuates regeneration in muscle.
In experimentally injured muscle, uPA is produced by SCs and by inflammatory cells (). Although T lymphocytes and neutrophils also infiltrate dystrophic muscles, infiltrated macrophages appear to be the major inflammatory cell type (Fig. S1 a, available at ; ; ; ; ). We first aimed to analyze the impact of uPA deficiency in the inflammatory response in muscular dystrophy. Before disease onset (i.e., at 14 d of age), Mac-1 macrophages and T-11 T lymphocytes were rarely detected in or uPA
muscles (). After disease onset (i.e., at 30 d of age), these inflammatory cells had infiltrated the dystrophic muscle of mice (). However, compared with mice, the number of infiltrated Mac-1 and T-11 cells in uPA
muscle was reduced up to ∼50% (). Consistent with this, the loss of uPA also reduced the number of infiltrated inflammatory cells in cardiotoxin (CTX)-injured muscle (Fig. S2 a). This was not the result of a genotypic difference in the number of circulating leukocytes in the peripheral blood (unpublished data).
This reduced infiltration and accumulation of inflammatory cells in uPA
dystrophic muscles was likely attributable to the fact that they lack the uPA needed to invade injured tissues. Indeed, when performing in vitro migration experiments, uPA
and uPA macrophages were found to migrate less compared with control cells ( and S2 b). Therefore, we evaluated whether the conditional restoration of uPA expression in the bone marrow (BM) of uPA
mice achieved via the transplantation of uPA
BM (termed uPA
mice from here on) could revert the deficient inflammatory response. As a negative control, we transplanted uPA
BM into uPA
mice (uPA
mice). We also transplanted WT BM into uPA recipient mice (termed uPA mice from here on) or into WT mice (WT mice) and induced muscle injury by intramuscular injection of CTX (supplemental material, available at ). In both experiments, we found that the transplantation of uPA-expressing BM increased the infiltration of inflammatory cells into dystrophic or injured uPA-deficient muscles. Indeed, compared with uPA
mice, muscles in uPA
mice became infiltrated with plenty of (uPA expressing) inflammatory cells (); likewise, inflammatory cells accumulated in the damaged muscle in uPA mice to the levels found in WT or WT mice (Fig. S2 c). Together, these data demonstrate that uPA is critical for inflammatory cells to infiltrate the degenerating myofibers of mice.
There is increasing evidence that the inflammatory response can promote both muscle injury and repair (; ; ; ). Therefore, we evaluated whether the transplantation of uPA-expressing BM also attenuated muscle degeneration in mice. Compared with uPA
mice, muscles in uPA
mice exhibited less severe signs of degeneration at 2 mo after transplantation (). Consistent with this, serum CK levels were lower in uPA
than in uPA
mice (). Thus, uPA-expressing BM-derived cells attenuate muscle degeneration in uPA
mice. Consistent with this notion, muscle damage was reduced and regeneration was rescued in uPA mice at 10 d and 25 d after CTX injury, respectively, whereas degeneration persisted in nontransplanted uPA-deficient mice ().
We previously showed that the persistent muscle degeneration in uPA mice after injury was mediated, at least in part, by the impaired dissolution of intramuscular fibrin/ogen deposits (). Therefore, we analyzed in and uPA
muscle the extent of fibrin/ogen accumulation before and after disease onset. Before disease onset (14 d of age), fibrin/ogen was undetectable by immunostaining or Western blotting in and uPA
muscles (unpublished data). However, at the first disease peak (30 d of age), fibrin/ogen deposits were readily detectable in muscles of both genotypes (). Importantly, however, compared with muscles, fibrin/ogen deposition was increased in uPA
muscles up to 2.5-fold (). Interestingly, the prior transplantation of uPA-expressing BM cells attenuated this increased deposition of fibrin/ogen in uPA
mice () and in uPA mice challenged with CTX (Fig. S2 d).
To directly prove that the increased accumulation of fibrin/ogen mediated the exacerbated dystrophic disease in uPA
mice, we depleted the circulating fibrinogen levels by administering the defibrinogenating snake venom ancrod to uPA
mice. Daily delivery of ancrod (1 U per day) starting at 12 d after birth and continuing for 18 d thereafter effectively reduced the accumulation of fibrin/ogen in uPA
muscles (). Importantly, compared with saline, the area of degenerated muscle in uPA
mice was significantly reduced (P < 0.05) by ancrod therapy, indicating that the increased deposition of fibrin/ogen mediated the severe muscle dystrophy in uPA
mice (). In addition, compared with saline, fewer muscle groups containing >10 degenerating fibers were found in ancrod-treated uPA
mice (). However, ancrod treatment in uPA
mice failed to completely rescue the exacerbated muscle dystrophy phenotype of uPA
mice. Indeed, compared with mice, the muscle degeneration area was still larger in ancrod-treated uPA
mice (compare with ). This incomplete rescue might be attributable to the finding that inflammatory infiltration remained halted in uPA
mice after ancrod treatment (). This result further underscores the importance of BM-derived uPA in the infiltration of inflammatory cells (). Thus, BM-derived uPA is required for dissolving fibrin/ogen deposits in dystrophic muscles, but it also mediates processes independent of fibrinolysis.
To further study the role of uPA during muscle regeneration, we used the model of CTX-induced muscle injury, wherein regeneration can be analyzed in a more time-controlled fashion. Consistent with the model, the loss of uPA impaired muscle regeneration in the CTX model ( and supplemental material). Notably, transplantation of WT BM improved the defective muscle regeneration in uPA mice (uPA mice), thereby highlighting the importance of BM-derived uPA in muscle repair (). However, we found no evidence of a relevant direct contribution of BM-derived uPA-expressing cells to regenerating myofibers (very few GFP-positive myofibers were detected after transplanting GFP-labeled WT BM cells; Fig. S2 e), suggesting that these cells likely promoted muscle regeneration via paracrine pathways.
During myofiber regeneration, resident SCs proliferate, migrate to, and fuse with the injured muscle fibers. As the loss of uPA in the and CTX models reduced the number of regenerating myofibers ( and ), we wondered whether uPA, which is expressed by SCs (Fig. S1 b; ), might also affect SC functions. Activation and proliferation rates of SCs were comparable in and uPA
muscles or in CTX-challenged WT and uPA muscles (). Consistent with this, although uPA/plasmin mediates the activation of hepatocyte growth factor (HGF)/scatter factor (SF) and TGF-β1 (i.e., modulators of SC activation and proliferation; ; ; ; ; ; ; ), the active levels of these factors were comparable in and uPA
muscles or in WT and uPA injured muscle (Fig. S3, available at ). Furthermore, SC–derived primary myoblasts from uPA muscle showed normal proliferation and migration in vitro (). Interestingly, however, the addition of murine recombinant (r-uPA) stimulated the migration of WT and uPA myoblasts in both scratch wounds and transwell assays (), although it failed to affect the proliferation rates (). Consistent with the promigratory effect of uPA, myoblast migration was increased in the presence of conditioned medium obtained from WT macrophage cultures (compared with nonconditioned control medium) but was only minimally stimulated by uPA macrophage conditioned medium (). The migration in response to WT macrophage conditioned medium was abrogated when the uPA inhibitor amiloride was added (). Moreover, the absence of migration in response to uPA macrophage conditioned medium was restored by supplementation with r-uPA (). Thus, our data suggest that macrophage-derived uPA might promote muscle regeneration by enhancing SC migration.
By binding to uPAR, uPA is capable of exerting its proteolytic effects at the pericellular level, but it also enables uPA to promote cell proliferation and migration via nonproteolytic pathways (; ). We found that uPAR expression was induced in muscle extracts of WT mice after CTX injury and of mice after disease onset (). Thus, we reasoned that the role of uPA in muscle regeneration might be dependent, at least in part, on its binding to uPAR. To directly evaluate this hypothesis, we performed CTX injury in muscles of WT and uPAR-deficient mice (uPAR), crossbred the mice into the uPAR- deficient background (uPAR
mice), and analyzed the consequences of uPAR deficiency on muscle regeneration in both models. CTX-induced muscle regeneration was indistinguishable between WT and uPAR mice after histological analyses at 2, 10, and 25 d after injury (). Consistent with this, the infiltration of inflammatory cells was also not affected in the absence of uPAR (). Most importantly, the muscle cross-sectional area and the extent of muscular dystrophy were also similar in and uPAR
mice (). Indeed, the percentage of muscle degeneration was not different between and uPAR
mice (). In addition, the number of infiltrated macrophages and T cells did not differ between and uPAR
mice (). Interestingly, the percentage of centrally nucleated fibers was slightly increased in uPAR
mice (); however, SC–derived primary myoblasts from uPAR mice presented normal proliferation and migration rates in vitro (Fig. S4, available at ). Altogether, these results demonstrate that uPAR is dispensable for muscle tissue remodeling during regeneration both after acute injury and in muscle dystrophy and suggest that uPA regulates key processes during muscle regeneration in a uPAR-independent manner.
Despite intense research efforts, DMD is still an incurable and fatal disease. The principal finding of this study is that uPA plays an important reparative role in muscular dystrophy. Indeed, uPA expression and activity increase during dystrophic disease, and the genetic loss of uPA exacerbated muscle dystrophinopathy and worsened muscle performance in the mouse model of Duchenne's disease. Importantly, these defects in the absence of uPA were largely rescued by the transplantation of uPA-expressing BM, thus highlighting the importance of uPA-secreting BM-derived cells in muscular dystrophy. Our data also indicated a critical role for fibrin/ogen deposits in dystrophic muscle and a crucial role for uPA to dissolve them. Notably, muscle dystrophinopathy was unaffected in the absence of uPAR, suggesting that uPA exerts its effect independently of its receptor. Thus, these results underscore the important role of muscle-extrinsic factors such as BM cell–derived uPA in DMD disease.
Our findings not only showed that uPA was produced by BM-derived cells but also that these cells required uPA for their infiltration into dystrophic muscle. Accordingly, macrophages showed reduced migration in vitro in the absence of uPA. It has long been proposed that inflammation exacerbates muscular dystrophy via the release of cytotoxic cytokines and free radicals, leading to myofiber necrosis (; ; ; ; ), although recently, evidence has been accumulating on a positive role for inflammatory cells during muscle regeneration (; ; ; ; ). Indeed, we found less inflammation but increased muscle degeneration in uPA
mice, whereas the transplantation of uPA-expressing inflammatory cells rescued these degenerative defects. Thus, it is conceivable that inflammatory cells require uPA to infiltrate degenerating muscles of dystrophic mice and initiate the repair process. Indeed, macrophages might require uPA for the activation and phagocytosis of necrotic debris and for extracellular matrix remodeling. It has been demonstrated that the activation and release of prorecovery cytokines by leukocytes is reduced in uPA mice (; ; ; ) and that uPA leukocytes have impaired phagocytosis capacity (). One potential mechanism underlying the uPA-mediated activation of leukocytes might involve mactinin, an α-actinin fragment that promotes monocyte/macrophage maturation, whose formation is mediated by uPA (; ). Moreover, our data indicate that uPA-expressing inflammatory cells are required for intramuscular fibrinolysis. Collectively, we propose that uPA drives the infiltration and function of inflammatory cells required to create a beneficial environment for the repair of dystrophic muscle.
Another prerequisite for the efficient regeneration of dystrophic muscle appears to be prevention of the excessive deposition of fibrin/ogen. Indeed, in muscle, fibrin/ogen accumulates as the disease progresses but is absent before disease onset. In the absence of uPA, both dystrophinopathy and fibrin/ogen accumulation were enhanced in mice. Importantly, depletion of fibrinogen by ancrod treatment attenuated the severe muscle degeneration in uPA
mice. Thus, removal of fibrin/ogen deposits appears to be required for the resolution of muscle damage in mice. Unpublished findings indeed indicated that fibrin/ogen promoted the persistent inflammation and degeneration of muscles. Thus, by preventing excessive fibrin/ogen accumulation, uPA produced by BM-derived inflammatory cells might attenuate muscle degeneration and persistent inflammation in mice.
Several studies have shown that both uPA and uPAR are expressed by a variety of cells of hematopoietic origin (; ; ) and that both molecules are up-regulated during severe infections, supporting a role for the uPA–uPAR system in inflammatory responses. Indeed, in uPAR-deficient mice, macrophages and neutrophils failed to infiltrate the lungs of mice in response to microbial infections (; ) or to migrate to the inflamed peritoneum of thioglycollate-treated mice (). Therefore, we reasoned that the critical role of uPA in driving the infiltration and function of inflammatory cells during muscle regeneration might involve uPAR. However, our results clearly showed that the loss of uPAR did not affect the degeneration/regeneration process nor did it impair the inflammatory response in dystrophic muscle, indicating that uPAR is not required for either process. Consistent with this notion, no degeneration or inflammatory phenotype was observed in uPAR-deficient mice after CTX injury. These results together with the reported observations that uPA and uPAR knockout mice have different susceptibilities to several pathogenic infections or biological processes (; , ; ; ) indicate that uPAR and uPA may operate at different steps and may even be independent of each other.
After the clearance of degenerating myofibers by uPA inflammatory cells, muscle regeneration also appears to require uPA. Indeed, in the absence of uPA, muscle regeneration was attenuated in and CTX-injured muscle; the transplantation of uPA-expressing BM rescued this defect. In addition, although the migratory capacity of primary myoblasts from uPA muscle was normal, myoblast migration was enhanced in the presence of recombinant or macrophage-produced uPA. In contrast, the supplementation of r-uPA failed to affect their ability to proliferate. Thus, our results suggest that uPA derived from inflammatory cells specifically promotes the migration of muscle cells. As uPA deficiency failed to affect the activation of latent growth factors (e.g., HGF/SF or TGF-β1) in regenerating muscle in vivo, uPA might affect SC migration via alternative pathways. Unpublished findings from our group suggest that the removal of fibrin/ogen deposits promotes SC migration. Notably, these data extend previous observations that uPA promotes the migration of C2C12 immortalized myoblasts and primary human myoblasts by regulating membrane ruffling or by binding uPAR (; ; ). However, we found that the genetic loss of uPAR did not affect primary myoblast migration.
DMD remains an incurable and fatal disease. No therapies correcting the primary defect in DMD (i.e., dystrophin replacement) are yet available, and current DMD therapies have a narrow therapeutic window (e.g., temporary efficacy and severe side effects). Our study shows that uPA activity, by providing an adequate inflammatory response and by promoting fibrinolysis and muscle regeneration, is beneficial in muscle dystrophy. Notably, we and others recently demonstrated that genetic loss of the uPA inhibitor PAI-1 accelerated the recovery of CTX-injured muscle (; ). Thus, stimulating uPA activity may constitute a novel potential alternative for DMD disease amelioration.
uPA and uPAR knockout male mice (; ) were crossed with female mice (Jackson ImmunoResearch Laboratories). Male F1 mice were bred with female mice, and their F2 heterozygous uPA and uPAR male and female offspring were intercrossed. The resulting F3 generation showed the expected Mendelian distribution of uPA-WT, uPAR-WT, and heterozygous and homozygous deficient genotypes, all of them in an background. The uPA and uPAR genotypes were confirmed by PCR of tail biopsy genomic DNA as previously described (; ). The genotype was confirmed by Western blotting of muscle biopsies using an antidystrophin antibody (1:200; Novocastra). All animal experiments were approved by the Catalan Government Animal Care Committee.
At selected times, muscles of WT, uPA, uPAR, , uPA
, and uPAR
mice were removed after cervical dislocation, frozen, and stored at −80°C before analysis. 10-μm sections were collected from the midbelly of muscles and stained with hematoxylin/eosin (HE). Images were acquired with a microscope (DMR; Leica) equipped with a camera (DFC300 FX; Leica) and using 10× 0.25 NA, 20× 0.40 NA, and 40× 0.75 NA objectives (Leica). The acquisition software was the IM1000 program (Leica). The cross-sectional areas of entire muscles and myofibers were measured using the computer-assisted morphometric measurement Image 1.62c program (Scion).
Serum CK was measured with the indirect CK colorimetric assay kit and standards (Thermo Electron). For the grip strength assay, forearm grip strength was measured as tension force using a computerized force transducer (grip strength meter; Bioseb) to measure the peak force exerted by a mouse's forelimbs as its grip was broken by the experimenter pulling the mouse by the base of the tail away from the transducer () of the grip strength meter (). Three trials of three measurements per trial were performed for each animal with a few minutes resting period between trials. The mean tension force (in newtons) was calculated for each group of mice. The 100% value was arbitrarily assigned to the recorded force of mice (). For the treadmill assay, the treadmill apparatus (Treadmill; Panlab) consisted of a belt set at a slope of 10° and varying in terms of rotational speed (5–150 rpm; ). At the end of the treadmill, an electrified grid was placed on which footshocks (0.6 mA) were administered whenever the mice felt off the belt. The latency to fall off the belt (time of shocks in seconds) and the number of received shocks in consecutive trials with increasing fixed rotational speeds (5, 10, 20, 30, 40, and 50 rpm) with a cut-off period of 1 min per trial were registered. Animals were trained to walk on a motor-driven treadmill belt at constant speed (5 rpm) to obtain baseline values for locomotion in the intact state.
Muscle sections were placed in a silver nitrate solution, exposed to strong light for 30 min, and rinsed in distilled water. Sections were treated with sodium thiosulphate, rinsed in distilled water, and counterstained with neutral red. Finally, preparations were covered with aqueous mounting media and photographed.
The following primary antibodies were used for immunohistochemistry: anti–Mac-1 (M1/70; Hybridoma Bank), anti-T11 conjugated with fluorescein (1:50; Coulter Immunology), anti-fibrin/ogen (1:100; Nordic), anti-F4/80 (1:200; Serotec), and anti-uPA (1:20; Santa Cruz Biotechnology, Inc.). Depending on the antibody, immunohistochemistry was performed with the tyramide signal amplification cyanine 3 system (PerkinElmer) or as previously described (; ). Control experiments without primary antibody demonstrated that the signals observed were specific.
Muscle extracts were prepared from gastrocnemius muscles in 100 mM Tris-HCl buffer, pH 7.6, containing 200 mM NaCl, 100 mM CaCl, and 0.4% Triton X-100. 50 μg of total protein was resolved by SDS-PAGE and transferred to polyvinylidene difluoride membranes. Antibody dilutions were anti-fibrin/ogen at 1:3,000 (provided by K. Dano, Finsen Laboratory, Rigshospitalet, Copenhagen, Denmark) and anti–α-tubulin at 1:4,000 (DM1A; Sigma-Aldrich).
Zymography of muscle extracts was performed as previously described (). An SDS-PAGE gel was laid onto a casein gel, incubated in a humid chamber at 37°C until caseinolytic bands (corresponding to uPA or/and tPA) were visualized, and photographed.
12-d-old uPA
mice were daily injected intraperitoneally with ancrod (1 U ancrod/day; Sigma-Aldrich) or with a saline solution for 18 d and killed at 30 d of age. Muscles were dissected and frozen before analysis.
Muscle degeneration was determined microscopically and expressed as a percentage of the total muscle area. The number of DGs (degenerating groups) that contained >10 degenerating fibers was counted in complete muscle cross sections of and uPA
mice. Muscle fiber regeneration was determined microscopically and expressed as the percentage of total muscle fibers containing central nuclei present in the entire cross section of the muscle.
Macrophage migration was assayed on transwells (3-μm pore size; Beckton Dickinson). BM-derived macrophages were obtained as previously described () from and uPA
mice (or from WT and uPA mice). 5 × 10 macrophages/transwell in RPMI 1640 containing 1% FCS were added to the upper chamber of the transwell, and the conditioned medium of muscle SCs, which was previously concentrated fivefold using the Centrifugal Filter Device (Millipore), was added to the lower chamber. SC migration was performed on 8-μm pore size transwells. SCs from WT or uPA mice (5 × 10 cells/transwell) in Hams F-10 containing 1% FCS were added to the upper chamber of transwells. Transwells were coated with matrigel before addition of the cells. When indicated, 10 nM recombinant murine uPA (Molecular Innovations) was added to the lower chamber of the transwell. Alternatively, conditioned medium of WT or uPA macrophages, which were previously concentrated 2.5-fold and supplemented or not supplemented with 10 nM of murine r-uPA (Molecular Innovations) or 1 mM amiloride (Sigma-Aldrich), was added to the lower chamber. After 16 h of incubation at 37°C, cells on the filter's upper surface were scraped off. Then, filters were fixed in cold ethanol and stained with 5% crystal violet. Cells on the filter's lower surface were counted (12 fields per filter). Experiments were performed in triplicate.
WT and uPA SCs (2 × 10 cells) were plated in 12-well plates coated with matrigel (BD Biosciences). Once cells were attached to the matrix, a wound was performed across the well using a sterile pipette tip with an outer diameter of 500 μm. When indicated, 10 nM recombinant murine uPA was added to the culture media. Cells were then photographed at 0, 8, and 24 h after wounding using a microscope with 10× magnification (DMR; Leica). Experiments were performed in triplicate.
Regeneration of skeletal muscle was induced by intramuscular injection of 300 μl of 10 M CTX (Latoxan) in the gastrocnemius muscle group of the mice (). This concentration and volume were chosen to ensure maximum degeneration of the myofibers. The experiments were performed in right hindlimb muscles, and contralateral intact muscles were used as a control. Morphological and biochemical examinations were performed at 0, 2, 10, and 25 d after injury.
Donor BM cells were obtained by flushing the femurs and tibiae of or uPA
mice with RPMI 1640 medium (Invitrogen) and were transplanted into 4-mo-old uPA
mice after lethal irradiation (9 Gy). Alternatively, donor BM cells were obtained from WT mice and transplanted into 8-wk-old WT or uPA mice. The reconstituting cells (5 × 10 cells) were injected intravenously into the tail of the recipient mice within 24 h after irradiation. Alternatively, donor BM cells were obtained from GFP mice (TgN-GFPU-5Nagy mice; provided by A. Nagy, Samuel Lunenfeld Research Institute, Mount Sinai Hospital, Toronto, Ontario, Canada) and were transplanted into 2-mo-old WT and uPA mice. The mice were placed in sterile cages and fed with sterile chow until the reconstitution of BM was completed 8 wk after the transplantation. No changes in general health status were noted in the recipient mice. Regeneration of skeletal muscle in WT and uPA mice was induced by the intramuscular injection of CTX as described in the previous section.
SCs were isolated from and uPA
mice of 2.5 mo of age as described previously (). For FACS analysis, 2 × 10 SCs were used. SCs were permeabilized with 70% EtOH for 1 h at −20°C after incubation with an anti-CD34 antibody (FITC anti–mouse CD34; Ram34; BD Biosciences). Next, SCs were incubated with an anti-MyoD antibody (MyoD; Santa Cruz Biotechnology, Inc.) followed by incubation with a phycoerythrin-labeled secondary antibody (donkey anti–rabbit phycoerythrin; Abcam). Activated SCs were cells double positive for CD34 and MyoD. Experiments were performed in triplicate.
WT, uPA, and uPAR SCs were cultured in Ham's F-10 medium containing 20% FBS. 3.5 × 10 cells were plated in 12-well plates. When indicated, recombinant murine uPA (Molecular Innovations) was added to the culture medium at different concentrations (2, 10, or 50 nM). After 18 h, proliferating cells were labeled with 1.53 μg/ml BrdU (Sigma-Aldrich) for 2 h. BrdU-labeled cells were detected by immunocytochemistry and counted microscopically. Antibodies used for immunodetection were monoclonal rat anti-BrdU (1:500; Oxford Biotechnology) and biotin-SP–conjugated donkey anti–rat IgG (1:250; Jackson ImmunoResearch Laboratories). Experiments were performed in triplicate.
Gastrocnemius, quadriceps, and tibialis muscles from WT and uPA mice were injected with CTX to induce muscle regeneration. 24 h after injury, 50 mg/kg BrdU was injected intraperitoneally. 18 h later, SCs were isolated as described previously () and cultured for 24 h in Ham's F-10 medium containing 20% FBS. The percentage of SCs that had been BrdU labeled in vivo was determined by immunocytochemistry using a monoclonal rat anti-BrdU (as described above) and counted microscopically. Experiments were performed in triplicate.
Total RNA was extracted from muscles or SCs using the commercially available Ultraspec RNA isolation system (Biotecx). For RT-PCR, 2 μg of total RNA were reverse transcribed using the first-strand cDNA synthesis kit (GE Healthcare). Amplification parameters were denaturation at 94°C for 30 s, annealing for 30 s at 50°C (uPAR) and 55°C (uPA and glyceraldehyde-3-phosphate dehydrogenase), and extension at 72°C for 30 s. Primers for the detection of reverse transcriptase products were derived from different exons to distinguish RT-PCR products from genomic DNA contaminations. Primer sequences were as follows: uPAR (5′-GTGACCCTCCAGAGCACAGAA -3′ and 5′-GCAGTGGGTGTAGTTGCAACA-3′), uPA (5′-GGCAGTGTACTTGGAGCTCCT-3′ and 5′-TAGAGCCTTCTGGCCACACTG-3′), and glyceraldehyde-3-phosphate dehydrogenase (5′-ACTCCCACTCTTCCACCTTC-3′ and 5′-TCTTGCTCAGTGTCCTTGC-3′). The expected product sizes were uPAR at 140 bp, uPA at 450 bp, and glyceraldehyde-3-phosphate dehydrogenase at 185 bp.
Muscle extracts were analyzed for the presence of activated HGF by Western blotting using a goat anti–αHGF antibody (1:100; Santa Cruz Biotechnology, Inc.), which recognizes the active form (60 kD) of mouse HGF.
Crushed muscle extracts from and uPA
mice and from WT and uPA mice after CTX injury were prepared as described previously (). The presence of activated TGF-β1 was analyzed using the Quantikine TGF-β1 immunoassay kit (R&D Systems) according to the manufacturer's instructions.
All quantitative data were analyzed by test. P < 0.05 was considered statistically significant.
Table S1 shows a comparison of the morphometric properties of gastrocnemius muscle of WT, , and uPA
mice. Fig. S1 demonstrates that macrophages and SCs express uPA. Fig. S2 shows that BM transplantation rescues abnormal inflammatory infiltration and fibrin/ogen deposition in uPA mice. Fig. S3 shows that uPA-deficient muscles present normal activated TGF-β1 and HGF/SF levels. Fig. S4 shows that uPAR SCs have normal proliferation and migration rates. Supplemental material contains a description of the impaired muscle regeneration in uPA-deficient mice after CTX injury. Online supplemental material is available at . |
The protein tyrosine kinase Pyk2 (also designated RAFTK, CADTK, and CTK) and FAK are two members of a distinct family of nonreceptor tyrosine kinases. PYK2 and FAK share ∼45% amino acid sequence identity and a common domain structure: an N-terminal FERM domain followed by a protein tyrosine kinase (PTK) domain, three proline-rich regions, and a focal adhesion targeting (FAT) domain at the C terminus. Although FAK is expressed in most cells (), Pyk2 exhibits a more restricted expression pattern with strongest expression in the central nervous system and in hematopoietic cells (). FAK is a major intracellular signaling component of integrin-mediated cell adhesion () and plays a role in signaling pathways mediated by growth factor receptors. PYK2, on the other hand, is activated by a variety of extracellular cues including agonists of G protein–coupled receptors, intracellular Ca concentration, inflammatory cytokines, and stress signals, as well as integrin-mediated cell adhesion (; ).
Pyk2 is highly expressed in osteoclasts, where it is primarily confined to podosomes (; ; ). Podosomes are highly dynamic, actin-rich structures that mediate cell attachment and migration of highly motile cells such as macrophages and osteoclasts. They are composed of a central actin-bundle core surrounded by integrins and integrin-associated adhesion molecules (). When plated on glass, mature osteoclasts organize their podosomes at the periphery of the cell in a large belt-like structure. The podosome belt is similar to the sealing zone, another podosome-containing structure that is formed in active bone-resorbing osteoclasts (). Both structures share the same molecular components and are stabilized by microtubules (, ). Reduction of Pyk2 expression in osteoclasts by adenovirus containing Pyk2 antisense RNA leads to impairment in integrin-mediated cytoskeletal organization and bone resorption (). In addition, macrophages from mice failed to become polarized and to migrate in response to chemokine stimulation in vitro and in vivo (). It has been proposed that a ternary Pyk2-Src-Cbl complex induced by integrin engagement plays an important role in the control of osteoclast migration and bone resorption (; ). However, it is not clear how signaling via Pyk2 is linked to changes in the osteoclast cytoskeleton, and in particular to podosome assembly and organization, processes required for bone resorption.
Here, we report that Pyk2 deficiency in mice leads to an osteopetrotic phenotype. Osteoclasts from mice are defective in cell polarization, fail to form proper sealing zones, and inefficiently resorb dentin in vitro. Furthermore, -null osteoclasts fail to form a podosome belt that is typically seen at the periphery of wild-type osteoclasts. We also demonstrate that Rho activity is increased and the cellular distribution and stability of microtubules are compromised in -null osteoclasts. Ectopic expression of wild-type or Pyk2 mutants in -null osteoclasts demonstrates that the FAT domain of Pyk2 plays a primary role in the control of podosome belt and sealing zone formation, as well as in bone resorption. These experiments show that Pyk2, by controlling Rho activity, regulates microtubule-dependent podosome organization in osteoclasts and thereby bone resorption.
We have previously described the generation of mice and demonstrated that Pyk2 deficiency results in impairment in multiple macrophage functions (). Because Pyk2 is abundantly expressed in osteoclasts, we have examined the possibility of whether deficiency in Pyk2 may result in bone abnormalities. Immunoblotting of osteoclast lysates with anti-Pyk2 antibodies revealed the presence of both the ubiquitous (110 kD) and hematopoietic (106 kD) Pyk2 isoforms, which have been shown to be generated by alternative RNA splicing () (). Histological and histomorphometric comparison showed that the size and shape of bones from mice are normal. The length and width of the femur at the metaphyseal mid-point of either 2- or 10-wk-old mice were similar, and no changes were detected in the proliferative or hypertrophic zones of growth plates (unpublished data). However, the density of bones of either 2- or 10-wk-old mice was substantially elevated throughout the skeleton as shown by X-ray analysis and by histology (). Histomorphometric analysis demonstrated higher trabecular bone volume in mice (). The increase in trabecular bone volume is largely due to increased trabecular number () and to a lesser extent to increased trabecular thickness, more evident in 10-wk-old mice (), whereas trabecular spacing was reduced at both time points ().
Although trabecular bone volume was high, the percentage of trabecular bone surface covered by osteoclasts (osteoclast surface) was substantially elevated in mice, by 60% in 2-wk-old mice, and by nearly 300% in 10-wk-old mice (). Elevated osteoclast surface in the presence of high trabecular bone volume suggests a defect in osteoclast function, as shown in several osteopetrotic mutants (; ; ). Gain of trabecular bone volume may also result from high bone formation caused by changes in osteoblast differentiation or function. A mildly elevated osteoblast surface (percentage of trabecular bone covered by osteoblasts) was detected in 2-wk-old mice, but both osteoblast surface and bone formation rates (determined by measuring incorporation of calcium-binding fluorochromes) were normal in 10-wk-old mice (, respectively), while their trabecular bone volume continued to increase. Although at this time we cannot completely rule out a minor effect of Pyk2 deficiency on osteoblast function, our results strongly support the notion that impairment in osteoclast function is the primary cause of the bone phenotype of mice. Indeed, and strongly supporting this conclusion, abundant cartilage remnants were detected in the trabeculae of 10-wk-old mice (). This may result from defective resorption of the growth plate cartilage, and is an important hallmark of osteopetrosis (; ).
To examine whether the osteopetrosis of mice is indeed caused by defective osteoclast function, osteoclasts isolated from wild-type or mice were plated on dentin, and their ability to form pits was compared. After 48 h, the area, depth, and volume of the pits were compared using three-dimensional scanning confocal microscopy. The volume of dentin excavated by -null osteoclasts was significantly reduced (). The decrease in pit volume resulted from a decrease in both area and depth; the pits formed by -null osteoclasts were more shallow (). Thus, in the absence of Pyk2, osteoclasts show a cell-autonomous decrease in bone-resorbing activity.
To gain further insight into the mechanism underlying the reduced resorption by -null osteoclasts, we compared the distribution and status of cytoskeletal proteins in osteoclasts from wild-type or mice. Immunofluorescence microscopy of osteoclasts plated on dentin and stained with fluorescently labeled phalloidin demonstrated that -null osteoclasts formed an abnormal, thinner and smaller sealing zone (). Furthermore, immunofluorescent labeling of osteoclasts plated on glass demonstrated that in wild-type osteoclasts, F-actin was primarily localized in a podosome belt at the periphery of the cells, whereas in osteoclasts from mice, actin was organized in podosome clusters and multiple small rings throughout the cell (). Cellular localization of actin, cortactin, vinculin, and paxillin in podosomes using fluorescence microscopy have demonstrated that their distribution is similar in wild-type and -null podosomes ( and Fig. S1, available at ). These experiments show that in absence of Pyk2, podosomes fail to organize in a belt-like structure at the periphery of the cell. The defect in organization of podosomes also correlates with the presence of abnormal sealing zones in -null osteoclasts, which may explain the shallower resorption cavities formed by -null osteoclasts in vitro.
To further determine whether the defect in -null osteoclasts is cell autonomous, we co-cultured different combinations of wild-type or -null bone marrow cells and calvarial osteoblasts, allowing marrow-derived osteoclast precursor cells to differentiate into osteoclasts (). As shown in , absence of podosome belt, revealed by both F-actin and vinculin staining, was detected in osteoclasts derived from marrow whether in the presence of wild-type or -null osteoblasts (). The number of osteoclasts generated from bone marrow was not significantly different from wild-type, regardless of the origin of the supporting osteoblasts (). Collectively, these results demonstrate that osteoclast differentiation is unaffected by the absence of Pyk2, and that the functional defect in -null osteoclasts is cell autonomous.
The dynamic nature of GFP-labeled actin that was expressed in osteoclasts from wild-type or mice was compared by fluorescence recovery after photobleaching (FRAP) measurements (). Characteristic photobleaching recovery time of GFP-actin in wild-type osteoclasts was ∼30 s, as described previously (). In contrast, the photobleaching recovery time of GFP-actin in -null osteoclasts was doubled (), indicating that the rate of actin flux in podosomes is reduced in the absence of Pyk2. The molecular mechanisms underlying the observed differences in FRAP measurements is currently unknown. Notwithstanding the decreased rate of actin flux in individual podosomes, the total podosome life-span (namely, the average time from the first appearance of a new podosome to its dissociation) in osteoclasts expressing GFP-actin was not significantly different (). Overall, these results suggest that in osteoclasts, Pyk2 contributes to the dynamic exchange of actin in podosomes, but not to podosome life span. The podosome organization defects in -null osteoclasts, which prevent podosome belt formation, are unlikely to be caused by a change in podosome life span in these cells.
Previous analyses have shown that the process of podosome patterning is controlled by the microtubule network in osteoclasts (). During podosome differentiation, podosomes initially organize into clusters that evolve into unstable small podosome rings by a mechanism of self-organization. Subsequently, the small rings fuse and expand by an oriented treadmilling process toward the periphery of the cell to form belts. Thus, the pattern of podosomes in -null osteoclasts resembles earlier stages of osteoclast differentiation, when podosomes are organized in clusters and small, dynamic podosome rings. It was demonstrated that the transition of clusters/rings to peripheral podosome belt requires an intact microtubule network (). We therefore analyzed microtubule distribution and found that -null osteoclasts lacked the characteristic circular microtubule network that is concentrated around the podosome belt in normal cells (; ), while maintaining the radial network of microtubules ().
Cellular microtubules are composed of two distinct pools with regard to their stability: a short-lived pool with a half-life of ∼5–10 min and a more stable pool that may last for as long as several hours. The stable microtubule pool is resistant to treatment with the microtubule-depolymerizing agent nocodazole and contributes to podosome belt stabilization in osteoclasts (; ). As expected, wild-type osteoclasts contained a nocodazole-resistant pool of stable microtubules. In contrast, very few nocodazole-resistant microtubules were seen in -null osteoclasts, demonstrating that the short-lived, rather than the stable pool, is the predominant form of microtubules in cells lacking Pyk2 (). Because the stable pool of microtubules that accumulates during osteoclastogenesis is highly acetylated (), we next analyzed microtubule acetylation by immunofluorescence microscopy of permeabilized cells labeled with antibodies specific for the acetylated form of tubulin. This experiment showed that microtubule acetylation is reduced in -null osteoclasts (). A similar conclusion was drawn by immunoblotting experiments demonstrating that acetylation of microtubules is, indeed, compromised in osteoclasts (). The reduced acetylation of microtubules in -null osteoclasts is consistent with their reduced stability, as revealed by nocodazole treatment. It was shown that microtubule acetylation is regulated by the mDia2-HDAC6 complex in osteoclasts (). However, analyses of mDia2 localization in wild-type and -null osteoclasts using fluorescence microscopy revealed similar cellular distribution (Fig. S4, available at ).
The small GTPase Rho has been implicated in the stabilization and post-translational modification of microtubules (). In addition, Rho inhibition leads to enhancement in the acetylation and stabilization of the microtubule network in osteoclasts (). To examine the possibility of whether the reduced stability and acetylation of microtubules in -null osteoclasts are caused by increased Rho activation, Rho activity was determined by using a GST fusion protein containing the Rho binding domain of Rhotekin (RBD) to pull down the active, GTP-bound form of Rho (). The results showed increased Rho activity in -null cells (). The potential role of Rho activation was further tested by examining the effect of the Rho inhibitor C3-toxin on the stability of microtubules, and on podosome belt formation. The experiment presented in shows that C3-toxin treatment increased the stable pool of microtubules and induced the formation of a belt-like structure at the cell periphery of -null osteoclasts. The stable pool of microtubules observed in -null osteoclasts after treatment with C3 was also highly acetylated (). These results further indicate that enhanced Rho activity in -null osteoclasts is responsible for the decrease in microtubule stability and acetylation, and impairment of podosome belt formation.
It was previously proposed that recruitment of Src by Pyk2 plays an important role in podosome belt formation and bone resorption in osteoclasts (; ). Examination of the expression and activity of Src in wild-type and -null osteoclasts revealed that, although the expression of Src was not altered in the absence of Pyk2 (), Src activity was reduced in both unstimulated and integrin-stimulated -null osteoclasts (). However, podosome belt formation was not rescued in -null osteoclasts that were microinjected with expression vector for activated Src (Src-Y527F) (), suggesting that the Pyk2-dependent pathway leading to microtubule-dependent podosome belt and sealing zone formation in osteoclasts is, by and large, a Src-independent process.
To determine which Pyk2 domain(s) regulate podosome belt formation, we tested the ability of several Pyk2 mutants to rescue podosome belt formation in -null osteoclasts. We tested Pyk2-Y402F, an autophosphorylation site mutant that does not bind Src; Pyk2-K457A, a kinase-negative mutant; Pyk2-ΔFAT (aa 1–868), a deletion mutant devoid of the FAT domain; Pyk2-ΔFERM (aa 381–1009), a deletion mutant that lacks the FERM domain; and, as a control, Pyk2-WT (). The different mutants were transiently expressed in HEK293 cells and their tyrosine kinase activity was analyzed by immunoblotting with antibodies specific for phosphorylated Tyr402 of Pyk2. All mutants except Pyk2-Y402F and Pyk2-K457A showed similar levels of autophosphorylation (Fig. S2 a, available at ). As expected, Pyk2-WT, Pyk2-ΔFAT, and Pyk2-ΔFERM formed a complex with Src, whereas the Pyk2-Y402F and Pyk2-K457A mutants failed to bind Src to a substantial degree (Fig. S2, b and c).
We next examined the ability of the different mutants to rescue podosome belt formation in -null osteoclasts. Expression vectors for the different mutants were microinjected into osteoclasts from mice, and the cells were stained with fluorescently labeled phalloidin and anti-Pyk2 antibodies. Microinjection of Pyk2-WT cDNA completely rescued podosome belt formation in -null osteoclasts (), providing further evidence that Pyk2 is required for the formation of the peripheral podosome belt in osteoclasts. Similar results were observed when -null osteoclasts were microinjected with Pyk2-ΔFERM, Pyk2-Y402F, or Pyk2-K457A cDNAs (). In contrast, podosome belt formation was not observed in -null osteoclasts that were microinjected with Pyk2-ΔFAT. Instead, podosome clusters and small rings were seen throughout the microinjected osteoclasts (). Examination of the cellular distribution of ectopically expressed Pyk2 molecules showed, however, that both wild-type and all mutant Pyk2 were confined to actin-containing podosomes, including the Pyk2-ΔFAT mutant (Fig. S3, available at ).
Similar results were obtained when -null osteoclasts were infected with adenovirus containing wild-type Pyk2 or the different Pyk2 mutants, and matched for expression levels of endogenous Pyk2 (). Comparison of the infected cells demonstrated that only Pyk2-ΔFAT failed to induce belt formation in comparison to belt formation induced by wild-type Pyk2 (). In agreement with that, expression of Pyk2-ΔFAT was not able to either restore Rho activation or stabilization of microtubules, as demonstrated by a nocodazole-resistance assay (, respectively). The lack of a podosome belt in -null osteoclasts infected with the Pyk2-ΔFAT virus correlated with the absence of sealing zone in these cells after replating on dentin, as shown in .
The effect of Pyk2 mutants on osteoclast function was then examined in a resorption assay. Expression of Pyk2-ΔFERM in -null osteoclasts completely rescued their bone-resorbing activity. On the other hand, the bone-resorbing activity was only partially restored by the Pyk2-Y402F and Pyk2-K457A mutants, notwithstanding the complete restoration of podosome belt formation by these mutants (). The area, depth, and volume of pits formed by -null osteoclasts that were infected with Pyk2-ΔFAT virus were significantly reduced (), further confirming the importance of the FAT domain in promoting osteoclast function. These results suggest that although Src recruitment and Pyk2 kinase activity contribute to bone resorption, the FAT domain of Pyk2 plays a major role in podosome organization and bone resorption.
We have previously reported that Pyk2 deficiency results in impairment of macrophage function (). Here, we report that Pyk2 deficiency also leads to osteopetrotic phenotype by impairment in the bone-resorbing activity of osteoclasts. In addition, experiments are presented demonstrating that Rho activity, microtubule stabilization and podosome organization are altered in -null osteoclasts. Although we cannot exclude the contribution of osteoblasts to the increased bone mass, the accumulation of cartilage remnants despite elevated osteoclast numbers confirm in vivo an impairment in osteoclast function. Furthermore, the in vitro experiments establish the fact that -null osteoclasts are deficient in cytoskeletal organization and bone resorption, in a cell-autonomous manner. Thus, Pyk2 plays a critical role in podosome organization and osteoclast function.
Analysis of the cellular distribution of Pyk2 by immunofluorescence microscopy showed that Pyk2 is preferentially localized in the periphery of osteoclasts in a region that overlaps with the podosome belt, when the cells are plated on glass, or the similar sealing zone, when plated on mineralized bone. Comparison of the cellular distribution of actin revealed that Pyk2 deficiency results in impairment in the formation of podosome belt and sealing zone in osteoclasts. Instead, -null osteoclasts contain multiple podosome clusters and small podosome rings throughout the cell.
Given that a link has been established between microtubules and actin organization (), we have explored the possibility of whether microtubule distribution and/or dynamics were altered in osteoclasts deficient in Pyk2. Immunofluorescence studies showed that the cellular distribution of microtubules is significantly altered in -null osteoclasts; while -null osteoclasts retained the radial network, they lost their circular microtubule network that is usually seen at the periphery of osteoclasts. The stable, nocodazole-resistant pool of microtubules that has previously been implicated in mediating the transition from podosome clusters and rings into the peripheral podosome belt (), was nearly eliminated in -null osteoclasts. Consistent with this finding, the acetylation of microtubules, which correlates with microtubule stabilization, was also markedly reduced in -null osteoclasts. Microtubule acetylation is controlled in part by the activity of histone and microtubule deacetylase HDAC6, which in turn is regulated by the small GTPase Rho (). We found that Pyk2 deficiency results in increased Rho activity, and established that this leads to reduced microtubule acetylation and reduced microtubule stability. We propose that Pyk2 may function as an upstream inhibitor of this signaling pathway in osteoclasts, which ultimately allows the transition of podosomes from the center of the cells, where they form clusters and small internal podosome rings, to the periphery, where they form a podosome belt and sealing zone, a structure necessary for efficient bone resorption.
During directional migration of fibroblasts, the stable pool of microtubules becomes oriented toward the leading edge of the cell (). Moreover, the microtubules of fibroblasts are post-translationally modified, and as in osteoclasts, the modifications are associated with enhanced microtubule stability (). In addition, FAK activation in the leading edge of fibroblasts leads to enhanced Rho activity and microtubule stability (). In osteoclasts, on the other hand, enhanced stimulation of microtubule stability and podosome organization are controlled by Pyk2-mediated reduced Rho activation. Although the reason for this difference is unknown, this could be explained by different effects of Pyk2 and FAK on Rho activation, or, as suggested before (), could be due to cell type–specific effects of Rho inhibition on microtubule stability.
Ectopic expression of wild-type Pyk2 in -null osteoclasts, at levels comparable to endogenous expression of the protein in wild-type cells, rescued the formation of a podosome belt and sealing zone as well as bone resorption. The cellular responses were also rescued by expression of a deletion mutant of Pyk2 lacking the FERM domain. Podosome belt formation was rescued by expression of a kinase-negative Pyk2 mutant (K457A) and by a point mutant in a tyrosine phosphorylation site (Y402F) that is responsible for mediating complex formation with Src. In contrast, a deletion mutant devoid of the FAT domain was not able to rescue normal responses in -null osteoclasts, emphasizing the importance of this domain in the regulation of microtubule stability and podosome organization in these cells.
It was previously shown that the FAT domain is responsible for Pyk2 localization in focal contacts of fibroblasts and HeLa cells (; ; ). However, a deletion mutant of Pyk2 devoid of the FAT domain remains localized around the actin-rich podosome core in -null osteoclasts, suggesting that the inability of mutant protein to rescue podosome belt or sealing zone formation and bone resorption are not caused by a defect in cellular distribution. The FAT domain of Pyk2 may participate in podosome belt formation by regulating the activities of guanine nucleotide exchange factors or GTPase-activating proteins (GAPs) that regulate Rho activity, resulting in the control of acetylation and stabilization of microtubules in the pathways mentioned earlier in the Discussion, thereby regulating osteoclast cytoskeletal organization and bone resorption.
The rescue of Pyk2-dependent podosome belt formation in -null osteoclasts by the kinase-negative Pyk2 mutant suggests that Pyk2 may function as a platform for recruitment and assembly of signaling proteins in addition to its function as tyrosine kinase. Accordingly, other tyrosine kinases may compensate for the loss of intrinsic tyrosine kinase activity of Pyk2 by trans-phosphorylation of key tyrosine residues that function as docking sites for signaling proteins. Pyk2 may also recruit signaling proteins in a phosphorylation-independent manner through interactions mediated by its proline-rich region with SH3 domain–containing signaling proteins.
The effect of overexpression of Pyk2-Y402F in wild-type osteoclasts was previously described (; ), demonstrating that osteoclasts that overexpress this mutant fail to form podosome belt and show reduced bone resorption in vitro. Consistently, as demonstrated in our paper, in the absence of endogenous Pyk2, Pyk2-Y402F can partially restore bone resorption, but completely restores the formation of a peripheral podosome belt. This, together with the finding that microinjection of Src-Y527F into -null osteoclasts cannot rescue podosome belt formation, suggests that Pyk2 may mediate two separate pathways: a Rho-dependent pathway that regulates microtubule-dependent podosome organization, and a Src-dependent pathway that may be involved in other functions related to bone resorption, such as actin dynamics, cell attachment, or ruffled border formation.
We also show that actin dynamics are altered in the absence of Pyk2; the rate of incorporation of GFP-actin into podosomes in photobleached areas is slowed down by an approximately twofold in the absence of Pyk2. Previous reports have interpreted slower actin fluxes as an indication of reduced actin polymerization (), which could result from the inhibition of actin polymerizing proteins or from enhanced actin severing along the filaments. The role of actin polymerization in podosome formation and function is still poorly understood, but it has been suggested that polymerization induces a force that pushes the cell membrane toward the substrate (; ), a process that may also be essential for attachment of osteoclasts to their substrate and therefore to bone resorption.
Finally, the ruffled border is a membrane-rich organelle that is surrounded by the sealing zone, and through which protons and proteolytic enzymes are secreted to resorb the bone matrix. We have observed in -null osteoclasts a shorter and irregular ruffled border structure and defective translocation of the proton pump to this region in -null osteoclasts (unpublished data). This may explain the shallow pits formed by -null osteoclasts, and may suggest additional, yet unknown roles of Pyk2 in osteoclast function and bone resorption.
In conclusion, this study demonstrates that Pyk2 is required for normal organization of the cytoskeleton in osteoclasts, and bone resorption. In the absence of Pyk2, Rho activity is increased, microtubule acetylation and stabilization are decreased, and transition of podosomes to the periphery of the cell is prevented.
Generation of mice was described previously (). Animals were handled in accordance with the guidelines of Yale University Institutional Animal Care and Use Committee.
Bone samples were collected for histomorphometric analysis at 2 and 10 wk of age. Double-fluorochrome labeling was performed in 10-wk-old mice as described previously (); animals were injected with calcein (20 mg/kg body weight) followed by the same dose of demeclocycline at 10 and 3 d before tissue collection, respectively. Tibiae and femora were collected, fixed in 3.7% formaldehyde in PBS, and embedded in methylmethacrylate as described previously (). 5-μm sections were stained with Toluidine blue or Alcian blue or by the Von Kossa method, or analyzed unstained for fluorochrome labels. Histomorphometric analysis was performed according to standard procedures using the Osteomeasure system (OsteoMetrics, Inc.) in the proximal tibiae. Tibial cortical thickness and periosteal mineral appositional rates were measured as described previously (). Femoral length and width were determined from contact X-rays that were scanned and measured using NIH Image 2.0.
Mouse MCSF and RANKL were obtained from R&D systems. Nocodazole was obtained from Sigma-Aldrich. Cell-permeable C3 was obtained from Cytoskeleton, Inc. Monoclonal anti-Pyk2 and anti-FAK antibodies were purchased from Transduction Laboratories. Anti-pY402 Pyk2 and anti-pY418 Src antibodies were from Biosource International. Anti-v-Src monoclonal antibody was from Calbiochem. Monoclonal anti-RhoA antibody (26C4) was purchased from Santa Cruz Biotechnology, Inc. Anti–α-tubulin was from Abcam. Anti-acetylated tubulin was from Sigma-Aldrich. Anti-actin antibody was from Chemicon. Texas red– and rhodamine-conjugated phalloidin and secondary antibodies for immunofluorescence were purchased from Invitrogen.
pEGFP-actin was obtained from CLONTECH Laboratories, Inc. pShuttle plasmids containing Pyk2-WT (wild-type Pyk2), Pyk2-Y402F (mutation in Src SH2-domain binding site), Pyk2-K457A (a kinase-negative mutant), Pyk2-ΔFAT (deletion mutant devoid of the FAT domain), and Pyk2-ΔFERM (deletion mutant devoid of the FERM domain) were used for transfection and microinjection experiments. Recombinant adenoviruses expressing the above mutants were prepared by recombination of the above plasmids using the Adenovator Vector System (Qbiogene) according to the manufacturer's instructions.
Authentic osteoclasts from 2–4-d-old neonatal mice and co-culture osteoclasts from 6–8-wk-old mice were prepared as described previously (). Spleen leukocyte cells were prepared and differentiated in culture using MCSF and RANK ligand as described previously ().
Mouse spleen cell–derived osteoclasts were transferred to observation medium (a-MEM without bicarbonate containing 10% fetal calf serum, 20 mM Hepes, 20 ng/ml M-CSF, and 20 ng/ml of soluble recombinant RANK-L). Intranuclear microinjection of cDNAs (0.2 mg/ml in water) was performed at room temperature on an inverted microscope (model IX 71; Olympus) using an InjectMan NI2 micromanipulator and a FemtoJet Microinjector (Eppendorf). After microinjection, cells were maintained at 37°C and 5% CO for at least 6 h in differentiation medium before imaging.
Osteoclasts were differentiated in 35-mm glass-bottom Petri dishes, then transferred to observation medium. After microinjection of DNA coding for GFP-tagged actin, the dishes were placed on a 37°C heated stage (Carl Zeiss MicroImaging, Inc.) and cells were imaged with a microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) containing a 40× (NA 1.0) Plan-Apochromat objective, a 63× (NA 1.4) Plan NeoFluor objective, and equipped with a MicroMax 5-MHz camera (Princeton Instruments, Inc.). Meta Imaging Series 7.0 (Universal Imaging Corp.) was used to mount AVI movies from image stacks. Extracted images from stacks were processed (brightness/contrast adjustment) with Adobe Photoshop 6.0 and ImageJ.
For immunofluorescence, cells were fixed with 4% paraformaldehyde diluted into PBS, pH 7.4, processed as described previously (), and imaged with a microscope (LSM Meta; Carl Zeiss MicroImaging, Inc.) using a 63× (NA 1.4) Plan NeoFluor objective. To prevent cross-contamination between fluorochromes, each channel was imaged sequentially using the multi-track recording module before merging.
FRAP experiments were performed on osteoclasts prepared as for regular videomicroscopy experiments using the same confocal setup described previously. Bleaching time (3.2 s), acquisition rate (1 image every 547 ms), and bleaching area were the same for all experiments. Image extraction was performed with Meta Imaging Series (Universal Imaging Corp.). The fluorescence recovery was measured only within podosome cores that existed during the whole recovery time. Fluorescence intensity at each time point was normalized to the starting fluorescence intensity (pre-bleach). To analyze recovery kinetics, FRAP measurements were fitted to a single exponential curve (performed with Igor Pro 4.0; WaveMetrics) as described in .
HEK293 cells were transfected using Lipofectamine (Invitrogen) according to the manufacturer's instructions. Immunoprecipitation and immunoblotting were preformed as described previously ().
GST fusion protein containing the Rho binding domain of rhotekin (RBD) was produced and used for pull-down experiments with osteoclast lysates as described previously ().
Pit resorption assay was preformed as described previously (). Three-dimensional profiles of resorbed pits were characterized by using a reflective confocal laser scanning microscope (RCLSM) (model VK8510; Keyence) under the 50× objective lens (NA 0.8) interfaced via a CCD camera to “Virtual view 3D (version 2.5)” for making the three-dimensional reconstruction image profile. The images were displayed at a resolution of 1024 × 768 pixels. Quantitative analysis of resorbed pit number, area, and volume were performed using Win ROOF image-analyzing software (version 5.5; Mitani Corp.) ().
Fig. S1 shows the distribution of paxillin in wild-type and -null osteoclasts. Fig. S2 shows expression of the different mutants of Pyk2, their autophosphorylation status, and their binding to Src. Fig. S3 shows the cellular distribution of actin and Pyk2 in osteoclasts microinjected with the different Pyk2 mutants. Fig. S4 shows localization of microinjected mDia2-GFP in wild-type and -null osteoclasts. Online supplemental material is available at . |
Microtubules are ubiquitous cytoskeletal polymers made of α/β-tubulin heterodimers that are required for cell motility, intracellular transport, and mitosis/meiosis. Little is known about how organelle-specific properties of microtubules are generated. The length of microtubules is an important parameter that determines the size of cytoskeletal organelles, and affects the distribution of forces generated by microtubule-dependent motors. The length of microtubules can be regulated either by addition/loss of tubulin subunits at the polymer ends, or by internal breakage.
Katanin, a dimer of p60 and p80 proteins (), has a strong microtubule-severing activity. p60 is a catalytic subunit that oligomerizes and breaks microtubules while hydrolyzing ATP (). Katanin has been implicated in the organization of microtubule arrays in vivo and is required for assembly of the meiotic spindle in (; ). Surprisingly, katanin deficiency led to a decrease in the microtubule polymer mass in the meiotic embryo of , probably because severing generates microtubule fragments that serve as seeds for nucleation of new microtubules (). Katanin was also implicated in generation of short microtubules destined for transport to neural extensions (). In plants, katanin is required for normal assembly of cortical bundles (; ). Furthermore, p80, the noncatalytic subunit of katanin, is required for assembly of central pair microtubules in ciliary axonemes of (). Thus, katanin has emerged as a positive regulator of the microtubule polymer mass, which is critical for assembly of diverse microtubule arrays.
On the other side, katanin activity increases during the mitotic prophase, suggesting that severing plays a role in the disassembly of interphase microtubules, to allow for reassembly of tubulin into the mitotic spindle (). Thus, katanin could also function as a negative regulator of the microtubule mass. The net effect of katanin on the polymer mass could depend on the size of microtubule products that are released by severing. Katanin could negatively regulate the polymer mass if the product of severing is a tubulin dimer or a short microtubule that cannot seed assembly.
Here, we show that katanin plays a negative regulatory role in the management of nonciliary microtubules in . We show that katanin reduces the polymer mass as well as increases the level of polymer dynamics in the cell body. However, in the same cell, katanin promotes assembly of ciliary microtubules, and therefore its effects are microtubule-type specific. We also show that katanin-mediated severing is nonrandom in vivo and that its activity is required to inhibit accumulation of post-translational modifications (PTMs) on microtubules. Furthermore, katanin mutations phenocopy a mutation of the domain of β-tubulin involved in polymeric PTMs, glutamylation () and glycylation (). We propose that katanin regulates the longevity of nonciliary microtubules by preferentially depolymerizing post-translationally modified segments of the polymer.
To identify microtubule-severing factor genes in , we searched the recently sequenced genome of this ciliate for sequences encoding AAA type ATPases. We identified two sequences encoding a p60 katanin subunit, representing the predicted genes and , and a sequence encoding a spastin-like protein, . Spastin is a related ATPase with a microtubule-severing activity (; ). The predicted Kat1p, Kat2p proteins have an AAA domain, but Kat2p also has an N-terminal LisH domain (). Kat1p and Kat2p belong to conserved clades and Kat1p is more closely related to the well-studied vertebrate p60s (). The genome also contains a sequence encoding p80, the putative noncatalytic subunit of katanin, .
To evaluate the role of microtubule severing in vivo, we constructed strains lacking either , , , or has two nuclei: the germline micronucleus, and the somatic macronucleus. The macronuclear genome determines the phenotype. Heterokaryon strains were constructed that were homozygous for gene disruption in the micronucleus and had wild-type genes in the macronucleus. Heterokaryons have a wild-type phenotype during vegetative propagation, but when two such cells conjugate, they produce progeny cells with a new macronucleus that express a gene disruption phenotype. Heterokaryon progeny cells lacking either or had a normal morphology (see Fig. S3, available at ) and gave rise to clones with normal vegetative growth rates (Fig. S4 I). While 95% of control mating pairs gave rise to vigorous clones that grew to maximal cell density in drops (∼10 cells/ml), the isolated knockout heterokaryon pairs produced on average only 12 cells per mating pair ( = 60) and thus had divided completely only 2–3 times. After the initial complete cell cycles (presumably resulting in depletion of parental Kat1p), the -null cells underwent 1–3 incomplete cell cycles during which cells grew in size, the nuclei had divided, but the cleavage furrow did not complete its ingression. This led to the formation of multinucleated cell chains composed of cortical “subcells” (). In a wild-type vegetatively growing population, 97% of cells are composed of a single subcell and 3% have 2 subcells (cells undergoing cytokinesis). At 40 h (after mixing of heterokaryons), 90% of Kat1p-null cells were composed of 2–4 subcells and had on average 2.7 subcells/chain ( = 172) (). Thus, the absence of zygotic Kat1p leads to an arrest in cytokinesis.
In there are two distinct modes of nuclear division: the micronucleus undergoes mitosis at an early stage of cell division, while the macronucleus divides by amitosis (a microtubule-dependent nuclear division that does not involve chromatin condensation) at the time of cytokinesis. The subcells of -null cell chains contained a macronucleus and a micronucleus, indicating that the nuclear divisions and segregation of nuclei are not disturbed (). Thus, both mitosis and amitosis are insensitive to the levels of Kat1p. The -null cell chains remained viable for 3–4 d. However, no proliferating clones were obtained from mass matings of 2 × 10
knockout heterokaryons using paromomycin selection that eliminates nonprogeny cells (see Materials and methods). Thus, Kat1p is essential. Although Kat1p is not required for nuclear divisions, its activity is required for completion of cytokinesis.
During cytokinesis of wild-type cells, cortical longitudinal microtubule (LM) bundles (see ) undergo partial depolymerization within the equatorial region as the contractile ring ingresses. In the advanced stage -null cell chains labeled with anti-tubulin antibodies, LMs appeared thicker then normal. While some LMs are twisted and show gaps in the contraction areas, other LMs continued to span adjacent subcells despite the completion of nuclear divisions (, arrows). Because LMs are the only microtubule bundles that run perpendicularly to the cleavage furrow, we speculate that in the absence of katanin, some cortical LMs do not depolymerize properly at the time of cytokinesis and physically obstruct the ingressing cleavage furrow.
Within the cell body, cell chains lacking zygotic Kat1p had an abnormally dense network of intracytoplasmic microtubules. Quantitative immunofluorescence detected about twofold increase in the mass of insoluble tubulin in the cell body of -null cell chains, which indicates that the density of internal cytoplasmic microtubules increases in the absence of zygotic Kat1p (). Furthermore, the internal microtubules of -null mutants are hyperstable and excessively modified post-translationally (see below).
While wild-type swim rapidly due to the beating of cilia, the knockout cell chains were nearly paralyzed. Immunofluorescence showed that the mutant cell chains had two types of cilia: a few normal-length cilia that were more abundant near the ends of cell chains, and a majority of excessively short cilia (; see ). In the ciliary rows that cover the cell surface of , new cilia-templating basal bodies form near old basal bodies, and old cilia are not resorbed (). Most likely, the normal-length cilia are parental units, while the short cilia were assembled in the absence of zygotic Kat1p. The pattern of normal-length cilia in the -null cell chains is consistent with the known preference for insertion of new basal bodies within the mid-posterior region (; ). Transmission electron microscopy (TEM) showed that mutants lacking zygotic Kat1p assemble incomplete axonemes. Although all ciliary cross sections analyzed in wild-type cells had a 9+2 axoneme ( = 55) (), 80% of mutant cross sections ( = 205) had peripheral doublets (associated with radial spokes and dynein arms) but lacked a central pair (9+0) (). The remaining 20% of normal 9+2 cross sections most likely represent cilia assembled before depletion of Kat1p.
In other organisms, katanin functions as a complex of p60 and p80, a noncatalytic subunit (; ). The progeny of heterokaryons lacking the (p80) gene developed into paralyzed cell chains that were remarkably similar to those produced by the lack of ( and unpublished data). The simplest explanation is that Kat1p and Kat3p work in the same pathway, and both are required for katanin activity.
Importantly, a strikingly similar paralyzed cell chain phenotype (with 9+0 axonemes) was earlier observed for the βDDDE mutation of β-tubulin ( and ). This mutation inhibits acquisition of polymeric PTMs, glutamylation and glycylation on the C-terminal tail of β-tubulin (, ). Thus, katanin and polymodifications on β-tubulin could function in the same pathway.
To test whether Kat1p has a microtubule-severing activity, we overproduced GFP-Kat1p using an inducible promoter at high copy number. After induction, GFP-Kat1p accumulated in cells () and cell multiplication was inhibited (). After 2–7 h, overproducing cells had fewer microtubules based on a decrease in the immunofluorescence signal of polymerized tubulin (unpublished data). Nearly all microtubules disappeared after 12 h (). We were intrigued by the rapid loss of ciliary axonemes () because katanin was earlier implicated in deciliation, a rapid shedding of cilia caused by low pH, based on severing of outer doublets at the transitional zone (; ). TEM showed that many cilia of GFP-Kat1p–overproducing cells had severed axonemes, but in all cases analyzed the breakage points were present outside of the transitional zone (). 44% ( = 91) of TEM cross sections of cilia in GFP-Kat1p–overproducing cells lacked one or more outer doublet microtubule profiles (), whereas 21% of cross sections had singlet peripheral microtubules, indicating a selective loss of the entire B-tubule (). Strikingly, on most cross sections the central pair microtubules were intact (). Thus, the A-tubule of the doublet and the central microtubules are either less accessible or more resistant to severing by overproduced katanin. Importantly, 31% of ciliary cross sections ( = 91) had two axonemes profiles adjacent to each other (). In most of these cilia, one axoneme profile was more fragmented. In one case only the central pair and a single singlet peripheral microtubule were present in the axoneme fragment (). All these images collectively suggest a likely mechanism of loss of ciliary axonemes in GFP-Kat1p–overproducing cells. Initially, one or more of the outer doublets could undergo nicking, introducing a point of structural weakness, and the beating of cilia could break the axoneme (). Next, the distal portion of the axoneme could slide past the proximal portion. The residual axoneme and the broken fragment could undergo further fragmentation ().
Overexpressed GFP-Kat1p also affected the triplet microtubules of basal bodies. By inspecting the undulating membrane (UM) of the oral apparatus, we found additional signs of selectivity in the GFP-Kat1p–severing activity. In the UM, the basal bodies are arranged into two rows with the same orientation, with only the outer row being ciliated. Strikingly, only the basal bodies of the outer row showed defects and the affected triplets were located at the same circumferential positions in adjacent organelles (). Thus, Kat1p is a genuine severing protein, but the enzyme has preferred sites of activity on a subset of microtubules in vivo.
Attempts at generating polyclonal antibodies against Kat1p were unsuccessful. We tagged Kat1p with GFP by rescuing progeny of heterokaryons, with a GFP-KAT1 gene targeted to the native locus. However, the rescue strains did not have detectable GFP expression (based on Western blotting and immunofluorescence with anti-GFP antibodies; unpublished data). Thus, the native promoter-driven Kat1p could be below the detection limit. Next, we rescued mating heterokaryons with a fragment encoding GFP-Kat1p under the control of a cadmium-dependent promoter maintained at a low copy number (see Materials and methods). The GFP-Kat1p rescue strains grew more slowly in the absence of exogenous cadmium, but had a normal rate of multiplication with cadmium (). In immunolabeling with anti-GFP antibodies, GFP-Kat1p was detected near the basal bodies (). In fixed cells that were not processed for immunofluorescence, we also detected GFP-Kat1p fluorescence as lines on a side of rows of basal bodies (), indicating that GFP-Kat1p also associates with LMs (see ). Thus, light microscopy studies detected GFP-Kat1p in the cell cortex (near basal bodies and along LMs) but not inside cilia. However, using post-embedding immunogold EM we detected GFP-Kat1p epitopes in two locations: near the basal bodies () and within cilia. Strikingly, GFP-Kat1p was seen exclusively near the outer doublets and not near the central pair (). Although control ciliary sections of wild-type cells lacked gold particles ( = 25), 71% of sections of cilia in GFP-Kat1p–expressing cells were labeled, and in all cases the particles were found near the outer doublets ( = 51).
Our observation that katanin-null mutations phenocopy a β-tubulin polymodification domain mutation ( and ) opened a possibility that katanin interacts with post-translationally modified microtubules. To further explore the relationship between microtubule PTMs and katanin, we examined the levels of specific PTMs in the Kat1p-null cells. In wild-type cells, tubulin polymodifications (glutamylation and glycylation) are present on most microtubules. However, on the relatively dynamic internal microtubules, the side chains are limited to a single G () or E (unpublished data). To evaluate the effect of Kat1p on the pattern of polymodifications, we used antibodies that recognize side chains with three or more units (ID5 for polyglutamylation () and AXO49 for polyglycylation ()) for immunofluorescence. Although all wild-type cells analyzed lacked internal polymodified microtubules ( = 58), most -null cell chains had detectable polyglycylated and polyglutamylated internal microtubules in the cell body (56% [ = 48] and 75% [ = 65], respectively) (). In the cell cortex of wild-type cells, polyglutamylation is detectable only in basal bodies and cilia (Fig. S2 C, available at ), whereas cortical bundles are only monoglutamylated (unpublished data). However, in most -null cell chains, cortical LMs were labeled with an anti-polyglutamylation antibody (Fig. S2 D). All cortical microtubules (basal bodies, transverse microtubule bundles, and post-ciliary microtubule bundles; see ) were labeled with the AXO49 anti-polyglycylation antibody in both wild-type and -null cells (unpublished data). Thus, based on the pattern of antibody labeling, we could not determine whether a polyglycylation side chain lengthening has occurred in the absence of Kat1p.
In wild-type cells, an assembling short cilium labels strongly with an anti-polyglutamylation antibody ID5, but the signal decreases as the cilium elongates and matures. The anti-polyglycylation antibody AXO49 gives a complementary pattern with a higher signal in mature cilia compared with short assembling cilia (). In the -null cell chains, the short cilia (presumably assembled in the absence of zygotic Kat1p) had higher level of polyglutamylation and lower level of polyglycylation as compared with normal-length cilia of parental origin concentrated mainly at cell extremities (). Thus, based on the pattern of microtubule PTMs, the cilia assembled in the absence of zygotic Kat1p resemble immature normal cilia. Although the pattern of polymodifications on most cilia in KAT1-null cells is abnormal, it is unclear whether katanin directly affects the levels of PTMs on ciliary microtubules or affects another process that is required for elongation and maturation of cilia.
In , the internal cytoplasmic microtubules are dynamic, judged by their sensitivity to microtubule-depolymerizing compounds () and lack of K-40 α-tubulin acetylation (a marker of stable microtubules []). Strikingly, based on immunofluorescence with the 6–11 B-1 antibody that recognizes α-tubulin acetylated at K-40, the Kat1p-null cell chains had strong acetylation on the internal cytoplasmic microtubules (). Although all wild-type cells analyzed did not have detectable acetylated internal microtubules ( = 50), such microtubules were detected in 99% ( = 76) of -null cell chains. Using quantitative immunofluorescence, a severalfold increase in the signal of acetylated tubulin was detected in the cell body of cell chains (). A Western blot of total cell chains () showed levels of tubulin acetylation similar to wild-type cells, but cell chains have excessively short cilia and therefore lack a major source of acetylated tubulin that wild-type cells have (). Thus, the Western blot result is consistent with an increase in acetylation on nonciliary microtubules in the absence of zygotic Kat1p. Cells lacking Kat3p (p80) also had hyperacetylated internal microtubules (unpublished data). However, elimination of either Kat2p or Spa1p did not lead to hyperacetylation of internal microtubules (Fig. S3, A–C). The accumulation of hyperacetylated microtubules is not an indirect consequence of absence of cytokinesis because cells blocked in cytokinesis by a distinct mechanism did not accumulate hyperacetylated cytoplasmic microtubules (Fig. S2, A and B). To summarize, a deficiency in katanin increases the levels of several PTMs on nonciliary microtubules. Specifically, internal microtubules show increased levels of glutamylation, glycylation, and acetylation, whereas cortical microtubules are hyperglutamylated in the absence of zygotic Kat1p. Thus, katanin functions as a negative regulator of microtubule PTMs in the cell body.
Internal cytoplasmic microtubules of depolymerize in the presence of nocodazole, whereas cortical and ciliary microtubules do not (). The accumulation of acetylation and polymodifications on internal microtubules in the Kat1p-null cell chains indicated that these microtubules were hyperstable. Indeed, after 1 h treatment with 40 μM nocodazole, all wild-type cells examined lacked detectable cell body microtubules ( = 60) (), whereas in 84% ( = 75) of Kat1p-null cell chains, cytoplasmic microtubules remained abundant (). Thus, the phenotype of deficiency could be caused to some extent by hyperstability of MTs. To test this hypothesis further, we incubated a population of -null cell chains with either 2 μM oryzalin or 10 μM paclitaxel to either destabilize or hyperstabilize MTs, respectively. Remarkably, oryzalin increased the number of complete cell divisions undergone by cells lacking zygotic Kat1p (). The treatment with paclitaxel blocked multiplication of Kat1p-null cells (). At these concentrations, neither of the drugs affected the growth of wild-type cells (). This pharmacological profile indicates that katanin-deficient cells have hyperstable microtubules and that katanin increases dynamics of microtubules in vivo.
Deciliation is a rapid shedding of cilia in response to chemical stresses (e.g., low pH), based on breakage of the axoneme within the transitional zone. Katanin was implicated in severing of axonemes in (). Surprisingly, the Kat1p-null cell chains deciliated in low pH. Nearly all cilia were lost, indicating that the short cilia that assemble under katanin deficiency also shed (Fig. S5, A and B; available at ). The deciliated Kat1p-null cells regenerated uniformly short cilia within 2 h (Fig. S5 C) and remained paralyzed, indicating that the regenerated cilia were 9+0. Double-knockout cell chains lacking and also shed (Fig. S5, D and E) and regenerated cilia (unpublished data), indicating that in deciliation is not dependent on p60 and p60-like proteins. Furthermore, the ability of katanin- deficient cells to regenerate cilia indicates that the absence of katanin does not lead to a general deficiency in the pool of unpolymerized tubulin in the cell body. This was confirmed by a Western blot that showed that cell chains had a normal level of soluble (Triton X-100 extractable) tubulin ().
We have investigated the functions of katanin- and spastin-like proteins in . The use of enabled us to establish the significance of severing in a cell type with diverse microtubules, including those forming internal networks, cortical arrays, and axonemes. We show that the absence of spastin-like protein Spa1p, and katanin-like protein, Kat2p, does not detectably change the phenotype of vegetative cells. It remains to be determined whether Kat2p and Spa1p play a role during conjugation, the sexual stage of ciliate life cycle. Furthermore, spastins are phylogenetically related to fidgetins and the genome contains a single sequence encoding a fidgetin-like protein (). In light of the recent demonstration that fidgetin contributes to shortening of spindle microtubules during mitosis in (), in the future, a potential functional redundancy between spastin and fidgetin will need to be addressed in .
We show that katanin affects the microtubule polymer mass differentially depending on the intracellular location. Specifically, we show that katanin decreases the polymer mass of nonciliary internal microtubules, and increases the mass of ciliary microtubules. In meiotic embryos, deficiency in katanin decreased the polymer mass, likely because in this cell type, katanin produces shorter microtubules that provide new free microtubule ends for polymerization (; ). The different outcome that we observed for the cell body of can be reconciled with the studies, if, in , katanin releases tubulin dimers or short microtubules that cannot prime assembly. Thus, the consequences of katanin presence could depend on the size of severing products as well as on the cellular context that determines the fate of microtubule fragments. Interestingly, the consequences of the loss of spastin in other organisms were similar to what we observed in for katanin. A mutation of murine spastin induced swellings in axons that were enriched in detyrosinated, stable microtubules (). A knockdown of spastin mRNA in caused stabilization of microtubules (; ), although another study reported a reduced number of microtubules in spastin-deficient (). Thus, most of the studies on spastin, collectively with our data, indicate that microtubule severing increases polymer dynamics and reduces PTMs on microtubules.
Our data indicate that katanin displays a high level of microtubule substrate selectivity in vivo. A moderately overproduced GFP-Kat1p localized with only a subset of microtubule locations. We detected tagged Kat1p around the basal bodies, near LMs and doublet microtubules inside cilia. The basal body–associated katanin may be involved in severing of minus ends of internal cytoplasmic microtubules and may contribute to the dynamic character of these microtubules. Cortical LMs are bundles of partly overlapping microtubules with a uniform polarity (). Katanin may be involved in the turnover of segment microtubules in LMs, and its deficiency could lead to abnormal persistence of LMs and block cytokinesis.
Although it is unclear whether katanin under physiological conditions severs stable microtubules of cilia and basal bodies, these structures were severed by an overproduced katanin. Remarkably, katanin preferentially severed microtubule triplets located at specific positions, confirming that basal bodies have radial asymmetry (; ). Inside cilia, katanin severed doublet microtubules, whereas the central pair microtubules were unaffected. Within the doublets, the B-tubule was more prone to severing compared with the A-tubule.
Two mechanisms could restrict katanin activity: differential binding of microtubule-associated proteins (MAPs) that sterically block katanin, or differential marking of microtubules with PTMs. It is already known that Tau MAP protects axonal microtubules against katanin severing (). However, binding of MAPs to microtubules is strongly affected by microtubule PTMs (; ; ). Thus, katanin could be directly or indirectly regulated by PTMs of microtubules.
Previously, we described the consequences of mutations that affect the sites of polymodifications on the tail domains of β-tubulin (, ). We now show that nearly all these defects are also present in the katanin-null mutants. A simple explanation is that katanin requires polymodified microtubules for its transport to proper sites of activity. Recently, acetylation of α-tubulin at Lys-40 was implicated in transport mediated by kinesin-1 (). Alternatively, polymodifications could mark preferred sites for katanin-mediated severing. Importantly, polymodifications occur on the C-terminal tail domains of tubulins and both katanin and spastin require tubulin tails for severing activity (; ; ). Furthermore, in , a substitution of a potentially polymodifiable glutamic acid in the tail domain of β-tubulin rescued lethality associated with overproduction of katanin (), supporting an idea that a polymodified tail of β-tubulin increases katanin activity. Importantly, the known pattern of glutamylation in vivo correlates with the preferred sites of katanin activity revealed by our study. For example, the axonemal B-tubules have higher levels of glutamylation, as compared with the A-tubules and the central pair (; ; ). The transitional zone appears to lack glutamylation (; ) and showed resistance to katanin severing in our study. Furthermore, microtubule types that become hypertrophic in the absence of zygotic katanin in (internal and LMs) are at least monoglutamylated in wild-type cells. The recent identification of tubulin glutamylases () should allow for a direct test of the role of tubulin glutamylation in katanin activity.
A recent study showed that a mutation of katanin p80 subunit inhibits the central pair formation in (). We now show that both subunits of katanin are required for assembly of motile normal-length cilia in and that lack of the catalytic katanin subunit blocks assembly of the central pair. Collectively, the results of and our observations uncover an evolutionarily conserved role of katanin in cilia biogenesis, and in particular its critical role in the central pair assembly. We show that Kat1p p60 has a strong microtubule-severing activity in vivo. It is therefore reasonable to propose that microtubule severing is required for assembly of cilia. Two (non-mutually exclusive) models can be proposed. First, the role of katanin in cilia biogenesis could be indirect and based on the dependence of cilia assembly on proper dynamics of microtubules in extra-ciliary locations. Furthermore, katanin-mediated severing could generate a pool of tubulin dimers or oligomers in the cell body, which could then be used as precursors for ciliary assembly. Although Kat1p-deficient cell chains have normal levels of soluble tubulin and regenerate 9+0 cilia after deciliation, katanin may generate a subpool of precursor tubulin required for the central pair formation. Alternatively, katanin could function inside cilia. Ciliary protein precursors are delivered by the intraflagellar transport (IFT) pathway (). It is possible that precursor tubulin transported to ciliary tips by IFT has a form of oligomers or very short microtubules, as has been proposed for tubulin transported inside neural extensions (; ). Katanin may be needed inside cilia for fragmentation of an oligomeric tubulin cargo before its assembly. It is intriguing, however, that overexpressed GFP-Kat1p preferentially bound and severed the outer doublets. This activity appears unrelated to deciliation response, as severing occurred outside of the transition zone, and cells lacking zygotic Kat1p p60 underwent deciliation. The association of katanin with doublets could be dismissed as an artifact of an abnormally high level of GFP-Kat1p and the potential high affinity of katanin for polymodified tubulin. In , however, antibodies detected p80 specifically in the outer doublet compartment (). Thus, the ciliary function of katanin could require its association with the outer doublets, possibly based on a novel form of severing of these microtubules that could provide supply of tubulin required for the central pair assembly. Interestingly, we recently found that several Nima A kinases (NRKs) promote depolymerization of axonemal microtubules when overproduced, but accumulate in assembling cilia when expressed at a normal level (). Thus, a form of turnover of axonemal microtubules mediated by katanin could play a role in the assembly of cilia.
Sequences of AAA domain proteins were obtained from National Center for Biotechnology Information (NCBI) databases. Gene accession numbers of sequences used for phylogenetic analyses are listed in the legend of Fig. S1 (available at ). The AAA domain sequences were aligned using ClustalX 1.82 and corrected manually in SEAVIEW (). A tree was calculated using the Phylip package (). 1,000 replicates of the sequence set were created using SEQBOOT. The distances were calculated in PROTDIST, and trees were reconstructed using NEIGHBOR. The Jones-Taylor-Thorton (JTT) substitution model was used. A consensus tree was obtained using CONSENSE and the tree was plotted using DRAWGRAM.
The plasmid constructs were made based on the macronuclear genome sequence of (), available at the Genome Database. To prepare a targeting fragment for disruption of , a 4.3-kb macronuclear genomic DNA fragment of was amplified with addition of SacI and ApaI sites and cloned into pBluescript (SK+). The primers used were: 5′-TTATAGAGCTCCTATGTATTTTGAGCAGGTC-3′ and 5′-TATAAGGGCCCGGCTTTTAATGTTCTCTTGA-3′. The pMNBL plasmid () carrying the gene was modified to reduce the size of the promoter to 0.9 kb. The shortened cassette of pMNBL starting at the AccI in promoter was subcloned, giving pTvec-neo3. pTvec-neo3 was digested with SpeI to release . The cloned genomic fragment was digested with SpeI and was inserted, resulting in pTvec-neo3R-KAT1. To prepare a plasmid for disruption of , 1.4 kb of 5′ UTR was amplified with addition of SacI and BamHI sites (primers: 5′-AATTTGAGCTCTGCAAAGCTACTACCAAGAT-3′ and 5′-ATATTGGATCCTTCATACGAGATTCACCTTC-3′) and cloned into p4T2ΔHindIII, a cassette plasmid, using SacI and BamHI sites. The resulting plasmid was digested with XhoI and ApaI and used to insert a 1.8-kb fragment of 3′ UTR of (primers: 5′-AATAACTCGAGGTAGACCAAAATAACACACT-3′ and 5′-TATATGGGCCCCCTTTGTTTCTTTGGATTTG -3′), to create pNeo2R-KAT2. To disrupt , a 1.3-kb fragment of 5′ UTR was amplified with addition of SmaI and ApaI restriction sites (primers: 5′-ATAATGGGCCCTACTTAAAATCTTCTTCTTCTA-3′, 5′-AATATCCCGGGTGTTTCTATTTAATGGTTTGTC-3′) and cloned into pTvec-neo3. The resulting plasmid was digested with ClaI and SacI and used to insert an amplified 1.5 kb of the 3′ UTR of (primers 5′-TAATAATCGATGTTTAACGTTGATGGAGAT-3′ and 5′-AATAAGAGCTCGCATCCATAACATAACAAGG -3′) to create pKAT3-neo3R. To disrupt , its 5′ UTR was amplified with addition of SmaI and ApaI sites (primers 5′-TATATGGGCCCAAAGTAGTAAACAAACCTCAAT-3′ and 5′-AATTACCCGGGATTACTTTTTACACTATTCAGC-3′), and cloned into pTvec-neo3. The resulting plasmid was digested with ClaI and SacI and used to clone a 3′ UTR of flanked with ClaI and SacI (primers 5′-TATATATCGATCCATAAGAAATAGACTCAGC-3′ and 5′-ATAAACTAAATAATTGATGTGAACTGA-3′).
For germline targeting, each disruption plasmid was digested with SacI and ApaI and used to transform mating CU428.1 and B2086 strains by biolistic bombardment. Heterokaryons were generated by bringing the micronucleus to homozygosity using a star cross while allowing the disrupted alleles to assort from the macronucleus ().
To overexpress Kat1p at a high copy number, the coding region of was amplified with primers carrying MluI (5′-TATATACGCGTCATGTCAAATTCAGATAAACAATTA-3′) and BamHI (5′-TAATTGGATCCCTATCAAACAGAACCAAATTCT-3′) sites and cloned into pMTT1-GFP to create pMTT1-GFP-KAT1. To overexpress GFP-Kat2p, the coding region of was amplified with primers carrying MluI (5′-ATTATACGCGTCATGAGTTATCTACTATCAAAA-3′) and BamHI (5′-TAATTGGATCCTCAAACTGAACCATGTTCCTT-3′) restriction sites and cloned into pMTT1-GFP to create pMTT1-GFP-KAT2. To overexpress GFP-Spa1p, the coding region of was amplified with primers carrying MluI (5′-TATATACGCGTCATGGATAGTATC AAAAAAAAAGAAG-3′) and BamHI (5′-TAATTGGATCCTCAAACCTATTT ATTATA CTCTTTG-3′) restriction sites and cloned into pMTT1-GFP to create pMTT1-GFP-SPA1.
To introduce transgenes, starved CU522 cells (from Donna Cassidy-Hanley, Cornell University, Ithaca, NY) were bombarded with a SacI- and XhoI-digested overexpression plasmid, and transformants were selected on SPP medium with 20 μM paclitaxel. Using this approach, the transgene integrates by homologous recombination into the nonessential gene that carries a mutation conferring sensitivity to paclitaxel. The copy number of the transgene was increased by allowing cells to assort the mutant allele during vegetative propagation in the presence of paclitaxel.
To express GFP-Kat1p as a sole Kat1p at a low copy number, we rescued mating heterokaryon progeny from death, by introducing a transgene but without applying any selection directly to increase the transgene copy number. The knockout heterokaryon strains were allowed to complete conjugation during 24 h, transformed biolistically with a BTU1-MTT1-GFP-KAT1-BTU1 fragment. Transformants that integrated the transgene into the locus were selected with paromomycin (120 μg/ml) and cadmium chloride (2.5 μg/ml) (based on cadmium-dependent resistance to paromomycin conferred by the gene inserted into the native locus).
We used the immunofluorescence protocol described in for analysis of Kat1p-null cell chains. The immunofluorescence localization of GFP-Kat1p in rescue cells was done as described in . The following primary antibodies were used: 12G10 anti– α-tubulin (1:10 dilution; mouse monoclonal; University of Iowa, Developmental Studies Hybridoma Bank), 20H5 anti-centrin (1:100; mouse monoclonal; a gift of J. Salisbury, Mayo Clinic, Rochester, MN), anti-total tubulins SG (1:600; rabbit polyclonal; a gift of M. Gorovsky, University of Rochester, Rochester, NY), AXO49 anti-polyglycylated tubulin (1:100; mouse monoclonal (a gift of M.-H. Bré and N. Levilliers, Université Paris-Sud, Orsay, France; ), ID5 anti-polyglutamylated tubulin (1:10; mouse monoclonal) (a gift of K. Weber, Max Planck Institute, Goettingen, Germany; ), anti-GFP (1:500; rabbit polyclonal) (Abcam), and 6-11B1 anti-acetylated α-tubulin (1:20; mouse monoclonal) (Sigma-Aldrich). FITC- or Cy-3– conjugated secondary antibodies (Zymed Laboratories) were used at 1:100. Nuclei were stained with DAPI (Sigma-Aldrich). Cells were viewed with a Leica TCS SP2 spectral confocal microscope (using 63× water immersion with 1.2 NA). Images were assembled in Adobe Photoshop 8.0.
For transmission electron microscopy of knockouts, ∼5,000 mutant cell chains were isolated and washed two times with 10 mM Tris HCl buffer at pH 7.5. The cells were fixed in 2% glutaraldehyde in 0.1M sodium cacodylate buffer at pH 7.2 for 1 h at 4°C. Fresh tannic acid was added to 0.01% for 1 h at 4°C. Cells were washed five times with 10 mM Tris pH 7.5 and postfixed in 1% OsO for 1 h at 4°C, washed five times in 4°C with water before dehydration through a graded ethanol series. Cells were transitioned to Epon 812 resin with acetone at 33, 66, and 100% intervals. Cells were infiltrated with 100% Epon for 8 h at 25°C. Fresh 100% Epon was added and allowed to polymerize at 60°C. Ultrathin sections of 50–60 nm were collected and post-stained with uranyl acetate and lead citrate. Sections were visualized on JEOL 100CXII, JOEL 1200 EX, or FEI Technai 20 transmission electron microscopes.
Immunogold labeling was performed on cells expressing GFP-Kat1p at a low copy number as a result of rescue of mating knockout heterokaryons. A post-embedding procedure was performed as described previously (). Ultrathin sections were immunolabeled with the polyclonal anti-GFP antibody (1:500; Abcam), followed by 10-nm colloidal gold-conjugated goat anti–rabbit IgG (GE Healthcare). Immunolabeled sections were stained with 2% aqueous uranyl acetate for 30 min and washed several times with distilled water until no signal was detectable in the control sections made for cells not expressing GFP.
To evaluate lethal phenotypes, pairs of mutant heterokaryon strains were allowed to mate and single pairs were isolated into drops of SPP medium 8 h later. The average number of cells and subcells (cortical domains of cell chains) per drop was determined for 48–96 drops at various time intervals using a dissecting scope. To measure the multiplication rate of vegetatively growing strains, cells were diluted to 10 cells/ml from a feeder culture of 2 × 10 cells/ml and grown without shaking in 10 ml of SPP. The cell densities were measured every 2 h. In some experiments, to induce MTT1-driven GFP-Kat1p expression, CdCl was added at 2.5 μg/ml.
To test the effects of oryzalin and paclitaxel on multiplication of cells lacking zygotic Kat1p, the -null phenotype was induced as follows. knockout heterokaryons were grown vegetatively to a density of 2 × 10 cells/ml. Cells were starved in 10 mM Tris-HCl for 24 h and 5 ml of each of the two heterokaryon strains were allowed to mate in a 50-ml conical flask for 24 h. Cells were spun down and suspended in 10 ml SPP to a final concentration of 10 cells/ml. Paromomycin (120 μg/ml) and cadmium chloride (2.5 μg/ml) were added to inhibit the growth of nonmating cells (note that between 5–10% of heterokaryon cells do not mate, and therefore retain a wild-type phenotype; these cells were prevented from overgrowing the population by addition of paromomycin based on drug resistance conferred by the selectable markers used for gene disruption). The resulting suspension, highly enriched in exconjugant cells lacking zygotic Kat1p, was incubated in SPP with or without 10 μM paclitaxel, 2 μM oryzalin, or 1% DMSO as a control for 80 h. 10 μl of cell suspension was scored in the DIC microscope and the total number of cells was determined (we scored cell chains as 1 cell regardless of the number of subcells) at 24, 36, 50, 65, and 80 h.
Fig. S1 contains a multiple sequence alignment corresponding to the phylogeny of AAA proteins shown in Fig. S2 shows a lack of a correlation between an arrest in cytokinesis and increased acetylation of microtubules in the cell body (A and B), and documents increased polyglutamylation on cortical microtubules in cells lacking Kat1p (C and D). Fig. S3 documents that cells lacking either Kat2p or Spa1p do not accumulate PTMs on internal microtubules. Fig. S4 documents lack of an effect of overexpression of Kat2p and Spa1p on microtubules and cell multiplication. Fig. S5 shows that cells lacking katanin p60 subunits (Kat1p and Kat2p) undergo deciliation and cilia regeneration. Online supplemental material is available at . |
The growth and maintenance of the axon is dependent on a continuous array of microtubules that extends from the cell body of the neuron into the growth cone. Individual microtubules assume a variety of lengths within the array, ranging from <1 to >100 μm (). Very short microtubules are able to move rapidly and in both directions (). In contrast, the longer microtubules in the axon are essentially immobile (; ). Twice as many short microtubules move in the anterograde direction, presumably to ensure that more tubulin enters the axon than is moved back to the cell body. In addition to supplying tubulin for axonal growth, mobility within the microtubule array promotes the morphological plasticity underlying events such as growth-cone motility and branch formation (; ). The longer microtubules are important too because they act as compression-bearing struts that prevent the axon from retracting under the myosin-II–based contractile forces imposed on the cortical actin (). We have proposed a model whereby the same molecular motors that transport the short microtubules impinge upon the long microtubules to regulate the degree to which they can resist the myosin-based contractile forces ().
We have begun to study the panoply of microtubule-based motors that generate forces against long and short microtubules. We have found that if we deplete cytoplasmic dynein or if we depolymerize actin filaments, the frequency of anterograde microtubule transport is reduced by about half (; ). These results are consistent with a scenario in which short microtubules move anterogradely via a “sliding filament” mechanism in which the cargo domain of the motor interacts (indirectly) with actin filaments (; ). The motor domain is then available to transport the short microtubules. Consistent with our model for long and short microtubules, depletion of cytoplasmic dynein also renders the long microtubules less capable of resisting the myosin-II–based contractile forces that cause the axon to retract (; ). The fact that not all of the mobility within the microtubule array depends on cytoplasmic dynein indicates that there are other participating motors, presumably kinesins.
The relevant kinesins are the so-called mitotic kinesins, which generate forces between adjacent microtubules rather than between microtubules and membranous cargo (). We have shown that many of these specialized kinesins continue to be expressed in terminally postmitotic neurons (; ; ). We are particularly interested in kinesin-5 (also known as Eg5), which forms homotetramers consisting of two sets of antiparallel motor domains (). In the mitotic spindle, the primary function of kinesin-5 is to maintain spindle bipolarity by generating centrosome-directed forces between antiparallel microtubules in the midzone (; ,; ). In neurons, pharmacological inhibition of kinesin-5 results in longer axons (; ), suggesting a key role for this motor in the development of the nervous system. In the present study, we have investigated the underlying mechanisms by which kinesin-5 influences the development of the axon.
#text
We previously speculated that kinesin-5 might be the motor that transports microtubules retrogradely in the axon, based on the premise that axonal growth rates may be determined by the ratio of anterograde/retrograde microtubule transport (; ). In this view, depleting kinesin-5 would cause axons to grow faster because of a marked increase in this ratio. However, the current results show something quite different, i.e., that the frequency of microtubule transport in both directions is markedly increased when kinesin-5 is depleted. These results indicate that kinesin-5 is not the motor that transports microtubules retrogradely. Instead, it appears that kinesin-5 restricts the transport of short microtubules by other motors. The results of the depletion studies () suggest that there is no directional preference on this effect, as removing kinesin-5 results in marked increases in both directions. In this sense, it may not be the ratio of anterograde/retrograde transport that is most critical to axonal growth, but rather, the overall vitality of the transport. The overexpression studies (), however, raise the possibility of a directional preference, suggesting that kinesin-5 may be more suited to opposing microtubule transport in the anterograde direction. This possibility is consistent with other studies in which kinesin-5 was shown to antagonize minus end–directed motors such as cytoplasmic dynein and kinesin-14a (; ; ; ). It may be that kinesin-5 restricts microtubule transport in both directions, but preferentially in the anterograde direction.
Additional observations suggest that the greater length of kinesin-5–depleted axons is not solely explicable (or even principally explicable) on the basis of the changes in microtubule transport. The axons depleted of kinesin-5 are also far less prone to retraction, and this appears to be a major factor in their capacity to achieve greater lengths. This phenomenon is particularly evident with regard to the greater number of branches displayed by axons depleted of kinesin-5, where the numbers of branches appear to be dependent on the propensity of newly formed branches to retract. This is important because the proclivity of the axon to retract is not regulated by short microtubules, but rather by the microtubules that are long enough to act as compression-bearing struts (; ; ). Thus, we would contend that kinesin-5's principal role in regulating axonal length may be to influence the balance of forces on the long microtubules, which is consistent with our overall view that the various motors that transport microtubules are not selective for short microtubules but rather impinge upon microtubules of all lengths ().
How does kinesin-5 elicits these effects? In the mitotic spindle, kinesin-5 is known to act between microtubules of opposite polarity orientation (interpolar microtubules). Although kinesin-5 was originally believed to function mainly to assist in the separation of the half spindles, newer studies indicate that it can alternatively function to resist separation of the half spindles by other motor proteins (; ; ). In this newer role, kinesin-5 has recently been likened to a brake on the microtubules within the spindle because it acts as a rate-limiting factor on the ability of the microtubules to slide relative to one another. The analogy of kinesin-5 acting as a brake on microtubule sliding is also quite applicable to our observations on the axon. The difference is that the mitotic spindle consists of regions of uniform and nonuniform microtubule polarity orientation, whereas nearly all of the microtubules in the axon are thought to be of the same orientation (; ). Thus it remains unclear whether the mechanism underlying the kinesin-5 brake is the same in the axon as in the mitotic spindle or one that is quite different.
It is important to note that kinesin-5 can also generate forces between neighboring microtubules of the same orientation within the mitotic spindle (). Indeed, some of the earliest studies indicating antagonism between kinesin-5 and minus-end directed motors were conducted on monoasters of uniform polarity orientation (). More recent studies suggest a complexity of possibilities for how kinesin-5 is able to generate different kinds of forces between microtubules of the same or different polarity orientations (Kapetein et al., 2005; ). One possibility is that the two ends of the homotetramer are not always in synchrony with regard to force generation, thus generating options for the manner by which forces impact the relevant microtubules. It appears that kinesin-5 is also quite different from other kinesins with regard to how it reacts to changes in load (Valentine et al, 2006), which would likely influence kinesin-5's force-dependent functions. Whatever the case, understanding the complexities of force generation by kinesin-5 may clarify how kinesin-5 is capable of opposing transport of microtubules in both directions within the axon.
In discussing its mechanism of action, it is also important to consider that not all of the properties of kinesin-5 are force dependent. Kinesin-5 has been observed to cross-link neighboring microtubules () and to do so in a passive manner that does not require ATP (). In this sense, kinesin-5 could antagonize microtubule sliding simply by cross-linking adjacent microtubules or by displacing other motors from the microtubule lattice, even when kinesin-5 is not undergoing a power stroke. Our studies using the rigor mutant indicate that the force-dependent properties of kinesin-5 are required for it to elicit its effects on axonal growth. Even so, we are reluctant to completely dismiss the possibility that the force-independent properties of kinesin-5 might contribute to its functions in the axon. For example, the growth cones of kinesin-5–depleted axons are clearly broader than those of control axons (), suggesting that the microtubules within these growth cones may be less cross-linked in the absence of kinesin-5. Thus it may be that the dual properties of kinesin-5 are manifested differently in the growth cone and in the axonal shaft. In this regard, it will be prudent in the future to conduct studies with the rigor mutant specifically on growth cone behaviors.
Another open question is whether kinesin-5 opposes all other motors equally, or if kinesin-5 specifically or preferentially antagonizes minus end–directed motors in the neuron. We have not yet studied kinesin-14a in neurons, but we have conducted a great deal of work on cytoplasmic dynein. As noted in the Introduction, these studies have enabled us to propose a model whereby cytoplasmic dynein generates forces between the longer microtubules in the axon and the cortical actin meshwork. These dynein-driven forces antagonize the contractile forces imposed on the cortical actin array by myosin-II (; ). Our finding that kinesin-5 promotes axonal retraction is consistent with a scenario by which forces generated by kinesin-5 antagonize the forces generated by cytoplasmic dynein. Our tentative thinking is that the microtubules nearest the cortex interact with the cortical actin via cytoplasmic dynein, and that these same microtubules sometimes interact with adjacent microtubules via kinesin-5. When the kinesin-5 is engaged there is an increase in the load on the cytoplasmic dynein, thereby attenuating its ability to antagonize the contractility of the cortical actin.
Clearly, an important goal for the future will be to further test the various possibilities of exactly how kinesin-5 elicits such a profound role in restricting the growth of the axon. In addition to elucidating its mechanism of action, an even more important question may be to ascertain why it is that developing neurons express such a potent brake on the growth potential of their axons. Also, of course, the question remains as to how kinesin-5's functions are regulated within the axon. In particular, local regulation of kinesin-5 would enable the axon to increase or decrease the relevant forces to comply with the needs of the axon. For example, branch formation would benefit both from a local increase in short microtubule transport and from a diminished propensity of the branch to retract once long microtubules have been established in the branch. This kind of regulation could be accomplished by signaling pathways, given that the association of kinesin-5 with microtubules is regulated by phosphorylation (; ). Finally, it will be important to ascertain whether our findings on kinesin-5 apply to adult axons as well as developing axons because the ability to enhance axonal growth so dramatically by inhibiting kinesin-5 could be a powerful tool for augmenting axonal regeneration after injury ().
Cultures of dissociated neurons from rat superior cervical ganglia were prepared as previously described (). For experiments using siRNA, dissociated neurons were transfected with 10 μM (final concentration) of either control siRNA or kinesin-5 siRNA ( KIF11 [Eg5] siRNA; Ambion). Once transfected, the neurons were cultured in L-15 medium on 35-mm plastic dishes that had been coated with 0.1 mg/ml poly--lysine (PDL) for 3 h and repeatedly rinsed in double-distilled HO. The following morning 5 μg/ml laminin and 5 μM arabinose C were added to the culture medium. For transport studies, the neurons were replated () at a density of ∼7,500 cells/dish at 60 h after transfection onto grided glass coverslips that had been coated with 0.1 mg/ml PDL and 25 μg/ml laminin. Replating was performed 72 h after transfection. The EGFP-tubulin construct (CLONTECH Laboratories, Inc.) and the mCherry-tubulin construct (provided by R. Tsien, University of California, San Diego, La Jolla, CA) were transfected at 15 μg. Wild-type and mutant kinesin-5 constructs were provided by M. Kress (Institut André Lwoff, Centre National de la Recherche Scientifique, Villejuif, France) and A. Blangy (Centre de Recherches en Biochimie Macromoléculaire, Centre National de la Recherche Scientifique Unité Propre de Recherche, Montpellier, France), and then engineered in our laboratory as fusions with EGFP. The mutant construct, termed T112N, is described in detail by . For kinesin-5 overexpression studies, 12 μg of either GFP (control) or EGFP–kinesin-5 was cotransfected with mCherry-tubulin.
In one study, we wished to determine if expression of wild-type and/or the mutant kinesin-5 could “rescue” the phenotype obtained with siRNA-based depletion of kinesin-5. For these experiments, neurons were first transfected with siRNA. 48 h after transfection with siRNA, either the wild-type EGFP–kinesin-5 or the rigor mutant EGFP–kinesin-5 (T112N) was introduced using Lipofectamine 2000 (Invitrogen) as previously described (), except that Lipofectamine treatments were performed for 4.5 h duration and were followed by rinsing and medium replacement with L-15 plating medium. Neurons were then replated 24–30 h after Lipofectamine treatment and morphological analysis was performed as described below for siRNA experiments. Expression was observed in ∼15–20% of neurons, and data analysis used neurons with expression levels greater than the mean expression value (determined by arbitrary fluorescence units) for that experimental group.
See the online supplemental Materials and methods.
For all morphological investigations, neurons were cultured and replated after the 72-h replating regimen. For axonal length and branching experiments, neurons were identified at 1 h after replating (before they had established axons). The neurons were then imaged sequentially at 3, 5, and 7 h after replating. Axonal lengths were measured using the “measure/curve” application of Axiovision LE 4.5 (Carl Zeiss MicroImaging, Inc.), and mean values were quantified. Analysis of branching frequency was performed by counting the total number of branches per neuron and dividing these values by total axonal length for individual neurons. For both total branching and categorical branching analysis, we used normalization of branch number to total axonal length to ensure accurate evaluation of branching frequency.
60 h after transfection with control or kinesin-5 siRNA, neurons were replated onto gridded PDL (0.1 mg/ml) and laminin-treated (25 μg/ml) glass-bottom culture dishes. By 72 h the neurons had generated extensive axons. The nitric oxide donor, noc-7 (Calbiochem), was prepared and applied as previously described (), except that it was used at a working concentration of 0.3 mM. DIC images of axons were recorded before and 30 min after addition of noc-7. Axonal lengths (from growth cone to cell body or to the first bifurcation point) were measured using the “measure/curve” application in Axiovision LE 4.5. Raw data were processed and graphs were produced using Excel (Microsoft Corp.).
The microtubule transport assay was performed essentially as described previously (), except that for experiments using pharmacological inhibition of kinesin-5 with monastrol, the neurons were plated directly onto glass coverslips in drug-containing medium and were imaged 24–36 h later. For siRNA-based studies, neurons transfected with either control siRNA or kinesin-5 siRNA were replated at 60 h as described in the previous paragraph. A total of 211 time-lapse images were taken at 700–900-ms exposure using 3-s intervals for each axon. Transport analysis included all microtubules observed to move continuously through the photobleached region during the imaging period. Transport frequencies were calculated by dividing the total number of movements by the total imaging time for individual movies.
See the online supplemental Materials and methods.
Neurons transfection with control or kinesin-5 siRNA (see Cell culture and transfection) were replated in L-15 medium (supplemented with FBS) onto PDL- and laminin (25 μg/ml)-coated glass coverslips. The L-15 medium was coated with 2 ml of mineral oil (Sigma-Aldrich) as a means to prevent medium evaporation and to maintain nutrient concentrations over time. Phase-contrast time-lapse images were acquired using either a 40 or 20× objective lens with 700-ms exposure time and 1 × 1 binning. Image acquisition was performed at 3-min intervals for 5 h using a heated stage apparatus to maintain the temperature at 37°C. Stepwise axon and branch growth were analyzed using the “Apps/Track Points” application of Metamorph software (Molecular Devices) by establishing a point of origin and then using automated tracking of axonal outgrowth. Data were exported from Metamorph to Excel and were analyzed for the relative change in distance to origin. A camera (Orca ER; Hamamatsu) was used. For additional details on statistical analysis and image processing, see online supplemental materials.
Video 1 shows that kinesin-5 depletion results in fewer bouts of retraction of axons and axonal branches. Online supplemental material is available at . |
My older sister got me into science. She studied biology at university. When I was deciding what to study, I was torn between medicine and biology. My sister told me that biology was really cool and that I should try it out. So it's my sister's fault. [laughs]
Since then, it has always been biology. As soon as I took my first genetics class at university (University of Vienna, Austria), I got really interested. Genetics provides an ideal tool for dissecting what humans or animals are about. As soon as I had my first lecture in genetics, I knew I definitely wanted to continue on that route.
I went to the Institute of Molecular Pathology (IMP, Vienna) to do my Master's thesis. I joined Erwin Wagner's lab. He studies bone development in mice. He makes knockout and transgenic mice to study the function of genes in bone development and cancer. Erwin's lab was where I got exposed to real science for the first time.
I got interested in cloning during my undergraduate course, when I learned about John Gurdon's classic frog cloning experiments. It fascinated me. Then in '97 the paper by Ian Wilmut came out on the cloning of Dolly the sheep. Before that, nuclear reprogramming hadn't been shown in mammals. It was thought that mammalian cells might be refractory to cloning.
The reason I then was drawn into epigenetics and stem cell biology was a lecture at the IMP by Rudolf Jaenisch. I was really fascinated by the data he presented on the role of epigenetics in cloning and reprogramming.
Yes, I decided to come to Cambridge (Massachusetts) to visit my sister, and I stopped by MIT and talked to Rudolf. I started in his lab in March of 2000.
For my Ph.D. thesis, I worked on nuclear transfer. I asked whether a terminally differentiated cell is still amenable to reprogramming and able to give rise to a cloned animal. This question had not been resolved unequivocally by the cloning of Dolly or other mammals.
They used adult cells, but it was possible that the cells that gave rise to successfully cloned animals were derived from rare adult stem cells. This also might have explained why cloning is inefficient: only 1–3% of cloned embryos eventually develop into an adult clone.
I took advantage of lymphocytes. These cells carry specific genetic marks that indicate their maturity—the genetic rearrangements responsible for antibody production. I was successful in cloning mice from lymphocytes and could show that the genetic marks were present in all the cells of the cloned mouse.
It was a very risky project. Rudolf has since told me that he thought I wouldn't pull it off. He thought it might not work at all, that you might not be able to clone from fully differentiated cells.
I stayed on in Rudolf's lab for another three years to do a postdoc, because I knew I wanted to stay in the reprogramming field. And at that point, there wasn't any other lab where I could really learn more. Rudolf's lab is one of the few that has all of these technologies together. Anything you can think of in mouse genetics or mouse embryology has happened in his lab, whether it's making mice from embryonic stem cells, making transgenic mice, nuclear transfer, embryonic stem cell biology, studying cancer models, or embryonic development.
It was very rewarding. I would say most of what I know now I learned in his lab.
It's also a very critical lab. You learn very quickly in the lab meetings that you have to justify your interpretation of data. You learn how to think critically. I think that's a very important part of the training that everyone goes through. Rudolf pushes you hard, and that's good.
Probably, subconsciously. I think you do things in the way you've been trained by your mentor. I try to be as critical as Rudolf used to be with me, and I hope it pays off with my people.
It was stressful in the beginning in terms of hiring people, getting your experiments to work again, your cells to grow, your mice to breed, getting adjusted to a new environment, and to new colleagues. There are new responsibilities. All of a sudden you have to manage people. You have to make sure that their salaries are paid, that they're happy, and that they are making progress in their experiments.
But the more I'm experiencing it, the more I like it. It's really very rewarding to see people excited about the work they do.
It was a little difficult at the beginning to optimize it and to make all the special culture media for growing the cloned embryos. But the system is working now.
But actually I've recently become interested in an alternative way of reprogramming: using defined factors or genes. This is based on a landmark study by Shinya Yamanaka, published last year, that showed that just four defined factors, when introduced into skin cells, are sufficient to turn these cells back into embryonic stem cells. We've recently reproduced this data and extended the original findings.
People were very skeptical that this was true and wondered whether there might be an alternative interpretation to the results. But recently, three groups, Yamanaka again, Rudy Jaenisch, and our group, independently reproduced the data. I think the field will now believe it. I certainly do.
It allows us now to study reprogramming at a molecular level, which was impossible to do with nuclear transfer or cell fusion, because you had such limited cellular material. You can take large numbers of cells, expose them to these four factors, and ask what happens at the molecular level, what genes are turned on.
We know very little about the process, about what exactly goes on at the level of DNA and chromatin. What is downstream of those four factors? That's something I'm very interested in pursuing.
We also want to ask whether different cell types can be reprogrammed by the same four factors. If it works in humans, there are therapeutic implications. It may circumvent the ethical and logistical limitations associated with nuclear transfer.
This finding has revolutionized the entire stem cell and reprogramming field and has opened up many new avenues of research. I think we'll see a lot more exciting research in that area in the next few years, or even months. |
The role of chromatin in the responses to DNA damage is currently the focus of intense study. On the one hand, the local modification or remodeling of histones at sites of DNA double-strand breaks (DSBs), such as the phosphorylation of the histone H2A variant H2AX (), has led to the proposal of a DNA repair–specific “histone code” () that, through combinatorial histone modifications, might coordinate the signaling and repair of the lesions. On the other hand, and besides the local changes at the break site, work done two decades ago showed that the presence of DSBs triggers a global chromatin relaxation process (). The interest in this phenomenon has now revived because of recent data showing that this DSB-induced chromatin decondensation is actively regulated by the DNA damage response (DDR; ). Likewise, a global increase in chromatin accessibility has been reported in response to UV damage, which is mediated by p53 and Gadd45a proteins (; ). These observations led the authors to propose that the relaxation of chromatin might facilitate genomic surveillance by enabling faster access of DDR factors to the DSBs. However, the precise effect that the overall compaction status of the chromatin exerts on the access, signaling, and repair of DNA damage is not known and remains a central issue for our understanding of the DDR.
One of the main factors involved in high-order chromatin compaction is the linker histone H1. Through its binding to the internucleosomal linker DNA, in vitro data showed that H1 can help stabilize DNA at the nucleosomal linker DNA interface, thus favoring the refolding of arrays of nucleosome core particles into more compact structures (for review see ). The existing in vivo knowledge comes from the selective elimination of linker histones in several organisms. Attempts to deplete H1 protein in mice led to embryonic lethality when total H1 levels were reduced to ∼50% of the wild-type level by inactivating three of the six somatic H1 genes (H1c, H1d, and H1e; ). However, triple-knockout murine embryonic stem (ES) cell lines (H1) also containing half the normal amount of H1 could be obtained. Analysis of these lines showed that the reduction in histone H1 is indeed associated with a less compact chromatin (). Taking advantage of this genetic system, we investigated the competence of the mutant cells in establishing a DDR in the context of a more “open” chromatin configuration.
To get a general view of how H1-depleted ES cells respond to DSBs, we analyzed the behavior of mutant and control cells in colony-survival assays after a brief exposure to various genotoxic agents (). Regardless of the source, mutant cells were consistently found to be more resistant to DNA damage than their wild-type counterparts, this being more pronounced in the case of alkylating agents such as methyl-methane sulfonate (MMS) than in response to ionizing radiation (IR). Interestingly, H1 ES cells also exhibited resistance to hydroxyurea (HU), which activates the DDR in replicating cells, and this behavior could not be attributed to the difference in the rates of replication between both genotypes (). In all cases, partial reconstitution of H1-depleted cells with exogenous H1 (H1; Fig. S1, A and B, available at ) led to an intermediate phenotype in the survival rate. Thus, diminishing the levels of the linker histone H1 renders ES cells hyperresistant to DNA damage.
The better performance of the H1-depleted cells in colony-survival assays is reflective of an enhanced cellular response to DSBs, which in eukaryotes is coordinated by a phosphorylation-based transduction cascade known as the DDR. Beyond individual phosphorylation events, cell cycle checkpoints can be used to measure the output of the DDR. Whereas wild-type (H1) and H1 cells arrested similarly on exposure to high doses of radiation, the response to low doses differed substantially among the two genotypes (). A detailed analysis showed that mutant cells activated the G2/M checkpoint at doses in which the cell cycle progression of the wild-type cells remained unaltered (). The mutant cells also maintained a more robust response overall (). The performance of H1 and H1 cells in the intra–S phase checkpoint assay was analogous to that of the G2 arrest, showing an enhanced checkpoint response in the mutant cells (). Again, the performance of H1 ES cells was found to be intermediate in all cases. Thus, lowering the level of H1 results in an enhanced DDR as measured by the cell cycle arrest induced by DNA damage.
The presence of the “hypercheckpoint” phenotype led us to evaluate the signaling cascade involved in the activation of the arrest. To this extent, and because of its cell cycle–restricted role in the activation of intra–S and G2 checkpoint responses (), we analyzed the dynamics of chaperoning checkpoint kinase 1 (Chk1) phosphorylation induced by DNA damage. Consistent with the checkpoint data, the levels of Chk1 in response to IR were found to be increased in H1 cells at all doses examined, being detectable at doses in which the levels of Chk1 in the wild-type cells were not considerably higher than those of untreated cells (). Similar observations were made in cells treated with HU or MMS (). In contrast to the Chk1 phosphorylation performed by the ataxia telangiectasia mutated (ATM) and Rad3 related (ATR) even in response to IR (; ), the strength of ATM signaling was not affected by H1 levels. These findings are in agreement with the colony survival data, in which the resistance to genotoxicity was more profound in the case of drugs that preferentially activate ATR, such as HU or MMS, than in the case of IR. Nevertheless, an ATR/Chk1-dependent DDR also operates in IR-induced responses (; ), and impairment of ATR activity radiosensitizes human cells (; ). Moreover, both the radiosensitivity and G2/M checkpoint defect of ATM-deficient human cells can be rescued by ectopic expression of yeast Chk1 (), demonstrating that a hyperactive ATR pathway is sufficient to stimulate checkpoint signaling and radioresistance in an ATM-independent manner.
Interestingly, phosphorylated H2AX (γ-H2AX) was also found to be increased in the mutant cells in the same conditions, and both Chk1 and H2AX phosphorylation were dampened by H1 reconstitution (). The increased γ-H2AX formation could be the consequence of either increased damage being generated at the same doses or of more signaling being generated per individual DSBs. In any case, the resistance to DNA damage will be contradictory with higher amounts of DSBs being formed in the H1 cells. Consistently, neither neutral comet analyses (Fig. S1, C and D) nor the quantification of the number of 53 binding protein 1 (BP1) and γ-H2AX foci that were generated by IR (Fig. S1, E and F) showed substantial differences between the amounts of DSBs generated in H1 or H1 ES cells. Of note, the same number of DSBs led to higher endogenous levels of γ-H2AX in replicating cells, which even if not sufficient to have an impact on cell cycle progression might prime the DDR. To evaluate the activation of the DDR at each lesion, we analyzed the distribution of 53BP1 and γ-H2AX foci in response to IR (). Consistent with the biochemical data, high-throughput image analysis showed that the mean intensity of γ-H2AX per nucleus was markedly higher in H1-depleted cells (). Furthermore, the amount of γ-H2AX per individual focus was found to be markedly higher in H1 than in H1 cells (), and both phenotypes were partially suppressed in H1 cells. In summary, the hyperactive DDR found in the H1-depleted cells is caused by enhanced signaling that is generated at each individual DSB. In regards to the effect of chromatin at the DSBs, kinetic analyses of the recruitment of 53BP1 to IR-induced foci (Fig. S1 E) or to a laser-induced path of DSBs (Fig. S1, G–I) showed that this relocalization was slightly fastened in cells with reduced H1 levels. This suggests that the increased signaling per DSB observed in H1-depleted cells might be stimulated by the enhanced accessibility of DDR factors to the lesion site.
Besides the cells' enhanced signaling per lesion and hypercheckpoint phenotype, the reduction of H1 levels only led to a moderate impact on the overall resistance of the mutant cells to DNA damage. However, this experiment represents the outcome of one single exposure to exogenous damage and it is therefore possible that the constant presence of a hyperactive DDR would have more acute consequences on the genome of H1-depleted cells. One of the roles of DNA repair is handling the breaks generated during the replication process. These type of DSBs are repaired through homologous recombination (HR) using the sister chromatid as the preferred partner. Instead of looking at an exogenous HR system, we evaluated the endogenous levels of sister chromatid exchange at a locus conserved among all chromosomes, the telomere. The analysis of telomeric sister chromatid exchange (T-SCE) of the H1 and H1 lines showed that the frequency of this recombination event was almost four times higher in the mutant lines (). One of the pathways responsible for the control of telomere length is based on recombination between telomeric sequences. Consequently, mutations that increase telomeric recombination rates (such as murine ES cells deficient in DNA methyltransferases DNMT1 or DNMT3a/b) have been found to lead to elongated telomeres (). We therefore evaluated whether the persistent increase in endogenous T-SCE levels also correlated with an increase in the telomere length of H1 cells. As shown in , quantitative FISH (Q-FISH) analysis demonstrated a remarkable increase in the telomere length of H1-depleted cells. Whereas we cannot disprove that telomere length might be also influenced by other factors such as the access of telomerase to its target DNA or telomerase activity per se, our data suggest that the enhanced phosphoinositide 3-kinase–related protein kinase signaling observed at each DSB could be stimulating telomeric recombination events at endogenously occurring sites of telomere damage. In agreement with this concept, spontaneously occurring T-SCE events are substantially diminished by ATM depletion in human cells (Fig. S2, available at ). In summary, our data demonstrate that decreasing H1 levels lead to a general increase in telomere length, an as of yet unrecognized role for mammalian H1 in the regulation of chromosome structure.
With the exception of a very limited set of genes (), the gene expression profile of wild-type and H1 mutant cells is almost identical, suggesting that the observed differences in the signaling of DSBs are more likely related to the role of H1 in chromatin structure. A more extensive transcriptional study performed confirms the previous findings and, specifically, fails to detect any major transcriptional change in components of the DDR (Fig. S3, available at ). To further investigate whether a more decondensed chromatin, regardless of H1 levels, would cause similar dynamics at DSB sites, we evaluated the activation of the DDR in cells that had been previously treated with the histone deacetylase (HDAC) inhibitor trichostatin A (TSA). Among the different reagents that affect chromatin compaction, TSA was chosen because it is one of the few chromatin-modifying compounds in which the actual effect on global chromatin compaction has been documented in detail (). Importantly, even though the addition of TSA has been previously reported to activate ATM in the absence of damage (), the doses and exposure times used in this analysis did not lead to any detectable phosphorylation of ATM in nonirradiated cells. Pretreatment of human cells (MCF7) with TSA mimicked the observations previously made in H1-depleted ES cells (). In this manner, TSA pretreatment led to a more robust phosphorylation of Chk1 and H2AX in response to IR without substantially affecting ATM or structural maintenance of chromosomes 1 (SMC1) phosphorylation. Moreover, whereas the amount of 53BP1 foci being generated after IR was not affected by the drug, the signaling emanating from each foci/DSB was higher in TSA-treated cells than in control cells (unpublished data). Equivalent observations were made with other HDAC inhibitors such as sodium butyrate and in several human and murine cell lines (mouse ES, mouse embryonic fibroblast, U2OS, HCT, A549, IMR90, and NIH3T3; unpublished data). Furthermore, the amount of H2AX phosphorylation per individual focus occurring in an I-SceI–transduced MCF7 cell line harboring a unique nuclease recognition site () was also found to be enhanced by TSA, formally demonstrating that chromatin condensation limits the signaling strength of each DSB. In summary, both H1 depletion and TSA data support the concept that the strength of the DDR derived from each DSB is modulated by chromatin configuration.
In addition to the use of H1-depleted cells in this model system of open chromatin, our findings on the role of H1 might have functional implications. In fact, the stoichiometry of H1 in higher eukaryotes is nearly one per nucleosome core particle (), suggesting that the fine tuning of H1 nucleosome interactions might be an active way of controlling chromatin structure. Several lines of evidence support this notion. For instance, Cdk-mediated phosphorylation of H1 in the S/G2 phases has been shown to weaken the interaction of H1 with chromatin (; ). In addition, a fraction of the histone H1c variant was shown to leave the nucleus on exposure to DSBs, leading to cytochrome release and activation of apoptosis (). However, we can also detect H1 removal in mouse embryonic fibroblasts that have only been briefly exposed to the radiomimetic drug neocarzinostatin, supporting additional roles of H1 clearance in the response to DSBs beyond apoptosis (unpublished data). Furthermore, a transient transfection of H1-GFP in MCF7 cells shows that the strength of the DDR is dampened by the overexpression of H1 (Fig. S1 J). Finally, there is evidence that the role of linker histones in modulating the responses to DNA damage might have been established early in evolution. mutants of the linker histone were shown to have increased levels of HR, which was also particularly accentuated at the telomeres (). Because ATM and ATR phosphorylation events also directly stimulate the repair of DSBs (; ), it will be interesting to test whether the increased recombination and resistance to MMS found in the yeast strains is also associated with a hyperactive DDR. Beyond its global role in chromatin compaction, the finding of a specific in vivo role for H1 in mammalian cells has remained elusive. Our data on the responses to broken DNA and on the regulation of telomere length now support additional roles for the linker histone H1 in the control of genome architecture.
In this manuscript we have evaluated the precise effect that global chromatin compaction has on the different stages (access, signaling, and repair) of the DDR (). Our data suggests that whereas chromatin compaction has a moderate impact on the access and repair of DSBs, it strongly stimulates the signaling of each DNA break. In any case, the differences on repair could be substantial in explaining the damage resistance phenotype. The preferential impact on ATR/Chk1 rather than on ATM/Chk2 might be reflective of the different requirements for chromatin remodeling activities of each pathway. One possibility is that chromatin compaction will limit the resection of each DSB and thus mainly impinge on ATR signaling and HR. Interestingly, the deficiencies in Rad51 loading and HR found in transformation-domain associated protein mutant cells were shown to be reversed by pretreatment with HDAC inhibitors (; ).
The relationship between chromatin and the DDR presented in this study could help to understand the heterogeneity of cellular responses to DNA damage in vivo. For instance, a recent study has shown that CD133 glioma cancer stem cells show a more robust activation of the DDR that is associated with radioresistance of the tumor-forming population (). Because overall chromatin accessibility is an innate property of stem cells that is lost on differentiation (), studying whether the increased DDR of glioma stem cells is just indicative of their chromatin status and thus evaluating whether chromatin compaction might be therapeutically exploited to radiosensitize cancer stem cells should be an interesting avenue for future research.
Generation of murine ES cells deficient in histone H1c, H1d, and H1e (and quantification of total H1 levels) has been described previously (). Two independently derived pairs of ES cell lines (F18(H1) and F9(H1), and F4(H1) and F1(H1)) at passage numbers 10–12 were used for this study. Generation of H1 cells is described in Fig. S1.
For G2/M checkpoints, cells were mock treated or irradiated, incubated at 37°C for the indicated times, and stained with antibodies recognizing phosphorylated histone H3 (1:50 dilution; Upstate Biotechnology), and the percentage of mitotic cells was quantified by flow cytometry. Intra–S phase checkpoint analyses were performed as described previously (). In brief, the analysis represents the incorporation of [H]thymidine in a short pulse after exposure to the genotoxic agent (normalized to the levels of incorporation in the nondamaged cells). In the case of MMS, cells were left with the drug for 45 min before the pulse.
Primary antibodies used in this work were γ-H2AX and Chk2 (Upstate Biotechnology), 53BP1 and ATM (Novus Biologicals), Chk1 (Cell Signaling), Chk1 (Vision BioSystems), SMC1 (Abcam), ATM (Rockland Immunochemicals, Inc.), and ATM (provided by A. Nussenzweig, National Institutes of Health, Bethesda, MD 20892). For immunofluorescence, secondary antibodies conjugated with Alexa 568 or 488 (Invitrogen) were used at 1:250. For high-throughput microscopy studies, cells were grown on μCLEAR bottom 96-well dishes (Greiner Bio-One) and analyzed with a bioimager (BD Pathway 855; Beckton Dickinson). Image acquisition was performed at room temperature using oil as an immersion media and a camera (ORCA 1394; Hamamatsu Photonics) with a 40× objective (HCX PL APO, 0.75 NA). Image analysis was performed with imaging software (AttoVision; BD Biosciences). The quantification of H2AX intensity per focus was performed with software (Metamorph 7.0; Molecular Devices). All the images for quantitative analyses were acquired under nonsaturating exposure conditions. All Western analyses shown in this study were performed on the LI-COR platform (LI-COR Biosciences) that allows linearly quantitative Western blot with the use of Alexa 680– and 800–conjugated secondary antibodies (Invitrogen).
T-SCE events were measured with the use of chromosome orientation FISH (CO-FISH). In brief, confluent mouse ES cells were subcultured in the presence of BrdU (Sigma-Aldrich) at a final concentration of 10 M and allowed to replicate their DNA once at 37°C overnight. Colcemid (Sigma-Aldrich) was added at a concentration of 1 μg ml for the last hour of incubation. Cells were then recovered and metaphase spreads were prepared as described previously (). CO-FISH was performed as previously described (), first using a (TTAGGG) probe labeled with Cy3 and then a (CCCTAA) probe labeled with Rhodamine green (Applied Biosystems). Metaphase spreads were captured on a fluorescence microscope (Leitz DMRB; Leica).
Metaphase cells were prepared for Q-FISH and hybridized as previously described (). To correct for lamp intensity and alignment, images from fluorescent beads (Invitrogen) were analyzed using the TFL-Telo program (provided by P. Lansdorp, British Columbia Cancer Research Center, Vancouver V5Z 1L3, Canada; ). Images and telomere fluorescence values were obtained from at least 10 metaphases for each data point as previously described ().
Retroviruses producing I-SceI (pMXPIE-ISceI; provided by A. Nussenzweig) were produced in 293 T cells using standard procedures. Retroviral supernanants were then used to transduce a line of MCF7 cells harboring a unique I-SceI site (provided by K.K. Khanna; Queensland Institute of Medical Research, Queensland, Australia). The presence of a considerable population of cells harboring a single DSB was first observed 12 h after transduction and was maximum at day 2 (with ∼60% of the cells containing a single 53BP1 focus).
A plasmid (pEBB-H1dHA) that directs expression of H1d under control of the elongation factor 1 α promoter was constructed by inserting an 804-bp PCR DNA fragment including the H1d coding sequence with an HA tag sequence. 10 μg pEBB-H1dHA was cotransfected with 100 ng pcDNA 6/TR (Invitrogen) into 5 × 10 H1 ES cells that were prepared without feeder cells, using the mouse ES cell nucleofector kit (Amaxa) according to the manufacturer's instructions. 2 d after transfection, selection with 10 μg/ml blasticidin was initiated, and single colonies were isolated 11 d later. Clones were propagated in the presence of blasticidin and screened by immunoblotting with an anti-HA antibody (Santa Cruz Biotechnology, Inc.). Total histones were prepared from ∼10 cells of the highest expressing clones and analyzed by high-performance liquid chromatography as described previously () to quantify the level of the exogenous H1d protein. There was no detectable difference in elution times of the exogenous HA-tagged H1d and endogenous H1d.
The neutral comet assay was performed with the Comet assay kit (Trevigen) according to the manufacturer's instructions.
53BP1-GFP–transfected ES cells were plated onto a gelatinized black μCLEAR bottom 96-well plates (Greiner Bio-One). The DNA intercalating dye Hoechst 33258 was added at 10 μg/ml and incubated for 20 min at 37°C. The plate was mounted on a microscope stage of a confocal microscope (TCS-SP2; Leica), and cells were irradiated with a 351-nm laser along a user-defined path to generate localized DSBs.
Retroviral constructs (pRETRO.SUPER backbone) containing control or ATM short hairpin RNAs were provided by Y. Shiloh (Tel Aviv University, Tel Aviv, Israel). Retroviral transduction of HCT116 cells was performed by standard infection procedures. Cells were selected in 2 mg/ml of puromycin for 3 d 24 h after infection.
Total RNA was extracted from H1 and H1 ES cells using the RNeasy midi kit (QIAGEN). Quality of the RNA was determined with a bioanalyzer (2100; Agilent Technologies) and the amount of RNA was established using a spectrophotometer (NanoDrop). RNA amplification and labeling was performed by using the Low RNA input linear amplification kit PLUS (two-color; Agilent Technologies). We used 4 × 44K mouse 60-mer oligo microarray slides (Agilent Technologies) containing four arrays, each with more than 41,000 unique mouse genes and transcripts represented, all with public domain annotations. All procedures of hybridization and slide and image processing were performed according to the manufacturer's instructions (two-color microarray-based gene expression analysis protocol). To account for biases in dye incorporation, dye-swap experiments were performed independently for each comparison; 825 ng of contrasting complementary RNA samples were fragmented at 60°C for 30 min and hybridized at 65°C for 17 h. The slides were then sequentially washed, dried, and scanned at a resolution of 5 μm with a DNA microarray scanner (G2565BA; Agilent Technologies). Signal quantification and data normalization were performed with software (Feature Extraction 9.1; Agilent Technologies) using default analysis parameters for 4 × 44K gene expression arrays (feature extraction protocol GE2-v5_91_0806). The assay was performed in duplicates with dye-swapping to compensate for labeling biases. The entire dataset can be found in Table S1.
10 μg H1-GFP (provided by T. Misteli, National Institutes of Health) was transiently transfected in MCF7 cells using Lipofectamine 2000 (Invitrogen). After 24 h of transfection, cells were seeded onto black μCLEAR bottom 96-well plates for irradiation and high-throughput microscopy studies (see Western blot and immunofluorescence).
Fig. S1 contains additional datasets regarding the role of H1 on the DDR. Fig. S2 shows the role of ATM on the regulation of telomere recombination. Fig. S3 contains an overview of the transcriptome analyses of H1 and H1 cells. Table S1 contains the entire dataset of the microarray analyses shown in Fig. S3. Online supplemental material is available at . |
To explain how genomic instability arises in cancer cells, one model posits that the formation of cells with tetraploid genomes represents a transient initiating event (see model in ; ; ). Consistent with this model, it has recently been observed that tetraploid genomes promote an increase in numerical and structural chromosomal aberrations in p53-deficient cells (; ; ). Although the precise mechanisms that lead to the chromosomal defects in these cells are unclear, the stress of maintaining tetraploid genomes may select for viable resolutions that give rise to a variety of genomic configurations. Such an increase in genomic instability would allow nascent tumor cells to sample a wide genetic space, increasing their chance of finding a genomic configuration to overcome their normal growth constraints. If tetraploid genomes play an initiating role in genomic instability in cancer, it is predicted that cancer-promoting lesions will result in the early appearance of tetraploid cells in sufficient numbers to ensure the selection of genetic winners. In this study, we provide evidence that dominant mutations in the tumor suppressor gene (), which is frequently observed in human colorectal cancer, drive tetraploid formation by causing failures in cytokinesis before the earliest steps associated with colorectal cancer progression.
Previous work has implicated APC in the regulation of mitotic events, including formation of the mitotic spindle and proper functioning of the spindle checkpoint (; ; ; ). APC associates with the mitotic spindle through its interaction with microtubules and with the microtubule plus end–binding protein EB1 (; ; ; ; ). In previous work, we have shown that APC mutations similar to those found in human cancer patients act dominantly to prevent the interaction of APC with EB1 and, thus, inhibit microtubule dynamics and proper chromosome alignment (Green et al., 2005). Although microtubule dynamics are important for chromosome alignment during metaphase, mitotic spindle defects can also interfere with cytokinesis in tumor cells (; ; ; ). In part, this reflects the role of multiple sets of microtubules in regulating initiation and ingression of the cleavage furrow; peripheral overlapping microtubules at the equatorial cortex converge to initiate furrow formation, whereas the spindle midzone is critical for completing furrow ingression (; ; ; ; ). The ability of APC mutants to dominantly interfere with mitotic spindle assembly led us to ask whether a single APC allele might induce failures in cytokinesis that contribute to genomic instability and cancer progression.
To address the possibility that APC mutants compromise cytokinesis, we examined the long-term effect of expressing the dominant-negative APC (amino acids 1–1,450) mutant on chromosomal ploidy. This allele is representative of mutations observed in human colorectal cancer (; ; ). We created HEK-293 cell lines stably expressing APC under an ecdysone-inducible promoter and control cells lacking the hormone receptor required for expression (labeled as the control in figures; ). Immunoblotting for the tagged APC allele showed relatively constant expression similar to endogenous levels over the course of the experiment (; ; and unpublished data). In control cells, we observed a constant and very low incidence of binucleate cells (1–2%) and few multinucleated (more than two nuclei) cells over 10 d (, A and B, control). In contrast, cells expressing APC exhibited a steady increase in the numbers of both binucleate and multinucleate cells during the course of the experiment (, A [binucleate and multinucleate] and B), with each category reaching 10% of the total cell population. Multiple stable cell lines all exhibited the same trend: after >3 d of APC expression, we observed a mean of 10.26% binucleate (SD = 6.34%) and 8.34% multinucleate (SD = 5.27%) compared with control cells with 0.63% bincucleate (SD = 0.37%) and 0.33% multinucleate (SD = 0.1%). These results are consistent with previous findings that cells expressing similar APC mutants become polyploid over time (; ).
To examine the possibility that APC expression gives rise to polyploid cells as a result of failed cytokinesis, we monitored the behavior of chromosomes in time-lapse videos using an E-YFP-histone 2B (H2B) fusion. Cells with chromosomes aligned in metaphase were identified and filmed as they proceeded through anaphase ( and Video 1, available at ). After chromosome segregation in control cells, cytokinetic ingression was observed in brightfield images (, 18 min; arrow). Control cells initiated a furrow, completed cytokinesis, and returned to their interphase state ∼30 min after anaphase began. In many cells expressing APC, we observed no evidence of furrow initiation after anaphase (, C and D; and Video 2), and chromosomes became tightly juxtaposed 10 min after segregation, which is consistent with a collapse of the anaphase spindle (, 28 min). Chromosomes began to decondense, and nuclei formed close to one another as cells returned to their interphase state. We believe that this behavior gives rise to the binucleate (i.e., polyploid) cells we observed in . Consistent with cytokinetic failure, staining of cells with antibodies against γ-tubulin revealed binucleate and multinucleate interphase cells with two or more centrosomes (unpublished data). We conclude that the expression of APC results in binucleated cells as a result of failures to carry out cytokinesis before exiting mitosis.
We suspected that defects in spindle microtubules were responsible for the failed cytokinesis in cells expressing APC for at least two reasons. First, an allelic series of APC mutants with increasingly severe defects in mitotic spindle organization correlated with the number of bi- and multinucleated cells (Fig. S1 A, available at ; and not depicted; ). The appearance of bi- and multinucleated cells is also increased in colorectal tumor cell lines with truncating mutations in APC (Fig. S1 A, SW480). Second, we observed that compared with control cells, large regions of the cortex of cells expressing APC lack growing microtubule plus ends marked with EB1-GFP, and these spindles undergo increased rotations with respect to the spindle axis (; arrows; and compare Video 4 with Video 3).
As furrow initiation is specified by the mitotic spindle, we were particularly intrigued by the observation that microtubules make less contact with the cell cortex, and spindles appear to undergo considerable rotation when EB1-GFP was monitored in cells expressing APC (Video 4). We reasoned that the rotation of anaphase spindles may impact the ability of cells to initiate a cytokinetic furrow. To test this possibility, we compared the behavior of the mitotic apparatus with respect to the onset of anaphase and furrow initiation in control cells with cells expressing APC. Metaphase cells expressing GFP fusions to tubulin, EB1, or histone 2B were filmed for 1–2 h. In many cases, cells initiated anaphase and formed a cytokinetic furrow during the video. In control cells, we observed the consistent behavior of mitotic spindles documented in videos but more completely depicted in the spindle histories. Spindle histories were created from traces of the cell cortex (circle) and spindle axis (a line) that were derived from each time frame of a video. The tracings were overlaid to provide an entire spindle history (; right). For example, in control cells during prometaphase and metaphase, spindles moved along their axis but infrequently exhibited large rotational deviations ( and Video 5, available at ). These movements decreased as cells entered anaphase and were quickly followed by furrow ingression and cytokinesis. We classified this stable behavior of control spindles as anchored (i.e., <15° of movement between time frames).
Spindles in cells expressing APC exhibited two distinct types of behaviors. In >80% of cells filmed, we observed unstable spindle positioning, with spindles more frequently sliding and undergoing large rotations relative to the spindle axis (, APC; and Video 6, available at ). The large movements of the spindle were observed equally in metaphase and anaphase cells. In many cases, anaphase was defective with the failed segregation of chromosomes (Video 2) or the reduced elongation of spindles. In <20% of cells, spindles behaved similarly to controls, exhibiting stable spindle anchoring, furrow ingression, and cytokinesis (, APC; and Video 7). Of the cells that initiated a cytokinetic furrow, there were no consistent changes in the timing of furrow completion between control cells and cells expressing APC (Fig. S2 A), arguing that APC mutants do not affect the timing of anaphase. To further examine the relationship between spindle anchoring, anaphase onset, and the initiation of furrow ingression, we collected >100 videos of mitosis for each cell type. We characterized the 68 videos in which we could measure each of the aforementioned parameters (see Materials and methods section Image and statistical analysis). More than 60% of control cells anchored their spindle, entered anaphase, and initiated furrow ingression. In stark contrast, cells expressing APC generally failed to anchor their spindles, and, although spindle elongation indicated entry into anaphase, <15% of cells initiated a cytokinetic furrow (). A direct tabulation of the relationship between spindle anchoring and furrow initiation more precisely confirmed the trends in : regardless of the genotype, we observed a perfect correlation between cells with partially anchored or unanchored spindles and failures to initiate a cytokinetic furrow (Fig. S2 B).
Although we consider the failure to anchor the mitotic spindle the most likely explanation for the failure to induce the cytokinetic furrow, it remains possible that defects in chromosome alignment previously reported in cells expressing APC could contribute to inhibition of the cytokinesis (). Indeed, findings in HeLa cells indicate that there is a connection between the rate of lagging chromosomes in anaphase and cytokinetic failure (). To address whether lagging chromosomes or spindle-anchoring defects contribute to cytokinetic failures in APC mutants, we turned to the previously characterized APC allele. This allele prevents the oligomerization of mutant and wild-type APC proteins, and spindles exhibit EB1-postive microtubule plus ends near the cell cortex (, arrows in APC). Consistent with stable spindle anchoring, videos of cells expressing APC and GFP-tubulin exhibit relatively normal spindle anchoring (, spindle history), and >50% of cells have normal cytokinetic ingression (compared with <15% of cells expressing APC and >60% for control cells; ). Despite stable spindle anchoring, cells expressing APC have a high rate of misaligned chromosomes in metaphase (>70%) and anaphase cells (>30%), possibly as a result of a decrease in the pause frequency of growing microtubules (; and unpublished data). The presence of stably anchored spindles in cells with lagging anaphase chromosomes allowed us to assess whether misaligned chromosomes are sufficient to contribute to a defect in cytokinesis. Despite the obvious presence of chromosomes near the midzone, we did not observe a high rate of bi- or multinucleate cells (Fig. S1 A). Moreover, examples of individual cells expressing H2B-YFP show that chromosomes left behind in the midzone are constricted by the ingressing furrow and clearly do not impede furrow ingression (, arrow in the 15-min time point). Thus, APC can be considered a separation of function allele (i.e., normal spindle anchoring and cytokinesis but defective chromosome alignment and segregation), and we conclude from this analysis that unanchored spindles but not lagging anaphase chromosomes correlate with a failure to induce cytokinesis in these cells.
Although genomic instability has been previously reported in early adenomas in both mice and humans (), our in vitro data suggest that mitotic defects could occur in intestinal epithelium before the appearance of dysplasia or adenomas (i.e., in histologically normal tissue). Specifically, we predicted that changes in microtubule structure and dynamics could result in misoriented mitotic spindles within the gut that could contribute to cytokinetic failure. To test these predictions, we performed a series of analyses of mitoses in the intestinal epithelium of APC mice (hereafter referred to as mice) and wild-type littermates at early time points after birth (7–13 wk before the appearance of polyps; ). Importantly, the allele (APC) gives rise to similar mitotic defects as observed for cells expressing APC. Specifically, we observed that cells expressing APC have an increased incidence of multinucleated cells (fivefold increase), unanchored mitotic spindles (<25% anchored compared with >80% in controls), and a high incidence of anaphase cells that fail to induce a cytokinetic furrow (<20% anchored compared with >60% in controls; Figs. S1 A and S3). We first characterized small intestines of wild-type mice using antibodies against the apical marker ZO-1, the basolateral marker β-catenin to identify cell boundaries, and DAPI to visualize chromosomes. We observed mitotic cells positioned in crypts as expected, and we observed a shift of mitotic chromosomes away from the basement membrane (Fig. S4, A–C; [inset]; available at ), a position that is consistent with the previous characterization of mitosis in MDCK cell monolayers ().
We next determined the orientation of mitotic spindles in wild-type mice. We used β-catenin staining to visualize the plasma membrane of the cells and a combination of DAPI and Numa (to visualize spindle poles in mitotic cells) staining to orient the spindle with respect to other cells in the crypt and to the basement membrane (). We measured the degree of spindle rotation with respect to the basement membrane (). In wild-type cells, 95% of metaphase and 85% of anaphase spindles were oriented at an angle of <30°. Using this value as a cut-off for proper spindle orientation, we observed a dramatic reduction in properly oriented mitotic spindles in mice (only 67% of metaphase cells and 53% of anaphase cells had a parallel orientation; P > 0.001; see multiple examples in ). This result is consistent with our in vitro observations that spindles in APC mutant cells fail to properly anchor at the cell cortex ().
Although we cannot clearly view microtubules in these preparations, we observed phenotypes consistent with microtubule defects. We observed that the ratio of the width to length of aligned metaphase chromosomes was increased in compared with wild-type mice (Fig. S4, D and E) and that chromosomes were often misaligned and failed to reach the metaphase plate (Fig. S4 F). These results are consistent with previous observations made in vitro and attributed to defective spindle microtubules (). Together, defects in chromosome congression, chromosome alignment, and spindle orientation in mice strongly argue that a truncating mutation in APC induces mitotic defects in vivo by a similar inhibition of microtubule dynamics as demonstrated in vitro ().
The mitotic defects we observed in mice take place in otherwise normal cells. To more directly address the state of these cells with respect to tumor progression, we determined the fate of the full-length (i.e., wild type) APC protein and the levels of β-catenin in mice intestines. We used a carboxy terminal–specific APC antibody that only recognizes the full-length wild-type APC protein (Fig. S5, A and B; available at ). In crypts from mice that exhibit mitotic abnormalities, we observed a characteristic staining of APC at the apical membrane and a basolateral enrichment of β-catenin, patterns that are indistinguishable from crypt cells in wild-type mice (). By way of comparison, we observed frequent dysplasias in 7-wk-old mice that are easily distinguishable from the normal organization of the crypts (, low magnification image). In contrast to normal crypts, we observed the loss of full-length APC and an increase in the cytoplasmic and nuclear β-catenin signal in every dysplasia analyzed (). This analysis shows that cells with a wild-type allele of APC and normal levels of β-catenin exhibit mitotic defects and, therefore, argues that APC mutants compromise mitosis independent of β-catenin deregulation. We further confirmed these conclusions by showing that a single mutant allele of APC does not alter β-catenin levels in vitro (Fig. S5 C) and that expression of stabilized β-catenin in normal 293 cells does not result in cytokinetic failures (Fig. S5, D–F). As expected, we also observed higher levels of multinucleate human colon tumor cells with APC mutations (SW480) but not in colon tumor cell lines with a stabilizing β-catenin mutation (HCT116; Fig. S1 A). We conclude that loss of wild-type APC protein is associated with the earliest stages of dysplasia in mice and, importantly, that mitotic defects precede the loss of the second allele of APC, β-catenin stabilization, and dysplastic growth.
Our in vitro data establish a link between defects in spindle orientation and failure to induce a cytokinetic furrow. Intriguingly, our observations in mice suggest that dividing crypt cells have similar spindle orientation defects, raising the possibility that histologically normal crypt cells in mice are prone to cytokinetic failures. We reasoned that these cytokinetic failures would give rise to tetraploid cells but that these events would be rare. To identify such rare events, we used FISH to determine the copy number of three chromosomes using probes to specific chromosomal regions (APC probe chromosome 8, p53 probe chromosome 11, and tex11 probe X chromosome; see Materials and methods FISH section). We used these probes individually and in pair-wise combinations on intestinal tissues from wild-type and mice to identify cells with four FISH signals/nucleus.
To ensure accuracy in identifying tetraploid cells, we used several methods. To distinguish between neighboring nuclei, we counterstained samples with DAPI and the membrane dye FM4-64. A representative example of a cross section through a Min mouse intestinal crypt is shown in . The FM4-64 dye was used to make the cell boundaries more apparent in the presentation of optical projections. To unambiguously assign FISH signals to individual nuclei, we analyzed stacks of z-sections that included both FISH signal and the DAPI counterstain. Using these approaches, we could easily identify diploid cells and G2 diploid cells by the close proximity of sister chromatid signals (<1 μm; ). Cells with four well-separated FISH signals were interpreted as being tetraploid, although we cannot rule out a near tetraploid genotype (). We applied stringent criteria for assigning a tetraploid designation, and, in some cases, we ignored potential tetraploid cells if it was not possible to assign their position in a single nucleus; this may have the effect of underestimating the total number of tetraploid cells. Despite this conservative approach, when a large number of crypts was analyzed, we observed a striking number of cells with four distinct FISH probe signals in mice (31 cells or 3.1%; ; see Discussion). In contrast, there only was one case observed in wild-type mice (0.1%; ). Similar trends were observed with all of the FISH probes regardless of whether they were used individually or in combination, arguing that these cells minimally contain four copies of chromosomes 8, 11, and X (, , and not depicted). These results are consistent with the single mutant allele of APC in mice promoting the formation of tetraploid cells.
Because the tetraploid cells we observe are isolated events and are predicted to occur only in the epithelial cells of the small intestine, it is not possible to assess their ploidy using more traditional flow cytometry techniques. Therefore, to confirm a tetraploid genotype, we looked for additional characteristics indicative of tetraploid cell formation. Cells harboring a 4N content of DNA are predicted to have larger nuclei than 2N cells. Analysis of DAPI-stained nuclei showed a consistent trend whereby cells with four FISH signals had nearly two times the cross-sectional area as cells believed to be 2N based on their FISH signals (). Second, failures in cytokinesis leave the resulting cell with two centrosomes; subsequent entry into the next cell cycle induces the duplication of centrosomes and the formation of tetrapolar spindles in the next mitosis. We used Numa staining of centrosomes to identify mitotic cells with more than two poles (). Importantly, we observed a large number of tetrapolar spindles only in mice (8/110 in or 0/105 in wild-type mitotic cells), which is consistent with our in vitro observations of multipolar spindles in cells expressing APC (unpublished data). Together, the FISH and cytological features we observed lead us to conclude that tetraploids form in the intestines of Min mice as a result of cytokinetic failures.
Our work shows that clinically relevant mutations in APC act dominantly to inhibit anaphase spindle anchoring, thus resulting in cytokinetic failures in dividing intestinal cells in mice. This finding implicates a known tumor-initiating mutation in the generation of tetraploid genomes (, red text), a genetic state linked to cancer progression and genomic instability (). The early appearance of tetraploid cells in a mouse model for colorectal cancer argues that these dramatic changes are important for cancer progression.
A major question that remains unanswered pertains to the fate of intestinal cells with tetraploid genomes. Although we expect that tetraploid cells are inherently unstable (minimally because of tetrapolar spindles), the tetraploid model for genomic instability predicts that these cells will give rise to aneuploid cells at the next stage of tumor progression. Unfortunately, it is not trivial to track the fate of a single tetraploid cell in the gut. However, one testable prediction is that dysplasias (Fig. S3) in Min mice will exhibit changes in chromosome ploidy. We find that in contrast to previously reported ploidies of Min tumors (), dysplasias in Min mice have a notable rate of chromosome gains (4.0% of cells; and see Discussion). This number is likely an underestimate of aneuploidy, as we were not confident assigning FISH signals for chromosome loss events in these regions. This finding supports a model () whereby APC mutants dominantly induce tetraploid cells in the crypt that resolve to populate dysplastic regions, thus contributing to genomic instability.
The mitotic defects we observe in mice are remarkably similar to those observed in human 293 cells conditionally expressing APC. Notably, the frequent occurrence of misoriented spindles in mice is consistent with an in vitro spindle-anchoring defect, and the appearance of tetraploid cells is similar to the formation of binucleate and multinucleate cells in vitro. Together, the collection of parallel phenotypes makes a compelling case that the mutation compromises microtubule dynamics in vivo, leading to spindle-anchoring defects and the formation of cells with tetraploid genotypes. Therefore, it is important to consider how changes in microtubule behavior that alter cell polarity and chromosome ploidy affect the earliest events in cancer progression.
Do the tetraploid cells we observe in mice contribute to tumor progression? We observe a remarkably large number of tetraploid cells in crypts: projections based on the number of crypts in the small intestine, the half-life of cells, and their division rate () suggest that there are nearly 10 tetraploid cells in the small intestine, resulting from one failure in 10 divisions. Notably, this failure rate is similar to previously published in vitro rates (). Obviously, the low number of tumors in mice (mean = 29 tumors/ mouse) means that not all tetraploid cells give rise to cancer cells. The ability of tetraploid cells to successfully complete another round of mitosis is dependent on their ability to resolve the abnormal configuration of centrosomes and chromosomes. The large number of mitotic cells with tetrapolar spindles in mice suggests that another round of division is attempted (). Such a resolution may give rise to the near diploid or hypotriploid cells observed in intestinal tumors in humans and mice ().
Although these results may appear to conflict with previous reports of stable genomes in Min tumors, there are at least two explanations for the differences. The first is that we analyzed tissue earlier in cancer progression than the adenomas reported by . It is possible that the stabilization of near diploid karyotypes occurs in late stages of tumor formation, when high chromosome number may become a liability to the cancer cell. A second potential factor relates to how the analysis was reported: report the total number of FISH signals divided by the total number of cells counted. Although reasonable for gross changes in tumor ploidy, such an approach will fail to identify the low frequency of aneuploid events that we observe in crypt cells or dysplasias. As proof of this point, a similar report of our data gives exactly the same numbers reported by ; unpublished data). The implication is that a constant low frequency of changes in ploidy is ongoing in dysplastic and tumor tissues. Although this rate of ploidy change may be specific for Min mice, its impact on tumor progression remains to be determined. It is interesting to note that Min tumors do not advance beyond the adenoma stage, suggesting a potential connection between modifiers of ploidy and disease progression.
Our observations are consistent with the models that posit the evolution of tumor cells () (; ). In this scenario, tetraploid cells in the small intestine form a pool of cells from which aneuploid and/or neoplastic cells sample a variety of genotypes, most of which will fail to give rise to a viable cell and, therefore, avoid the cancer phenotype. These observations argue for a direct path from mitotic defects to cancer and raise important questions as to the molecular and cellular state of these cells at this early stage. It is abundantly clear from our analyses that aneuploidy is a constant and early feature of intestinal epithelium expressing clinically relevant mutations in . This is the first evidence that a mutation that predisposes patients to colorectal cancer induces chromosomal aneuploidy before any other changes associated with cancer, therefore placing genomic instability in a position to play a causative role in tumor progression.
Regions encoding the indicated fragments of human APC and a 13-myc epitope tag were subcloned for expression, and stable human 293 cell lines were generated as previously described (). Cells were transfected with EB1-GFP (pIK131; provided by J. Tirnauer, University of Connecticut Health Center, Farmington, CT), enhanced GFP–tubulin (pFM258; provided by F. McNally, University of California, Davis, Davis, CA), or enhanced YFP (EYFP)–histone 2B (provided by G. Kops, University Medical Center, Utrecht, Netherlands). Antibodies against c-myc (9E10; Santa Cruz Biotechnology, Inc.), tubulin (Tub2.1 or FITC-conjugated Tub2.1; Sigma-Aldrich), human centromere antigens (anticentromere antibody; Antibodies, Inc.), γ-tubulin (GTU-88; Abcam), or β-catenin (BD Biosciences) were used according to the manufacturers' recommendations. Antibodies against Numa were used at a 1:200 dilution for immunohistochemistry. The c-APC antibody was made by injecting rabbits with a purified recombinant fragment of APC encoding amino acids 2,038–2,843 of the human protein. Specificity of the serum was confirmed by immunoblotting and on APC mutant human tumor cells.
Cells were cultured in DME high glucose medium (Cellgro) supplemented with 10% FCS, 2 mg/ml -glutamine, 1 mM sodium pyruvate, 200 μg/ml penicillin, and 50 U/ml streptomycin. Cells were maintained at 37°C and 5% CO. Plasmid expression vectors for the indicated fusion proteins were transfected into the indicated cell lines using the FuGene 6 reagent (Roche). Cells were treated with ponasterone (Invitrogen) to induce expression of the APC allele 24 h after transfection and were maintained in ponasterone-containing media for 24–48 h before analysis. In the long-term expression experiments, the media was changed every 3–4 d, and fresh ponasterone was added to the media.
Cells were grown on coverslips and processed for immunofluorescence microscopy as previously described (). Fluorescence was collected using an inverted microscope (IX70; Olympus) fitted to an imaging system (DeltaVision RT; Applied Precision). Images were recorded using a 60× NA 1.4 oil immersion lens (for fixed images, live cell imaging, and intestinal sections; Olympus) or a 20× NA 0.75 oil immersion lens (for intestinal sections; Olympus) and a camera (CoolSnap HQ; Roper Scientific). For live cell imaging, cells were grown on 50-mm round coverslips before being transiently transfected with the indicated expression constructs, were allowed to recover for 24 h, and then were induced with ponasterone (as indicated) for 24 h. The coverslip was fitted to a live cell imaging chamber (FCS2; Bioptechs) and maintained at 37°C on the microscope stage. The pH of the media during filming was buffered by the addition of 10 mM Hepes, pH 7.4. An open shutter system was used to rapidly collect 2–4 μm of data at each time point of the video. Deconvolution and image analysis were performed using tools provided in the SoftWorx v 3.5.1 software package (Applied Precision, Inc.).
and wild-type littermate controls were obtained from Jackson ImmunoResearch Laboratories. Mice were housed and maintained according to approved University of California Davis animal protocols. After killing, intestines were removed, fixed as Swiss rolls in 4% formalin overnight, and embedded in paraffin. Paraffin-embedded tissue were cleared and rehydrated according to standard protocols. Tissues were placed in CitraPlus antigen retrieval (Biogenix), heated in a microwave for 4 min, allowed to cool for 1 min, heated for 4 min, cooled for 1 min, and heated for 4 min. Room-temperature slides were rinsed twice in TBS-Tween. Tissues were blocked for 1 h in 50 mM TBS, pH 7.5, 0.1% Tween, 1% BSA, and 2.5% horse and goat serum (Vector Laboratories). Primary antibody was diluted in blocking buffer and incubated for 1 h in a humid chamber and room temperature. Tissue was rinsed twice in blocking buffer. Secondary antibodies were diluted 1:100 in blocking buffer and incubated for 1 h. Tissues were rinsed twice in TBST and once in PBS. Tissues were stained with DAPI for 10 min, rinsed in PBS, and mounted.
Bacterial artificial chromosomes containing defined portions of the mouse genome were obtained from the BACPAC Resources Center (; APC, RP24-73J15; Trp53, RP23-150N14; Tex11, RP24-140K17). FISH was preformed as described previously (). The probe mixture consisted of nick-translated labeled probe (digoxigenin-labeled dUTP; Roche) and biotin-labled dUTP (Roche) in LSI/WCP hybridization buffer (Vysis, Inc.) and was denatured on the tissue for 2.5 min at 85°C. When a single biotin-labeled probe was used, the membrane dye FM4-64 (Invitrogen) was incubated at 1:1,000 for 10 min just before adding the DAPI stain and coverslipping the slides.
Analysis of spindle anchoring, anaphase, and furrow initiation parameters were performed on >100 videos for each cell line presented (i.e., control and APC). Videos in which one parameter could not be determined were discarded (32 videos discarded). Linear regression was used to determine the statistical difference between wild-type and spindle orientation. A test was used for all other analysis.
Fig. S1 is an analysis of binucleate and multinucleate cells in a series of APC mutants and in several colon cancer cell lines. Fig. S2 is an analysis of anaphase timing in wild-type and APC mutant cells, and Fig. S3 presents an expanded analysis of anaphase and spindle histories of cells expressing APC. Fig. S4 provides more detailed views of mitotic cells in intestinal crypts and of the associated mitotic defects found in mice. In Fig. S5, to validate the carboxy-terminal APC antibody, we provide immunoblot analysis. We also provide evidence that the expression of stabilized β-catenin does not contribute to the accumulation of binucleate and multinucleate cells. Videos 1 and 2 show chromosome behavior in a mitotic control cell (Video 1) and in a mitotic APC cell (Video 2). Videos 3 and 4 show EB1-GFP dynamics in control mitotic cells (Video 3) and in APC mitotic cells (Video 4). Videos 5 and 6 show mitotic spindle behavior in control cells (Video 5) and in APC cells (failed cytokinesis). Video 7 shows mitotic spindle behavior in APC cells (successful cytokinesis). Online supplemental material is available at . |
The nuclear envelope (NE) separates the contents of the nucleus and cytoplasm and is a physical barrier for the exchange of macromolecules. The only known mechanism for nuclear import and export is via nuclear pore complexes (NPCs; ; ). Thus, the NPC is a central player in controlling gene expression and regulating nucleocytoplasmic signaling. Specifically, the NPC precludes molecules larger than ∼30–40 kD from freely diffusing through its central aqueous channel. Larger macromolecules use transport receptors to pass through the NPC in a signal-dependent process (). The karyopherin (Kap) β proteins (also termed importins, exportins, and/or transportins) are a major family of transport receptors. There are 14 Kapβs in budding yeast and >20 identified in higher eukaryotes (; ). Each Kap binds a specific nuclear localization signal (NLS) or nuclear export sequence (NES) on a cargo, with Kap cargo release and transport directionality triggered by the small GTPase Ran (; ). There are non-Kapβ transport receptors for RanGDP import (Ntf2; ; ) and for mRNA export (the heterodimer Mex67-Mtr2 [TAP/NXF1-p15/NXT1 in vertebrates]; ; ; ; ). With the potential for at least 16 different receptors transporting thousands of distinct cargoes, the NPC is a complex machine. Indeed, it is not fully understood how such a myriad of distinct transport receptors use the NPC structure for presumably simultaneous translocation.
The ∼40–60-MD NPCs are formed by the assembly of multiple copies of ∼30 individual proteins called nucleoporins (Nups; ; ). Nups associate in discrete subcomplexes and localize in specific substructures of the NPC, including the cytoplasmic filaments, the central core structure in the pore, and a nuclear basket structure (; ; ; ). Movement of cargo-bound Kapβs, Ntf2, or Mex67-Mtr2 through the NPC requires interactions between the given transport receptor and a specialized subset of NPC proteins termed the FG (phenylalanine-glycine) Nups (). The FG Nups are defined by domains with numerous, clustered repeats of the core dipeptide FG flanked by characteristic spacer sequences (). Nearly half of the Nups contain these FG domains, each with predominant FG subtypes (FG, FXFG [phenylalanine–any residue–phenylalanine-glycine], or GLFG [glycine-leucine-phenylalanine-glycine]), defined NPC substructural locations, and corresponding orthologues across species (; ; ). Some FG Nups are exclusively on the cytoplasmic (C) NPC fibrils (in Nup159 and Nup42), and some are exclusively on the nuclear (N) NPC basket (in Nup1, Nup2, and Nup60); together, these are collectively defined as the asymmetric FG Nups (). The remaining FG Nups are distributed on both sides and through the central NPC channel and are termed the symmetric Nups (in Nup49, Nup57, Nsp1, Nup100, Nup116, and Nup145; ; ).
The physical interactions between transport receptors and FG peptides have been structurally analyzed for Kapβ1, Ntf2, and Nxt1. In these receptors, the Phe of an FG repeat is found in hydrophobic pockets on the protein surface (,, ,; ). Indeed, transport receptor mutants with impaired FG binding are defective for NPC translocation (). Thus, each transport receptor serves as a molecular bridge between FG Nups and a signal-containing cargo. With multiple FG repeats per FG domain and multiple FG Nups in each NPC, the pore displays thousands of individual FG repeats, each of which is a potential binding site for a transport receptor. The abundance of FG repeats and sequence redundancies between FG Nups have made understanding the sequence of molecular interactions between the NPC and transport receptors a formidable task.
Given their critical role in the translocation mechanism, the FG Nups have been the focus of intense study. Models for the mechanism of NPC translocation have as their tenets the unfolded nature of the FG domains, the huge number of FG repeats per NPC, and the intrinsic binding affinities of transport receptors for FG domains. Localization of FG domains in the NPC and the physiological constraints of NPC translocation rates are also key considerations. Two of the fundamental models proposed contrast the FG domains as forming either a primarily physical or energetic barrier for selective translocation. As a physical barrier, weak interactions between FG domains are proposed to form a hydrophobic gel into which transport receptors selectively partition as a result of their FG interaction capacity (; ). The hydrophobic gel would form a selective phase and exclude macromolecules larger than the physical barrier generated by the FG interaction meshwork. As an energetic barrier, the interaction of a transport receptor with an FG Nup would allow the transport receptor to overcome an entropic threshold for diffusion through the NPC central channel (). The FG domains would also function as repulsive bristles to entropically exclude nontransport receptor molecules (). As such, the NPC would be governed by a virtual gate. From the analysis of individual FG domains in vitro, there is independent data to support both the selective phase and virtual gate models.
To analyze the requirements for FG domains in the context of the intact NPC, we have used a large-scale genetic strategy in (). By combinatorial in-frame deletions in genes encoding the FG Nups, we showed that the asymmetric FG domains are dispensable for facilitated transport, whereas the symmetric FG domains are sufficient. Interestingly, although the selective-phase model predicts that the abundance or mass of FG repeats is critical to transport function (; , ; ), we found that the number or mass of FG repeats does not correlate with in vivo transport capacity. We also found that for a given FG deletion (designated FGΔ) mutant, only a subset of the Kapβ transport receptors were perturbed. This suggests that different transport receptors require distinct combinations of FG domains for function (). In support of this, biochemical studies have demonstrated that different Kaps have different relative in vitro binding levels for the same FG Nup (; ). There is also evidence that Kap95 might use different FG-binding sites than those used by Mex67 (; ; ). Collectively, these studies suggest that the NPC may harbor multiple translocation pathways for different transport receptors.
To further investigate the FG-dependent transport pathways through the NPC, we generated a new collection of FG domain deletion mutants. We specifically compared Kapβ versus non-Kapβ translocation pathways by dissecting the requirements for Mex67-Mtr2–dependent mRNA export. Multiple laboratories have identified -null or temperature-sensitive alleles that cause mRNA export defects, and overproduction of the Nup116 GLFG domain inhibits mRNA export (; ; ). However, our new mutants have allowed the first global analysis of specific FG domain requirements in mRNA export. We have found striking differences in the requirements for Mex67-mediated mRNA export versus Kapβ-mediated transport. These results impact models for the in vivo NPC translocation mechanisms and support our hypothesis that multiple FG pathways exist for receptor-mediated translocation across the NPC.
In our previous study, we generated an mutant that lacked all of the asymmetric FG domains on the N and C faces of the NPC, which is designated the mutant (). The mutant has a slight rate delay in import via Kap95 and Kap104; however, it has no marked steady-state defect for any transport receptor assayed. Thus, the asymmetric FG domains do not serve essential functions. However, we speculated that the asymmetric FG domains could be key to maximal transport efficiency. In addition, because the FG domains can presumably occupy multiple topological positions in the NPC (; ; ), it is possible that the asymmetric FG domains functionally compensate when individual symmetric FG domains are deleted. Therefore, we selected the mutant as a foundation for studying the transport roles of individual symmetric FG domains. In frame, internal chromosomal deletions of the sequence encoding individual symmetric FG domains were constructed in the background. If lethality was observed when a symmetric FG domain was removed in the background, control complementation experiments were conducted with plasmids expressing the full-length or FGΔ mutant versions (see Plasmids and yeast strains section in Materials and methods). This generated a series of () FGΔ mutant strains. Specifically, the mutant was combined with individual deletions of the GLFG regions in Nup49, Nup57, Nup145, Nup100, Nup116, or the FG and FXFG regions in Nsp1. We found that all of the FGΔ mutant strains with only one symmetric FG domain removed were viable (; ). Additionally, the mutant was viable despite having only four FG Nups intact (Nsp1, Nup49, Nup57, and Nup116).
The strains in this new FGΔ mutant collection were characterized for growth properties at a range of temperatures. As shown in , the mutant showed robust growth at all temperatures tested. In comparison, the mutant had inhibited growth at 37°C, whereas the mutant was cold sensitive at 16°C. The mutant showed both temperature sensitivity at 37°C and cold sensitivity at 16°C. Overall, the mutant and the mutant strains had the most severe growth phenotypes with both temperature sensitivity at 34°C and cold sensitivity (). The mutant generated in our previous study is cold sensitive at 23°C and also inhibited at 37°C ().
We speculated that the temperature-dependent growth defects were linked to perturbations of an essential transport receptor. To test for defects in transport, the FGΔ mutants were transformed with a panel of GFP-based reporters for different Kapβ transport receptors. Each transport reporter was based on a Kapβ- or Kapα-specific NLS fused to GFP or a tandem NLS-NES fused to GFP. In wild-type cells, all of the NLS-GFP reporters are predominantly nuclear, whereas NLS-NES-GFP is mostly cytoplasmic. The basic classic NLS (cNLS) of SV40 large T antigen is imported by the Kap95–Kap60 heterodimer (; ), and Nab2 and the Nab2-NLS-GFP reporter are imported by Kap104 (; ). Spo12-NLS is recognized primarily by Kap121/Pse1 (). The NLS-NES-GFP reporter includes a cNLS for Kap95–Kap60 import and a leucine-rich NES for Xpo1/Crm1 export (). Steady-state transport assays in the wild-type and FGΔ mutants were conducted at both the permissive temperature and after shifting to growth at 37°C for 1 h. The results are summarized in . For all of the mutants, no defects at steady state were detected with either the cNLS (Kap95–Kap60) or NLS-NES-GFP (Crm1/Xpo1) reporters ( and not depicted). However, several of the mutants showed altered Spo12-NLS-GFP (Kap121) import. This included the mutant combined with either the , , , or alleles ( and ; ). At 37°C, the Spo12-NLS-GFP reporter showed a coincident increased cytoplasmic signal and decreased nuclear intensity in the mutant and mutant cells (). This indicated that these strains had defects in Kap121 transport.
Interestingly, only one of the FGΔ mutant strains, , showed a strong perturbation in steady-state Nab2 import by Kap104, with diminished nuclear localization and increased cytoplasmic signal at all growth temperatures. The defect was apparent using either the Nab2-NLS-GFP reporter (not depicted) or via indirect immunofluorescence for Nab2 localization (). Steady-state transport defects for Kap104 or Kap121 were not observed in the mutant, the mutant, or the mutant strains ( and ). When comparing the Kap104 and Kap121 transport defects, it was especially striking that the mutant showed differential perturbations. The Kap104 cargo Nab2 was efficiently imported (, far right), whereas the Kap121 reporter accumulated in the cytoplasm at 23 and 37°C (, far right). This is the first reported in vivo separation of FG-domain requirements for Kap104 and Kap121 NPC translocation. Overall, the FGΔ mutant strains showed distinct defects for transport by specific Kaps.
To understand the contributions of FG domains to mRNA export, we screened a subset of our existing FGΔ mutant strains and our new FGΔ mutant strains for mRNA export defects. This was evaluated using in situ hybridization with an oligo d(T) probe, which detects poly(A) RNA. All of the viable FGΔ mutant strains with three symmetric FG domains deleted showed the nuclear accumulation of poly(A) RNA after a 1-h shift to 37°C (, , and not depicted). However, the mutant cells did not show the nuclear accumulation of poly(A) RNA. We also did not observe mRNA export defects in the mutant, the mutant, the mutant, or the mutant cells. For mutants that showed no nuclear poly(A) RNA accumulation, we also used an independent assay for mRNA export capacity and analyzed the effect on heat shock protein production. After heat shock in wild-type cells, elevated levels of Hsp104, Hsp82, Ssa4, and Ssa1 are a direct reflection of proper export and translation for the respective heat shock–induced mRNAs (; ). The mutant and the mutant were competent for heat shock protein production (unpublished data). We concluded that FG domains of the asymmetric FG Nups (Nup159, Nup42, Nup1, Nup2, and Nup60) and three specific symmetric FG Nups (Nup100, Nup116, and Nsp1) were not individually essential for mRNA export. In contrast, the and the mutant strains showed strong perturbations in mRNA export with marked nuclear accumulation of poly(A) RNA ( and ). This indicated that Nup57 and/or Nup49 were preferentially required for mRNA export.
To further probe the requirements for the GLFG domains of Nup57 or Nup49, we examined a double mutant strain. The mutant was assayed for mRNA export defects. Nuclear poly(A) RNA accumulation was observed in 9.9 ± 0.9% of the cells. Although this defect is significantly different from the level observed in wild-type cells (P = 0.0031), it is not as penetrant as the defect in either the mutant or mutant cells (30.3 ± 2.5% and 26.7 ± 6.1%, respectively). Thus, the GLFG domains of Nup57 and Nup49 are not essential for mRNA export, either individually or in combination. This suggested that other symmetric FG domains (Nup116, Nup100, Nup145, and Nsp1) functionally compensate in the absence of the Nup57 and Nup49 GLFG domains. However, when the asymmetric FG domains were removed (), the GLFG domain of Nup57 or Nup49 was specifically required, and the FG domains from Nup116, Nup100, Nup145, and Nsp1 were not sufficient. Collectively, these results revealed a combinatorial requirement in mRNA export for specific GLFG domains with the asymmetric FG domains. Moreover, such differential requirements for FG domains in mRNA export were unanticipated. Previous studies have reported that Mex67 interacts in vitro with several of the asymmetric FG domains (Nup159, Nup42, Nup1, and Nup60) and with three symmetric FG domains (Nup100, Nup116, and Nsp1; ; ; ; ). Although the GLFG domains of Nup57 and Nup49 have not previously been reported to bind Mex67, these results suggested that the FG domains of Nup57 and Nup49 are key sites in vivo for mRNA export.
Nup57 and Nup49 are both GLFG Nups that assemble in a heterotrimeric complex with Nsp1 (; ; ). Given this shared NPC localization, the common FG types (GLFG), and the growth and transport phenotypes in the FGΔ analysis, we concluded that the mutant and mutant strains were functionally comparable. We selected the mutant for further analysis, as it was genotypically less complex (see Plasmids and yeast strains section in Materials and methods). To pinpoint which of the FG domains in the mutant were most critical for mRNA export, we systematically generated strains with fewer FGΔ combinations. Each mutant strain was assayed for poly(A) RNA localization by in situ hybridization with the oligo d(T) probe, and the percentage of cells in the population showing nuclear accumulation of poly(A) RNA was scored (). The single mutant and the mutant did not have defects, as the percentage of cells showing nuclear poly(A) RNA accumulation was not significantly different from wild type (P > 0.0602). The mutant strain also did not have a poly(A) RNA export defect. In contrast, mutant cells had a strong export defect after shifting to growth at 37°C for 1 h, with nearly 80% of the cells showing the nuclear accumulation of poly(A) RNA. It was striking that the defect in the mutant (in 79.9 ± 9.4% of the cells at the assay time point) was more severe than that in the mutant (in 26.7 ± 10.6% of the cells; see Discussion).
To further dissect the mutant phenotype, we assayed mutants with all possible FGΔ combinations of nuclear face FG domains (Nup1, Nup2, and Nup60) with the allele. The triple mutant had a poly(A) RNA export defect with penetrance similar to the mutant (). This indicated that the allele did not contribute considerably to the mutant phenotype. In fact, addition of the mutant allele to any single or double Δ mutant did not result in a statistically significant difference in the level of nuclear poly(A) RNA accumulation (P > 0.07 for all comparisons). The double mutant and the double mutant strains also had defects; however, the percentage of cells with nuclear poly(A) RNA accumulation was significantly less in the double mutant and double mutant strains than in the combined triple mutant (P = 0.0018 and P = 0.0011, respectively). Overall, these results suggested that the export of mRNA requires both a symmetric GLFG domain (Nup57 and Nup49) and the FXFG domains on the nuclear face (Nup1 and Nup2). This is the first evidence for an in vivo role for the specifically asymmetric FG domains in active NPC translocation.
We speculated that the deletion of FG domains critical for Mex67 docking at the NPC was the mechanistic basis for the mRNA export defects in the respective FGΔ mutants. Specifically, the in vivo results suggested that Mex67 required binding sites in the FG domains of Nup57 or Nup49 and Nup1 or Nup2. Previous studies have documented that Mex67-Mtr2 can bind representative FG, FXFG, and GLFG domains (; ; ). The FXFG domain of Nup1 has been directly analyzed (); however, tests of the Nup57 GLFG region have not been reported. We conducted studies to verify this interaction biochemically with recombinant proteins and a soluble binding assay. Clarified bacterial lysates from cells expressing GST alone or GST fused with the GLFG regions of Nup57 or Nup116 (GST-GLFG-Nup57 or GST-GLFG-Nup116) were incubated with glutathione- Sepharose. Purified maltose-binding protein (MBP)–Mex67 was then applied to the resin with the respective immobilized GST fusion proteins. As shown in , GST-GLFG-Nup57 bound MBP-Mex67, whereas GST alone did not bind MBP-Mex67. Binding was also detected between MBP-Mex67 and GST-GLFG-Nup116, as has previously been shown (Strawn et al., 2001). Thus, the GLFG domain of Nup57 directly binds Mex67 in vitro.
An mRNA export defect in an FGΔ mutant could result from either a direct effect on Mex67–NPC interactions or an indirect perturbation on Kap-mediated import of an essential mRNA export factor. We speculated that FGΔ mutants with primary defects in Mex67-mediated mRNA export would have decreased rates of Mex67-GFP recruitment to the NE/NPC as a result of the lack of critical FG-binding sites. To directly examine the dynamic properties of Mex67-GFP, we developed a live cell assay (). This strategy was based on the well-established assay for monitoring NLS-GFP import in live yeast cells (). Wild-type parental or FGΔ mutant cells expressing chromosomally tagged Mex67-GFP were incubated in glucose-free media in the presence of 10 mM 2-deoxy--glucose and 10 mM sodium azide for 45 min. This treatment results in cellular energy depletion and inhibits active nuclear transport (). The process of mRNA export is energy dependent (), at a minimum requiring the ATPase Dbp5 (; ). As shown in , before energy depletion, all strains showed a strong Mex67-GFP signal at the nuclear rim. After energy depletion in all of the strains, Mex67-GFP was no longer concentrated at the NE/NPC, and the cytoplasmic and nuclear signals increased. Coexpression of a dsRed-HDEL (histidine-aspartate-glutamate-leucine; fusion protein with amino acid signal sequence for the ER retention) was used to facilitate visualization of the NE/ER. Localization of the dsRed-HDEL protein was not altered by energy depletion. As a control, we monitored the localization of two structural non-FG Nups, GFP-Nic96 and Nup170-GFP (), and found that a strong punctate NE/NPC signal was present both before and after energy depletion. Nuclear rim localization of Nup49-GFP was also not altered by energy depletion in wild-type cells or in mutant cells ( and not depicted, respectively). This indicated that energy depletion results in the mislocalization of Mex67-GFP without a general perturbation of NE/NPC structure.
Using this assay, NE/NPC reassociation kinetics was determined by fluorescence microscopic monitoring of Mex67-GFP localization. At the start of the assay, the energy-depleted cells were washed and resuspended in 23°C glucose-containing media. The cells were then incubated until the NE/NPC signal recovered to pretreatment levels. Individual cells ( > 150) in a population were scored for normal continuous NE/NPC signal and relative levels of nucleoplasmic and cytoplasmic staining (). By plotting the percentage of cells with normal continuous NE/NPC signal as a function of time, relative association rates were determined. We then compared the association kinetics wherein a single variable was changed (e.g., the FGΔ mutant background).
After restoring energy to the system, Mex67-GFP in the wild-type cells returned to the pretreatment phenotype with Mex67-GFP predominantly at NE/NPCs (). The mutant cells recovered more slowly than wild-type cells, and, at intermediate time points, an increased frequency of cells had elevated intranuclear signal relative to cytoplasmic. The recovery process in the mutant was substantially more delayed. After 15 min, the cells showed only a minimal recovery of Mex67-GFP localization to the NE/NPC. Moreover, at the intermediate time points, Mex67-GFP localization in the cells was mostly intranuclear with no distinct NE/NPC staining (). This phenotype was also observed in the mutant, in which >50% of the cells accumulated Mex67-GFP in the nucleus and concentrated nuclear rim localization was not achieved over the time course of the assay ().
Again, as in the assays of poly(A) RNA accumulation, the rate of Mex67-GFP localization to the NE/NPC was clearly more inhibited in the mutant than in the mutant (see Discussion). Overall, we concluded that Mex67-GFP recruitment to the NPC in the mutant and mutant was impaired. The intranuclear localization before distinct NE/NPC staining might reflect the efficient import of Mex67-GFP with specific mRNA export inhibition. These results correlate with our assays for poly(A) RNA export and suggest that the mutant and mutant are blocked for poly(A) RNA export as a result of altered Mex67 recruitment to and/or translocation through the NPC.
Many approaches have been used to study the mechanism by which transport receptors cross the NPC and the requirements for transport receptor interactions with FG Nups. We have used a genetic strategy in to generate extensive collections of mutants with specific combinations of FG domains removed and have conducted direct tests of the in vivo roles of putative FG-binding sites for transport receptors in the intact NPC (). In the present study, we report the analysis of new FGΔ mutants wherein the symmetric FG domains were removed in the absence of all asymmetric FG domains (). In some cases, the FGΔ phenotypes correlate directly with reported in vitro binding results. For example, previous studies have shown in vitro binding of Kap104 to the Nup116 GLFG region (; ), and, indeed, the mutant has defects in Kap104-mediated transport, whereas the mutant does not. This confirms that the Nup116 GLFG domain is a critical Kap104-binding site. On the other hand, we found that not all in vitro binding events are essential in vivo. Although Mex67 interacts with the GLFG region of Nup116 in vitro (; ), the mutant has no mRNA export defect. As a result, we conclude that in vitro binding between a transport receptor and an FG domain does not necessarily correlate with a requirement for that FG domain in vivo. Rather, the substructural location and physiological context of each FG domain is likely a key determinant in the organization of transport pathways through the NPC.
We have also identified binding events that were not previously recognized as important. We found that distinct combinations of both symmetric and asymmetric FG domains are needed for efficient nuclear export of poly(A) RNA and recruitment of Mex67-GFP to the NE/NPC. This includes a GLFG domain from the symmetric Nup57 or Nup49 plus the asymmetric FXFG domains of Nup1 and Nup2 on the nuclear NPC face. Surprisingly, import by Kaps does not require these same FG domains. These results support a model wherein different transport receptors use distinct FG domains, allowing for multiple, preferred, and independent transport pathways through the NPC.
Plasmids and yeast strains used in this study are listed in Tables S1 and S2 (available at ). Plasmid cloning was performed according to standard molecular biology strategies. Yeast strains were grown in YPD (1% yeast extract, 2% peptone, and 2% glucose) or in synthetic complete (SC) media with 2% glucose and lacking appropriate amino acids. New yeast FGΔ mutants were generated using a Cre-Lox system as previously described (; ), with the exception of the strain. Using the Cre-LoxP system, deletion of the sequence encoding amino acids 2–236 from was coincident with insertion of the sequence for a T7 epitope tag and a LoxP site fused in frame with the sequence encoding the C-terminal region of Nup49. The lethality of this strain was rescued by transformation with a plasmid (pSW3261). All assays were conducted with the strain.
Yeast strains carrying pGAD-GFP (cNLS-GFP), pNS167 (Nab2NLS-GFP), pKW430 (NLS-NES-GFP), or pSpo12 -GFP (Spo12NLS-GFP) were grown to early log phase in SC media lacking the appropriate amino acid and supplemented with 2% glucose. Cells were examined from culture at 23°C or after 1-h shift to 37°C. All images were acquired using a microscope (BX50; Olympus) with a UPlanF1 100× NA 1.30 oil immersion objective (Olympus) and a camera (CoolSNAP HQ; Photometrics). Within each experiment, all images were collected and scaled identically. Images were collected using MetaVue version 4.6 (Molecular Devices) and processed with Photoshop 9.0 software (Adobe).
Yeast cells were grown in YPD to early log phase at 23°C, and aliquots were shifted to 37°C for 1 or 3 h. Cells were fixed for 10 min and processed as previously described (; ). For indirect immunofluorescence, cells were incubated overnight with affinity- purified rabbit anti-Nab2 antibodies (1:4,000) and were detected with fluorescein-conjugated donkey anti–rabbit IgG (1:200; Jackson ImmunoResearch Laboratories). For in situ hybridization, cells were incubated overnight with a digoxigenin-dUTP–labeled oligo d(T) probe and were detected with fluorescein-labeled antidigoxigenin Fabs (1:25; Boehringer). DNA was stained with 0.1 μg/ml DAPI, and samples were mounted for imaging in 90% glycerol and 1 mg/ml -phenylenediamine (Sigma-Aldrich), pH 8.0. Images were acquired and processed as described in the previous section.
GST, GST-GLFG-Nup57, and GST-GLFG-Nup116 were expressed in Rosetta (DE3) cells (EMD Biosciences). Clarified lysates of GST fusion proteins were prepared in 20 mM Hepes, pH 7.5, 150 mM NaCl, and 20% wt/vol glycerol. MBP-Mex67 was expressed in Rosetta cells, affinity purified over amylose resin according to the manufacturer's protocol (New England Biolabs, Inc.), and dialyzed into binding buffer of 20 mM Hepes, pH 7.5, 150 mM NaCl, and 20% wt/vol glycerol. Clarified GST fusion protein lysates were bound to glutathione-Sepharose (GE Healthcare) and washed in binding buffer. MBP-Mex67 was applied to beads and incubated at 4°C for 30 min. Samples were washed twice in binding buffer and eluted on ice for 20 min in binding buffer, pH 7.5, with 20 mM glutathione. Equal fractions of bound protein were analyzed by SDS-PAGE and Coomassie blue staining.
was chromosomally tagged with the sequence encoding GFP in haploid wild-type and FGΔ yeast by amplification of the region from the yeast GFP collection strain YPL169C (Invitrogen). Integrants were selected on SC-histidine and verified by PCR and immunoblotting with rabbit anti-GFP (1:1,000). To allow integration of the gene for expression of dsRED-HDEL, YIplac204/TKC-DsRed-HDEL () was linearized with EcoRV and transformed into yeast cells. Cells were selected on SC-tryptophan, and integrants were verified by live cell microscopy. For energy depletion assays, cells were grown to early log phase in YPD at 23°C. A culture aliquot of 2.5 A U was used, and the cells were pelleted, washed, and resuspended in 1 mL YP (without glucose) with 10 mM NaN and 10 mM 2-deoxy--glucose. Cells were treated for 45 min at 23°C and were pelleted, washed, and placed on ice before microscopy. At time = 0, cells were resuspended in 23°C YPD, mounted on a glass slide, and visualized as described in Microscopy and analysis of live cell GFP reporters. Images of the GFP and dsRED signals were acquired every 30 s for 15 min. Cells were scored for the recovery of Mex67-GFP to the nuclear rim and the relative nuclear to cytoplasmic GFP signal. Control strains SWY734 and SWY3302 were examined before and immediately after energy depletion.
Table S1 lists the strains with genotypes and sources that are used in this study. Table S2 lists the plasmids used in this study and designates plasmid backbone and source. Online supplemental material is available at . |
A balance between synthesis and decay determines the steady-state level of mRNA. For most mRNAs, synthesis and decay take place in different cellular compartments, and little is known about cross talk between these two processes.
The yeast RNA polymerase II (Pol II) is composed of 12 subunits termed Rpb1p-Rpb12p (). The crystal structure of yeast Pol II reveals that the enzyme comprises two distinctive parts (; ); a ten-subunit core carrying the catalytic active site and a two-subunit heterodimer composed of Rpb4p and Rpb7p (for review see ; ). Rpb7p is an essential Pol II subunit that is conserved from archea to humans (; ; ). It carries two highly conserved RNA binding domains (; ; ) that likely bind RNA cooperatively (). In vitro studies indicate that Rpb7p can bind RNA in a sequence- independent fashion (; , ). Importantly, Rpb7p interacts with a transcript during in vitro transcription as soon as the latter emerges from the Pol II core (). Therefore, Rpb7p is most likely among the first proteins to interact with nascent Pol II transcripts.
Unlike Rpb7p, Rpb4p is a nonessential protein (; ; ). Notably, Rpb4p is involved in mRNA export from the nucleus to the cytoplasm only during stress () and plays a direct role in cytoplasmic degradation of a specific class of mRNAs encoding protein biosynthetic factors (PBFs) (). PBFs include ribosomal proteins, translation initiation factors, aminoacyl tRNA synthetases, and ribosomal biosynthetic proteins. Because cells can proliferate in the absence of , it is clear that the indispensable Rpb7p can function independently of Rpb4p and can interact with Pol II in the absence of Rpb4p (). However, it is not known whether Rpb7p plays a role outside Pol II context.
In yeast, major pathways of mRNA degradation (for review see ; ; ) initiate with shortening of the mRNA poly(A) tail. When the length of the poly(A) tail reaches 10–12 bases or less, one of two alternative pathways is initiated (or both pathways are initiated simultaneously). One pathway involves removal of the mRNA 5′ cap [mGpppN] by the Dcp1p/Dcp2p heterodimer (; ; ). Several proteins regulate this decapping process, including Pat1/Lsm1-7 (; ; ; ). Pat1p is recruited to mRNA while it is still associated with translation factors (). Subsequently, Pat1p recruits the hepta-heterodimer Lsm1-7 complex and this event is associated with the transition from translation to decay (; ). In vitro, Pat1-Lsm1-7 complex binds at or near the 3′ end of the mRNA () and is cable of protecting the 3′ ends of mRNAs in vivo from trimming (). The deadenylated mRNA can be degraded by alternative pathway that degrades the mRNA from its 3′ terminus. A well-studied 3′ to 5′ exonuclease is a large complex known as exosome (for review see ). The relative contribution of each mechanism remains a subject of debate. In , knocking out components of either the 3′ to 5′ () or the 5′ to 3′ pathway () had minimal effects on the transcriptome, which implies tight regulation of mRNA level involving cross talk between the two decay mechanisms. Nonetheless, very little is known about any possible dialogue between the two decay mechanisms.
Recently, we proposed that Rpb4p is the first mRNA decay factor to be recruited to mRNA in the nucleus. After transport of the mRNP to the cytoplasm, Rpb4p is involved in recruiting the Pat1/Lsm1-7 complex, or stimulating Pat1/Lsm1-7 function, and stimulates mRNA decay (). However, because Rpb4p does not have an RNA-binding domain (; ; ), it was not clear how this protein is recruited to the mRNP.
Here, we report novel functions for Rpb7p in the two cytoplasmic mRNA decay pathways. Mutation analyses reveal the existence of temperature-sensitive (ts) Rpb7ps that are transcriptionally functional but fail to stimulate mRNA decay. We have identified mutant forms of Rpb7p that are specifically defective in only one of their three functions, raising the possibility that the three functions are distinct. Further genetic analyses suggest that Rpb7p functions in the context of Rpb4/7. Nevertheless, whereas Rpb4p plays a role in the decay of specific mRNAs, Rpb7p has a more general role in the mRNA decay pathway. Collectively, our observations suggest that the Rpb4/7 is involved in a cross talk between the two cytoplasmic mRNA decay mechanisms and between the nuclear and cytoplasmic stages of gene expression.
Previously we have shown that Rpb7p shuttles between the nucleus and the cytoplasm by a transcription-dependent mechanism. Moreover, Rpb7p shuttles as a heterodimer together with Rpb4p (), which plays a role in the cytoplasmic mRNA decay pathway (see Introduction). In addition, we observed that Rpb7p-GFP localizes in discrete foci that resemble processing bodies (PBs) (), the site where decapping and degradation of mRNAs occurs (; ; ; ; ). These observations prompted us to investigate whether Rpb7p is involved in mRNA decay. To determine the nature of Rpb7p cytoplasmic foci, we examined whether known PB proteins colocalize in these areas. Indeed, two PB markers (; ), RFP-tagged Dcp2p () and Lsm1-RFP (unpublished data), colocalize with Rpb7p-GFP. Colocalization was observed in response to starvation and ranges between 10 and 60% of the cases, depending on the strain background and on the environmental conditions. Rpb7p-GFP foci were also colocalized with Rpb4-RFP foci that were recently shown to represent PBs (). Moreover, like genuine PBs, Rpb7p-GFP foci disappear in response to pretreatment with the translation inhibitor cyclohexamide (). These results confirm our suspicion that Rpb7p is localized in PBs and raise the possibility that Rpb7p is involved in the deadenylation-dependent mRNA decay that takes place there.
To determine more directly whether Rpb7p is involved in mRNA decay, we mutagenized randomly the essential gene and selected ts alleles (see Materials and methods). Consistent with previous results (), we could not identify ts alleles that carry only one mutation. We selected 13 ts mutants that carry 2–6 point mutations (see Fig. S1, available at ). The mutations are scattered throughout the proteins, suggesting that the overall conformation of this small protein is important for its function. First, we determined the transcriptional capacity of the ts mutants at their nonpermissive temperature by monitoring de novo synthesis of various heat-shock (HS) mRNAs in response to abrupt HS. As expected for an essential pol II subunit, some of the 13 mutants exhibit transcriptional defects at nonpermissive temperatures (Fig. S2; ). Notably, however, some of the mutants do not exhibit diminished transcription capacity (Fig. S2; and ). Consistently, levels of these mutant forms of Rpb7p, as detected by Western analysis, are comparable to that of the wild-type (WT) Rpb7p (unpublished data). Yet, these mutants are ts. These results suggest that Rpb7p has an essential posttranscriptional role(s) at 37°C. Moreover, because the levels of HS mRNAs in the transcriptionally functional ts mutants are higher than in WT, these data support the premise that the essential posttranscriptional role of Rpb7p is in mRNA decay. It is worth noting that although most mRNA decay factors are nonessential at 30°C, some of them become essential at high temperatures (; ).
Therefore, we determined whether any of the ts mutants exhibit defects in mRNA degradation. To monitor mRNA decay, transcription was blocked either synthetically by adding 1, 10, phenanthroline () or naturally by shifting cells from 30°C (our standard permissive temperature) to 42°C (). The latter method relies on physiological responses to HS, which comprises aborting transcription of non-HS genes and enhancing decay of non-HS mRNAs (; see also ). Using this HS assay, which simultaneously inactivates the ts Rpb7p and blocks transcription, we found that >50% of the ts mutants exhibit detectable defects in the decay of and mRNAs (unpublished data). We analyzed in more detail yRL26, yRL28, yRL29, and yRL34 strains (carrying , and ts alleles, respectively), which show the most severe defects in mRNA decay.
Next, we examined whether the mutant strains are defective in mRNA decay at 37°C, instead of 42°C. Because at 37°C transcription is not blocked, we monitored decay of Tet-Off- transcript after blocking transcription by doxycycline. As shown in Fig. S3 (available at ), the initial decay of mRNA was similar in both strains. However, after 25 min post-temperature shift-up, mRNA degradation was slower in the mutant as compared to its degradation in the WT. This behavior shows that, already at 37°C, mRNA decay is defective in the mutant cells, but raises the possibility that inactivation of Rpb7p ts at 37°C was delayed and requires ∼25 min. In our subsequent analyses we monitored mRNA stability at 42°C, as discussed below.
To systematically examine mRNA decay in our mutant strains, we first examined mRNA turnover at the permissive temperature. At 30°C, decay kinetics in the mutants are comparable to those in WT (; unpublished data). In agreement with this, at the permissive temperature the steady-state levels of various mRNAs in all our mutant strains are comparable to those in WT. In contrast, at 42°C, the mutants exhibit defective decay (). This defect is observed independently of the method that was used to block transcription (compare with ; and unpublished data). Importantly, at 42°C, the mutants exhibited defective decay as soon as cells were shifted to the high temperature (). A good example for rapid inactivation of Rpb7 ts is the decay of mRNA, which becomes very unstable at 42°C (). Already 5 min after the temperature shift, more mRNA is observed in the mutant than in the WT (, ). A quantitative example is the decay of ().
Rpb4p is specifically involved in the decay of PBF mRNAs (see Introduction). Unlike Rpb4p, mutations in Rpb7p affect the decay of both PBF and non-PBF mRNAs. Among the non-PBF mRNAs are short-lived mRNAs, such as and (see ) and mRNAs, as well as relatively stable mRNAs, such as mRNA (). Thus, it appears that the role of Rpb7p in mRNA decay is more general than that of Rpb4p (see Discussion).
To gain insight into the mechanism by which Rpb7p functions in the major mRNA decay, we examined what step(s) in mRNAs turnover is/are stimulated by Rpb7p. Cells were shifted to 42°C to inactivate both transcription and the function of Rpb7p ts form, and then the levels and poly(A) tail lengths of and mRNAs were monitored using the polyacrylamide gel electrophoresis Northern (PAGEN) technique (), as shown in . These mRNAs were chosen because their short length enables detection of small differences in the size of their poly(A) tail. Three conclusions can be made concerning the results shown in . First, the decay kinetics of these mRNAs is fast in WT cells and slower in mutant cells, consistent with our results obtained by standard Northern analysis shown in . Second, deadenylation is abnormally slow in the mutant cells. For example, whereas complete deadenylation of mRNA is observed in WT after 45 min (, lane 5), it takes 70 min for this to occur in (, lane 13). We conclude that Rpb7-28p is defective in stimulating rapid deadenylation. As shown in , after its complete deadenylation 25 min post-transcription arrest, it is only a further 20 min before mRNA is almost completely degraded in WT (between 25 to 45 min post-transcription arrest; see lanes 4 and 5). In contrast, to degrade comparably the deadenylated mRNA in takes 75 min (lanes 12–14). A third conclusion, then, is that Rpb7p stimulates not only deadenylation, but also post-deadenylation step.
The major post-deadenylation step that is defective in cells might be either decapping or 5′ to 3′ degradation or 3′ to 5′ degradation of the decaying mRNA (; ; ). To distinguish between these processes we took advantage of a construct designed to identify degradation intermediates. Specifically, a previous study has revealed that a (G)18 tract, placed in the 3′ UTR of mRNA, serves as a barrier for exonuclease activity. Consequently, 5′ to 3′ degradation of pG mRNA by Xrn1p produces a ∼190- base degradation intermediate fragment (, “Frag.”) that stretches from the poly(G) tract to the 3′ end (; ). This fragment can be degraded by the 3′ to 5′ pathway (). Because Xrn1p degrades only uncapped RNA, accumulation of this intermediate is usually indicative of efficient decapping (; ; , ). We introduced Tet-Off- in WT and cells and used this assay to determine which post-deadenylation step is defective in the mutant cells. As shown in , fragment level relative to the full-length mRNA is threefold higher in cells than it is in wild-type cells. Because the fragment is degraded by 3′ to 5′ exonuclease (), high fragment level is often indicative of efficient decapping and poor 3′ to 5′ exonuclease activity.
To examine if 3′ to 5′ decay is defective in mutants, we first took advantage of a genetic assay based on the observation that strains lacking both the major deadenylation and the alternative 3′ to 5′ are inviable (; ). Thus, any double-mutant that contains a block to the 5′ decay pathway (Δ) with a block to the 3′ decay pathway (e.g., Δ, Δ) is synthetically lethal (). The synthetic relationship between and ts mutants is shown in . can replace in Δ cells. This result indicates that cells lacking and carrying as the only source of Rpb7p can support 3′ to 5′ decay. In contrast, cells and cells cannot survive in the absence of cells proliferate very slowly without (), consistent with the abnormal accumulation of the fragment (). These results suggest that, indeed, Rpb7p is required for execution of the 3′ to 5′ decay pathway, and that some of our mutants are defective in this function. Interestingly, the synthetic lethality was observed at 30°C, suggesting that the mutants are defective in 3′ to 5′ decay at the permissive temperature.
To corroborate the synthetic lethality results biochemically, we introduced pTet-Off- in cells, which showed the most severe synthetic lethality, and determined the fragment stability. To this end, we blocked transcription of Tet-Off- by doxycycline and simultaneously blocked decapping (and hence Xrn1p capacity to act) by cyclohexamide () while shifting cells to the nonpermissive temperature. These two inhibitors blocked supply of new fragments, permitting us to follow 3′ to 5′ decay of the fragment that had preexisted in the cytoplasm (because it is a product of Xrn1p that acts in the cytoplasm) prior drugs addition (; ; ). Whereas in WT cells the fragment is rapidly degraded (
= 30 min), no degradation can be detected throughout the experiment duration ( and D). Thus, the 3′ to 5′ decay mechanism is defective in the mutant cells. In cells, which do not show synthetic lethality with (), the fragment half-life was 30 min (unpublished data). Note that in cells, the steady-state level before the temperature shift (time 0) is not higher than it is in WT cells (unlike the case of cells). It is possible that Rpb7-29 is also defective in decapping. Collectively, the results shown in indicate that Rpb7p is required for efficient 3′ to 5′ degradation of mRNAs.
Previously it was shown that mutations in some mRNA decay factors impact the number of PBs (; ; ; ; ). The novel roles of Rpb7p in mRNA decay and its localization to PBs () prompted us to examine whether mutations in similarly affect PB number, using Dhh1p-GFP or Pat1p-GFP as the PB marker (). As expected, at the permissive temperature when Rpb7p ts forms function normally (see ), PBs are readily detected in both WT and mutant cells (; 30°C). To rapidly inactivate the ts Rpb7p, we exposed cells to nonpermissive temperatures (42°C) for 45–60 min before visualizing them microscopically. In response to temperature increase of WT cells, the number (and size) of PBs decreases (; WT). Thus, PBs respond to HS differently than they respond to some other stress conditions. Specifically, during starvation and other stresses, the number (and size) of PBs increases because mRNA decay is repressed (). However, during HS, mRNA decay is not repressed (see ). Moreover, because both transcription initiation (see ) and translation initiation () are repressed at high temperatures, there is no new supply of mRNPs to PBs. Consequently, at least one mechanism for disassembly of PBs during HS is mRNA degradation. In contrast with WT, no decrease in the number (and size) of PBs is observed in the ts mutant in this time frame (). These results indicate that Rpb7p is an active component of PBs. They are consistent with the defective mRNA degradation that characterizes mutants at nonpermissive temperatures, as such a defect is likely to result in more mRNA molecules remaining in PBs after the temperature increase and hence less dissociation of PBs.
We have shown previously that Pat1p can be immunoprecipitated with Rpb4p and Rpb7p (). To corroborate the interaction between Rpb7p and Pat1p by other means, we used both genetic and two-hybrid approaches. Taking advantage of the temperature sensitivity (ts) of Δ cells (; ), we determined whether overexpression of and could suppress this phenotype. Overexpression of has a small, yet significant, suppression effect on the ts phenotype of Δ (). In contrast, overexpression of enhances the proliferation defect at of Δ cells 36°C, but has little effect on proliferation of WT cells (). We also evaluated the consequence of co-overexpressing both and , including their own 5′ and 3′ noncoding regulatory sequences, on the same high copy plasmid. Overexpressing both genes in this way suppresses the proliferation defect of Δ more substantially than overexpressing alone (). The genetic interaction between and supports our conjecture that their products share a similar biochemical pathway and that Rpb7p plays a direct role in the major mRNA decay pathway that involves Pat1p. The synergistic suppression effect of simultaneously over-producing both Rpb4p and Rpb7p suggests that they interact optimally with Pat1p in the context of the Rpb4/7 heterodimer. Consistent with this possibility, we found that ts alleles (including those that are defective only in mRNA decay), but not WT , are synthetically lethal with deletion of (Fig. S4, available at ). These results indicate that the mutant forms of Rpb7p require Rpb4p to carry out their essential function.
Two-hybrid analysis suggests that Rpb7p forms a direct contact with Pat1p (). In addition we examined two-hybrid interaction between Rpb7p as the bait and, as prey, each of 20 proteins known to be involved in the mRNA decay pathway (listed in the legend to ). Only Pat1p exhibits a strong and reproducible interaction with Rpb7p. Lsm2p showed weak interaction as well (unpublished data).
To further study the mechanism by which Rpb7p functions in mRNA decay, we evaluated the capacity of our Rpb7p ts mutants to interact with Pat1, using the two-hybrid assay. As shown in , the various mutant forms of Rpb7p that are defective in stimulating mRNA decay are also defective in interacting with Pat1p. This defect is more severe at 36°C than at 30°C. The liquid β-gal test, which examines an additional reporter of the two-hybrid interaction (), yields similar results (). These data suggest that the defective capacity of Rpb7p mutant forms to stimulate mRNA decay is related to their poor binding of Pat1p.
RNA polymerase II is composed of two distinct structural domains that can be readily dissociated from each other; one of them is the Rpb4/7 heterodimer. Thus, within the context of pol II, Rpb7p functions in concert with Rpb4p (). Previously, we discovered that Rpb4p is involved in mRNA decay in the cytoplasm. Overexpression of Rpb7p could not suppress the defective capacity of Δ cells to support mRNA decay (), suggesting that Rpb4p has a distinct role in the decay of these mRNAs, which cannot be replaced by Rpb7p. Thus, an unresolved issue has been whether Rpb7p has any role in the decay pathway. Here, we show that Rpb7p plays key roles in mRNA decay.
We have uncovered two novel roles for Rpb7p in two major mRNA decay pathways. Thus, Rpb7p stimulates the deadenylation stage required for execution of both pathways. Rpb7p is required for the major pathway via its interaction with Pat1/Lsm1-7 complex (see below) and for efficient execution of the second pathway. These two functions can be uncoupled genetically as some Rpb7p mutant forms are defective and some are not defective in stimulating 3′ to 5′ decay (). We also identified a ts mutant that is not defective in stimulating deadenylation and is defective in supporting decay after deadenylation (unpublished data). Thus, Rpb7p plays distinct roles in either pathway and is probably involved in a cross talk between the two. Although it is quite possible that the two major decay pathways are linked, detailed mechanistic understanding of this possible cross talk remains to be determined.
Consistent with its role in the major decay pathway, Rpb7p is a constituent of PBs. Moreover, defects in Rpb7p can affect the capacity of PBs to dissociate when cells are exposed to HS (which is also a nonpermissive temperature for ts alleles). Thus, like some other known mRNA decay factors (; ), Rpb7p seems to be an active PB effector. In addition, we show here that Rpb7p interacts with components of the Pat1/Lsm1-7 complex. Moreover, mutations in Rpb7p that compromise the capacity of Rpb7p to function in mRNA decay also diminish its capacity to interact with these components. All these data strongly suggest that Rpb7p plays a direct role in the major mRNA decay pathway.
As Rpb7p is both a nuclear and cytoplasmic protein, it is important to discern whether its role in mRNA decay is performed in the nucleus or in the cytoplasm. We contend that Rpb7p is involved in the cytoplasmic decay pathways for the following reasons. First, Rpb7p is involved in the major deadenylation-dependent decay pathways; these pathways operate in the cytoplasm (; ; ). Second, Rpb7p is a constituent of PBs, which are cytoplasmic loci where mRNA decapping and degradation are performed, and Rpb7p is a functional effector of PBs. Third, Rpb7p interacts with components of the Pat1p/Lsm1-7 complex, known to be involved in the cytoplasmic decay pathway (; ; ). Last, Rpb7p is required for 3′ to 5′ degradation of fragment. Because this fragment is the product of Xrn1p that acts in the cytoplasm (; ; ), the stimulation of the 3′ to 5′ fragment degradation by Rpb7p must occur in the cytoplasm.
Several observations have led us to propose that Rpb7p role in mRNA decay, at least in the decay of some mRNAs, is performed in the context of Rpb4/7. First, Rpb4p and Rpb7p shuttle between the nucleus and the cytoplasm as a heterodimer (), suggesting that their role in the cytoplasm is also performed in the context of the heterodimer. Compromising this interaction by mutating Rpb7p (e.g., in cells) slows down import of both partners to the nucleus (). Second, both proteins interact with Pat1p and Lsm2p, but not with other known decay factors. Third, overexpression of both and , but not one of them at a time, has a strong suppressive effect on the ts phenotype of a Δ mutant. Fourth, both proteins are colocalized in the same PBs. It is worth noting, however, that we have observed cases in which Rpb4p-containing foci were not colocalized with Rpb7p-containing ones, and vice versa. Last, , which is otherwise a nonessential gene, is synthetically lethal with the alleles whose products are specifically defective in mRNA decay (Fig. S4).
Unlike the influence of Rpb4p on mRNA decay that is limited to PBF mRNAs (), Rpb7p appears to play a more general role in mRNA decay. It is important to emphasize that Rpb7p can function normally in the decay of several non-PBF mRNAs (e.g., mRNAs) in the absence of Rpb4p (). We propose that Rpb7p plays at least two roles in mRNA decay, only one of which (and not necessarily in all cases) involves recruiting Rpb4p. Furthermore, only in the case of PBF mRNA degradation, is this recruitment functionally important. Thus, each partner contributes differently to the mRNA decay pathway. We find that both proteins interact with the basal decay factor Pat1p/Lsm1-7p via direct interactions with Pat1p and Lsm2p. However, the two-hybrid approach cannot indicate the relative contribution of each partner to these interactions. Nevertheless, as overexpression of both proteins has a synergistic effect on genetic interactions with , it is likely that each protein contributes to the capacity of the heterodimer to recruit the Pat1p/Lsm1-7p complex. Thus, it is not likely that the interaction of the heterodimer with Pat1p/Lsm1-7p accounts for the specific role of Rpb4p in the decay pathway of PBF mRNAs. Therefore, we posit that there may exist other partners that interact specifically with Rpb4p or with Rpb7p that are required for modulating mRNA decay. For example, we propose that Rpb4p specifically interacts with a factor that modulates the decay of the very unstable PBF mRNAs. This scenario can explain why overexpression of cannot correct the abnormally slow decay kinetics of PBF in Δ cells (), reinforcing the different functions of the two partners in mRNA decay. A search for these putative factors is underway.
Rpb7p interacts with Pat1p and (to a lesser extent) with Lsm2p, two subunits of a key factor in the 5′ to 3′ mRNA decay pathway, the Pat1/Lsm1-7 complex. It is likely that this interaction represents one way that Rpb7p functions in mRNA decay, as defects in this interaction correlate with defective mRNA decay. The Rpb7p–Pat1p interaction might be important for recruiting the Pat1/Lsm1-7 complex to the mRNP, as Rpb7p is among the first proteins that interact with Pol II transcripts (; , ; ). Alternatively, interaction of Pat1p with Rpb7p might be required for subsequent Pat1p activity. Rpb7p may activate the capacity of Pat1/Lsm1-7 to stimulate decapping or may direct the complex to appropriate locations within the mRNP complex or the mRNA sequence. Further experiments are required to obtain detailed mechanistic understanding of the role of the Rpb7p-Pat1/Lsm1-7 interaction in mRNA decay.
ts mutant cells efficiently recruit Pat1p to PBs at the permissive temperatures and this Pat1p does not dissociate from PBs after shifting to the nonpermissive temperature. We interpret these results to indicate that once bound to the mRNP at the permissive temperature, Pat1p does not dissociate from mRNP even if its association with Rpb7p is diminished. Like Dhh1p-GFP–containing PBs, Pat1p-GFP–containing PBs do not disappear rapidly during HS in the mutant cells, probably because mRNAs are degraded slowly.
The transcription and mRNA decay machineries are equally responsible for setting the mRNA steady-state levels; however, probable communication between these two processes has yet to be characterized. Being both Pol II subunits and mRNA decay factors, Rpb4p and Rpb7p represent promising candidates for mediators of such cross talk. Two of the four Rpb7p mutants that we analyzed in detail are not defective in transcription at the nonpermissive temperature (see Fig. S2). This suggests that the functions of Rpb7p in transcription and in mRNA decay can be uncoupled. It remains to be determined whether the nuclear and the cytoplasmic roles of Rpb4p and Rpb7p are related, and whether the two proteins are involved in any dialogue between mRNA synthetic and decay machineries.
Yeast strains are listed in . pMC116 is a high copy (2μ) plasmid carrying under its own natural promoter bordered by NotI–HapI sites. pMC257 is a derivative of pMC116. p 2μ is a high copy plasmid and was described previously as pMC4 2μ (). To construct p 2μ, a fragment containing including 428 bp of the 5′ noncoding region and 436 bp of the 3′ noncodong region was inserted into SacII–BamHI sites of p 2μ.
was mutagenized randomly, using a PCR mutagenesis scheme as described previously (), using oMC367 (5′-GCATGCATGTGCTCTGTATG-3′) and oMC368 (5′-GTTACATGCGTACACGCG-3′) with pMC116 as the template. Both primers recognize vector sequences flanking which are identical in pMC116 and pMC257. The mutagenized -containing fragment was inserted into NotI–HapI-linearized pMC257 by cotransformation in yMC140. ∷ present in yMC140 was evicted using 5-fluoroorotic acid (5-FOA) to select for cells that lose , resulting in cells that carry only the mutagenized whose transcription is controlled by promoter. 5-FOA selection was done twice. 5-FOA–resistant strains were allowed to form colonies at 30 and 37°C. Mutants unable to form colonies at 37°C were selected.
was deleted from a strain whose sole Rpb7p was expressed from a plasmid to create yMC414. Next, yMC414 was transformed with plasmid carrying or mutant . Transformants were spotted in a tenfold serial dilution (starting with 10 cells per spot) on a 5-FOA–containing plate (this drug kills cells expressing ). Plates were incubated at 30°C for 3 d. In case of synthetic lethality, cells could not lose p and died. Note that all our mutants can proliferate well at 30°C if is present (see previous section).
To rapidly inactivate Rpb7p ts forms and simultaneously block transcription of non-HS genes, we shifted cells from 30°C (the permissive temperature) to 42°C. Northern analysis was done as described previously (). mRNA decay assays were done as described previously (). The specific probe of mRNA was prepared as follows: MFA2pG was cut with BamH1 (56 bp upstream of (G)18). oMC609 (AATGAAAGGGTAGATATTGATT) was used in a primer extension reaction in the presence of 20 μCi of radiolabled dATP. The resulting fragment was 108 bases (the poly(G) was in the middle). This fragment was hybridized at high stringency as previously published (), except that the hybridization temperature was 72°C. Membrane was then washed at 75°C. Under these conditions, the natural mRNA is undetected.
Images of GFP-containing cells suspended in water were acquired with Nikon ACT-1, using a Nikon Eclipse E400, with Nikon Plan Apo 100×/1.40 oil objective. They were collected at room temperature as 1280 × 1024-pixel files with a digital camera (DXM 1200F; Nikon).
The bait 7-AD or AD tagged mutants were constructed according to . A two-hybrid interaction was determined by growth on plates lacking leucine, tryptophane, adenine, and histidine, supplemented with 5 mM 3-amino-1,2,4-triazole and also by β-gal liquid test, as reported previously ().
Fig. S1 shows the sequence of the Rpb7p mutant forms. Fig. S2, transcription capacity of mutants at 42°C. Fig. S3, decay of Tet-Off-MFA2 at 37°C. Fig. S4, deletion of displays synthetic lethality in combination with ts alleles. Online supplemental material is available at . |
Nonsense-mediated mRNA decay (NMD) is a quality-control process found in all eukaryotic organisms studied to date (; ). One role of this process is to degrade mRNA harboring a premature termination codon (PTC) to prevent the synthesis of truncated proteins that could be nonfunctional or whose function may be deleterious to cells. The NMD pathway has been shown to be involved in the regulation of gene expression in yeast, , and mammals (; ; ; ; ).
In mammalian cells, NMD takes place after pre-mRNA splicing and in most cases is mediated by a protein complex deposited 20–24 nucleotides upstream of exon–exon junctions (; ; ). This protein complex called the exon junction complex (EJC) is thought to recruit the evolutionarily conserved UPF proteins that play an essential but still not fully characterized role in NMD. During what is referred to as the “pioneer round of translation” (), PTCs are recognized and the targeted mRNAs are degraded by both 5′–3′ decay involving decapping and 5′–3′ exoribonucleases such as hXRN1 and hXRN2/hRAT1, and by 3′–5′ decay involving deadenylation and the exosome (; ; ).
NMD implicates the participation of hUPF proteins such as hUPF1, hUPF2, hUPF3 (also named hUPF3a), and hUPF3X (also called hUPF3b). The function of these hUPF proteins is still unclear. However, it has been proposed that they are sequentially recruited by the EJC: hUPF3/3X first, followed by hUPF2, and finally hUPF1 in mammalian cells (; ). Interestingly, the function of hUPF2 has been demonstrated to be dispensable in some NMD cases, suggesting the existence of different pathways to elicit NMD ().
UPF1 is a phosphoprotein that undergoes phosphorylation/dephosphorylation cycles during NMD (; ; ). UPF1 has been shown to interact with release factors in yeast () and mammals (), and could link the EJC and translation termination complex. A direct interaction between hUPF1 and the cap-binding protein CBP80 has also recently been demonstrated in mammalian cells (), indicating that hUPF1 establishes a complex interaction network either before or during the pioneer round of translation. Phosphorylation of hUPF1 has been shown to be performed by hSMG1, a phosphoinositide 3-kinase–related kinase (; ; ), and requires the presence of hUPF2 and hUPF3 (). In contrast, dephosphorylation of hUPF1 requires the presence of a multiprotein complex composed of hSMG5, hSMG6, hSMG7, and protein phosphatase 2A (; ). For the most part, hSMG5 and hSMG7 proteins are distributed evenly throughout the cytoplasm, but a fraction is also present in processing bodies (P-bodies; ). hSMG6 is a cytoplasmic protein that concentrates in cytoplasmic foci distinct from P-bodies and whose nature is still unclear ().
P-bodies have been described in lower and higher eukaryotic cells (; ; ; ). In mammals, these cytoplasmic structures contain many factors involved in mRNA decay, including components of the decapping machinery such as decapping protein 1a (DCP1a; ), DCP2 (; ), GE1 (also called Hedls; ; ), p54/RCK (), the deadenylase CCR4 (), XRN1 (), the LSM1-7 complex involved in different aspects of RNA processing (; ), and the hUPF1, hSMG5, and hSMG7 components of the NMD machinery (; ). The function of P-bodies is still unclear but they may serve as a storage compartment for both untranslated RNAs and proteins involved in RNA decay (; ; ; ), and/or as a site for RNA decay (; ).
In a recent work, we showed that hydrophobic tetracyclic indole derivatives block the function of specific splicing factors (). In light of these findings, we decided to look further at this collection to determine if certain of these compounds also inhibit NMD. The underlying idea was that such small molecule inhibitors could represent powerful tools to decipher the NMD process. In this paper, we report the identification of one such molecule, nonsense-mediated mRNA decay inhibitor 1 (NMDI 1), that inhibits nucleus- associated as well as cytoplasmic NMD. The inhibitory mechanism appears to be caused by the loss of the interaction between hSMG5 and hUPF1, thereby leading to the stabilization of the hyperphosphorylated forms of hUPF1 and to its concentration in P-bodies. Interestingly, NMDI 1–mediated inhibition revealed that other NMD factors and PTC-containing mRNA can traffic through P-bodies as is the case in yeast ().
In a previous paper we identified a series of polycyclic indole derivatives that block the function of specific splicing factors (). Because certain molecules from this family inhibit the function of proteins involved in mRNA maturation, we decided to assess their capacity to inhibit NMD. HeLa cells were transfected by two test plasmids coding for β globin (Gl) and glutathione peroxidase 1 (GPx1) mRNA, either with (Ter) or without (Norm) a PTC. Gl mRNA was subject to nucleus- associated NMD in nonerythroid cells (; ) whereas GPx1 mRNA was subject to cytoplasmic NMD (). Additionally, a reference plasmid coding for the mouse major urinary protein (MUP) mRNA was also introduced into the cells (). 24 h after transfection, cells were incubated for 20 h with 5 μM of indole compound (Table S1, available at ) or DMSO(−) as a control. Then, total RNAs were purified and analyzed by RT-PCR () to measure NMD efficiency. Among the 25 indole derivatives tested, only compound 70 (NMDI 1) stabilized the Gl Ter mRNA level about threefold, indicating that this molecule is an inhibitor of nucleus-associated NMD ( and not depicted). Interestingly, NMDI 1 also stabilized the level of GPx1 Ter mRNA by approximately twofold ( and not depicted). To confirm these results, we measured the NMD inhibition by RNase protection assay (RPA), as it represents a more reliable approach for RNA quantification. The results are presented in Fig. S1 (A and B) and confirm the two- to threefold NMD inhibition by NMDI 1. Altogether, these data allowed us to conclude that NMDI 1 is an inhibitor of nucleus-associated as well as cytoplasmic NMD. Notably, the inhibition level obtained with NMDI 1 is similar to that observed with other NMD inhibitors such as cycloheximide (CHX) or to the down-regulation of hDCP2 or hPARN (; ). To show a more direct correlation between NMDI 1 and NMD inhibition, we measured NMD efficiency on PTC-containing Gl or GPx1 mRNA in cells that were treated with an increasing amount of NMDI 1 (). Interestingly, we observed a progressive NMD inhibition from 0 to 5 μM NMDI 1 for both Gl and GPx1 constructs. At >5 μM, we were unable to get a substantially stronger inhibition, suggesting that NMD cannot be 100% eliminated in our experimental conditions or that the 20–30% of mRNA that escaped from NMD inhibition represents the fraction of mRNAs already engaged in the NMD process at the time of NMDI 1 treatment. In all subsequent experiments, we used 5 μM NMDI 1 as our working concentration. Notably, NMDI 1 does not exhibit any cellular toxicity, as measured by trypan blue staining, even at concentrations as high as 125 μM (unpublished data).
At this stage, some controls were performed to investigate the specificity of inhibition mediated by NMDI 1. First, NMDI 1 failed to have any effect on splicing of several pre-mRNA reporter transcripts () and did not affect the level of pre-mRNA (Fig. S1, A and B). Second, unlike CHX, which inhibits translation, NMDI 1 does not alter the expression of transfected firefly luciferase (Fluc; ), suggesting that NMDI 1 is not a general translation inhibitor. To further demonstrate the absence of any effects of NMDI 1 on translation, we performed metabolic labeling of proteins with [S]methionine in HeLa cells and showed that treatment with NMDI 1 had no effect on S incorporation (Fig. S1 C). Third, to assay the integrity of the microRNA (miRNA) decay pathway in the presence of NMDI 1, we used a Renilla luciferase (Rluc) construct that is subject to degradation by miRNA (pRL-Perf) or immune to miRNA decay pathway (pRL-3XBugleMut; ). Our results indicate that NMDI 1 does not increase Renilla activity, which is under the control of miRNA, confirming that targeted mRNA degradation by miRNA is not altered by NMDI 1 (). Finally, we also tested whether NMDI 1 could induce the formation of the stress granules that provides a sensitive assay for proper mRNA metabolism. Indeed, these structures are aggregates of messenger RNPs that form when cells are subjected to several stresses, including mild translational inhibition. Unlike sodium arsenite treatment that is commonly used to induce stress granule formation (), NMDI 1 treatment did not change the localization of G3BP protein, a well-characterized marker of stress granules (Fig S1 D; ). Collectively, these results indicate that NMDI 1 is a new and specific NMD inhibitor.
To gain insight into the mode of inhibition of NMDI 1, we analyzed its effects on a tethering system that mimics the sequential recruitment of NMD factors on mRNA (; ). Cells were transfected with two types of constructs. The first codes for a Fluc mRNA containing eight binding sites for the MS2 protein in its 3′ untranslated region and the second codes for either the MS2 protein or one of the following fusions: MS2-hUPF1, MS2-hUPF2, or MS2-hUPF3X. Additionally, we transfected HeLa cells with a construct coding for the Rluc mRNA to normalize the amount of analyzed RNA. Cells were then incubated for 20 h with NMDI 1 or DMSO(−) as a negative control, and Rluc as well as Fluc mRNA levels were measured by RT-PCR as described previously (). The expression of each MS2 fusion was controlled by Western blot to verify that the observed effects were not caused by a variation in protein expression (). In each case, the compound did not affect expression of the MS2 fusion, which was itself never higher than the level of the endogenous protein. As expected, the control experiment performed in the presence of DMSO revealed that the level of Fluc mRNA was lower in cells expressing one of the MS2-hUPF fusion proteins compared with cells expressing only MS2 (). Remarkably, NMDI 1 counteracted the degradation induced by MS2-hUPF2 or MS2-hUPF3X but had no effect against MS2-hUPF1 (). Notably, the inhibition levels obtained with NMDI 1 were very similar to those observed when NMD was inhibited through down-regulation of hCBP80 (). To obtain a more accurate measure of the NMD inhibition, Rluc and Fluc mRNA levels were also measured by RPA. The results are presented in Fig. S2 A (available at ) and reproduce the quantification of mRNA levels by RT-PCR (). Altogether, these results indicate that NMDI 1 inhibits NMD downstream of hUPF3X or hUPF2 recruitment and upstream of hUPF1 functions.
In the light of the results described in the previous paragraph, we hypothesized that NMDI 1 could prevent the recruitment of hUPF1 to the EJC via its interactions with the other hUPF proteins. To test this, we immunoprecipitated hUPF1 from HeLa cell extracts under conditions that preserve the integrity of messenger RNPs (). NMDI 1 or DMSO(−) was added to the cell culture 20 h before immunoprecipitation (IP). Because hUPF2 was shown to be dispensable in some NMD cases (), we focused our analysis on the presence of hUPF3X protein in each IP (). As a control for IP specificity, we did not detect tubulin protein in any of the hUPF1 IPs and no proteins were present in the IP performed with normal rabbit serum. The results show that hUPF3X was present in hUPF1 IP even when cells were incubated with our NMD inhibitor. Thus, these data demonstrate that the interaction between hUPF1 and hUPF3X is not abolished by NMDI 1 and suggest that NMDI 1 would not prevent the recruitment of hUPF1 to the EJC.
Because hUPF1 requires a cycle of phosphorylation and dephosphorylation during NMD (), we next investigated the possibility that NMDI 1 may affect hUPF1 function by interfering with its phosphorylation level. We thus measured the level of hUPF1 phosphorylation in cells incubated with NMDI 1 or, as a control, with DMSO(−) by 2D gel analysis. Because hUPF1 phosphorylation is influenced by serum (), we used 293T cells that, unlike HeLa cells, can be synchronized by serum deprivation. Cells were transfected with the expression vector pCI-neo-FLAG-hUpf1 () and synchronized for 24 h by serum deprivation 12 h after transfection. Finally DMSO or 5 μM NMDI 1 was added for 3 h before adding back serum for 1 h. Our results show that when serum was not added back to the cell culture, the FLAG-hUPF1 protein electrofocalized in one spot corresponding to the unphosphorylated protein (; ). After serum addition, we observed a mild phosphorylation of FLAG-hUPF1 protein in the presence of DMSO and the stabilization of hyperphosphorylated isoforms of FLAG-hUPF1 when cells were incubated with NMDI 1 (). We concluded that NMDI 1 stabilized hyperphosphorylated isoforms of hUPF1.
As it has been proposed that hUPF1 would localize to P-bodies when hyperphosphorylated (; ), we analyzed the cellular localization of FLAG-hUPF1 in HeLa cells in the absence or presence of NMDI 1 (). With the exception of coexpression experiments with hSMG7, which induces the recruitment of hUPF1 to P-bodies (; ), exogenous hUPF1 was equally distributed through the cytoplasm when cells were incubated with DMSO(−) as previously reported for untreated cells (; ; ). When cells were treated with NMDI 1, we observed cytoplasmic concentrations of FLAG-hUPF1 in some structures that colocalize with the three commonly used markers of P-bodies: GFP-GE1, YFP-hSMG7, or CFP-hDCP1a (). We also verified that FLAG-hUPF1 accumulated in P-bodies in the presence of NMDI 1 in 293T cells under the same experimental conditions used to study the phosphorylation level of FLAG-hUPF1 (Fig. S2 B). We used hXRN1 protein as a P-body marker to avoid any additional transfected DNA. After addition of serum, we observed some FLAG-hUPF1 cytoplasmic concentrations that colocalize with hXRN1 only when cells were treated with NMDI 1 but not in its absence. To definitively demonstrate that hyperphosphorylated isoform of hUPF1 accumulates in P-bodies, HeLa cells were treated for 20 h with either DMSO(−) or NMDI 1, and the cellular localization of endogenous phosphorylated hUPF1 was determined using a specific antibody raised against a phosphoepitope of this protein (). The results presented in Fig. S3 (available at ) indicate that in the presence of NMDI 1, phosphorylated hUPF1 isoforms colocalize with CFP-hDCP1a foci. We conclude that NMDI 1 induces the accumulation of hyperphosphorylated hUPF1 isoforms in P-bodies. This may occur either via stimulation of phosphorylation or by blocking dephosphorylation.
To distinguish between these two possibilities, we subsequently investigated the association of hUPF1 with other NMD partners in HeLa cells treated or untreated with NMDI 1 (). We first analyzed the interaction of hUPF1 with its dephosphorylation complex. Immunoprecipitation of hUPF1 allowed recovery of hSMG5, hSMG6, and hSMG7 from DMSO-treated cells. However, after treatment of HeLa cells with NMDI 1 only hSMG6 and hSMG7 but not hSMG5 were still associated with hUPF1 (). Thus, we conclude that NMDI 1 destabilizes the interaction between hUPF1 and hSMG5. The fact that NMDI 1 does not alter the association of hUPF1 with hSMG1 and hUPF3X strongly suggests that NMDI 1 does not influence the interactions between hUPF1 and its phosphorylation complex (). Altogether, our results indicate that the hyperphosphorylation of hUPF1 is most likely caused by a failure in dephosphorylation because of the loss of interaction between hUPF1 and hSMG5 rather than an activation of phosphorylation. This conclusion is consistent with findings showing that hSMG5 is essential for hUPF1 dephosphorylation ().
Because NMDI 1 induces the localization of hUPF1 in P-bodies, hUPF1 hyperphosphorylation, and the destabilization of interactions between hUPF1 and hSMG5, we assessed the cellular localization of the hUPF1 dephosphorylation complex during NMDI 1 treatment. hSMG5 and hSMG7 have been shown to localize mainly in the cytoplasm and particularly in P-bodies as shown by colocalization experiments with the endogenous LSM4 for hSMG7 and with hSMG7 for hSMG5 (). hSMG6 similarly localizes mainly in the cytoplasm and also in some cytoplasmic foci that do not contain endogenous LSM4 (). We transfected HeLa cells with expression vectors encoding YFP-hSMG5, YFP-hSMG6, or YFP-hSMG7 () and CFP-hDCP1a as a P-body marker. After 24 h, we added DMSO(−) or 5 μM NMDI 1 to the cells. As previously shown, in the absence of the inhibitor, YFP-hSMG5, YFP-hSMG6, and YFP-hSMG7 were concentrated in cytoplasmic foci (; ), which, for a substantial fraction of them (33, 72, and 100%, respectively), colocalized with CFP-hDCP1a (). The fact that we observed hSMG6 in P-bodies unlike what was previously observed () was likely caused by the different markers used for detection of P-bodies and may reflect heterogeneity of P-bodies in their protein composition (see Discussion). In the presence of NMDI 1, the cytoplasmic foci containing YFP-hSMG6 or YFP-hSMG7 perfectly colocalized with CFP-hDCP1a P-bodies. Interestingly, hSMG5 was no longer observed in cytoplasmic foci and became evenly distributed in the cytoplasm in cells treated with NMDI 1 ().
Endogenous hSMG5, hSMG6, or hSMG7 cannot be detected in cytoplasmic foci because of their weak expression (). Because NMDI 1 inhibits NMD and induces the accumulation of hUPF1 in P-bodies as shown with exogenous as well as endogenous hUPF1 (, S2 B, and S3), we tested whether the three hSMG proteins would also accumulate in cytoplasmic foci of treated cells. The results shown in Fig. S3 indicated that these three proteins were not detected in cytoplasmic foci in DMSO-treated cells. However, when cells were incubated with NMDI 1, both hSMG6 and hSMG7 colocalized with CFP-hDCP1a in P-bodies. In agreement with the transfection experiment () under the same conditions, hSMG5 was not detected in cytoplasmic foci. Altogether, our results indicate that NMDI 1 modifies the cellular localization of hSMG5 by excluding it from P-bodies. This is consistent with the failure of hUPF1 antibodies to immunoprecipitate hSMG5 from NMDI 1–treated cells ().
Because some NMD factors such as hUPF1, hSMG5, hSMG6, or hSMG7 localize to P-bodies (, this paper) we envisaged that other NMD factors may pass through P-bodies in a transient manner. As NMDI 1 is able to block NMD at a step where hUPF1 is confined to P-bodies, we investigated the cellular localization of hUPF3 and hUPF3X in both treated and untreated cells. These two proteins have been previously shown to be primarily nuclear proteins in untreated cells (). We transfected HeLa cells with expression vectors that code for hUPF3-FLAG or hUPF3X-FLAG together with one of the P-body markers YFP-hSMG6, YFP-hSMG7, GFP-GE1, or CFP-hDCP1a. The cells were treated with DMSO or NMDI 1 before we performed indirect immunofluorescence experiments ( and not depicted for CFP-hDCP1a). As for untreated cells (; ), hUPF3 or hUPF3X localized primarily in the nucleus when cells were incubated with DMSO(−) (). However, when cells were grown in the presence of NMDI 1, we observed a cytoplasmic localization of hUPF3 as well as hUPF3X, with some accumulation in foci. To further characterize these foci, we analyzed their colocalization with cotransfected P-body markers ( and not depicted). hUPF3-FLAG as well as hUPF3X-FLAG proteins colocalized with YFP-hSMG6 (68 and 57%, respectively) or YFP-hSMG7 (50 and 51%, respectively) in cells treated with NMDI 1. Surprisingly, unlike with FLAG-hUPF1, GFP-GE1 did not colocalize with hUPF3-FLAG or hUPF3X-FLAG. We conclude that hUPF3/3X proteins can translocate to the cytoplasm, which is consistent with their shuttling properties (Lykke- Andersen et al., 2000; ), and can reach a subset of P-bodies.
Because NMD factors accumulate in P-bodies when NMD is inhibited, we were interested in determining whether NMD substrates also accumulate in P-bodies. In yeast, it has recently been shown that PTC-containing mRNAs accumulate in P-bodies when NMD is blocked (). We speculated that our NMD inhibitor would allow us to reach the same conclusion in mammalian cells. HeLa cells were transfected with pmCMV-Gl Ter or pmCMV-GPx1 Ter and the localization of the resulting mRNAs was analyzed with several P-body markers: GFP-GE1, YFP-hSMG6, YFP-hSMG7, CFP-hDCP1a, GFP-hCCR4, FLAG-hUPF1, hUPF3-FLAG, or hUPF3X-FLAG ( and S4, available at ). As expected, in the absence of the inhibitor we were unable to detect PTC-containing mRNAs, most likely because of their rapid decay by NMD. However, in the presence of NMDI 1, PTC-containing mRNAs were stabilized and detected mainly in cytoplasmic aggregates. Interestingly, our results indicated a substantial colocalization between PTC-containing Gl or GPx1 mRNA and each tagged version of the tested hUPF proteins CFP-hDCP1a, YFP-hSMG6, or YFP-hSMG7 ( and S4). Although the colocalization between Gl Ter or GPx1 Ter mRNAs and hUPF3/3X was total ( and not depicted for hUPF3X), it was only partial with other P-body components such as FLAG-hUPF1 (63 and 74%, respectively), YFP-hSMG6 (76 and 81%, respectively), or YFP-hSMG7 (71 and 94%, respectively), and infrequently with GFP-hCCR4 (11 and 33%, respectively; and S4). Notably, we often observed that RNA foci and P-bodies did not overlap perfectly. Additionally, we were unable to observe a colocalization between PTC-containing mRNAs and GFP-GE1 ( and S4). Altogether, our results show that PTC-containing mRNAs were present in and adjacent to P-bodies when NMD was inhibited by NMDI 1. In addition, these data indicate heterogeneity in the composition of P-bodies because some markers colocalize with PTC-containing mRNAs while others do not.
The accumulation of PTC-containing mRNAs in P-bodies when NMD is blocked in mammalian cells was confirmed by a more resolutive approach in U2OS cells. In this setting, we tagged the mRNA with a 24× MS2 binding site repeat because this approach can efficiently detect single mRNA molecules by in situ hybridization (; ). In control cells, PTC-containing mRNAs were mostly detected in the nucleus, and the cytoplasmic molecules that were detected did not accumulate in P-bodies labeled with CFP-hDCP1a. When NMD was inhibited with NMDI 1, higher levels of Gl-Ter MS2 mRNAs were detected in the cytoplasm, and these molecules accumulated in structures that colocalized with P-bodies. As previously observed, mRNAs did not perfectly colocalize with CFP-hDCP1a, but were instead adjacent and formed a ring at the periphery of P-bodies, similar to what was found with miRNA targets (). These data confirmed that mRNAs subjected to NMD accumulate in P-bodies when their degradation is inhibited, and this conclusion seems not be cell type specific.
We also analyzed the localization of Gl or GPx1 Norm mRNAs in HeLa and U2OS cells treated with NMDI 1 (). Our results showed no specific accumulation of these mRNAs in cytoplasmic bodies typified by YFP-hSMG6, GFP-GE1, or CFP-hDCP1a in the presence of NMDI 1. In contrast, we did not see any wild-type mRNAs in P-bodies either because Norm mRNAs do not go to P-bodies or because their degradation pathway is not affected by either DMSO or NMDI 1. These results support the idea that NMDI 1 is an NMD inhibitor rather than a general RNA decay inhibitor.
To determine whether the accumulation of NMD factors in P-bodies can be triggered when any step of the NMD process is inhibited, we decided to interfere with the NMD process in three different ways. The first one relies on the down-regulation of hUPF2 using siRNA (; ). According to the current model of NMD in mammalian cells (), depletion of hUPF2 will block NMD at an earlier step than the one induced by NMDI 1. Interestingly, hUPF2 down-regulation does not induce the accumulation of FLAG-hUPF1, hUPF3-FLAG, or hUPF3X-FLAG P-bodies. Additionally, whereas hUPF2 depletion induces a stabilization of Gl-Ter mRNA because of the inhibition of NMD, Gl-Ter mRNA was homogenously distributed in the cytoplasm with no accumulation in P-bodies (). Thus, inhibition of NMD by hUPF2 depletion does not trigger accumulation of NMD factors and substrates in P-bodies.
The second one is based on the down-regulation by siRNA of a protein involved in a late step of the NMD process such as hXRN1 (; ). As for the down-regulation of hUPF2, the cellular localization of NMD factors including FLAG-hUPF1 or hUPF3-FLAG was not modified by the cellular lack of the hXRN1 protein (), whereas Gl-Ter RNA was detected in P-bodies. This result may indicate that the recycling of NMD factors had already occurred before the function of hXRN1. We conclude that the presence of NMD factors in P-bodies depends on the NMD step, i.e., only some steps in the NMD process occur in P-bodies.
According to our results, hUPF1 dephosphorylation is one of the NMD steps that occur in P-bodies. In the presence of NMDI 1, NMD is inhibited because hUPF1 is stalled in a hyperphosphorylated form caused by the release of hSMG5 from P-bodies. To further confirm this model, we aimed to mimic the effect of NMDI 1 by analyzing the cellular localization of NMD factors in the presence of the hUPF1 mutant protein (HA-hUPF1dNT) that has been shown to prevent its interaction with hSMG5 because it lacks an N- terminal, or in the presence of hSMG5 mutant proteins (HA-hSMG5dCT and HA-hSMG5DA) that cannot dephosphorylate hUPF1 because of either the lack of a C-terminal or the substitution of aspartic acid 860 by an alanine (). As shown in , when HeLa cells express HA-hUPF1dNT protein, hUPF3-FLAG, hUPF3X-FLAG proteins, Gl Ter, and GPx1 Ter mRNA localize to P-bodies as shown by the colocalization with CFP-hDCP1a. This result is similar to what we observed when cells were treated with NMDI 1. Interestingly, YFP-hSMG5 was not detected in P-bodies, suggesting that by destabilizing the interaction between hSMG5 and hUPF1, hSMG5 is excluded from P-bodies. Additionally, expression of either HA-SMG5dCT or HA-SMG5DA induced accumulation of FLAG-hUPF1, hUPF3-FLAG, hUPF3X-FLAG proteins, Gl Ter, and GPx1 Ter mRNA into P-bodies (, respectively). Thus, by specifically blocking the dephosphorylation of hUPF1, NMD factors and substrates concentrate into P-bodies.
In this study, we have characterized an indole derivative, NMDI 1, as an NMD inhibitor. This molecule has allowed us to study specific steps of NMD. The power of this approach lies in the ability to freeze NMD at a step when hUPF1 and hUPF3/3X are detected in P-bodies ( and ). Using biochemical and cellular biology approaches, we have determined the precise event blocked by NMDI 1 and established that this inhibitor prevents the interactions between hUPF1 and hSMG5, resulting in the subsequent exclusion of hSMG5 from P-bodies and the stabilization of hyperphosphorylated isoforms of hUPF1. Unlike other approaches, such as transfection-mediated down-regulation of NMD factors, where only a fraction of cells are subject to the inhibition of NMD, small chemical molecules have the ability to diffuse across the cell membrane and affect most cells in culture. Therefore, such inhibitors should enable NMD to be inhibited in more physiologically complex environments such as tissue or multicellular organisms to study NMD mechanism in vivo and to evaluate their potential therapeutic capacities ().
As in yeast (), PTC-containing mRNAs, hUPF1, and hUPF3/3X proteins are found in P-bodies of mammalian cells when NMD is prevented (, , and ). Undoubtedly, yeast and human P-bodies share some similarities in their protein compositions and functions but clear differences can also be seen. Unlike in yeast, in mammalian cells PTC- containing mRNAs accumulate in P-bodies, or more precisely at the periphery of P-bodies, suggesting that P-bodies can be formed by several compartments. This observation is consistent with a recent paper () showing that RNA can be localized at the periphery of P-bodies where it might be stored before being degraded or released from the P-bodies. Another difference between yeast and mammalian P-bodies is that a down-regulation of hUPF2 in mammals does not lead to the accumulation of hUPF1 into P-bodies () as it does in yeast (). Our results suggest that hUPF2 is involved in the NMD process before the transit of NMD factors and substrates through P-bodies. These differences likely reflect a divergence in the process of NMD in yeast and in mammalian cells. Surprisingly, when we blocked NMD at a late step when RNAs are going to be degraded, (i.e., by down-regulating hXRN1) we did not detect NMD factors in P-bodies but we did observe PTC-containing mRNAs in P-bodies (). This confirms that NMD involves trafficking to P-bodies and suggests that by inhibiting the RNA degradation step, we did not prevent the recycling of NMD factors from P-bodies.
Interestingly, we did not observe any differences in the cellular distribution and in the protein composition between P-bodies containing Gl Ter mRNA and those containing GPx1 Ter mRNA, even though these two mRNAs are subject to nucleus-associated or cytoplasmic NMD, respectively. However, we cannot exclude the possibility that NMDI 1 would freeze a series of dynamic events that occurs during NMD and that this would induce a drift of nucleus-associated P-bodies to the cytoplasm. It is also possible that P-bodies with nucleus-associated NMD substrates and P-bodies with cytoplasmic NMD substrates have different biochemical or physical properties that would lead to the cosedimentation of one with the nuclear fraction and of the other with the cytoplasmic fraction. Further investigations will be necessary to clarify this point.
NMDI 1 allowed us to show that P-bodies display a large degree of variability in their NMD factor composition. For example, whereas hSMG6 colocalizes with CFP-hDCP1a or FLAG-hUPF1 when NMD is inhibited ( and not depicted), it only shows a partial overlap with CFP-hDCP1a in DMSO-treated cells () and no colocalization with LSM4 (). As another example, we detected hUPF3/3X-FLAG proteins in P-bodies positive for YFP-hSMG6 or YFP-hSMG7 but not in P-bodies containing GFP-GE1 (). Similarly, PTC-containing mRNA is found in all P-bodies holding hUPF3/3X-FLAG proteins and in most P-bodies containing FLAG-hUPF1, YFP-hSMG6, or YFP-hSMG7, but rarely or never in P-bodies stained with GFP-hCCR4 or GFP-GE1 ( and S4). Because the factors that do not colocalize upon NMDI 1 treatment, such as hUPF3/3X (or PTC- containing mRNAs) and GE1, still form foci that colocalize with other P-body markers (such as hSMG7), these data suggest that P-bodies can exist in several “flavors” or forms that would differ in protein composition (at least in mammalian cells). These variations in P-body composition could reflect different functional states. In this view, hUPF3/3X, hSMG6, hSMG7, hUPF1, and hDCP1a would be involved at early steps of RNA processing in P-bodies or, as has been recently proposed, would nucleate the formation of P-bodies on the PTC-containing mRNA (). hCCR4 would then join the structure, followed by GE1 that would induce degradation of the PTC-containing mRNA and recycling of hUPF3/3X. Then other NMD factors are recycled for a new turn of NMD (). This evolution in P-body composition could arise by fusion of different subcategories of P-bodies, by the shuttling of individual components, or by a combination of these processes. An attractive approach to answer these questions would be to characterize NMD inhibitors that target other steps than the dephosphorylation of hUPF1.
l
r
e
s
u
l
t
s
p
r
e
s
e
n
t
e
d
i
n
t
h
i
s
a
r
t
i
c
l
e
a
r
e
r
e
p
r
e
s
e
n
t
a
t
i
v
e
f
r
o
m
a
t
l
e
a
s
t
t
h
r
e
e
i
n
d
e
p
e
n
d
e
n
t
e
x
p
e
r
i
m
e
n
t
s
. |
Mitochondria have elaborate translocation machinery for the import and assembly of nuclear-coded proteins. Translocons of the outer membrane (TOM) and inner membrane (TIM) coordinate protein translocation across the outer and inner membranes (; ; ). The TIM23 translocon mediates the import of proteins with a typical N-terminal targeting sequence. The TIM22 import pathway facilitates the TIM proteins, including the carrier family and the import components Tim22p and Tim23p. Members of the TIM22 pathway include two soluble complexes in the intermembrane space, Tim9p-Tim10p and Tim8p-Tim13p, as well as a 300-kD complex at the inner membrane, which consists of Tim12p, Tim18p, Tim22p, Tim54p, and a fraction of Tim9p and Tim10p (). The soluble complexes aid movement of the hydrophobic precursors across the aqueous intermembrane space, and the 300-kD complex mediates insertion into the inner membrane.
The essential components in the TIM22 translocon are Tim9p, Tim10p, Tim12p, and Tim22p (, ; ; ,), which all bind directly to the imported substrate as shown by chemical cross-linking. Tim18p and Tim54p, however, seem to play secondary roles in protein translocation because a direct interaction with a translocating TIM22 substrate has yet to be demonstrated. Tim18p is an accessory protein, and the 300-kD TIM22 complex is present, albeit decreased in molecular mass, in strains lacking (; ). Tim54p was identified in a two-hybrid screen using the cytosolic domain of Mmm1p, an outer membrane protein involved in the maintenance of mitochondrial morphology and mitochondrial DNA, as the bait (). This interaction may be reflected by Mmm1p's role in assembly of β-barrel outer membrane proteins (). Jensen and colleagues specifically showed that Tim54p partners with Tim22p in the 300-kD complex and is essential for viability (), suggesting that Tim54p might assemble or stabilize the 300-kD complex. However, subsequent studies by Pfanner and Jensen showed that Tim54p is not essential under certain conditions (), and raised a general question about the role of Tim54p in mitochondrial biogenesis.
Surprisingly, cells lacking have a petite-negative phenotype, which is revealed by inviability on glucose medium in the presence of ethidium bromide (). In , mitochondrial respiration is not essential for viability and is referred to as petite-positivity because of the ability to grow on fermentable carbon sources when the mitochondrial genome is lost (ρ) or contains large deletions (ρ; ; ). However, most yeast strains, including and are petite-negative and require a functional mitochondrial genome for growth (). Petite-positivity is achieved because a mitochondrial membrane potential (Δψ) can be maintained in the respiration-deficient state by the import of ATP (via the ADP/ATP carrier, AAC) in exchange for ADP produced by the F-ATPase, which acts as an ATP hydrolase ().
In addition to , mutations in several nuclear genes can render petite-negative (). These nuclear genes include those for the F portion of the ATPase (; ), AAC (), and the mitochondrial inner membrane protease, Yme1p (; ). AAC is required for the aforementioned exchange of ATP and ADP between the cytosol and the matrix (). Yme1p may affect ATPase function by catalyzing the turnover of protein inhibitors of the ATPase () or altering F-ATPase activity, possibly by inducing structural changes (). The petite-negative phenotype of cells was suppressed by several genes coding for cytosolic proteins including translation components and chaperones (). Jensen and colleagues suggested that this set of suppressors might compensate for a defect in protein import, allowing the mitochondrion to maintain a membrane potential (). In addition, a mutation in results in a petite-negative phenotype that has been linked to a defect in transcription (), and petite-negativity for cells lacking is dependent on the strain background (). Thus, the specific mechanism(s) resulting in the synthetic lethality between loss of the mitochondrial genome and mutations in this wide array of nuclear genes is not well understood.
In this paper, we investigate the function of Tim54p in mitochondrial biogenesis and find that the cells are petite-negative. In contrast, strains harboring mutations in , , , , , and , all import components, are petite-positive in our strain background. From a systematic analysis, we show that Tim54p is not required for maintenance of mitochondrial DNA or nucleoid morphology. Rather, Tim54p is required for the stability/assembly of the TIM22 complex and, second and specifically, for the assembly of Yme1p into a proteolytically active complex. Because Yme1p is imported through the TIM23 translocon and Tim54p is a stabilizing component of the TIM22 translocon, Yme1p import and assembly represent a novel collaborative effort between the two translocons of the inner membrane. Given the role of Yme1p in protein turnover, Tim54p effectively links pathways of import, assembly, and turnover in the mitochondrion.
Whereas Tim9p, Tim10p, Tim12p, and Tim22p (, ; ; ,) all bind directly to an imported TIM22 substrate, we have not been successful at identifying a direct biochemical interaction between a translocating TIM22 substrate and Tim54p (unpublished data). These observations prompted us to investigate the specific role of Tim54p in mitochondrial biogenesis.
Because Jensen and colleagues originally identified Tim54p through a two-hybrid interaction with Mmm1p () and because Mmm1p has a role in mitochondrial DNA maintenance (; ), we probed whether Tim54p might function in mitochondrial DNA stabilization. We generated a strain deleted for () using plasmid shuffling in the parental strain GA74 (genotypes for strains are listed in Table S1, available at ) (; ; ). We also generated a temperature-sensitive (ts) mutant using error-prone PCR in GA74 (; ). Whereas the and strains grew with a doubling time of ∼3 h at 25°C on glucose media, the strains arrested growth after 6 h at 37°C. The strains also grew slowly on ethanol-glycerol media at 25°C (Fig. S1). As expected, Tim54p was not detected by immunoblot analysis in mitochondrial extracts derived from the strain (). Tim54p also was not detected in mitochondria purified from the mutant grown at 25°C; the mutant protein may be present at steady-state levels that are not detectable by the antibody, or the protein may be turned over at an increased rate compared with wild-type Tim54p.
We investigated whether Tim54p might function in a pathway for the maintenance of mitochondrial DNA. We analyzed growth at 25°C by serial dilution in glucose medium supplemented with ethidium bromide, which causes cells to lose their mitochondrial DNA (referred to as ρ; ) (). The and strains, like the control strain, were not viable on glucose media containing ethidium bromide. In contrast, ts mutants , , and (; ) grew on media containing ethidium bromide, indicating that strains defective in protein import in the GA74 background typically do not require the mitochondrial genome for growth. Therefore, the and strains have a petite-negative phenotype, whereas the and mutants are petite-positive.
We checked if Tim54p was required for maintenance of mitochondrial morphology. Mitochondrial and nucleoid structure were probed in a strain by transforming the strain with a mitochondrial matrix-targeted GFP construct and staining with 2 μg/ml DAPI, respectively (). Compared with the parental strain with many small punctate nucleoids (25 ± 10, = 32), cells contained fewer nucleoids (6 ± 3, = 30), which is similar to that observed in the isogenic strain (6 ± 2, = 34). In contrast, and the parental strain had a similar mitochondrial network.
Because morphology defects can arise as a secondary defect in mitochondrial function, as has been demonstrated for mutants in mitochondrial outer membrane assembly (), we evaluated the nucleoid quantity in the mutant at the restrictive temperature of 37°C in a time course assay. At the initial time point and after 12 h at 37°C, the cells showed normal nucleoid number (20 ± 9, = 33) (). Only after 18–24 h at 37°C were structural changes—from abundant, punctate nucleoids to fewer, coalescent nucleoids (5 ± 2, = 20)—evident in cells, in contrast to the parental strain that maintained a normal nucleoid phenotype at all time points analyzed (). The aberrant nucleoid morphology thus is a secondary consequence associated with loss of functional Tim54p, indicating that Tim54p does not play a direct role in regulating mitochondrial or nucleoid structure.
A comprehensive study by Jensen and colleagues illustrated a link between mitochondrial protein import and cytosolic protein translation and trafficking pathways (). We assessed the petite-negativity of import components in different strain backgrounds available in our laboratory by serial dilution on glucose media supplemented with ethidium bromide (). Whereas the Jensen group reported previously that their and strains were petite-negative, the BY4742 strain deleted for was the only strain that showed partial sensitivity to ethidium bromide () (). Instead, our GA74 strain, as well as other backgrounds (; ; ), did not display petite-negativity when deleted for and . In contrast, the petite-negative phenotype associated with was independent of strain background () (). Additionally, mutations in , , and did not cause petite-negativity () (,). The studies by Jensen and colleagues show that their background choice was important for deciphering the threshold effects caused by a defect in protein import and subsequent compensatory mechanisms, whereas the GA74 strain background does not seem to display these threshold effects. Thus, petite-negativity in the GA74 strain is a unique phenotype specifically associated with mutations in and not other import components.
Because cells lacking Yme1p are not viable in the absence of mitochondrial DNA, we investigated the steady-state levels of Yme1p and other mitochondrial proteins in , , , and wild–type (WT) mitochondria grown at 25°C (). Increasing amounts of a mitochondrial protein lysate were separated by SDS-PAGE and immunoblotted for mitochondrial proteins including Yme1p, Tim23p, Tim22p, AAC, Fα-ATPase, Hsp70, Tom40p, porin, and Hsp60. The abundance of Yme1p was decreased in the and mitochondria. The steady-state levels of Tim22p were also decreased in the mutants, but not in mitochondria. In addition, Tim23p and AAC levels (substrates of the TIM22 pathway) were lower in mutant mitochondria because the TIM22 import pathway is compromised in the absence of Tim54p (; ). In contrast, the steady-state levels of porin, Tom40p, Hsp70, Hsp60, and the Fα-ATPase were similar among the tested strains. Therefore, mitochondria lacking functional Tim54p show a decreased abundance in Yme1p, as well as the previously characterized TIM22 translocon substrates, Tim23p, Tim22p, and AAC.
A reason for the decreased abundance of the aforementioned proteins in the mitochondria lacking functional Tim54p might in part be caused by a general defect in protein import (). We tested the import of Hsp60 and Tim23p, which use the TIM23 and TIM22 import pathways, respectively, into mitochondria isolated from the and strains (). Both Hsp60 and Tim23p were imported into and strains, albeit at a slower rate than WT mitochondria. The steady-state level of Hsp60 in the and mitochondria, however, was not markedly reduced (). We have observed this scenario previously in which the in vitro import rate with isolated mitochondria is impaired compared with the steady-state levels, indicating that either import in vivo is more efficient than in vitro (; ) or protein turnover is decreased to compensate for a reduced rate/abundance of import.
We investigated the function of Tim54p in the TIM22 pathway in greater detail with a combined genetic and biochemical approach. We transformed the and mutants with a high-copy plasmid (designated [TIM22]) in which Tim22p was overexpressed (). Overexpression of specifically restored growth to the mutant at 37°C on glucose media; however, the petite-negative phenotype was not reversed (). In contrast to studies by Pfanner and Jensen and colleagues (), overexpression did not suppress the growth defect or the petite-negativity in the strain (), indicating that suppression is dependent on the presence of the gene and that the mutant protein (albeit undetectable by immunoblot) is required for growth at 37°C. Also, overexpression did not suppress the growth defect on rich ethanol-glycerol media at 37°C, demonstrating that phenocopies the mutant (). We also tested cold-sensitivity at 15°C on rich glucose media (Fig. S2, available at ). Like the mutant, and strains displayed cold sensitivity, which was not restored upon overexpression (Fig. S2). In addition, the abundance of Tim22p, Tim23p, AAC, and Yme1p increased in mitochondria overexpressing Tim22p (), and import was restored to near-WT levels for substrates of the TIM22 pathway (; Fig. S3). This analysis indicates that overexpression of Tim22p restores defects associated with protein import and growth in the mutant, supporting the hypothesis that Tim54p functions as a stabilizing scaffold/assembly factor for the TIM22 complex.
This set of experiments also suggests that Tim54p has a separate function associated with petite-negativity that overexpression cannot suppress. Because overexpression of , identified in the screen by Jensen and colleagues was a suppressor of (; ), we investigated whether might suppress the mutant (Fig. S4, available at ). Interestingly, overexpression failed to restore growth of the mutant on ethidium bromide medium (Fig. S4). Therefore, the petite-negativity of the mutants is not seemingly caused by a defect in protein import or translation.
Because assembled ATPase and AAC complex are required for maintenance of a membrane potential (), we investigated the assembly state of the ATPase and AAC (); impaired assembly could lead to inviability in the presence of ethidium bromide. As assessed by one-dimensional blue-native PAGE, ATPase assembly was not impaired in mitochondria lacking functional Tim22p, Tim23p, Tim54p, and Yme1p (). To determine if AAC assembled normally as a dimer, mitochondrial extracts were resolved by blue-native PAGE in the first dimension and SDS-PAGE in the second dimension. AAC assembled into a dimer in mitochondria and WT mitochondria () as well as the mutant (unpublished data). In previous studies, we have shown that AAC levels are essentially undetectable in our , , mutant mitochondria (; ; ). However, given that these mutants are not petite-negative as shown in and , even almost undetectable levels of AAC seem adequate to support the minimal membrane potential required for growth in the absence of respiration. Thus, the petite-negative phenotype associated with the strains is not due to a lack of assembly of the ATPase or AAC complexes.
Given that the petite-negative phenotype and the respiratory-deficiency at 37°C could not be restored by overexpression of , we focused on Yme1p assembly and function. First, we evaluated Yme1p import into mitochondria defective in , , and , as well as the mitochondria with overexpressed (Fig. S5 A, available at ). Yme1p contains a typical N-terminal targeting sequence that is cleaved, presumably by the matrix processing peptidase, upon import (; ); as such, Yme1p is predicted to use the TIM23 import pathway because of its typical presequence (). The Yme1p import rate was decreased in and mutant mitochondria (Fig. S5 A). Despite the decreased rate of import, however, Yme1p was present at WT levels in and mitochondria (). In addition, the rate of Yme1p into mutant mitochondria was impaired (Fig. S5 A), but Yme1 was detected in mutant mitochondria ( and ). Finally, overexpression of Tim22p restored the rate of import (Fig. S5 A). Whereas Yme1p import is impaired, Yme1p still accumulates to levels that are detectable in WT mitochondria.
Yme1p assembles into a large complex with a predicted molecular weight of ∼1 MDa (). Using blue-native gel analysis, we tested whether Yme1p assembled in mitochondria defective in Tim54p function. Mitochondria were solubilized in 1.0% (wt/vol) digitonin and separated on a 5–10% blue-native gel (). In WT,
, , and mitochondria, Yme1p assembled into a large complex migrating well above the highest molecular weight standard, consistent with the predicted molecular weight of 1 MDa obtained by size-exclusion chromatography (). Yme1p thus assembled in mutant and mitochondria. In addition, Yme1p assembly was not impaired in mitochondria lacking functional small Tim proteins or Tim18p. In contrast, Yme1p assembly was impaired in and mitochondria as well as mitochondria with overexpressed Tim22p (); when the gel was overexposed to film, a signal at the molecular mass of 80 kD—that of unassembled Yme1p (Yme1p monomer)—was detected (). We followed the assembly of Yme1p by coupling import assays and blue-native gels (). As expected, assembledYme1p accumulated into a high molecular weight complex in WT and mutant mitochondria but not in mitochondria, even after import for 20 min. Note that the small amount of Yme1p that is detected in the mitochondria accumulates in both the presence and absence of a membrane potential and might reflect a small amount of Yme1p that can assemble transiently in the import assay, but fails to accumulate in vivo (). Together, these data indicate that Tim54p is required for the assembly of Yme1p.
To confirm that Yme1p was not functional in mutant mitochondria, we investigated the import and degradation of a Yme1p model substrate, Yta10-DHFR (), which consists of a fusion between the N-terminal 161 amino acids of Yta10p and a loosely folded mutant of DHFR (). Importantly, the Yta10-DHFR construct could be imported into WT, , , and mitochondria, although import was decreased in the mutant mitochondria (). To test if Yme1p was assembled into a proteolytically active complex, we investigated the degradation rate of imported Yta10-DHFR in the presence of an ATP-regenerating system at 37°C (). Whereas ∼50% of the Yta10-DHFR was degraded in WT mitochondria, Yta10-DHFR was essentially stable in mitochondria lacking functional Tim54p (even when Tim22p was overexpressed) and Yme1p (). Thus, Tim54p is required for assembly of a proteolytically active Yme1p complex.
The observed impairment in Yme1p assembly in mitochondria may be caused by a mislocalization of Yme1p within mitochondria (). Initially, Yme1p localization was investigated by carbonate extraction and osmotic shock in the presence and absence of proteinase K in and mitochondria. However, and mitochondria were resistant to osmotic shock (unpublished data). We reasoned that the composition of the inner membrane might be altered, so we used the [] mitochondria in which was overexpressed. Indeed, these mitochondria were amenable to osmotic shock. As expected, Yme1p was an integral membrane protein because it was recovered in the pellet fraction like the integral membrane protein Tom70p after carbonate extraction (). To confirm that Yme1p was not mislocalized to the mitochondrial matrix, we used osmotic shock, which ruptures the mitochondrial outer membrane, in the presence and absence of protease to determine the location of Yme1p (; the control reaction in WT mitochondria are presented in Fig. S5 B). As expected, Yme1p and Taz1p () localized to the mitochondrial intermembrane space because both were degraded by the added protease upon osmotic shock. In contrast, the matrix marker α-ketoglutarate dehydrogenase (Kdh) was protected from protease because the inner membrane remained intact and the outer membrane marker Tom70p was protease susceptible in both mitochondria and mitoplasts (MP). Yme1p thus resides in the intermembrane space in [] mutant mitochondria.
Because of Tim54p's role in Yme1p assembly, we predicted that Tim54p might transiently interact with Yme1p to facilitate assembly. We used a chemical cross-linking/immunoprecipitation approach coupled with the in vitro import assay () to trap an interaction between Tim54p and imported Yme1p (). Specifically, when the cross-linked partners were reduced with β-mercaptoethanol before separation by SDS-PAGE, cross-linked Yme1p was co-immunoprecipitated with antibodies against Tim54p and Tim23p.
failed to immunoprecipitate cross-linked Yme1p. The addition of reductant was required to detect the interaction because the cross-linked product migrated at a large molecular mass. Using another approach, we constructed a GA74 strain with a hexa-histidine tag on Yme1p (Yme1-His; ); this strain grew like the WT strain, indicating that the tag did not disrupt Yme1p function. After solubilization of mitochondria, a small amount of the Tim54p, but neither Tim44p nor Kdh, interacted with tagged-Yme1p. In control reactions with WT and mitochondria, Tim54p did not copurify. Additional partner proteins may assist Tim54p with Yme1p assembly, because in a genetic approach, overexpression of Yme1p did not complement the petite-negativity or growth phenotype in the mutant (unpublished data). These studies—import/coimmunoprecipitation assays and in organello complex purification—indicate that a fraction of Tim54p interacts directly with Yme1p.
Collectively, this investigation shows that Tim54p, possibly with assistance of additional components (), is required for the assembly of Yme1p into a functional complex after import via the TIM23 pathway. Thus, Tim54p, a component of the TIM22 translocon, integrates functions of the two inner membrane translocases by providing an assembly activity that is independent from the translocation properties of the TIM22 translocon. As such, Tim54p serves as a link between the import, assembly, and proteolysis pathways in the mitochondrion.
In this study, we analyzed the function of Tim54p in mitochondrial biogenesis. Previous experiments suggested that Tim54p was an essential protein of the 300-kD TIM22 translocon of the mitochondrial inner membrane. Tim22p and Tim54p coimmunoprecipitated and expressed from a multicopy plasmid, suppressed a growth defect in a mutant strain (). We therefore expected Tim54p to play a direct role in protein translocation. However, more recent studies suggested that Tim54p might have other functions in mitochondrial biogenesis. Specifically, was not essential for viability and a functional translocon consisting of only Tim22p was subsequently purified (). Moreover, a translocation intermediate between a TIM22 pathway substrate and Tim54p has not yet been demonstrated (). These studies suggest that, although an integral subunit of the TIM22 complex, Tim54p most likely functions in an alternative aspect of mitochondrial biogenesis.
This study confirms that Tim54p does not play a direct role in import of the known TIM22 translocon substrates. In vitro import of Tim23p and Hsp60, however, was decreased in Δtim54 and tim54-3 mitochondria. This decrease in import is potentially caused by a decrease in Tim22p and Tim23p levels (). Indeed, Tim54p is required for stabilization of Tim22p, even though a functional translocon can be assembled in the absence of Tim54p (; ). Specifically, overexpression of from a high-copy plasmid restored growth on rich glucose medium and the import of TIM22 substrates (). Tim54p therefore serves as a scaffold/assembly factor for the TIM22 complex, potentially by stabilizing the 300-kD TIM22 complex in the inner membrane.
However, overexpression of did not restore the petite-negativity of the and strains. Rather, our studies show that Tim54p plays a direct role in the assembly of a proteolytically active Yme1p complex. Two specific defects associated with mitochondria lacking functional Tim54p that are not observed with mitochondria defective in other components of the TIM22 pathway are that mutant mitochondria require mitochondrial DNA for viability and are defective in Yme1p assembly into a large functional complex. Thus, Tim54p seems to play a direct role in Yme1p biogenesis independent from the classical function assigned the TIM22 translocon, namely, translocation and insertion of polytopic membrane proteins into the inner membrane.
In addition to mutations in , several defects in mitochondrial function contribute to petite-negativity in yeast. Other components involved in mitochondrial biogenesis are inviable when the mitochondrial genome is lost in particular genetic backgrounds (; ); certain strains lacking functional Tim18p, Tom70p, and Tim9p cannot tolerate loss of the mitochondrial genome. For Tom70p and Tim18p, Jensen and colleagues suggested that the decreased import efficiency in combination with a decreased membrane potential resulted in lethality when the mitochondrial genome was absent and overexpression of cytosolic proteins that improved import efficiency could suppress the defect (); these studies suggest that the pathways to maintain a membrane potential are obviously complex. Surprisingly, overexpression of in mutant mitochondria did not suppress the petite-negativity or the respiratory deficiency at 37°C. In addition, overexpression of () suppressed petite-negativity in the strain, but not . These results argue that the petite-negativity caused by a defect in is not the result of a combined impairment in import efficiency and a decrease in membrane potential. AAC function and abundance also is critical for maintaining a membrane potential in the absence of a mitochondrial genome (). Because AAC depends on the TIM22 import pathway for biogenesis and AAC levels are lower in the and mutants, defects in AAC function may contribute to petite- negativity in these mutant strains. However, because our , , and mutants, all with severely lowered levels of AAC, are petite-positive ( and ) and AAC levels are increased when is overexpressed in the mutant, our data that mutants are petite-negative support the postulation that Tim54p has an alternative function in mitochondrial biogenesis, namely assembly of Yme1p.
We therefore propose the following model in which Tim54p mediates Yme1p assembly (). In step 1, Yme1p is imported via the TIM23 translocon. After cleavage of the presequence by the matrix processing protease, the monomer is released into the inner membrane (step 2). Subsequently, Tim54p mediates assembly of the Yme1p monomers into a functional complex (step 3). Indeed, in mitochondria defective in Tim54p function, the Yme1 monomer is detected in blue-native gels (). Our cross-linking studies in which Tim54p binds to both processed and unprocessed Yme1 suggests that the TIM22 and TIM23 translocons are proximally associated in the inner membrane. In addition, Tim54p and Tim23p may be simultaneously interacting with Yme1p in transit; Tim54p might serve as a tether for the Yme1 substrate as it enters the intermembrane space, similar to a role for Yme1p in the import of PNPase (). Our studies suggest that the interaction between Tim54p and Yme1p is transient because freezing a stable interaction required cross-linking and only a small fraction of the Tim54p interacted with Yme1p-His. We suggest that Tim54p is a bonafide member of the TIM22 complex, because Tim54p comigrates in the 300-kD complex with Tim22p in blue-native gels and coimmunoprecipitates with Tim22p (; ). However, it is formally possible that Tim54p assembles in other complexes (of similar size on blue-native gels) and might partner with additional proteins to assist in the assembly of Yme1p independent of its association with the TIM22 complex. As an example, a recent study by Jensen and colleagues suggests that Mgr1p might be such a candidate because deletion of resulted in a defect in Yme1p assembly (); however, Yme1p still assembled into a high molecular weight complex, in contrast to our studies in which Yme1p completely failed to assemble.
Our present study shows that Tim54p functions in a new pathway for assembly of Yme1p. Because Tim54p is a component of the TIM22 complex and Yme1p is imported via the TIM23 translocon, Yme1p, after import via the TIM23 pathway, requires an assembly activity provided by Tim54p of the TIM22 translocon for generating a functional complex. Thus, Tim54p functions to integrate the activities of the two major translocons of the mitochondrial inner membrane. Moreover, given that Yme1p is the -AAA protease involved in basic quality control of the inner membrane, Tim54p functionally links pathways of protein import, assembly, and turnover within the mitochondrion. Does this pathway function in other organisms? Tim54p homologues have been identified in other fungi, but close homologues have not been identified in higher organisms including worms, fly, mouse and humans. However, the SIMAP database (Similarity of Matrix Proteins, ; ) suggests that Tim54p is similar to a mitochondrial multi-substrate lipid kinase (identity 21%, similarity 41%; ), suggesting that Tim54p might have multiple functions. Additional investigations into the mammalian TIM22 pathway will be required to identify the true components.
Standard techniques were used for growth, manipulation and transformation of yeast strains (; ; ). Details on the strains used in this study are listed in Table S1. A 1,830-bp fragment containing the gene and its promoter was cloned into pYCPlac33, using the KpnI–XbaI sites, to form pC. The yeast strain deleted for () was constructed as previously described by plasmid shuffling (). The gene for was replaced with the gene flanked by the promoter and terminator regions by homologous recombination. The haploid yeast strain containing the disruption and pC was grown at 25°C on minimal media with 2% glucose and 5-fluoroortic acid (5-FOA) to select for loss of the plasmid. The was disrupted in strain BY4742 (ResGen) and the strain was generated by deletion of with in GA74. The Yme1-His strain was generated by integrating a hexahistindine tag in frame to the C terminus of Yme1 in strain GA74. The ts strains has been described previously and was isolated from the same study (; ). The strain was provided by Dr. Pfanner (University of Freiburg, Freiburg, Germany; ), and the Yta10-DHFR construct was provided by Dr. Langer (University of Cologne, Cologne, Germany; ). For in vitro transcription/translation, the DNA fragments encoding the substrates were cloned into pSP65 (Promega).
A ts strain was constructed using error prone PCR as described previously (). Amplified and gapped pRS314 (, ) were co-transformed into the strain. After auxotrophic selection, the mutant plasmid was selected by plasmid shuffling in the presence of 5-FOA. Resulting colonies were then screened for ts growth arrest at 37°C. The mutant plasmid was recovered and used to reconstruct the ts mutant, confirming that the ts phenotype was plasmid dependent. For microscopy experiments, plasmids containing Su-GFP fusion were transformed into the WT, , and strains.
Mitochondria were purified from yeast cells grown in YPEG at 25°C () and assayed for protein import as previously described (; ). Proteins were synthesized in a rabbit reticulocyte lysate in the presence of [S]-methionine after in vitro transcription of the corresponding gene by SP6 or T7 polymerase. The reticulocyte lysate containing radiolabeled precursor was incubated at 25°C with isolated mitochondria in import buffer (1 mg/ml bovine serum albumin, 0.6 M sorbitol, 150 mM KCl, 10 mM MgCl, 2.5 mM EDTA, 2 mM ATP, 2 mM NADH, and 20 mM Hepes-KOH, pH 7.4). Where indicated, the potential across the mitochondrial inner membrane was dissipated using 1 μM valinomycin and 25 μM FCCP. Non-imported radiolabeled precursor was removed by treatment with 100 μg/ml trypsin or 50 μg/ml proteinase K for 15–30 min on ice. Trypsin was inhibited with 400 μg/ml soybean trypsin inhibitor and proteinase K with 1 mM phenylmethylsulfonyl fluoride (PMSF).
For osmotic shock treatment to disrupt the mitochondrial outer membrane, the mitochondria were pelleted by centrifugation, suspended to 1 mg/ml in breaking buffer (0.6 M sorbitol and 20 mM Hepes–KOH, pH 7.4) and then diluted 20-fold with 20 mM Hepes-KOH, pH 7.4, and incubated on ice in the presence or absence of 10 μg/ml proteinase K. Mitoplasts were collected by centrifugation at 21,000 for 10 min. For alkali extraction, the mitochondria were pelleted by centrifugation, suspended to 0.1 mg/ml in 0.1 M NaCO, and incubated for 30 min on ice (). The membrane fraction was collected by centrifugation at 21,000 for 15 min.
The degradation assay for Yta10-DHFR was performed as described previously (; ). After import of Yta10-DHFR into mitochondria, non-imported precursor was removed by protease treatment. Mitochondria were then incubated at 37°C in the presence of an energy regenerating system (200 μg/ml creatine phosphokinase and 10 mM creatine phosphate) and aliquots were removed over a 60-min time period. Samples were subjected to alkali extraction and separated by SDS-PAGE. Data was collected using a BioRad FX Molecular Imager and the affiliated Quantity 1 software. Data from three independent assays was pooled and statistical comparisons were performed using SigmaStat 3 software (Jandel Corp.).
Mitochondria (2.5 mg/ml) were solubilized in 20 mM Hepes–KOH, pH 7.4, 50 mM NaCl, 10% glycerol, 2.5 mM MgCl, 1 mM EDTA, and 1% digitonin for 30 min on ice. Insoluble material was removed by centrifugation at 21,000 for 15 min. Solubilized proteins were analyzed by blue-native gel electrophoresis on the linear polyacrylamide gels as indicated in the figure legends (). After transfer to polyvinylidene fluoride membranes, proteins were detected by immunoblotting with the indicated primary antibodies. Detection was performed either using HRP-conjugated secondary antibodies and ECL (Pierce Chemical Co.) or [I]-protein A and autoradiography. Cross-linking reactions and immunoprecipitation assays were done as reported previously ().
Yeast strains were grown at their permissive temperature to mid-log phase in YPD and then incubated with 2 μg/ml DAPI for 15 min at room temperature (). The cells were then washed in water, immobilized on poly-lysine/concanavalin A coated coverslips, and mounted with the Prolong AntiFade kit (Molecular Probes). The cells were visualized at 25°C immediately on a Deltavision Spectrics Applied Precision model 52–000067-002 Olympus IX71 microscope. Images were taken at 0.2-μm steps through the sample using an oil immersion 100× NA1.35 objective. The fluorochromes were DAPI (visualized with the standard DAPI filter set), and GFP (visualized with the standard GFP filter set). Images were taken with the Photometric Coolsnap HQ camera using the Deltavision Softworx version 3.3.5 program. The images were processed with constrained iterative deconvolution at 10 iterations, using the software's proprietary algorithms. Representative sections containing mitochondria were converted to TIFF format and exported to Adobe Photoshop. In a time-course assay, cells were grown at 25°C to mid-log phase and then shifted to 37°C. Aliquots were removed at = 0, 12, 18, and 24 h after shifting to 37°C and immediately visualized. Nucleoids were counted in at least 20 separate cells. Ten focal planes were analyzed for each cell.
Table S1 lists the strains that were used in this study. The supplemental figures include additional control experiments for the results. Fig. S1 presents growth curves for the mutant strains. Fig. S2 analyzes cold sensitivity and show that mutants are cold sensitive, which was not rescued by overexpression. Fig. S3 shows the import of Hsp60 into [TIM22] mitochondria. Fig. S4 presents a growth analysis and shows that overexpression suppresses the petite-negativity of but not . Fig. S5 control reactions for the import of Yme1 (import is decreased, but not absent, in mutant mitochondria.) and localization of Yme1 during osmotic shock in WT mitochondria. Online supplemental material is available at . |
Nuclear movement in animal cells can be driven by pulling forces applied to astral microtubules (MTs) emanating from centrosomes bound to the nuclear envelope (). In the one-cell zygote, asymmetric cleavage is preceded by a series of stereotyped nuclear movements (; for review see ). Female and male pronuclei invariably meet in the posterior cytoplasm and move as a unit with the paired centrosomes to the cell center during mitotic prophase; this centration of the nucleus–centrosome complex (NCC) is accompanied by a 90° rotation that aligns the axis of the centrosome pair with the zygote's longer anteroposterior (AP) axis (). Shortly after centration and rotation are completed, nuclear envelopes break down, and the mitotic spindle forms.
Laser ablation experiments have revealed that during centration, centrosomes are pulled strongly toward the anterior and weakly toward the posterior (). These forces are likely transmitted by astral MTs, which extend from each centrosome to the cell cortex and are required for centration and rotation (for review see ). The molecular motors that pull MTs are unknown, as is the molecular basis of their attachment to the cell cortex. Cortically enriched filamentous actin (F-actin) could provide such anchorage; however, disruption of F-actin by cytochalasin D does not prevent NCC centration () or rotation (), leading to suggestions that cortical F-actin is unlikely to be involved (). The latter conclusion is surprising, as centrosome movement in other systems is clearly F-actin dependent ().
A related problem is how the forces that drive centration and rotation are spatially regulated. Although spatial regulation is not in principle necessary for centration (), the zygote is a polarized cell, and an asymmetrically distributed DEP domain–containing protein encoded by the gene is required for centration (; ). The LET-99 protein is enriched in a posterior cortical band coinciding with the initial position of the NCC before centration. In − mutant zygotes, centration fails, and the NCC undergoes a series of abnormally rapid and extensive oscillatory rocking movements. DEP domains have been implicated in the regulation of heterotrimeric G proteins (; ), with LET-99 regulating a redundant pair of Gα subunits called GOA-1 and GPA-16 (). These two Gα proteins promote forces that pull on spindle pole centrosomes during anaphase in the embryo (; ; ), as in other systems (for review see ), but their importance for centrosome movement at other stages is unclear. GOA-1/GPA-16 depletion in − mutant zygotes eliminates NCC rocking, indicating that during wild- type centration, LET-99 restricts excessive force stimulation by Gα signaling (). Strikingly, GOA-1/GPA-16 depletion also rescues the centration defect in − mutants, indicating that in the absence of LET-99, inappropriate Gα signaling promotes a net force on the NCC directed away from the cell center toward the posterior pole. However, GOA-1 and GPA-16 have not been shown to influence centration in wild-type zygotes.
In this study, we address two related questions about the control of NCC centration and rotation: the molecular basis of the driving force and the spatial regulation of this force. We show that Gα function is required for wild-type rates of centration and rotation and that the nonmuscle myosin II NMY-2 is required to a similar degree for both processes. Like Gα, actomyosin opposes centration in − mutants, suggesting that Gα and actomyosin act together to generate a force on the NCC whose direction is determined by LET-99. Finally, we show that cortical NMY-2–GFP aggregates anterior to the NCC move with an anterior bias in wild-type zygotes; this bias is strongly reduced in Gα-depleted zygotes and is reversed in − mutant zygotes, supporting our conclusion that polarized actomyosin contraction, which is promoted by Gα and spatially directed by LET-99, generates part of the force driving centration and rotation.
Centration, the movement of the NCC to the cell center, occurs during mitotic prophase in wild-type one-cell zygotes (Video 1, available at ). To determine whether the redundantly acting Gα proteins GOA-1 and GPA-16 contribute to centration, we used RNAi to deplete both simultaneously and measured the rate of NCC movement along the AP axis during centration (see Materials and methods section DIC video microscopy and analysis…). Although not absolutely required for centration, our measurements showed that Gα activity is required for the wild-type rate of centration ( and ). In wild-type zygotes, the speed of the NCC (v, plotted as incremental velocity in , with peak and mean velocity values in ) increased during the first half of centration to a peak value of 7.9 μm/min and a mean value of 6 μm/min followed by a gradual decrease to zero. In contrast to wild-type zygotes, v in Gα-depleted zygotes remained relatively constant during centration (), and the peak values of v were reduced relative to wild type by a factor of approximately two (4.2 μm/min [ test; P < 0.00001] compared with 7.9 μm/min in wild type; ). The time-averaged value of v was also reduced to 2.0 μm/min compared with 2.9 μm/min in wild type (P < 0.0001). These results suggest that GOA-1 and GPA-16 contribute to an anterior-directed net force on the NCC during centration. The lack of an absolute requirement could reflect a failure to fully deplete these proteins using RNAi or the presence of multiple additive force-generating mechanisms. Further depletion of GOA-1/GPA-16 results in sterility (see Materials and methods section strains and maintenance), complicating our ability to distinguish between these two explanations (see Discussion).
The molecular basis of force production downstream of Gα signaling during mitotic spindle positioning is unknown. We tested the role of NMY-2, a nonmuscle myosin II required for zygote polarity and cytokinesis (; ). A requirement for NMY-2 in centration has not been tested previously because the pronuclei meet near the cell center in most zygotes (). To assess the effect of NMY-2 depletion on the centration velocity profile, we examined zygotes in which fertilization occurred at a site near the oocyte meiotic spindle rather than the typical fertilization site opposite the meiotic spindle (see Materials and methods section strains and maintenance). In such cases of reversed fertilization, both pronuclei form near one pole and then move together to the cell center (; ). In comparison with + zygotes with reversed fertilization, NMY-2–depleted zygotes with reversed fertilization showed substantial reductions in both peak and time-averaged values of v (4.8 μm/min and 1.3 μm/min, respectively [P < 0.0001 for both], compared with 10.0 μm/min and 2.6 μm/min in +; ). As observed in GOA-1/GPA-16–depleted zygotes, v remained at a constantly low level throughout centration in NMY-2–depleted zygotes (), and the highest values of v were similar to those measured in GOA-1/GPA-16– depleted zygotes (P = 0.08; ). Still lower values of v were measured in NMY-2–depleted zygotes after normal fertilization, in which the NCC moved only a short distance after pronuclei met near the cell center ().
The reduced NCC velocity in zygotes could result indirectly from previously documented defects in cell polarity that occur as a result of NMY-2 depletion (; ). To address this possibility, we examined centration in zygotes, which exhibit extensive polarity defects (; ; ; ). In mutant zygotes with reversed fertilization ( and ), time-averaged (2.5 μm/min) and peak NCC velocity measurements (6.8 μm/min) were intermediate between wild-type and NMY-2–depleted zygotes (P < 0.0001 for both comparisons; ). We also examined NCC movement in normally fertilized zygotes, in which pronuclei met near the center of the zygote, at 58% egg length (EL; compared with 66% EL in wild type). The NCC moved toward the anterior at speeds similar to those observed after reversed fertilization (), invariably moving past the center into the anterior cytoplasm (unpublished data). Finally, the rate of NCC movement in − zygotes was reduced more than twofold by NMY-2 depletion to levels slightly below those measured in NMY-2–depleted wild-type zygotes (), indicating that NMY-2 does not act merely by restricting PAR-3 localization. We conclude that although PAR-3 makes a minor contribution to the rate of centration, the requirement for NMY-2 is at least partially independent of PAR-3 and cell polarity.
We next examined the rate of NCC rotation to determine whether actomyosin and Gα signaling also contribute to this movement. In wild-type zygotes ( = 30), the NCC rotated through an angle of 24 ± 11° during the 1-min period preceding nuclear envelope breakdown (NEB; see Materials and methods section DIC video microscopy and analysis…). Similar to the effects on centration, NCC rotation during this time interval was reduced by a factor of approximately two in both GOA-1/GPA-16–depleted and NMY-2–depleted zygotes (12 ± 6°/min for both; = 13 and = 9, respectively). We conclude that the same actomyosin- and Gα-dependent forces that promote centration also contribute to NCC rotation.
To further investigate the relationship between NMY-2, GOA-1/GPA-16, and NCC movement, we altered actomyosin function in − mutants, in which an abnormal GOA-1/GPA-16– dependent force prevents centration and causes excess NCC rocking (). Worms homozygous for a temperature- sensitive allele of that we isolated and named ts (see Materials and methods section strains and maintenance) produced zygotes (hereafter referred to as mutant zygotes) that exhibited an absence of centration and excessive NCC rocking at the restrictive temperature of 26°C (, A and B; and Video 2, available at ). Strikingly, RNAi-mediated depletion of NMY-2 in the ts mutant rescued centration () and reduced NCC rocking by more than fivefold ( and Video 3). In most NMY-2–depleted ts zygotes ( = 7), the pronuclei met posteriorly at a mean position of 63% EL, which is similar to wild-type (66% EL; = 10) and ts single mutants (67% EL; = 8). NCC position at the time of NEB in the NMY-2– depleted ts zygotes was significantly further anterior than in ts single mutants (51% vs. 63% EL; P < 0.0001) and was not significantly different from wild type (48% EL; P = 0.06). NMY-2 depletion also rescued centration and reduced NCC rocking in another − mutant, ts (unpublished data). We conclude that NMY-2, like GOA-1/GPA-16, is required for the abnormal forces that prevent centration and promote NCC rocking in − mutant zygotes.
To further assess the role of myosin II activity in the − phenotype, we inactivated the Rho-binding kinase called LET-502, which is required for wild-type levels of NMY-2 activity during cytokinesis and morphogenesis (; ). In embryos from homozygous tsts double mutant worms (see Materials and methods section strains and maintenance), centration was again rescued (), and NCC rocking was strongly reduced (Video 4, available at ; and unpublished data). Thus, Rho-binding kinase, presumably acting through NMY-2, is required to prevent centration and generate NCC rocking in the absence of LET-99. We also examined ts zygotes after the depletion of profilin (PFN-1), an F-actin assembly factor that is required for the accumulation of cortical F-actin in the zygote (). In PFN-1–depleted − mutant zygotes, the position of pronuclear meeting was highly variable, but, in each case, centration was rescued (). The rescue of centration was again accompanied by a marked reduction of NCC rocking compared with the ts single mutant (Video 5 and unpublished data). We also used RNAi to simultaneously deplete two components of the Arp2/3 complex, another F-actin assembly factor required for cortical stability in the early embryo and for gastrulation but not for the myosin II–dependent processes of cell polarization and cytokinesis (). Depletion of Arp2/3 in − zygotes destabilized the cell cortex but did not restore centration (unpublished data), suggesting that the abnormal forces opposing centration in − mutant zygotes are mediated specifically by profilin-dependent F-actin in association with NMY-2 but are not affected by a different perturbation of the F-actin cytoskeleton. We conclude that in the absence of LET-99 function, actomyosin-dependent contractile forces oppose centration and contribute to excessive NCC rocking.
Because myosin II activity opposes centration in − mutants, we reasoned that overactivating NMY-2 might drive the NCC further toward the posterior pole. To test this prediction, we used RNAi to deplete MEL-11, the orthologue of MYPT (myosin phosphatase-targeting subunit), a protein phosphatase that inhibits NMY-2 activity (). Strikingly, MEL-11 depletion in ts mutants caused the NCC to move toward the posterior in seven out of eight cases ( and Video 6, available at ). To our surprise, the depletion of MEL-11 also reduced NCC rocking by more than threefold (), suggesting that the amount of rocking is not a simple function of NMY-2 activity.
To summarize, we conclude that although NMY-2 normally acts to promote centration, loss of LET-99 reverses the mechanical output of NMY-2 activity to oppose centration. To account for the suppression of NCC rocking by either the depletion or overactivation of NMY-2, we speculate that the misdirected actomyosin-dependent force in − mutants is counteracted by an independently generated centration-promoting force and that a balance of roughly equal but oppositely acting forces promotes excess rocking (see Discussion).
Seeking to understand how NMY-2 contributes to NCC movement, we used spinning disc confocal microscopy to image cortical NMY-2–GFP in transgenic zygotes during centration. In agreement with previous studies (; ), we found the onset of centration to coincide with a sudden reorganization of cortical NMY-2. At the time of pronuclear meeting in wild type, an anteriorly enriched cortical network of relatively large actomyosin bundles is replaced by a more finely reticular F-actin network in which NMY-2 is focused in small, isolated aggregates. Both F-actin and NMY-2 localize most densely to a cap covering the anterior half of the zygote, and a separate, less dense cap of actomyosin also is observed over the posterior pole. The lowest density of NMY-2 aggregates is consistently found in an intermediate zone (50–75% EL) that encircles the position of the NCC at the beginning of centration. Interestingly, the highest NMY-2 density is always found at the boundary between this intermediate zone and the anterior cap; until a late stage of centration, this myosin-dense ring is several micrometers anterior to the advancing NCC (, Fig. S1, and Video 7, available at ). Although aggregates appeared to be highly mobile, the overall NMY-2 distribution pattern persisted past the end of centration. Thus, cortical actomyosin is well positioned to play a role in generating forces that pull centrosomes strongly toward the anterior and weakly toward the posterior (). Although the relative enrichment of NMY-2 in the anterior cap at NEB was slightly reduced in Ga-depleted and − mutant zygotes (Fig. S1), the significance of this is unclear. Our finding that centration occurs at a near wild-type rate in zygotes, in which NMY-2 aggregates are distributed uniformly throughout the cortex (), suggests that a polarized distribution is unnecessary for the action of NMY-2 in centration. Therefore, rather than focus on the overall distribution of NMY-2, we examined the behavior of individual NMY-2 aggregates for clues to actomyosin's mechanical role.
If cortical actomyosin is directly involved in generating a pulling force on centrosomes, the mechanism of force generation should be reflected by the movements of NMY-2 aggregates. For example, pulling force might be entirely generated by an MT-based motor such as dynein, with actomyosin simply providing cortical anchorage for the motor. In this case, every component of the cortical attachment site, including NMY-2, should tend to move toward the centrosomes if at all. Alternatively (or additionally), NMY-2 itself could itself act as a force generator by pulling the actomyosin cortex and attached MT plus ends away from centrosomes. This model predicts that NMY-2 aggregates in at least some region anterior to the NCC will tend to move away from the NCC during centration. To test these ideas, we used automated particle tracking () to measure displacements of brightly labeled cortical NMY-2 aggregates in the cortical plane of wild-type zygotes ( = 10) during the early stage of centration (; and Video 8, available at ; see Materials and methods section Particle tracking for details of tracking procedure).
To detect local behaviors of NMY-2 aggregates, we divided the region of the cortical plane from 25 to 75% EL into eight subregions (octants), which were defined with respect to the cell poles and the direction of NCC rotation. Each octant comprised a lateral half of one AP segment with length equal to 12.5% EL (6–7 μm), and the zygotes were oriented such that NCC rotation was clockwise. This convention positioned centrosomes initially near the anterior edge of the posterior-most pair of octants, with the presumptive anterior spindle pole centrosome subsequently swinging through the full breadth of octant g ().
We first asked whether NMY-2 aggregate movement in any region of the cortex is directionally biased. For every cortical octant, we obtained the frequency distribution of all displacement sizes at a given time scale projected onto the AP axis () or the transverse axis (). At short time scales (0.25–2 s), these distributions resembled smooth Gaussian curves with peaks very close to zero, indicating that random thermal forces influence NMY-2 aggregate movement (unpublished data). At longer time scales (4–16 s), the appearance of directed nonrandom movements along one or both axes became increasingly prominent in some octants (Fig. S2, available at ).
At the time scale of 12 s, wild-type NMY-2 aggregate displacements throughout much of the cortex were biased toward the anterior (). This bias was evident in the major peaks of displacements in the central octants (b, c, f, and g) near and anterior to the NCC. Large displacements (>1 μm/12 s) were biased toward the anterior on the side of the presumptive anterior spindle pole centrosome (octants e, f, and h) and toward the posterior on the opposite side (octants a–c), suggesting a positive correlation between large displacements of NMY-2 aggregates and the angular displacement of the nearest centrosome.
The correlated movements of NMY-2 aggregates and the NCC suggested a mechanical link. To test the possibility that NMY- 2 aggregate movement is driven by NCC movement, we blocked centration and rotation by inactivating ZYG-9, a protein required for astral MT stability and, consequently, for centrosome movement. In ts mutant zygotes ( = 8), centration and rotation failed as reported previously (), but NMY-2 aggregates in most octants still showed a pronounced anterior bias in peak displacement (). In octants c and g, large displacements in both directions occurred more frequently in the − mutant than in wild type; in octants b and f, large displacements occurred more frequently toward the posterior. We conclude that NMY-2 aggregate motility shows an intrinsic anterior bias throughout much of the cortex that is independent of aster movement. Interestingly, astral MTs appear to restrict large movements of actomyosin in the equatorial region of the cortex, suggesting that astral MTs and cortical actomyosin are indeed linked.
We also examined NMY-2 aggregate displacements along the transverse axis. In wild type, large displacements (>1 μm/12 s) occurred much more frequently along the transverse axis than along the AP axis (). In every octant, these large displacements were biased in the upward direction, which is coincident with the trajectory of the forward- and upward-moving anterior centrosome. Such biased transverse displacement was strongly reduced in the − mutant, suggesting that transverse aggregate movements are normally stimulated by NCC centration/rotation or by some other function of astral MTs.
In Gα-depleted and − mutant zygotes, NMY-2 was similarly localized during prophase to a large, dense anterior cap and a small, sparse posterior cap (Fig. S1, Videos 9 and 10, available at ; and unpublished data). Although the relative density of NMY-2 in the anterior cap was slightly reduced in both mutants, we thought it was unlikely that LET-99 and Ga exert their disparate influences on centration through similar requirements for polarized NMY-2 enrichment. Instead, a slightly flattened profile of NMY-2 intensity could result in each case from an abnormal pattern of NMY-2 aggregate displacements. Therefore, we examined these displacements in ( = 9) and ts zygotes ( = 11).
In striking contrast to wild type, the AP displacements in cortical regions near and anterior to the NCC in Gα-depleted zygotes were distributed symmetrically at about zero, suggesting random movement (). However, in the two posterior octants (d and h), the major peaks were biased toward the posterior, indicating that actomyosin movement is not completely randomized by GOA-1/GPA-16 depletion. Also reflecting nonrandom behavior, large displacements (>1 μm/12 s) were prominently biased toward the NCC in regions both to its anterior (b and f) and its posterior (d and h; see Discussion).
A different abnormal pattern of biased displacement along the AP axis was detected in − zygotes (). In all four central octants (b, c, f, and g), displacements showed a characteristically abnormal profile, with peaks offset further toward the anterior than in wild type. In three of these four central octants, however, the largest displacements (>1 μm/12 s) were strongly biased toward the posterior (to a more subtle degree, this bias was also present in the two anterior-most octants, a and e). An abnormally polarized trend in large displacements occurred in octants b and f: large displacements occurred more frequently toward the posterior and more rarely toward the anterior compared with wild type. The resulting degree of asymmetry was beyond any found in wild-type or Gα-depleted zygotes, although similar profiles of anteriorly biased small and posteriorly biased large displacements were found in −
Surprisingly, the inactivation of LET-99 caused a striking reduction in both the magnitude and directional bias of transverse displacements (). Similarly narrow and symmetrical displacement profiles were measured in most regions of Gα- depleted zygotes, with the two posterior octants again showing distinctively nonrandom trends. Thus, both GOA-1/GPA-16 and LET-99 seem to play essential roles in generating the directionally biased movement of NMY-2 aggregates along the transverse axis.
For each population of aggregates, we measured the projection of the MSD onto the AP and transverse axes. In all wild-type and mutant/RNAi conditions, mobility in some octants increased and/or decreased over time scales ranging from 4 to 16 s (), suggesting that diverse factors or conditions locally modify NMY-2 aggregate mobility. Interestingly, throughout most of the regions examined, mobility over time scales of 1–2 s was higher in wild type than in the corresponding octants of Gα-depleted or − zygotes, suggesting that wild-type aggregates move relatively fast and/or in a more directed manner at these short time scales. At longer time scales as well, wild-type NMY-2 aggregates in most regions of the cortex showed relatively high AP mobility, which is consistent with a normal predominance of directed transport that is lacking or constrained in the absence of Gα or LET-99.
Both short-time and long-time mobilities varied substantially with position, being highest in wild type near the NCC (octants c and g) and lowest in the anterior octants (a and e), with more than a twofold difference between these regions. In general, mobility along the AP axis was markedly diminished in Gα-depleted zygotes, with the exception of the most posterior region, where asymmetric displacement peaks had been observed (). Together, these data suggest that GOA-1/GPA-16 are required for directed movement in anterior regions, but only for the spatial regulation of directed movement in the posterior. In − zygotes, an interesting asymmetry was detected: in anterior regions, mobility did not increase over time but rather decreased over time scales of 10–15 s. This time-dependent decrease was expected as a reflection of oscillatory cortical flows that occur in approximate synchrony with NCC rocking (unpublished data). Strikingly, mobility in posterior regions of − zygotes tended to increase over time, showing only a slight decrease at the time scale of NCC oscillation. This regional difference in aggregate mobility could account for the abnormal distribution of NMY-2–dependent forces that act on the NCC in − mutant zygotes (see Discussion).
Transverse mobility increased over time in most regions of the wild-type cortex, but the pattern of regional variation was roughly opposite that of AP mobility, indicating local preferences for aggregates to move along one axis or the other (). Strikingly, time-dependent mobility increases along the transverse axis were largely absent in both −− and − zygotes. Together with the symmetrically distributed displacements along the transverse axis in both conditions (), this result suggests that the directionally biased active transport of NMY-2 aggregates along the transverse axis depends on both GOA-1/GPA-16 and LET-99.
We have shown that the Gα proteins GOA-1/GPA-16 and the myosin II NMY-2 act similarly to promote NCC centration and rotation in wild type and to oppose centration and provoke excessive NCC rocking in − mutants. In addition to NMY-2, the F-actin assembly factor PFN-1 and the Rho-binding kinase LET-502, which activates NMY-2 by phosphorylating its regulatory light chain, are also required to prevent centration in the absence of LET-99. Moreover, depletion of the inhibitory myosin II light chain phosphatase MEL-11 in a − mutant promotes posterior movement of the NCC. Based on these findings, we propose that the Gα proteins, acting either through or in parallel to myosin regulatory light chain phosphorylation, activate an actomyosin-dependent force that promotes NCC movement and that the direction of this force is regulated by LET-99 (). Our findings also suggest that the actomyosin-dependent force acts in parallel with a separate, unidentified centration-promoting force that is independent of actomyosin, Gα, and LET-99. Finally, we have shown that elements of the actomyosin cortex undergo intrinsically polarized movement that could directly generate a pulling force on centrosomes, and we propose that Gα and LET-99 act by regulating this movement.
Our findings challenge the view that actomyosin is not required for centrosome movement in the early embryo (see Introduction). Analysis of direct roles of actomyosin in centrosome positioning in the embryo has previously been complicated by an earlier requirement for the polarization of cortical PAR (partitioning) proteins, which, in turn, are thought to act upstream of Gα and LET-99 to control centrosome movement (; ). However, F-actin appears to play PAR-independent roles in the Gα-dependent processes of anaphase spindle pole flattening and NCC rotation (). Our finding that NMY-2 promotes centration in zygotes likewise indicates a direct role for actomyosin and further suggests that centration is independent of cortical polarity, as neither NMY-2 nor LET-99 is asymmetrically distributed in zygotes at this stage (; ). In contrast, NCC rotation appears to depend on PAR-3 as well as on Gα, LET-99, and NMY-2 ().
N2 Bristol was used as the wild-type strain and was maintained according to . Alleles listed by chromosome number were used as follows: tsts I; ts II; tsts III; IV; V; and X. mutant embryos were obtained by picking long adults from the strain KK571 ( III). To visualize cortical NMY-2–GFP, we used the strain JJ1473 ( III; ++ V). NMY-2–GFP localization throughout early cleavage in live embryos is indistinguishable from the immunolocalization of endogenous NMY-2 ().
ts and ts were isolated in screens for temperature-sensitive embryonic-lethal mutations affecting early cell divisions (). Both were outcrossed to N2 males through six generations before phenotypic analysis. In both mutants, the defective centration phenotype was strongly expressive immediately after shifting worms from 15°C to room temperature, suggesting that the mutant proteins rapidly inactivate. ts and ts mutant worms were maintained at 15°C and shifted to 26°C for at least 1 h before beginning experiments. Broods of embryos produced by hermaphrodites homozygous for either ts or ts were nearly 100% viable at 15°C and >99% lethal at 26°C. Both alleles were mapped to linkage group IV by mating with strains carrying visible mutations. ts is strictly recessive: 914/945 (97%) embryos from heterozygotes grown at 26°C hatched. ts may be weakly dominant: 818/898 (91%) embryos from heterozygotes grown at 26°C hatched. ts failed to complement a known recessive and nonconditional allele, (), and further failed to complement ts. To identify the molecular lesions in ts and ts, genomic DNA encompassing the K08E7.3 locus was amplified from each mutant by PCR using primer sequences not found in the paralogous locus F55H2.4 Sequences from pooled PCR reactions were obtained using a genetic analysis sequencer (CEQ 800; Beckman Coulter) at the University of Oregon Sequencing Facility and were compared with published wild-ype sequences as well as the background strain used for mutagenesis, CB1309 ( X) Sequence analysis revealed that ts carries two missense mutations at the locus: a T®A transition that converts leucine 409 to histidine and a G®A transversion that converts arginine 515 to histidine. ts also carries two missense mutations in : a G®A transversion that converts aspartate 371 to asparagine and a T®C transversion that converts methionine 405 to threonine. None of these four mutations was present in CB1309. The phenotypes of ts or ts are indistinguishable from other − mutants described previously ().
A double mutant strain of the genotype ts I;ts IV was constructed by mating ts males with ts hermaphrodites. The mutation confers a visible phenotype with fewer oocytes than normal present in the hermaphrodite uterus; from F2 self-progeny that were Unc with few oocytes, we recovered a strain homozygous for both ts and ts. Double mutant worms were maintained at the restrictive temperature of 26°C for at least 6 h before collecting embryos. We confirmed the presence of the ts mutation by outcrossing the double mutant strain with N2 males to recover an ts strain that was wild type at the locus.
To obtain wild-type embryos with reversed fertilization, ts hermaphrodites were shifted to the restrictive temperature of 26°C for several hours and were mated with wild-type males ().
All RNAi was performed by feeding worms on lawns of RNase III–deficient –expressing double-stranded RNA from inducible promoters. Most bacterial strains were obtained from the library of and distributed by Geneservice Ltd. The double RNAi feeding strain was provided by P. Gonczy (Swiss Institute for Experimental Cancer Research, Lausanne, Switzerland; ). Bacterial lawns were grown overnight at 37°C on feeding RNAi plates containing 1 mM IPTG + 75 μg/ml ampicillin () from ∼200 μl of a twofold concentrated stationary-phase liquid culture (uninduced) and were stored for up to 2 wk at 4°C. Worms were transferred to feeding RNAi plates as suspensions of unfed L1 hatchlings and grown at 15°C until adulthood. With the exception of , all RNAi treatments caused a proportion of animals to become sterile and produce no eggs. In cases in which treatment from L1 produced sterility, worms were plated at later stages of development. All RNAi treatments were predicted to affect phenotypic traits other than NCC movement; the presence of these phenotypic markers in zygotes or their siblings was used to assess the efficacy of gene inactivation. For example, to ensure that NMY-2 was strongly depleted, we collected data only from zygotes showing severe defects in the NMY-2–dependent processes of cytokinesis and interphase cortical contractility. In the case of double RNAi, zygotes were used whose older siblings showed large multinucleate cells, which is a result of cytokinesis defects during late cleavage that indicates the effective depletion of both GOA-1 and GPA-16.
Zygotes were dissected out in M9 salts, placed on 3% agarose pads, and overlaid with a coverslip, under which was placed enough M9 to delay evaporation from the agar pad. For Nomarski differential interference contrast (DIC) microscopy, 63× 1.4 NA oil immersion objectives (Carl Zeiss MicroImaging, Inc.) were used with microscopes (Axioskop; Carl Zeiss MicroImaging, Inc.). DIC images were collected at room temperature (23 ± 1°C) at frame rates of 0.1–2.0 Hz using a digital video camera (307,200 pixels; CCDX; Dage-MTI) connected to a frame-grabber card controlled by a computer (G4; Macintosh). The frame-grabber software was Image 1.62 (Scion) run on an OS9 operating system (Macintosh). These image sequences were collected as TIFF stacks and analyzed using ImageJ (National Institutes of Health [NIH]) and Excel (Microsoft). Image stacks were rotated to align the zygote's AP axis horizontally with anterior to the left. The NCC position, relative to the zygote's anterior pole at the time of NEB, was tracked at 10-s intervals beginning just before pronuclear meeting and terminating at NEB. The center of the NCC (midpoint between the diametrically opposed centrosomes) was approximated by eye. NCC velocity was calculated for each 10-s interval. The mean velocity and SD for each genotype/RNAi treatment were computed for the period between pronuclear meeting and NEB. To estimate peak velocities, incremental velocity measurements for each zygote of a given genotype/RNAi treatment were sorted in descending order; the three highest values obtained for each zygote were selected, and the mean of these values was computed for all zygotes. Net NCC rotation in wild-type, , and zygotes was measured during the 60 s preceding NEB. NCC angular position was measured as the angle of a line connecting the two centrosomes relative to the AP axis, and the change in this angle over the 60-s period was calculated. NCC oscillations in − single and double mutants were also measured during the 60 s preceding NEB, when the NCC in − single mutants is abnormally rotating in place. At each reversal of NCC rotation, angular position was measured as the angle of a line connecting the two centrosomes relative to the AP axis. To compare any two sets of measurements, we used the test (two-tailed, assuming unequal variance).
To quantitate cortical NMY-2–GFP intensity, we imaged the entire thickness of the zygote cortex in a z-series of eight images 0.2 μm apart and used ImageJ to make a two-dimensional projection by summing the signal at each point. We selected a rectangular area corresponding to the median 50% of the zygote (25% egg width on each side of a line corresponding to the AP axis) and plotted the distribution profiles of fluorescence intensity at pronuclear meeting and NEB by adding the signal at each point along the normalized length of the AP axis. To eliminate variation arising from curvature of the cortex at the cell poles, we analyzed only a segment from 0.2 to 0.8 EL. For each zygote, relative distribution of NMY-2–GFP along the AP axis was calculated as the fraction of total cortical signal at each point along the axis. Data presented in Fig. S1 A are mean values for each genotype/RNAi condition. A test (two-tailed, assuming unequal variance) was used to compare values between genotypes at each axial position. To simultaneously visualize cortical NMY-2–GFP and centrosomes (Fig. S1 B), we collected a cortical fluorescence z-series and a DIC image of the NCC every 10 s. Fluorescence z-series corresponding to each time point were projected as a sum, and stacks were normalized for EL. Each stack was cropped to a median 50% of egg width, and then the reslice command was used to generate a three-dimensional kymograph. This was flattened by projecting the average (mean) of all slices, thus reducing the AP axis to a row of pixels and generating a kymograph that showed change in the AP intensity profile over time. All kymographs were autoadjusted for brightness and contrast, aligned with reference to the time of NEB, and averaged. Corresponding DIC stacks were likewise normalized for EL, and the centrosomes were marked at each 10-s time point; these marked stacks were used to generate kymographs using the same method as used for fluorescence.
To quantify the movements of actomyosin aggregates, we measured NMY-2–GFP displacements occurring at time scales ranging from 0.25 to 16 s. (C and D) shows the trajectories of all cortical NMY-2–GFP aggregates in a wild-type zygote imaged over 60 s using 1- and 12-s time scales. The trajectory of each aggregate is described by a single track; sampling the aggregates' displacements at time intervals of 1 s (time scale = 1 s) will yield a set of displacements (in this case, up to 59; shown in ). Sampling at a longer time scale (12 s) yields a smaller number of displacements of more widely distributed magnitude (). Importantly, sampling at this longer time scale reveals directional trends in aggregate movement that may not be apparent at shorter time scales. Note that from the same set of trajectories, a different set of displacements over a 12-s interval would be obtained by shifting the initial sampling time point. Therefore, when measuring displacements over a time scale longer than the acquisiton frame rate, the most complete description is obtained by sampling all possible overlapping time frames in each image sequence. Thus, we extracted all displacements at a given time scale and analyzed frequency distributions of displacement size and direction.
Zygotes were observed using DIC optics until the time of pronuclear meeting; the objective was then focused on the upper cortex, and images of a single focal plane were collected at intervals of 0.25–1 s beginning 30 s after pronuclear meeting. After recording for 30–60 s, we determined the direction of NCC rotation and confirmed that the cell successfully divided and that its daughters reentered interphase. Although we could not simultaneously record movements of NMY-2–GFP aggregates and the NCC, we found that during the longer 60-s period of data collection, the NCC in wild-type zygotes moves anteriorly a mean distance of 12 ± 2% of EL and rotates through a mean angle of 28 ± 14° ( = 25). In Gα- depleted zygotes, the NCC moves 5.5 ± 2% of EL toward the anterior and rotates through 14 ± 13° ( = 16). In ts zygotes, the NCC moves 1 ± 2% of EL toward the anterior and changes its direction of rotation four to five times ( = 16).
Image sequences were opened using ImageJ, background was subtracted, and the stack was rotated to align the zygote horizontally with anterior to the left before saving as a series of 8-bit TIFF images. Further image processing and analysis were performed using IDL (Research Systems, Inc.). NMY-2–GFP aggregates were tracked as described previously (; ; ) The first image of the sequence was imported and digitally filtered using a Gaussian filter of 11-pixel (0.7 μm) diameter. A set of features was defined () corresponding to the brightest regions distinguishable through the Gaussian filter; most of these were too faint to be discerned by eye and were removed by selecting for brightness. The remaining set of features was overlaid on the original image to confirm identity with observed NMY-2–GFP aggregates. The features identified were compiled into a time-dependent trajectory that linked features corresponding to the same aggregates in successive frames. Trajectories were plotted on the x and y coordinates to visualize aggregate displacements (). Histograms of aggregate displacements as a function of time scale were constructed from the aggregate trajectories. Distributions of aggregate displacements were analyzed at the time scale of 12 s. This choice represented a compromise between the objectives of detecting biologically driven behaviors (increasingly prominent with increasing time scale) and maintaining a representative dataset: at longer time scales, considerably fewer data points were available from a subset of recordings that lasted 30 s.
MSDs were calculated as a function of the time scale τ from the aggregate trajectories according to the method outlined by . The two orthogonal projections of the MSD—parallel and perpendicular to the AP axis—are given by the formulas and , respectively. The angle brackets indicate that the x- and y-component MSDs are averaged over all aggregates within a designated spatial region (i.e., an octant) and over all possible starting times () of the image frame sequence.
Fig. S1 quantitatively shows the spatial distribution of cortical NMY-2–GFP in wild-type, , and ts at the times of pronuclear meeting and NEB (A). It also shows for each genotype a time course of averaged NMY-2–GFP intensity profile over the AP axis during 240 s before NEB together with a time course of centrosome positions in the same zygotes (B). Figs. S2–4 show histograms of NMY-2– GFP aggregate displacements at time scales of 4, 8, 12, and 16 s for wild type (Fig. S2), −− (Fig. S3), and − (Fig. S4). Videos show wild-type centration (Video 1), defective centration in − and its suppression or enhancement after actomyosin manipulation (Videos 2–6), wild-type distribution of cortical NMY-2–GFP from pronuclear migration through cytokinesis (Video 7), and high speed image sequences used for particle tracking analysis of cortical NMY-2–GFP during centration in wild-type, −−, and − zygotes (Videos 8–10). Online supplemental material is available at . |
Cell polarity creates unique functional domains within cells, exemplified by axons and dendrites of neurons and the apical and basolateral surfaces of epithelial cells (; ; ). Cells generate polarity, in part, through directed transport of secretory vesicles and mRNAs by cytoskeleton-based motor proteins (; ; ; ; ). Motor proteins involved in the transport of vesicles and mRNAs have been identified, and much is known about how these motor proteins move along actin filaments and microtubules in vitro (). Less clear is how in vitro motility translates into long distance transport of cargo in vivo. In particular, the obvious physical differences between cargo, such as membrane-bound vesicles and RNAs, would seem to necessitate different mechanisms to generate sustained transport in vivo.
To address this question, we analyzed two class V myosins in the budding yeast : Myo2p and Myo4p. Together, the two myosins transport most of the cellular material into the bud (). Myo2p is required for cell viability and is essential for the transport of secretory vesicles into the bud to facilitate growth (). Myo2p also transports other membrane-bound cargo, including vacuoles, late Golgi, peroxisomes, and mitochondria (; ; ; ; ). In contrast, Myo4p is nonessential, but its primary biological function is to transport mRNA to the bud tip, generating cell fate differences between mother and daughter cells (; ; ). Myo4p has also been implicated in the movement of cortical ER into the bud (). Myo2p and Myo4p are closely related and are descended from a common ancestral gene, but they cannot functionally substitute for each other, indicating that they have evolved to transport different types of cargo.
To move cargo over appreciable distances, motors must bind tightly to membranes and nucleic acids. Myo2p binds membrane-bound vesicles and organelles through separate domains in its globular C-tail. For example, amino acids 1389–1491 facilitate association with secretory vesicles, whereas amino acids 1131–1345 mediate association with vacuoles (; ). These sequences do not appear to bind directly to membranes; rather, protein intermediaries transiently link Myo2p to different membrane-bound cargo. For example, Vac8p and Vac17p mediate the binding of Myo2p to vacuoles, whereas Sec4p likely facilitates the binding of Myo2p to vesicles (; ). The interaction between Myo2p and vacuoles is dissolved by the proteolysis of Vac17p (). In contrast, Myo4p binds mRNA through She3p (). Myo4p directly binds the N terminus of She3p, and the C terminus of She3p associates with mRNA (). Another RNA-binding protein, She2p, facilitates a specific interaction between She3p and sequences in mRNA called zipcodes that are sufficient to localize an RNA (; ).
In , vesicles and mRNA move rapidly and continuously into the bud. Myo2p moves secretory vesicles at rates around 3 μm/s, and Myo4p transports RNA at 0.25–0.4 μm/s (; ). How Myo2p and Myo4p facilitate long distance transport in vivo is unclear, as an initial analysis of Myo2p and Myo4p motility indicated that both are nonprocessive (). Processive motors have a high probability of remaining associated with a filament through many enzymatic cycles, allowing them to move considerable distances before falling off a filament. In contrast, nonprocessive motors have a low probability of maintaining an association with a filament during their enzymatic cycles, preventing individual nonprocessive motors from moving along a filament. Groups of nonprocessive motors bound to the same cargo might generate sustained movement by providing multiple attachments to a filament, thereby increasing the probability that the cargo remains associated with a filament. Vesicles and organelles contain ample surface area to bind multiple motors, and recent work has shown that vesicles and some organelles use multiple motors to generate bidirectional transport, as well as increase the distance of sustained movement (; ; ). Whether a similar mechanism works for mRNA transport is unclear. mRNA contains four zipcodes, but a single zipcode is sufficient to generate continuous transport of an RNA to the bud tip (). Moreover, active zipcodes can be as small as 25 nucleotides, clearly limiting the number of Myo4p motors that could bind (; ). The spatial restriction of zipcode suggests that Myo4p likely utilizes a different mechanism than Myo2p to generate continuous movement of RNA.
Here, we have compared the structure and motility of Myo2p and Myo4p and discovered two substantial differences. First, Myo4p is a monomeric motor, whereas Myo2p, like all other class V myosins, is a dimer. Second, the in vitro motility of Myo2p and Myo4p were very different. In an actin filament gliding assay, as Myo2p density decreased, filament velocity decreased, indicative of a nonprocessive motor. In contrast, as Myo4p density decreased, filament velocity increased, suggesting that Myo4p is more efficient in smaller groups. Single- molecule analysis demonstrated that Myo2p motors are weakly processive, and although individual Myo4p motors are nonprocessive, ensembles of Myo4p can move processively. To determine whether the motor domain of Myo4p is responsible for the unique motility of Myo4p, we constructed chimeras of Myo2p and Myo4p. We found that a Myo2p-motor/Myo4p-tail chimera (Myo2/4p) is a monomer with motility similar to Myo4p, whereas a Myo4p-motor/Myo2p-tail chimera (Myo4/2p) is a dimer with motility similar to Myo2p. Moreover, Myo2/4p restored localization of mRNA to the bud tip in cells, and Myo4/2p restored vesicle transport in a mutant. Our results suggest that the monomeric structure of Myo4p is important for regulating its motility and transporting mRNAs in vivo.
Class V myosin heavy chains are predicted to dimerize through coiled-coil interactions in their proximal tail domain (). For vertebrate myosin Va, dimerization allows the two motor domains to coordinate their enzymatic cycles to produce fast and processive movement along actin filaments (, ; ; ). The pair-coil algorithm predicts a robust region of coiled-coil in the myosin Va C-tail, and a similar stretch of coiled-coil is found in the C-tail of Myo2p (). In contrast, Myo4p contains a limited stretch of amino acids (aa 990–1027) predicted to form a coiled-coil, and such a short coiled-coil domain may be insufficient to facilitate dimerization of Myo4p heavy chains (). Because of the importance of dimerization in motor function, we investigated whether Myo4p exists as a monomer or dimer.
We first tested for Myo4p dimers using a coimmunoprecipitation assay. Two versions of Myo4p with different C-terminal tags were expressed in the same cell. was integrated and under the control of its native promoter, whereas was expressed from a low copy CEN/ARS plasmid. Anti-GFP antibodies were used to immunoprecipitate Myo4p-GFP, and the amount of Myo4p-HA in the immunoprecipitates was determined by Western blot. We did not detect Myo4p-HA in immunoprecipitates of Myo4p-GFP, suggesting that the two heavy chains do not interact. (). As a control, we probed the Myo4p-GFP immunoprecipitates for She3p-HA, which binds directly to Myo4p. She3p-HA was readily detected in Myo4p-GFP immunoprecipitates (), indicating that most Myo4p heavy chains associate with She3p, but not with another Myo4p heavy chain.
To demonstrate that the coimmunoprecipitation assay can detect myosin dimers, we analyzed Myo2p. or were tagged at their C termini with a 1/2TAP module (TEV cleavage site and IgG-binding domain) and coexpressed with or , respectively. Myo2p-1/2TAP or MYO4p-1/2TAP were adsorbed from cell extracts onto IgG resin, and the amount of Myo2p-HA or Myo4p-HA in the immunoprecipitates was determined by Western blot. Myo2p-HA was clearly present in the immunoprecipitates of Myo2p-1/2TAP, but we failed to detect Myo4p-HA in the immunoprecipitates of Myo4p-1/2TAP (). We achieved similar results using other combinations of tagged Myo2p and Myo4p (unpublished data). The results demonstrated that the coimmunoprecipitation assay can detect myosin dimers, and therefore, our inability to coprecipitate two Myo4p heavy chains suggests that Myo4p does not form dimers in cell extracts.
One limitation of the coimmunoprecipitation assay is that weak or transient Myo4p dimers may dissociate during the assay. To capture these potential dimers, we added the zero-length chemical cross-linker 1-Ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride (EDC) to yeast lysates, and then determined by Western blot whether EDC increased the molecular weight of Myo4p to a size consistent with a dimer. We performed the cross-linking in cell lysates from cells because GFP increases the molecular weight of the Myo4p heavy chain to ∼15 kD more than Myo2p. Therefore, a cross-linked Myo4p dimer should be larger and migrate more slowly than a cross-linked Myo2p dimer. cell lysates were incubated with or without 50 mM EDC for 1 h at room temperature. Subsequent Western blot analysis revealed that EDC-treated Myo4p-GFP migrated as a larger complex compared with untreated Myo4p-GFP (). Reprobing the same blot for Myo2p revealed that EDC-treated Myo2p showed a dramatic increase in molecular weight compared with untreated Myo2p (). Importantly, EDC-treated Myo4p-GFP was smaller than EDC-treated Myo2p, indicating that the cross-linked Myo4p-GFP complex was not Myo4-GFP dimers. Instead, the EDC-treated Myo4- GFP complex probably reflects Myo4p-GFP cross-linked to She3p. In support of this conclusion, She3p was detected at the same location as EDC-treated Myo4p on Western blots (). Second, in cells, EDC-treated Myo4p-GFP did not exhibit a considerable size shift, whereas in the same extracts EDC-treated Myo2p still migrated as a large complex ( and unpublished data). Thus, cross-linking failed to capture Myo4p dimers and indicated that the bulk of Myo4p in cell extracts exists as a monomer coupled to She3p.
To test further the oligomeric state of Myo4p, we partially purified native Myo2p and Myo4p from cell extracts and performed velocity sedimentation analysis. Myo2p-1/2TAP and Myo4p-1/2TAP were adsorbed onto IgG resin from cell lysates and then eluted with TEV protease. Both eluates were enriched for either Myo2p or Myo4p and showed robust activity in an actin filament gliding assay (; and Videos 1 and 2, available at ). The TEV eluates were analyzed by velocity sedimentation on 5–20% sucrose gradients. Silver stain and Western blot analysis of the sucrose gradient fractions revealed that Myo2p sedimented as a larger complex than Myo4p (). Based on the sedimentation of four protein standards, we estimate sedimentation values of 9.1 ± 0.6S for Myo2p and 7.6 ± 0.3S for Myo4p (). Interestingly, the sedimentation value of Myo2p is similar to myosin V (10S), which forms a dimer, and the sedimentation value of Myo4p is similar to myosin IXb (7.3S), a monomeric myosin (; ).
To confirm visually the oligomeric state of Myo2p and Myo4p, we imaged both motors by rotary metal shadowing. Images of Myo2p contained many examples of a Y-shaped molecule with globular domains at each end (). Previously published images of myosin V dimers revealed a similar structure and suggest that two of the globular domains in our images correspond to the motor domains of Myo2p and the third to the C tail (). The images thus confirm that Myo2p is a dimer. In contrast, most images of Myo4p revealed a single globular domain, occasionally with an extended tail (). Rarely, we observed two globular domains in proximity, but the spacing of the heads was noticeably different than Myo2p. The density of images of double-headed motor in several fields of Myo2p and Myo4p was 3.8 /μm and 0.4/μm, respectively, and the density of images of single globular domains of Myo2p and Myo4p was 12/μm and 10.8/μm, respectively (Fig. S2, available at ). The number of single globular domains may be inflated, as dust and debris often appear similar to a single globular domain. Thus, evidence from coimmunoprecipitation, cross-linking, hydrodynamic analysis, and electron microscopy indicated that Myo4p is a monomeric myosin, sharply distinguishing it from other class V myosins.
Like other class V myosins, Myo4p moves cargo, in this case mRNA, rapidly and continuously within cells. To gain a better understanding of how monomeric Myo4p transports ASH1 mRNA in vivo, we purified native Myo4p and analyzed its motility in vitro. We also purified Myo2p to compare the motility of Myo4p to a known dimer. TEV-eluted Myo2p and Myo4p fractions were highly enriched for each myosin, but contained additional proteins (). Most of these were not associated with the myosins, as they did not cofractionate in the sucrose gradients (). TEV-eluted Myo4p contained both She3p and She2p, but only She3p cofractionated with Myo4p on sucrose gradients, suggesting that She2p had dissociated from Myo4p/She3p. The weak interaction between She2p and Myo4p/She3p indicated that only a fraction of purified Myo4p/She3p was associated with She2p. To avoid analyzing two different species of Myo4p, we purified Myo4p from cells.
To further purify Myo4p and Myo2p, we used an actin bind and release step. A similar method using microtubules was recently used to purify dynein from (). TEV-eluted myosins were incubated with phalloidin-stabilized F-actin in the absence of ATP to promote binding to actin filaments. Actin filaments and attached myosins were sedimented by centrifugation and then incubated with buffer containing ATP to release myosins from actin filaments. The actin filaments were again sedimented by centrifugation, and the supernatant (ATP-release) was saved. Silver stain analysis of the ATP-releases revealed that Myo4p and Myo2p were highly enriched and pure (). Both fractions contained myosin light chains (∼15 kD), and the Myo4p fraction contained She3p. In addition, a small amount of actin was present representing short filaments that had not completely sedimented. Myo2p appeared to be clipped during purification, as the protein bands between 75 and 150 kD that were apparent by silver stain were detected with anti-Myo2p-tail antibody by Western blot (unpublished data). Thus, the two-step purification yielded very pure preparations of native Myo4p and Myo2p.
To demonstrate that purified Myo4p and Myo2p maintain their oligomeric status of monomer and dimer, respectively, we incubated both purified myosins with EDC. EDC-treated and untreated samples were run on the same gel and detected by Western blot with antibodies against the C-tails of Myo4p and Myo2p (). EDC-treated Myo2p migrated as a large band between 400 and 450 kD, which fits well with the theoretical molecular weight of two cross-linked Myo2p heavy chains and some of the associated myosin light chains. EDC-treated Myo4p migrated between 270 and 325 kD, which approximates a single Myo4p heavy chain cross-linked to She3p and the light chains. Interestingly, when the film was overexposed, a second higher molecular weight Myo4p band was detected. This band migrated slower than cross-linked Myo2p, suggesting that a small amount of Myo4p had formed into dimers. Reprobing the blot with anti-She3p antibodies revealed that She3p co-migrated with EDC-treated Myo4p (). We also tested cross-linkers with different reaction groups and longer spacers than EDC, but they cross-linked Myo2p less efficiently than EDC and none generated Myo4p dimers (Fig. S1, available at ). We conclude that the majority of purified Myo4p exists in a monomeric state bound to She3p, but a small fraction of Myo4p forms dimers when concentrated.
We analyzed the motility and processivity of purified Myo2p and Myo4p in an actin filament gliding assay. Nonprocessive motors exhibit decreasing filament velocity as motor density decreases, whereas processive motors show little change in velocity over several orders of motor density (). In addition, processive motors generate filament movement at densities of 1–2 motors/μm. Purified Myo2p behaved as a nonprocessive motor (). At high motor densities, Myo2p moved actin filaments at velocities of 2–2.5 μm/s, which decreased as motor density decreased. Below ∼50 motors/μm, actin filaments did not land on coverslips. The maximum velocity of purified Myo2p is consistent with the velocity measured for secretory vesicles in vivo. Our analysis of the motility of purified Myo2p confirms previous work on Myo2p motility in cell lysates ().
In contrast, Myo4p produced an actin velocity curve that was the inverse of Myo2p (). At the highest Myo4p densities, actin filaments moved at 0.45–0.65 μm/s, but as Myo4p density decreased, filament velocity increased to a maximum of 1.5–1.75 μm/s. This unique velocity curve is unlikely to result from inactive or damaged Myo4p, as only active myosin should be purified with the actin affinity step. Further, we detected very few nonmotile actin filaments on the coverslips, which indicates healthy active motors (Videos 1–4). In contrast to our results, an earlier study of Myo4p showed that Myo4p generated a velocity curve similar to Myo2p in actin gliding assays (). Two differences in methodology may account for this discrepancy. First, studied Myo4p-GFP in cell lysates, whereas we used purified Myo4p. Second, in the actin gliding assay, Reck-Peterson et al. tethered Myo4-GFP to coverslips with anti-GFP antibodies, whereas we directly adsorbed Myo4p to coverslips. Bivalent antibodies have the potential to bind two Myo4p-GFP motors and perhaps stabilize Myo4p dimers. Although the velocity curve of Myo4p is different from the nonprocessive curve of Myo2p, we suspect that Myo4p is also nonprocessive, as below Myo4p densities of ∼100 motors/μm, filaments did not land on the coverslips.
Because Myo4p co-purifies with She3p, we tested whether She3p affected the ability of Myo4p to move actin filaments. We purified Myo4p-1/2TAP from cells and analyzed Myo4p motility. We observed the same inverse relation between filament velocity and Myo4p density, indicating that She3p was not affecting Myo4p activity (). The slight difference in the lowest density of Myo4p that supports filament gliding between figures is likely due to variability in the amount of nonfunctional motor in different preparations of Myo4p and errors in calculating motor density in the flow cells.
To test more definitively whether Myo2p and Myo4p move processively, we analyzed the motility of purified Myo2p-GFP and Myo4p-GFP on actin bound to the coverslip surface by total internal reflection fluorescence (TIRF) microscopy. First, we measured the total fluorescence intensity and the photobleaching of individual spots to determine how many GFP molecules exist in each. The initial intensity of a spot minus the background after photobleaching was measured and plotted as a histogram (Fig. S3, available at ). The data for Myo2p-GFP and Myo4p-GFP were each plotted as separate histograms and each were fitted with a two-peak Gaussian function corresponding to a population with a large intensity (25500 a.u.) and a smaller intensity (13,000 a.u.), probably representing spots with two and one GFP molecules, respectively. This result revealed that most Myo2p-GFP spots contained two GFPs, whereas the majority of Myo4p-GFP spots contained a single GFP, indicating that Myo2p-GFP is a dimer and Myo4p-GFP is a monomer (Fig. S3). In the presence of ATP, most Myo2p-GFP spots failed to move along actin filaments, and instead either released from the filaments or bleached. At low (300 nM) ATP, a small fraction of Myo2p-GFP (∼2%) moved processively for distances less than 0.5 μm, and fluorescence imaging with one-nanometer accuracy (FIONA) analysis revealed a trend toward step of 36 and 72 nm (Fig. S4). One interpretation of the results is that Myo2p-GFP is an intermediate duty ratio motor with a low probability of remaining associated with a filament over a few enzymatic cycles, and thus, in a population of Myo2p motors a small fraction will move processively for short distances. This conclusion is supported by previous published work, and we suggest that Myo2p is a weakly processive motor (). In contrast, none of the Myo4p-GFP spots with fluorescent intensity equivalent to one or two GFPs moved processively at any ATP concentration. Interestingly, we observed several Myo4p-GFP spots with fluorescent intensity equivalent to at least 3 GFPs, and of these, ∼20% moved processively (), taking a range of step sizes (Fig. S4). The variability in step size in this case may be related to the presence of multiple, simultaneously interacting myosin molecules with actin, which might dampen the measured step size (). Back steps of Myo4p-GFP were occasionally observed at 100 nM ATP, and back-and-forth movement of Myo4p-GFP was also observed at high (1 mM) ATP. The results suggest that Myo4p is nonprocessive as a monomer, but at a low frequency assembles into ensembles capable of processive movement.
The strikingly different velocity curve of Myo4p as compared with other myosins could be due to unique properties of the Myo4p motor domain or in its oligomeric state (i.e., monomer versus dimer). To distinguish between these possibilities, we constructed chimeras consisting of the motor domain from one myosin fused to the IQ and C-tail domains of the other myosin. Myo2/4p (Myo2 motor − Myo4 IQ + C-tail) was expressed with a C-terminal 1/2TAP tag in cells and purified as described for Myo4p-1/2TAP. The Myo4/2p chimera (Myo4 motor − Myo2 IQ + C-tail) was expressed with a C-terminal 1/2TAP tag in cells and purified as described for Myo2p-1/2TAP with one additional step. Because Myo4/2p contains the coiled-coil domain of Myo2p, a small amount of Myo4/2p-1/2TAP may dimerize with Myo2p-GFP. To eliminate these mixed dimers from purified Myo4/2p, we incubated the TEV eluate with beads coated with anti-GFP antibodies. The unbound fraction was then subjected to the actin affinity step to further purify Myo4/2p.
The structure and protein composition of the purified chimeras resembled the myosin from which they obtained their C-tail. The protein profile of purified Myo2/4p contained She3p and myosin light chains (). EDC cross-linking of purified Myo2/4p increased its molecular weight similarly to purified Myo4p, and She3p was evident in the cross-linked complex (). Myo2/4p had a velocity sedimentation coefficient of 7.3S, similar to Myo4p (). These results indicated that the Myo2/4p chimera, like Myo4p, is a monomer that binds directly to She3p, confirming previous results that She3p interacts with the C-tail of Myo4p (). In contrast, the Myo4/2p chimera did not co-purify She3p, and EDC-treated Myo4/2p migrated at a similar molecular weight as EDC-treated Myo2p (). In addition, velocity sedimentation analysis of purified Myo4/2p revealed a sedimentation coefficient of 9.1S, identical to Myo2p (). Thus, the Myo4/2p chimera, like Myo2p, is a dimer.
To determine whether the motor domain or oligomeric state of Myo4p limits its ability to move actin filaments at high density, we analyzed Myo4/2p and Myo2/4p in the actin filament gliding assay. Myo4/2p behaved similarly to Myo2p: as the density of Myo4/2p decreased, the velocity of actin filaments decreased (). This profile suggests that Myo4/2p, like Myo4p, is a nonprocessive motor. Interestingly, engineered a myosin chimera that contained the Myo4p motor and myosin V C-tail. In contrast to Myo4/2p, their myosin chimera was processive. In addition, they found that their chimera exhibited longer run lengths in lower ionic strength buffers. To determine whether Myo4/2p was processive at lower salt concentrations, we tested the motility of Myo4/2p in the actin filament gliding assay in buffer containing 25 mM KCl. There was no difference in the velocity curve at 25 mM KCl compared with higher salt buffers, indicating that the difference between the chimeras is not due to buffer conditions (unpublished data). Instead, the results suggest that subtle differences in the coiled-coil domains of Myo2p and myosin V may be crucial for generating processivity. Similar to Myo4p, Myo2/4p showed an inverse relationship between motor density and velocity: as Myo2/4p density decreased, actin filament velocity increased (). Importantly, the maximum velocity of Myo2/4p was similar to Myo2p, and Myo4/2p velocity was similar to Myo4p, suggesting that the basic enzymatic properties of the motors had not substantially changed in the chimeras. Thus, the results indicate that the slowing of actin filaments at higher Myo4p density is not due to the properties of the Myo4p motor domain. Rather, they suggest that the monomeric status of Myo4p allows it to move filaments more efficiently than Myo2p at low motor densities.
We next examined whether the myosin chimeras could functionally replace the wild-type myosins. To determine whether Myo2/4p localizes Ash1p to the daughter cell nucleus, we used a growth assay that is sensitive to the localization of Ash1p. cells grow in the absence of adenine, because Ash1p localizes exclusively to the daughter cell, and its absence from the mother cell allows for expression of and synthesis of adenine. cells fail to grow without adenine because Ash1p accumulates in the nuclei of both mother and daughter cells where it represses expression of . We expressed from low copy plasmids either or in cells. Compared with cells containing only an empty vector, both and restored growth in the absence of adenine, suggesting that Myo2/4p was sufficient to localize Ash1p to the daughter cell (). To confirm that Myo2/4p localizes mRNA to the bud tip, we examined the same cells by FISH with probes against mRNA. The results showed that mRNA localized to the bud tip in 66% of cells expressing and 62% of cells expressing (). In contrast, none of the cells containing only an empty vector showed localized ASH1 mRNA. Thus, Myo2/4p can functionally substitute for Myo4p to transport mRNA to the bud tip.
To determine whether Myo4/2p could substitute for Myo2p, we expressed either or on low copy plasmids in cells, which contain a temperature-sensitive mutation in the motor domain, rendering the motor inactive at elevated temperature (). Because is essential, cells do not grow at elevated temperatures. All strains, , + , and + grew well at 25°C, but only and grew robustly at 30°C (). Although rescued growth at 30°C, it grew poorly at 37°C compared with . To determine whether Myo4/2p mediated vesicle transport to the bud at restrictive temperatures, we examined the distribution of Sec4p, which accumulates on secretory vesicles (). All strains, , + , and + , localized GFP-Sec4p to buds at 25°C (unpublished data). But only Myo2p and Myo4/2p restored localization of GPF-Sec4p to buds of myo2-66 cells at 30°C (). At 37°C, GFP-Sec4p was detected in buds of cells expressing Myo4/2p, but GFP-Sec4p localization was stronger in cells expressing Myo2p. The velocity of Myo4/2p was about half of Myo2p (), which may explain the weaker localization of GFP-Sec4p in cells. The results demonstrate that Myo4/2p mediates vesicle transport, though less efficiently than Myo2p.
Although a large body of information exists on the mechanism of molecular motor movement, limited work has been done to address how the activity of a motor translates to successful cargo transport in vivo (). In particular, the physical and biochemical differences between membrane-bound vesicles and mRNAs would seem to require different properties in a motor protein. We have analyzed two closely related class V myosins, Myo2p and Myo4p, which transport secretory vesicles and mRNA, respectively, in . We find two important differences between Myo2p and Myo4p. First, unlike Myo2p and other known class V myosins, Myo4p is a monomer (). Second, Myo2p and Myo4p differ in their movement along actin filaments. Myo2p behaves as a nonprocessive motor in actin gliding assays and as a weakly processive motor in single-molecule assays. In this regard, it is worth noting that myosin V from was also shown to be a low duty ratio motor that probably does not move processively (). The motility of Myo4p is unique. In an actin gliding assay, the velocity of Myo4p increases as motor density decreases. Further, single-molecule analysis suggests that although individual Myo4p motors are nonprocessive, ensembles of Myo4p move processively. Finally, the motor domain of Myo2p can substitute for Myo4p motor to localize mRNA, indicating that the unique motility of Myo4p is not due to differences in its motor domain.
How do the properties of Myo2p and Myo4p allow them to transport vesicles and mRNA, respectively, in vivo? Myo2p is at best a weakly processive motor, and the run lengths of the few processive Myo2p motors were at most ∼0.5 μm, far shorter than the length of a yeast cell. Thus, a single Myo2p is unlikely to maintain transport of cargo to the bud tip. The key, therefore, to generating sustained movement of cargo is likely the presence of multiple Myo2p motors on a vesicle or organelle, increasing the probability that the vesicle or organelle remains associated with a filament during transport to the bud. That Myo2p is weakly processive suggests that as few as two or three motors may be sufficient to sustain transport to the bud, and even the smallest vesicle should provide ample surface area to bind multiple motors. Although the number of molecules of Myo2p on a vesicle or vacuole has not been quantified, the expression levels of Myo2p (∼4,000/cell) and Vac8p (∼5,000/cell), which links Myo2p to vacuoles, offer sufficient amounts of both proteins to bind multiple Myo2p motors to every vacuole (). An important question is whether motors bound to the same cargo and filament coordinate their enzymatic cycles to ensure sustained and rapid transport in vivo. We found that at high densities of Myo2p, when multiple motors are attached to a filament, filament velocity is comparable to the rate of vesicle transport in vivo, suggesting that Myo2p motors can coordinate their activities in vitro to generate rapid transport.
How Myo4p generates sustained movement of RNA in vivo is likely more complicated. Myo4p is clearly nonprocessive as a single-headed molecule, implying that multiple motors are needed to sustain transport of RNA in vivo. The small size of an active zipcode, however, seems to preclude the attachment of multiple motors. Our results suggest a potential solution. Ensembles of three or more Myo4p motors move processively, so instead of Myo4p motors binding to multiple sites along an mRNA, like Myo2p on a vesicle, Myo4p may assemble into a multimotor complex that can be linked to a single zipcode within an mRNA. How Myo4p assembles into ensembles is unclear, as there is no evidence that other class V myosins form multimotor complexes. The primary structural difference between Myo4p and other class V myosins is that Myo4p is a monomer, suggesting the possibility that as a monomer Myo4p is capable of assembling into larger order complexes. She3p may be an important component, as it binds tightly and directly to Myo4p. She3p binds the C-tail of Myo4p, including the coiled-coil region (). She3p also contains an extensive stretch of sequence with a strong probability of forming a coiled-coil. Intriguingly, the multicoil program predicts six heptad repeats to form a three-stranded coiled-coil. One explanation, then, why Myo4p is a monomer is that the C-tail evolved to form association with She3p rather than with itself. Finally, She2p, which links Myo4p/She3p to mRNA, is a dimer and could potentially bind two copies of Myo4/She3p to facilitate assembly of Myo4p ensembles.
Another unusual aspect of Myo4p motility is that the velocity of actin filaments was threefold higher at low density of motors compared with high density. Interestingly, the in vitro velocity of Myo4p at high density is similar to the in vivo velocity of RNA particles (). What generates this unique motility of Myo4p is unclear. A chimera of the Myo2p motor and C-tail of Myo4p exhibited a similar motility to Myo4p, suggesting that either the C-tail of Myo4p regulates the activity of the motor or that the monomeric state of the motor protein affects its activity. Recent work on myosin Va has shown that the C-tail folds over to bind the motor domain and repress its activity, but our results question whether such a mechanism explains the behavior of Myo4p (; ). First, velocity sedimentation analysis of myosin Va reveals a 14S version, representing the C-tail bound to the motor, and an 11S species, which is myosin Va extended. Our analysis of Myo4p and Myo2p did not reveal more compact versions of either myosin, suggesting that both are fully extended. Second, the slowing of Myo4p velocity is concentration dependent, whereas the velocity of myosin Va is not affected by concentration but instead regulated by Ca (). Although our results do not rule out a role for the C-tail in regulating the activity of the Myo4p motor, we suggest that the monomeric state of Myo4p is the likely reason for its unusual motility.
To determine the molecular mechanisms of active transport complexes, future work will need to be directed at reconstituting and analyzing active myosin–cargo complexes. Toward this aim, we have developed methods to purify native and active Myo2p and Myo4p. RNA transport is especially ripe for reconstitution, as the components that link Myo4p to RNA have been identified, and the cargo can be readily synthesized and modified in vitro. Thus, we can address questions about the minimal number of Myo4p motors needed to generate sustained transport and whether multiple zipcodes increase the robustness of RNA transport. Similarly, it is now possible to attempt to reconstitute vesicle or vacuole transport in vitro. In addition to the number of Myo2p motors needed for cargo movement, an important question is whether an even distribution or clustering of Myo2p on the surface of a vesicle facilitates sustained transport. Finally, the use of myosin chimeras will enable us to dissect further those features of Myo2p and Myo4p that are unique for vesicle and RNA transport, respectively, and will lead to a broader understanding of how motor proteins have evolved to transport different types of cargo.
All strains used in this study are listed in . , , , , , and were made by PCR-mediated gene modification. and were made by gene replacement of endogenous and , respectively. The Myo4-HA plasmid was made by inserting a single HA sequence into an engineered BamHI site just upstream of the stop codon. The She3-HA plasmid was made by inserting a single HA sequence in an engineered NheI site just upstream of the stop codon. Both Myo4-HA and She3-HA were subcloned into CEN ARS vectors to allow for expression from a low copy plasmid. - 3′ UTR has been described elsewhere (), and - 3′ UTR was made by PCR-mediated gene disruption. To make the Myo2/4 plasmid, PCR products of 5′ UTR and promoter (-579 to +3), motor domain (+4 to +2334), IQ + C-tail (+2347 to +4722), and 3′ UTR (stop codon + 370 nucleotides of downstream sequence) were ligated using unique restriction sites and inserted into pRS315. To make the Myo4/2 plasmid, PCR products of 5′ UTR and promoter (−687 to +3), motor domain (+4 to +2346), IQ + C-tail (+2335 to +4413), and 3′ UTR (stop codon + 193 nucleotides of downstream sequence) were ligated using unique restriction sites and inserted into pRS315. 1/2 TAP versions of both proteins were made by inserting a PCR product containing the sequence encoding a TEV protease site and IgG-binding domain into unique restriction sites just upstream of the stop codon in and
9E10 (anti-myc) and HA.11 (anti-HA) were purchased from Covance Research Products. JL8 (anti-GFP) was purchased from CLONTECH Laboratories, Inc. Antisera against She2p, She3p, Myo4p, GFP, and HA were prepared by Cocalico Biologicals. Antibodies were affinity purified using either purified 6× His She2p, 6× His She3p 197–425, or GST-Myo4p C-tail (1075 to end). GFP antibodies were affinity purified on a membrane containing GFP. HA antibodies were affinity purified on a column containing HA peptide. ECL reagents, protein molecular weight standards, IgG Sepharose 6 Fast Flow, and protein G Sepharose 4 Fast Flow were purchased from GE Healthcare. 1-Ethyl-3-[3-dimethylaminopropyl]carbodiimide hydrochloride (EDC) and -hydroxysuccinimide (NHS) were purchased from Pierce Chemical Co. Cross-linked Phosphorylase b, glucose oxidase, catalase, phalloidin, and TRITC-phalloidin were purchased from Sigma- Aldrich. AcTEV protease was purchased from Invitrogen. Actin was purchased from Cytoskeleton, Inc. Anti-Myo2p tail antibody was a gift from Mark Mooseker (Yale University, New Haven, CT).
Yeast cultures were grown in YPD or selective media to mid-log phase. Cells were harvested by centrifugation, pellets were washed twice with cool HO, and resuspended in 3× cell pellet volume of lysis buffer A (20 mM Imidazol, pH 7.2, 150 mM KCl, 2.5 mM MgCl, 1 mM EGTA, 1 mM EDTA, 2 μg/ml of aprotinin, leupeptin, and pepstatin, 400 μM AEBSF, 5 mM ATP, and 2 mM DTT). Lysates were made by bead-beating 4× 45 s, with 60 s in ice water between cycles. The lysate was centrifuged for 5 min at 2,000 rcf, and supernatant was collected and centrifuged for 20 min at 85,000 rpm in a TLA120.2 rotor. The high speed supernatant (S3) was used for coimmunoprecipitation experiments. S3 was incubated at 4°C for 1 h with either free antibody or antibody coupled to beads. Free antibody was subsequently linked to protein G beads for 1 h at 4°C. Beads were isolated by centrifugation, washed 4× with 50× bead volume of wash buffer (150 mM KOAc, 25 mM Hepes-KOH, pH 7.5, 2 mM MgOAc, and 0.1% IGEPAL), resuspended in 1× SDS-PAGE sample buffer, and boiled. Supernatants were mixed with 5× SDS-PAGE sample buffer and boiled.
EDC and NHS were dissolved in HO just before use at a concentration of 1 M. For crude S3 lysates, only EDC was added to a final concentration of 50 mM. For TEV and ATP-released actin affinity-purified myosin, both EDC and NHS were added to a final concentration of 40 mM. Samples were incubated for 1 h at room temperature and reactions were quenched by addition of 5× sample buffer. The samples were boiled and run on either 3.5 or 5% denaturing gels and then analyzed by Western blot. Bio-Rad all-blue precision plus and Sigma-Aldrich cross-linked Phosporylase b were used as size standards. We detected Phosphorylase b by staining Western blot membranes with ponceau S before blocking.
A modified TAP tag (1/2TAP) lacking the calmodulin binding domain was used for myosin purification. The tag was added to the C terminus of all myosins. 1/2TAP-tagged Myo2p and Myo4p were integrated, whereas 1/2TAP-tagged MYO2/4 and MYO4/2 were expressed from low copy plasmids. Cultures were grown to mid-log phase in YPD and harvested by centrifugation. Cell pellets were washed twice with cool HO, resuspended in 1/5 pellet vol lysis buffer A, frozen in liquid N2, and stored at −80°C. Frozen cells were lysed by blending in a Waring blender with liquid nitrogen. Lysed cells were thawed in 1.5× cell pellet vol of ice-cold lysis buffer A containing 0.3 mg/ml RNase A on a hot plate set to low with constant stirring. Thawed lysates were centrifuged for 10 min at 20 , supernatants were collected, and centrifuged for 1 h at 50,000 rpm in a Ti70 rotor. Cleared lysate was passed over IgG sepharose columns at 4°C. Columns were washed with 250× column vol of lysis buffer without protease inhibitors or ATP. Washed beads were resuspended in 1× column vol lysis buffer without protease inhibitors or ATP containing 50–100 U TEV protease and incubated overnight at 4°C. Columns were eluted and washed with lysis buffer A without protease inhibitors or ATP. Elution fractions were used in the assays described or further purified. TEV eluted myosin was further purified by incubating with phalloidin-stabilized actin filaments (final actin concentration 0.5 μM) for 10–15 min at 4°C. The myosin bound to actin was added onto two-step sucrose gradients (100 μl 60% and 300 μl 10% sucrose) and centrifuged for 45 min at 45,000 rpm in a SW 50.1 rotor. The top of the gradient was discarded, leaving the bottom 250 μl containing the myosin bound to actin. ATP was added to the myosin and actin (final ATP concentration 10 mM) and then immediately centrifuged for 25 min at 85,000 rpm in a TLA 120.2 rotor. The supernatant containing myosin released from actin was collected and used for motility assays and EDC cross-linking.
TEV-eluted myosin was added to the top of 5–20% 5-ml sucrose gradients. Sucrose was dissolved in buffer A. Gradients were spun for 16–20 h at 40,000 rpm in a SW50.1 rotor. Fractions were collected from the bottom of the gradients with a peristaltic pump. Parallel gradients containing ovalbumin, albumin, aldolase, and catalase protein standards were used to calculate S values for each myosin.
Myo2p-GFP and Myo4p-GFP were diluted to 0.1 mg/ml in 0.5 M ammonium acetate and mixed thoroughly with glycerol. The samples were sprayed onto a mica chip and loaded into the vacuum chamber of RFD-9010 CR Freeze Etching Equipment (RMC Products). Platinum was deposited by rotary shadowing at an angle of 9° with 3 KV and 70 mA. Carbon was deposited to reinforce the replica at an angle of 90° with 2.5 KV and 90 mA. The replicas were prepared by floating on clean water and lifted onto formvar-carbon coated cupper grids and observed in an electron microscope (JEM-1200EX II; JEOL).
Actin gliding assays were preformed as described in with the following changes. Grease-coated coverslips were used as spacers to make the motility chambers. Buffer A without protease inhibitors or ATP was used for all washes. Coverslips were not coated with nitrocellulose. Chambers were pre-blocked with 0.01 mg/ml BSA in buffer A, washed, and purified myosin was added. The myosin was incubated in the chamber for 30+ min on wet ice. Motility buffer (10 mM imidazole, pH 7.2, 75 mM KCl, 1 mM EGTA, 2.5 mM MgCl, 10 mM Mg-ATP, 8 mM DTT, 0.2 mg/ml glucose oxidase, and 0.1 mg/ml catalase) was mixed with TRITC-phalloidin–stabilized actin filaments (8 nM final) and dextrose (0.25% final). Unlabeled actin was not used to block dead myosin. Actin filaments were observed on a fluorescence microscope (TE2000; Nikon) equipped with a 100×/NA 1.4 lens (Nikon), and images were acquired with an ORCA-ER CCD camera (Hamamatsu) using IPLab software (Scanalytics, Inc.). Videos were taken within the first 5 min of adding actin in motility buffer, except at low motor densities when actin landing was slower. Filaments were tracked manually from frame to frame to compute velocity.
Single-molecule motility assay and FIONA analysis using TIRF microscopy was performed and quantified as described previously (; ; ). Fluorescence intensities of individual spots were measured as follows. Myosin solutions without ATP in single-motility buffer were added into a flow-chamber that had actin filaments bound to the surface, and were incubated for 2 min at room temperature. The actin filaments were a copolymer with 10% biotin-labeled actin + 90% unlabeled actin (). They were bound to the coverslip surface via a biotin-avidin sandwich. The free myosin was washed out with 600 μl single-molecule buffer (SMB: 20 mM MOPS, pH 7.2, 5 mM MgCl, 0.1 M EGTA, and 50 mM KCl) containing 1 mg/ml bovine serum albumin, 50 mM DTT, and an oxygen scavenging system (125 mg/ml glucose, 1,665 units glucose oxidase (G-7076; Sigma-Aldrich), and 26,000 units Catalase (106810; Roche). The flow-chamber was then placed on a microscope (IX70; Olympus) and observed through a 60×/NA 1.45 lens (Olympus) in TIRF mode at 25°C. Two relay lenses of 1.5 and 2.5× were also used to achieve a total magnification of 225. Images were captured with either a Micromax CCD (Roper Scientific) or an Andor EMCCD DV 897 using Andor iQ software. The surface of the flow chambers was first checked to determine whether myosin molecules were bound to actin filaments. A manual shutter to block the light was closed and the stage was moved to a different field. To prevent drifting of stage, the shutter was closed 2–4 min. Image acquisition started when the shutter was reopened.
To analyze spot intensity and photobleaching, an 8 × 8 pixel region of interest (ROI) was applied to each spot and the integrated intensities for all frames were determined with MetaMorph (Molecular Devices). Background signal intensities were subtracted from the measured intensities following photobleaching of each 8 × 8 pixel ROI.
YPT148, YPT149, and YPT150 were grown overnight in minimal media lacking leucine. The overnight cultures were diluted into fresh minimal media lacking leucine and incubated at 30°C until achieving OD ∼0.5. Cells were fixed in 4% formaldehyde for 45 min at room temperature, washed, and then spheroplasted. Spheroplasts were adsorbed onto poly--lysine coated glass coverslips and incubated in hybridization mix (50% formamide, 5× SSC 1 mg/ml, yeast tRNA,100 μg/ml Heparin, 1× Denhardt's Solution, 0.1% Tween 20, 0.1% Triton X-100, and 5 mM EDTA, pH 8.0) for 1 h at room temperature. Samples were incubated in hybridization mix containing an anti-sense, Dig-labeled RNA probe against mRNA overnight at 37°C. Samples were washed in 0.2× SSC and then incubated with mouse anti-Dig in 0.05 M Tris-Cl, pH 7.5, 0.15 M NaCl, and 5% fetal bovine serum for 30 min at 37°C. Samples were washed in 0.05 M Tris-Cl, pH 7.5, 0.15 M NaCl, and 0.05% Tween 20 and then incubated with goat anti–mouse Alexa488 in 0.05 M Tris-Cl, pH 7.5, 0.15 M NaCl, and 5% fetal bovine serum for 1 h at room temperature. Samples were then washed in 0.05 M Tris-Cl, pH 7.5, 0.15 M NaCl, and 0.05% Tween 20 and then mounted on glass slides. Samples were observed at room temperature on a fluorescence microscope (TE2000; Nikon) equipped with a 100×/NA1.4 lens (Nikon), and images were acquired with an ORCA-ER CCD camera (Hamamatsu) using IPLab software (Scanalytics, Inc.). Cropping and brightness/contrast adjustments were made in Adobe Photoshop.
For Ash1p localization, YPT148, YPT149, and YPT150 were grown overnight in synthetic media lacking leucine (SC −leu) The cultures were adjusted to OD ∼0.3 and then four tenfold serial dilutions were made from the culture. 5 μl of the starting culture and each dilution were spotted onto SC −leu plates with or without 25 mg/l adenine. The plates were incubated at 30°C for 3 d. For rescue of growth, YPT168, YPT171, and YPT172 were grown overnight in SC −leu. The cultures were diluted to OD ∼0.3 and four tenfold serial dilutions were made from the culture. 5 μl of the starting culture and each dilution were spotted onto SC −leu plates and incubated for 2 d at either 25, 30, or 37°C.
YPT173, YPT174, and YPT175 were grown overnight in SC −leu. Cultures were diluted into fresh SC −leu and incubated for 6 h at 25, 30, or 37°C. Cells were fixed in 4% formaldehyde, washed, and mounted on glass slides. Samples were observed at room temperature on a fluorescence microscope (Nikon TE2000) equipped with a 63×/NA1.4 lens (Nikon), and images were acquired with an ORCA-ER CCD camera (Hamamatsu) using IPLab software (Scanalytics, Inc.). Cropping and brightness/contrast adjustments were made in Adobe Photoshop.
Videos 1–4 are representative actin gliding assay movies. Videos 1 and 3 are TEV-eluted and ATP-released actin affinity-purified Myo2p, respectively. Videos 2 and 4 are TEV-eluted and ATP-released actin affinity-purified Myo4p, respectively. Collection and display rates are both 2.5 frames/s. Online supplemental material is available at . |
The coordinated movement of crawling cells depends on the proper spatial and temporal regulation of the actin cytoskeleton. Actin polymerization is biased toward the cell front, enabling protrusion of the leading edge, and myosin contraction is biased toward the cell rear, enabling traction of the cell body toward the front. In addition to the polarization of actin network dynamics, crawling cells exhibit a morphological polarization, with the cell front and rear being easily distinguishable.
In neutrophils, which respond to external gradients of chemoattractants, front-back polarization is initiated by protrusion of the leading edge closest to the chemoattractant source. However, stationary neutrophils in a uniform concentration of chemoattractant can also polarize spontaneously, initiate motility, and move in a random direction (; ). In both cases, polarization is accompanied by the recruitment of PIP to the cell membrane () and the formation of actin ruffles at the leading edge (). Other cell types have also been reported to spontaneously break symmetry in the absence of any external cues, including fibroblasts, which polarize after plating onto glass coverslips coated with polylysine and ConA (), and Walker carcinosarcoma cells (). This suggests that polarization and motility initiation do not require an external directional cue. Where one is present, it only induces a preferred directionality, but there must be an intrinsic mechanism for cell polarization.
In contrast to neutrophils, where polarization is initiated by protrusion, stationary lamellipodial fragments from fish epidermal keratocytes, which are circular, can be pushed at the rear by a stream of media from a micropipette. This induces rear retraction followed by protrusion of the front and the initiation of persistent motility (). Similarly, in already polarized chick heart fibroblasts, rear retraction precedes front protrusion (; ). So, although neutrophil polarization in response to a chemoattractant is initiated by protrusion at the front, lamellipodial fragments can polarize when physically pushed at the rear. However, it has not been determined how polarization arises and propagates in the absence of an external stimulus and what the events throughout spontaneous symmetry breaking may be.
The goal of this study is to establish quantitatively, for the first time, the sequence of structural events during spontaneous cell polarization. We chose rapidly moving fish epidermal keratocytes as a model system for our analyses. These cells are not known to respond to chemotactic stimuli; thus, the underlying mechanism of spontaneous symmetry breaking can be studied without cross talk from chemotaxis. Once keratocytes are polarized, they migrate in a persistent manner (). Furthermore, keratocytes are flat and, thus, are particularly amenable to high resolution live cell microscopy of morphodynamic events.
Our data indicate that the first signs of polarization do not emerge at the cell front or periphery but at the cell rear and perinuclear region. Rearrangements of the actomyosin network near the cell periphery and cell front occurred later in the polarization process, indicating that they were a consequence rather than a cause of symmetry breaking. This symmetry breaking was driven by contraction and was dependent on Rho kinase–mediated reorganization of the actomyosin network. Thus, we have identified an alternative paradigm for spontaneous cell symmetry breaking and motility initiation.
Keratocytes isolated from cichlid scales migrated from the scale as large epidermal sheets that could be disaggregated by incubation in 2.5 mM EGTA/85% PBS to form a mixture of smaller islands, individual motile polarized keratocytes, and individual stationary nonpolarized keratocytes. Stationary keratocytes were circular and radially symmetric (), which is in contrast to motile keratocytes, which were crescent shaped and bilaterally symmetric (). The network of actin filaments (filamentous actin [F-actin]) of stationary keratocytes was denser around the cell body than at the periphery and often formed circular bands around the cell body (). Although stationary keratocytes did not exhibit net translocation, transient protrusion and retraction occurred around the cell edge.
In the lamellipodia of motile cells, F-actin generally moves from the leading edge to the cell body. To image F-actin network movement in the lamellipodia of stationary keratocytes, we used fluorescent speckle microscopy (FSM) with low levels of labeled phalloidin (; ; ), which binds specifically to F-actin and yields a higher signal/noise level than G-actin probes (; ; ). We measured the rate of F-actin network movement by adaptive multiframe correlation tracking of fluorescent speckle motion (; ). In stationary keratocytes, F-actin flowed centripetally from the cell edge to the cell body ( and Video 1, available at ) at a rate of ∼25–60 nm/s. The flux was fastest at the cell periphery, decreasing gradually toward the cell body. This radial inward flow was consistent with previous studies of centripetal actin flow in circular, stationary lamellipodial fragments () and circular sea urchin coelomocytes (). In motile keratocytes, it has been established that the lamellipodial F-actin network moves slowly rearwards with respect to the substratum as the cell moves rapidly forward (; ). In our experiments, we measured retrograde F-actin flow of ∼10 nm/s relative to the substratum in the lamellipodia ( and Video 2) and faster inward flow at the rear sides of the cell ( and Video 2) as observed previously (; ; ), where myosin-dependent contraction gathers the F-actin network toward the cell body (; ).
Stationary keratocytes could spontaneously break symmetry, polarize, and initiate motility in the absence of external cues at a frequency of ∼15% per 30 min. Given that the F-actin flow pattern and speed were different in stationary compared with motile keratocytes (), mechanisms that lead to symmetry breaking and morphological polarity should also alter F-actin dynamics within the cell. These changes could occur at the prospective cell front, prospective cell rear, or simultaneously at the front and rear.
Not all stationary cells initiated motility, and stationary cells had variable protrusion and retraction at the cell edge. To determine the sequence of events leading to cell polarization and movement, we acquired time-lapse videos of keratocytes spontaneously initiating motility ( and Videos 3 and 4, available at ) and first examined the movement of the cell edge. We measured the position of the cell edge in each frame of a video using the gradient vector flow variation of the active contours fitting method (see Materials and methods and Fig. S1; ; ). Time sequences of cell edge outlines indicated that the prospective rear edge moved inwards before forward advancement of the front edge (). To quantify this, we used consecutive cell outlines to measure the extent of protrusion or retraction at points along the cell edge for each time point () and generated maps of cell edge movement over time (). A region of continuous retraction at the cell rear (, blue boxes) could reliably be identified in cells before the overt initiation of directed motility. There was also a region of continuous protrusion at the cell front (, red boxes), but this generally started after rear retraction was initiated (9/11 cells).
Although motility initiation was a continuous process, we typically observed three characteristic phases during motility initiation (). Phase I was a period of slow rear retraction (∼10– 40 nm/s) of highly variable duration, ranging from 20 to 430 s. In addition, by imaging cells with a fluorescent volume marker, we found that the lamellipodial thickness decreased at the prospective rear during phase I, whereas no changes in lamellipodial thickness were observed elsewhere (Fig. S2, available at ). The types of transients observed in phase I were not unique to cells initiating motility but were also within the range of edge fluctuations observed in stationary cells. Phase II was a period of fast (∼50–180 nm/s) and sustained rear retraction in which the rear edge rapidly moved inwards toward the cell body for 40–170 s, leading to morphological polarization. In contrast to phase I, phase II seemed specific to those cells initiating motility. This was followed by phase III, in which the cell rear advanced persistently. Persistent protrusion of the cell front coordinated with movement of the cell rear usually began during phase III (8/11 cells).
Next, we examined transients in F-actin network flow during polarization and motility initiation using FSM and multiframe correlation tracking. We performed detailed tracking of F-actin network flow in five cells undergoing motility initiation. For each F-actin flow vector, we calculated the radial component of velocity (v) and the centripetal deviation (φ), which was defined as the angular displacement between the radial component of velocity and the original velocity vector (). In stationary cells, the radial velocity generally decreased from the cell periphery to the center, and the centripetal deviation fluctuated around zero ( and Video 5, available at ). During phase I, slow edge retraction at the prospective cell rear was accompanied by a slight increase in radial velocity and an increase in the magnitude of centripetal deviation in the perinuclear region (, left; and Video 6), indicating that instead of flowing centripetally toward the cell body, F-actin flow was biased in the direction of prospective cell movement. Phase II was characterized by a large increase in the radial velocity at the cell rear and continued alignment of the F-actin flow in the perinuclear region toward the eventual direction of cell movement (, right; and Video 6).
To measure when and where changes in F-actin flow occurred during motility initiation, we distinguished F-actin dynamics in the perinuclear and peripheral regions of the lamellipodia based on the AF546-phalloidin fluorescence pattern. The F-actin network in the perinuclear region was denser and often formed circular bands, whereas the peripheral region had less dense labeling (). We also defined four sectors of equal angular size—front, back, and two sides—based on the eventual direction of cell movement. In the five cells tracked while initiating motility, the first sign of polarization was an increase in radial velocity (∼15–40%) at the rear perinuclear and peripheral regions and a reorientation of F-actin flow in the perinuclear region toward the prospective direction of movement, which was manifested in an increased magnitude of centripetal deviation (). These changes coincided with phase I. The changes in actin dynamics in the rear and side perinuclear regions were often accompanied by a decrease in the radial velocity in the front perinuclear region, suggesting a high coordination of F-actin dynamics in the perinuclear ring. However, no changes occurred in the front peripheral region during phase I. Notably, flow pattern variations similar to those observed in phase I were also observed in some cells while stationary ( at 140–180s shows extreme variation before motility initiation). This suggests that flow transients of phase I are not specific to motility initiation and probably reflect inherent variation in F-actin flow in stationary cells. However, changes in F-actin flow characteristic of phase I always preceded phase II during motility initiation.
Approximately 5–20 s before the start of phase II, the rate of change of radial velocity increased in the rear perinuclear and peripheral regions, leading to the high radial velocity at the cell rear characteristic of phase II, approximately two- to threefold faster than basal speeds (; top). These changes were specific to the cell rear and were not observed at the prospective cell sides (, top). The increased magnitude of centripetal deviation in the perinuclear region was maintained during phase II and also occurred in the side peripheral regions (; bottom).
In phases I and II, no changes in flow speed were detectable in the front peripheral region. However, the speed of F-actin retrograde flow with respect to the substratum in lamellipodia of fully polarized cells is lower than that of the centripetal flow in stationary cells (), implying that F-actin flow at the prospective cell front must decrease at some point. We followed cells that had initiated motility (), and, during phase III, in which the cell front and rear moved persistently in a coordinated fashion, we observed a period of maturation of the polarized form. The cell speed increased, accompanied by a decrease in F-actin retrograde flow speed relative to the substratum at the cell front (). Thus, flow changes at the front occurred only during the later stages of motility initiation and were a consequence, not a cause, of motility initiation. In addition, during maturation, the perinuclear F-actin bands collapsed toward the cell rear (Fig. S3 and Video 8, available at ), where they formed an actin axle characteristic of motile cells ().
Microtubules and myosin II were heterogeneously distributed in the lamellipodium of stationary keratocytes (), and asymmetries in these distributions might influence cell polarization. Microtubules can regulate actin dynamics and are implicated in the establishment and maintenance of cell polarity during the migration of fibroblasts and epithelial cells (; ), whereas myosin has a polarized distribution in motile keratocytes (). In addition, for neutrophils, polarization requires the recruitment of PIP to the cell membrane (). To determine whether any of these molecules might contribute to motility initiation, we developed a quantitative assay for motility initiation based on the observation that the frequency of motility initiation was enhanced by increasing temperature (see Materials and methods Pharmacological treatments section).
Under control conditions, ∼45% of stationary cells spontaneously broke symmetry and initiated motility within 30 min of a temperature shift from 20 to 30°C (). Depolymerization of microtubules with nocodazole had no effect on the frequency of motility initiation, implying that microtubules were not required for motility initiation of keratocytes. Likewise, inhibition of PIP production by the addition of LY294002, a phosphatidylinositol phosphate 3-kinase (PI 3-kinase) inhibitor, had no effect (). To confirm that PIP was not required for motility initiation of keratocytes, we used a GFP-tagged pleckstrin homology (PH)–Akt construct (), which binds to the 3′ phosphorylated lipid products generated by PI 3-kinase, to visualize the PIP localization in keratocytes. GFP-PH-Akt was uniformly localized to the cell membrane. Notably, there was lack of enhancement of GFP-PH-Akt to the leading edge of motile keratocytes or to the periphery of stationary keratocytes. Furthermore, when circular stationary keratocytes were forced to polarize and initiate motility by pushing of the cell rear, no redistribution of GFP-PH-Akt was observed (unpublished data). Together, this suggests that PI 3-kinase and the lipid products of PI 3-kinase are not required for polarization and motility initiation of keratocytes.
However, the inhibition of myosin II with blebbistatin () decreased the fraction of cells initiating motility to 10% (), suggesting that myosin II was required for motility initiation. Nonmuscle myosin II is primarily regulated by Rho kinase, myosin light chain kinase (MLCK), and myosin light chain phosphatase. Used at concentrations effective in keratocytes, the Rho kinase inhibitor Y-27632 but not the MLCK inhibitor ML-7 () mimicked the effect of myosin II inhibition by blebbistatin (). Rho kinase phosphorylates myosin light chain phosphatase, inhibiting its phosphatase activity and, thereby, increasing phosphorylation of the myosin II light chain. Calyculin A, an inhibitor of Ser/Thr protein phosphatases types I and IIA and a known inhibitor of myosin light chain phosphatase, inhibits dephosphorylation of the myosin II light chain, potentiating myosin II activity and leading to constriction of the central actin–myosin ring in sea urchin coelomocytes () and increased F-actin flow in newt lung epithelial cells (). In motile keratocytes, calyculin A increases myosin-dependent inwards F-actin flow at the cell rear (unpublished data). Therefore, we hypothesized that calyculin A might facilitate symmetry breaking in keratocytes by increasing myosin II activity. Indeed, we found that 100% of cells initiated motility with calyculin A treatment (). Importantly, these drugs did not influence the morphology, polarity, or F-actin organization of motile keratocytes (Fig. S4, available at ) and had little effect on the steady-state speed. Together, these data implicate myosin II activity regulated by Rho kinase as a critical factor in symmetry breaking and motility initiation of keratocytes.
To investigate the effect of myosin II or Rho kinase inhibition on F-actin organization, drug-treated cells were fixed and stained with fluorescent phalloidin to visualize the F-actin network. Cells treated with blebbistatin and Y-27632 were larger and flatter, and F-actin tended to accumulate around the cell periphery, whereas circular F-actin bands were absent from the perinuclear region (). In contrast, ML-7 had no detectable effect on the F-actin network (), which is consistent with its lack of effect on motility initiation. Myosin phosphatase inhibition by calyculin A at low concentrations also had no detectable effect (). In some cases, increased phalloidin staining in the perinuclear region was observed at higher concentrations or longer incubations with calyculin A. These results suggest that perturbation of F-actin network organization, particularly in the perinuclear region, inhibits motility initiation.
We also measured the effect of the various drugs on F-actin network flow. Cells were imaged by FSM before and after drug treatment, and F-actin movement was calculated by multiframe correlation tracking. We characterized the F-actin flows by radial velocity and directional coherence, a measure of how similar the orientations of the F-actin flow vectors in a local region were to each other. Perfectly coherent flow had a value of 1 (see Materials and methods section Analysis of displacement fields generated…). All drug treatments affecting symmetry breaking altered F-actin flow. Myosin II inhibition by blebbistatin decreased the radial velocity and coherence in the perinuclear and peripheral regions (), with a greater effect in the perinuclear region. The Rho kinase inhibitor Y-27632 decreased F-actin flow coherence, with a stronger effect in the perinuclear region than the peripheral region (). These changes in F-actin dynamics were consistent with the F-actin staining in cells showing perinuclear F-actin network disruption by myosin II or Rho kinase inhibition. In contrast, ML-7, which had no effect on the frequency of motility initiation or F-actin organization, also had no effect on F-actin flow (). Calyculin A, which increased the frequency of motility initiation, increased the radial velocity both in the perinuclear and peripheral regions (), supporting the notion that calyculin A promotes motility initiation by increasing actomyosin contractility.
Based on these results, we hypothesized that symmetry breaking could be induced by the local activation of myosin II. First, we tested whether calyculin A treatment of keratocytes promoted motility initiation via the same mechanism as in normal untreated cells. We imaged a cell initiating motility spontaneously in the presence of calyculin A. Like untreated cells undergoing spontaneous motility initiation, the action started by retraction of the prospective rear edge accompanied by an increase in F-actin flow at the prospective cell rear and reorientation of the perinuclear F-actin flow toward the prospective direction of movement (). Again, no change in F-actin flow was observed at the front peripheral region. Although the transients in F-actin network flow were faster, motility initiation by calyculin A stimulation appeared to operate through the same pathway as identified for untreated cells.
Next, we sought to stimulate asymmetric contraction with the expectation that we might direct symmetry breaking and motility initiation. We locally applied calyculin A to one side of a stationary cell and indeed found that the cell would break symmetry and move away from the source of calyculin A ( and Video 7, available at ). 39% (15/38 cells) initiated motility away from the calyculin A source within 5 min, with 29% (11/38 cells) polarizing without initiating motility, and only 26% (10/38 cells) remained stationary. In contrast, when cells were exposed to local perfusion with medium lacking drug, 13% (3/24 cells) initiated motility away from the pipette and 13% (3/24 cells) polarized without initiating motility, whereas most cells (67%; 16/24) remained stationary. We were unable to induce motility initiation by a local decrease in contractility at the presumptive front (via local application of blebbistatin or Y-27632); this may have been the result of unfavorable kinetics of drug entry or intracellular diffusion of the drug or drug–target complex. Thus, a local increase in actin–myosin contraction mediated by the local application of calyculin A could induce cell polarization from the cell rear.
This study establishes the sequence of morphodynamic events and spatio-temporal reorganization of the F-actin network during spontaneous symmetry breaking of fish epidermal keratocytes. In contrast to polarization and motility initiation of neutrophils in response to chemoattractants, where initial morphological changes occur at the prospective cell front and periphery (; ), here, the first events were detected at the prospective cell rear and perinuclear region (). Changes at the cell front occurred later in motility initiation and appeared to be consequences rather than causes of cell polarization.
In unpolarized stationary keratocytes, the F-actin network flows centripetally from the cell periphery to the cell body. Based on changes in cell morphology, we defined phases of motility initiation: phase I (slow rear retraction) and phase II (fast rear retraction). These phases were associated with characteristic changes in F-actin dynamics, suggesting that morphological changes are directly mediated by reorganization of the F-actin network (). During phase I, F-actin flow increased at the prospective cell rear and reoriented in the perinuclear region such that it was biased along the eventual direction of cell movement. The flow speed increase appeared to be tightly coupled to the change in flow orientation, as we were unable to distinguish whether one event occurred before the other. The radial velocity usually decreased in the front perinuclear region during phase I, but F-actin flow was never altered in the front peripheral region. The changes in F-actin flow observed during phase I is most likely a consequence of fluctuations in actin polymerization kinetics and increased perinuclear actomyosin contraction, similar to variations that can occur in stationary cells. In fact, cells exhibiting transient F-actin dynamics similar to those in phase I did not necessarily initiate motility and often remained stationary. Phase I appeared to be necessary but not sufficient for motility initiation.
The requirement for Rho kinase and myosin II activity for symmetry breaking in our data is consistent with experiments identifying a requirement for Rho, Rho kinase, and myosin light chain phosphorylation in the spontaneous polarization and migration of Walker 256 carcinosarcoma cells (; ) and in the polarization of neutrophils in response to a chemoattractant (). In stationary keratocytes, Rho kinase and myosin II were required for perinuclear F-actin band formation and normal F-actin flow. We suggest that the perinuclear F-actin bands in stationary keratocytes assemble by dynamic network contraction (), a model proposed for the behavior of the actomyosin network in lamellipodia of motile keratocytes (). The dynamic network contraction model postulates that at the periphery of the lamellipodium, myosin bipolar filaments cross-link a dendritic F-actin network without contraction. Toward the rear of the lamellipodium, network contractility is enhanced as the size of the myosin clusters increases, and the myosin clusters are able to align F-actin into bundles. In stationary keratocytes, this bundling activity is dependent on Rho kinase. We propose that during symmetry breaking, asymmetric actomyosin contraction is transformed into the morphological events of cell polarization via the consequential asymmetry of F-actin flow. According to the network contraction model, this mechanical signal from the cell center to the cell periphery triggers an autocatalytic positive feedback loop in that the more asymmetric the flow, the more effective the contraction and bundle alignment leading to further flow asymmetry.
The implication of Rho kinase in perinuclear actin dynamics but not peripheral actin dynamics is consistent with Rho kinase activity in neuronal growth cones, which is responsible for the maintenance and movement of actin arcs in the center (C domain) of the growth cone lamellipodium but not for peripheral (P domain) F-actin retrograde flow (). Similarly, Rho kinase has been implicated in regulating myosin activity and stress fiber formation in the interior and rear of fibroblasts (), and, in neutrophils, Rho, Rho kinase, and activated myosin are restricted to and define the cell rear (). However, unlike keratocytes, activation of backness signals alone (e.g., through the expression of a constitutively active myosin light chain mutant or the addition of fMLP (formyl-methionyl-leucine-phenylalanine) to pertussis toxin–treated cells) in neutrophils does not lead to correct polarization and cell migration. Instead, pertussis toxin–treated neutrophils exposed to a gradient of fMLP polarize in the opposite direction to normal, with their rear closest to the fMLP source, and do not initiate movement (). This and other results indicate that frontness is normally dominant over backness in neutrophils and that chemotactic signals at the front are required for normal polarization. Likewise, neutrophil polarization depends on PIP localization to the cell front (), whereas spontaneous keratocyte polarization is independent of PIP. Backness signals alone in keratocytes can trigger polarization possibly because, unlike neutrophils, keratocytes are not known to respond to chemoattractants and can polarize in the absence of external chemical cues.
It remains to be determined whether the propagation of organizational information from the cell rear to the cell front during polarization also involves chemical signals besides mechanical processes. It has been proposed for neutrophils and lamellipodial fragments that the functional incompatibility of the F-actin assemblies characteristic of the front and back causes them to segregate into separate domains (; ). Furthermore, changes in actomyosin contractility around the cell body during motility initiation could generate hydrostatic pressure, which could affect protrusion at the leading edge (). In simpler model systems of actin-based motility, such as polystyrene beads coated with the actin nucleating proteins ActA or N-WASP, spontaneous symmetry breaking and motility initiation depend on purely mechanical effects. For small beads in which actin turnover is rapid, symmetry breaking results from cooperative interaction among a large population of polymerizing actin filaments, amplifying small stochastic events to generate large-scale polarity (). In large beads, symmetry breaking results from the accumulation of strain in the actin gel followed by release of elastic energy via breakage of cross-links or filaments (; ).
Keratocytes were cultured as described previously (). To obtain individual cells, sheets of keratocytes were disaggregated by incubating in 85% PBS and 2.5 mM EGTA, pH 7.4, for 4.5–5 min. YFP myosin regulatory light chain (a gift from A.F. Straight, Stanford University, Stanford, CA) and GFP-PH-Akt (a gift from T. Meyer, Stanford University, Stanford, CA) was transfected into keratocytes as described previously ().
AlexaFluor546 phalloidin (AF546-phalloidin; Invitrogen) was used to visualize F-actin dynamics in live keratocytes using FSM (; ). AF546-phalloidin was introduced into keratocytes with a small volume electroporator for adherent cells (provided by M.N. Teruel and T. Meyer, Stanford University, Stanford, CA; ; ). 2 μM AF546-phalloidin was premixed with 7.5 μM deoxy-ATP, 7.5 μM deoxy-GTP, and 5 μM deoxy-CTP in water for ∼15 min at room temperature to prevent phalloidin aggregation. Cells were electroporated with 20 μl of the phalloidin mixture with three pulses at 150 V and were allowed to recover for ∼10 min before viewing on an inverted microscope (Diaphot-300; Nikon). Phalloidin staining of F-actin and staining of microtubules were performed as described previously (; ).
Phalloidin was conjugated to AlexaFluor546 or tetramethylrhodamine for live cell FSM or fixed cell labeling, respectively. Myosin regulatory light chain and PH-Akt were visualized in live cells by conjugation to YFP or GFP, respectively. FITC-conjugated secondary antibodies were used to detect the primary antibody in tubulin labeling of fixed cells. Live cell imaging was performed at room temperature or at 25°C by mounting the coverslips on a temperature-controlled chamber, with the exception of temperature shift experiments (see next section) in which the temperature was shifted from 20 to 30°C. Cells were imaged in culture media ( Leibovitz's L-15 medium without phenol red supplemented with 14.2 mM Hepes, pH 7.4, 10% FBS, and 1% antibiotic-antimycotic). Time-lapse phase contrast and epifluorescent images were acquired using an inverted microscope (Diaphot-300; Nikon) with a 40× NA 1.3 oil Fluor or 60× NA 1.4 oil plan-Apo objective (Nikon). For FSM of AF546-phalloidin, images were acquired every 2 or 3 s with the 60× oil objective. A 20× NA 0.4 air phase-contrast objective was also used for some image acquisition. Images of fixed cells were collected with a microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) using a 63× NA 1.4 oil plan-Apochromat objective (Carl Zeiss MicroImaging, Inc.). All time-lapse and fixed images were collected with a cooled back-thinned CCD camera (MicroMax 512BFT; Princeton Instruments) with a 2× optovar attached using MetaMorph software version 6 (Molecular Devices). Adjustments to brightness and/or contrast were performed with MetaMorph or Photoshop (Adobe), and pseudocolor overlays were made with Photoshop.
Cell outlines were calculated using a variation of the active contours algorithm () derived from the gradient vector flow method of . The method was custom written in MATLAB 7 (The MathWorks). First, a nonlinear sigmoidal scaling of pixel intensities was applied to the image to stretch the contrast of the cell margin while reducing the contrast of edge responses inside the cell, such as those around the cell body (Fig. S1). This was followed by bandpass filtering to reduce noise and fine detail. Then, the edge map was calculated as the squared magnitude of the image gradient. A vector field oriented toward the steepest edges and zero elsewhere was generated from the gradient of the edge map. This vector field was then diffused iteratively as described previously () and normalized. After diffusion, the vector field was oriented toward the edges of interest even in relatively smooth regions of the image. An active contour representation of the cell outline was initialized by the cell outline from the previous time point. For the first frame of a time-lapse sequence, the contour was initialized manually. Subsequently, the contour was deformed iteratively by minimizing an energy function that used the edge-oriented vector field to attract the contour to the position of image edges while controlling the internal tension and curvature of the contour (). The contour minimizing this energy function was taken as the cell outline for that image (Fig. S1).
Using these cell outlines, the movement of the cell boundary was determined according to the mechanical model described by using software provided by M. Machacek (The Scripps Research Institute, La Jolla, CA). The cell boundary positions were translated to polar coordinates, and the scalar map function (see Scalar map function section) was used to generate continuous space-time plots of protrusion and retraction.
An adaptive multiframe correlation technique () was used to measure the movement of fluorescent image features from the FSM image sequences and, thus, the velocity of F-actin network flow. We implemented modifications to the technique to remove stationary background features before correlation for tracking of F-actin flow in stationary cells. Signals from these background features were very weak but contributed to the integrated correlation score because of their intensity stability and consistency. In many cases, they created a false global maximum at zero displacement (Fig. S5 A, available at ). To eliminate signal contributions from stable background features, we calculated the time-averaged intensity of each pixel in the correlation window and subtracted the average map pixel by pixel from both the original and the shifted templates before correlation. This maintained the characteristic signal variation in a template for cross-correlation while removing stationary signals. Consequently, the global correlation maximum at zero displacement was suppressed, whereas the secondary local maximum associated with the true velocity was preserved (Fig. S5 B).
Once cells entered rapid motility, they moved too quickly to still assume stationary F-actin network flow over the 7–10 frames required for the application of multiframe correlation tracking. Therefore, the videos were transformed into a cell frame of reference using a rigid body approximation of whole cell motion (). Multiframe correlation tracking was then performed on the transformed videos using sequences of overlapping five-frame windows and a template size adjusted between 13 and 23 pixels. The flow fields generated were transformed back into the original lab frame of reference for comparison with measurements from stationary cells.
To confirm the results obtained from adaptive multiframe correlation tracking, we analyzed a subset of our stationary cell FSM data with spatio-temporal image correlation spectroscopy (STICS). STICS was performed on three consecutive nonoverlapping 20-frame time windows with partially overlapping subregions of 16 × 16 pixels as previously described (; ). STICS averaged the actin network flow over longer time scales compared with multiframe correlation tracking, producing qualitatively similar results and trends with slightly lower absolute speeds.
For analysis of flow patterns in stationary cells before and during symmetry breaking, each flow vector was transformed into polar coordinates with the center of the polar coordinate system defined as the cell center. The radial component of velocity and the centripetal deviation (i.e., the angle between the radial velocity component and the original velocity vector) were calculated (). These data were presented as continuous scalar maps of the radial flow speed and centripetal deviation. The directional coherence of the flow vectors was defined as the similarity in orientation among vectors inside 1 of 12 radial sectors defined over the lamellipodium. For each sector, the flow vectors were normalized to unit speed and averaged. The magnitude of the mean vector was taken as the coherence score for the sector. The maximum coherence score was one (all vectors in the sector aligned along the same direction).
For continuous space-time plots of protrusion and retraction and scalar field visualizations of radial velocity and centripetal deviation, it was necessary to resample data that were not measured regularly in the dimensions along which they were being plotted. A custom scalar map function implemented in MATLAB 7 (The MathWorks) was used to resample data on a regularly spaced grid. The value at each grid point was interpolated from measurements within a given radius and weighted using a Gaussian window centered at the grid point. Then, bicubic interpolation between grid points was performed to generate continuous images. Color maps were then applied to the continuous maps.
Fig. S1 illustrates the automated cell outline determination method. Fig. S2 shows that the prospective rear edge thickness decreases early during motility initiation. Fig. S3 illustrates the formation of an actin axle at the cell rear during maturation of the polarized form. Fig. S4 shows that perturbation of myosin II activity does not perturb the polarity of motile cells. Fig. S5 illustrates the removal of stationary background features in adaptive multiframe correlation tracking. Videos 1 and 2 show F-actin network movement in stationary and motile keratocytes, respectively, visualized with FSM. Videos 3 and 4 depict stationary keratocytes spontaneously breaking symmetry and initiating motility. Video 5 shows the radially symmetric F-actin flow field in a stationary keratocyte. Video 6 shows changes in the F-actin flow field in a keratocyte initiating motility. Video 7 shows a stationary keratocyte initiating motility in response to the local application of calyculin A. Video 8 illustrates the formation of an actin axle at the cell rear during the maturation of polarity. Online supplemental material is available at . |
Epithelial–mesenchymal transition (EMT) is fundamental during early development for the reorganization of groups of progenitor cells within a preexisting epithelium into complex sets of juxtaposing tissues, thus allowing new inductive patterning events necessary for organogenesis. In addition to the formation of the three germ layers and the neural crest, EMTs give rise to, amongst others, sclerotome from ventral somites; endocardial cushions of the atrio-ventricular (A-V) canal of the heart from the endocardial epithelium; and bones, cartilage, and musculature from cells of the neural plate (for review see ; ).
The process of EMT enables epithelial cells, which are normally constrained by cell-to-cell adherence, to delaminate and take on the invasive and migratory properties that are a feature of mesenchymal cells. Aberrant reactivation of this process in the adult can give rise to undesirable pathological states such as tumor progression, whereby delamination of the tumor cells and acquisition of invasive and migratory properties is the forebear of metastatic disease (for review see ). Loss of E-cadherin expression facilitates the loss of cell-cell adhesion. This is often achieved during EMT by the induction of repressors of E-cadherin transcription, such as Snail, ZEB1/δEF1 (), ZEB2/SIP1 (), E47 (), Twist (), and Slug/Snail2 ().
In vivo, the importance of E-cadherin transcriptional repressors in EMT is demonstrated by the role of Snail in EMT. Snail-null mice cannot complete gastrulation because they fail to down-regulate E-cadherin to undergo EMT (). Evidence is also accumulating that indicates a role for Snail and other transcriptional repressors of E-cadherin in cancer progression. In malignant tumors in general, E-cadherin expression correlates with more highly differentiated cancers that exhibit cell–cell adhesion and are less invasive (). Elevated expression of Snail1, ZEB1, and Twist have been found in metastatic cancer cell lines and invasive human cancers, with concomitant loss of E-cadherin expression in these cell lines and tissues (; ). Although these observations clearly illustrate that regulation of E-cadherin and its transcriptional repressors are key elements in embryonal and pathological EMT, the molecular events that initiate the EMT are ill defined.
The TGFβs are pleiotrophic cytokines that regulate cellular functions ranging from embryonal development to maintaining homeostasis in the adult. The three mammalian TGFβ isoforms (TGFβ1, TGFβ2, and TGFβ3) show high sequence homology and have overlapping functions. Most cells, including epithelial cells, make TGFβs and express TGFβ receptors (); hence, the TGFβs can act in a paracrine, autocrine, and/or endocrine fashion (). The role of TGFβ in EMT, where it regulates cell–cell adhesion, cytoskeletal remodelling, and cell-matrix adhesion, is well documented (for review see ). In embryonal development, TGFβ plays a role in organogenesis in the late stages of embryogenesis. The requirement for TGFβ in EMT in the formation of the endocardial cushions during cardiac valve development () and in EMT involved in palate fusion () support a role for TGFβ in EMT in vivo. TGFβ has also been shown to play opposing roles in oncogenesis and metastasis (; ). In the early stages of carcinogenesis, its epithelial growth suppressive properties liken its role to that of a tumor suppressor. However, in tumor cells that have progressed to evade its anti-proliferative effects, apoptosis and anoikis, TGFβ can promote oncogenic EMT leading to an increase in invasiveness and motility, the beginnings of metastasis.
Phosphorylation of proteins on tyrosine residues by tyrosine kinases and the reversal of the process by protein tyrosine phosphatases (PTPs) regulate many diverse cellular functions from proliferation to differentiation. The PTP Pez is an intracellular PTP that is located at the adherens junctions (AJ) in endothelial and epithelial cells (, ). Dominant-negative mutants of Pez induce an increase in tyrosine phosphorylation at the AJ, which together with the observation that Pez directly dephosphorylates β-catenin () suggests that Pez plays a role in regulating the tyrosine phosphorylation status of the AJ. However, Pez can be localized elsewhere in the cytoplasm and in the nucleus (), suggesting that it may have other targets and functions apart from regulating the phosphorylation status of the AJ. Here, we show that Pez expression in zebrafish embryos is developmentally regulated and that knocking down its expression results in embryos with defective organ development. In vitro and in vivo studies revealed a role for Pez in regulating TGFβ and EMT, processes critical for the formation of new tissues and organs during embryonic development.
The temporal and spatial expression profiles of a gene in the context of a whole organism can shed light on its physiological function. To assess whether Pez has a developmental role, we examined its temporal and spatial expression profiles during zebrafish embryonal development up to 48 h post-fertilization (hpf) by whole-mount in situ hybridization. With the exception of maternally derived mRNA present up to 4 hpf (unpublished data), Pez expression was not detectable until 16 hpf, at which time it became faintly detectable throughout the embryo (unpublished data). Between 22–24 hpf Pez expression was strongest in the ventricular/subventricular zone of the developing brain (), a region rich in proliferating and migrating neuronal progenitors. Pez expression in the brain subsequently declined, with very low levels remaining at 36 hpf (). Pez expression was also detected in the somites at 24 hpf, but was no longer detectable by 36 hpf (; Fig. S1, available at ), suggesting that Pez may play a role in the later stages of somite maturation. Transient Pez expression was also observed during cardiac development. Pez expression was visible in the early developing heart tube at 24 hpf, reaching a peak between 42–44 hpf and then rapidly down-regulated, so that by 48 hpf little Pez expression remained (). Another tissue in which Pez expression was observed is the pectoral fin, where little or no Pez expression was detected in the fin bud at 36 hpf, but more sustained expression was observed, between 42 to 48 hpf, the latest time point investigated. These examples of transient Pez expression at specific stages of tissue development suggest that Pez plays a regulatory role in development.
To determine whether Pez plays a crucial role in the regulation of developmental processes, three different antisense morpholino- oligonucleotides (MOs) (), designed to inhibit translation of Pez protein, were microinjected into zebrafish embryos at the one- to four-cell stage. Embryos injected with a control MO were indistinguishable from uninjected controls, whereas injection with each of the anti-Pez MOs produced identical morphant phenotypes.
Pez morphants displayed a number of defects in several organs and tissues (summarized in Table S1, available at ) that correlated with regions of Pez expression (see ). Most notable was a severe pericardial edema obvious by 2–3 dpf (, B and C; Fig. S2). From in situ hybridization (), we had noted that Pez expression generally peaks after the rudimentary tissues and organs have formed, suggesting that Pez may play a role in the development of specific structures within the organs rather than in initiating organ development. In concordance with this hypothesis, the organs were formed in the Pez morphants, but with architectural defects in multiple organs and tissues. Somites and somite boundaries were formed in Pez morphants, but the boundaries were irregular (). In the ventricular/subventricular zone of the brain, a region with strong Pez expression in wild-type embryos at 24 hpf (), there was an increase in cellular density and expansion of the ventricular zone at the expense of the cortex () in the Pez morphants compared with uninjected or control MO- injected embryos. Similarly, structural defects were observed in the formation of the pharyngeal arches, which included a higher cellular density, shortening, and poor formation of the cartilage ().
The severe pericardial edema observed in Pez morphants (, B and C; Fig. S2) is symptomatic of defects in cardiac function. To further evaluate cardiac function in Pez morphants, we used fli1-GFP transgenic zebrafish, which express GFP in the endothelial cells comprising the endocardium (), to visualize structure and function of the atrium, ventricle, and the atrio-ventricular (A-V) valve in live embryos. By 53 hpf, the developing heart of control-injected embryos had begun to loop and the atrium and ventricle were partially superimposed (). The A-V valves between the two chambers were well developed and opened and shut in synchrony with rhythmic contractions (; Video 1, available at ). In contrast, the heart of Pez morphants remained tubular, suggesting a defect in looping, and they lacked patent A-V valves (; Video 2), which normally function to prevent regurgitation of blood back into the atrium during rhythmic contractions. We also observed a loss of unidirectional blood flow in the atrium of Pez morphants when the movements of blood cells in the atrium were tracked (), supporting a lack of functional A-V valve.
The expression profile of Pez during zebrafish development indicates that its expression is low or absent until required, whereupon it is highly up-regulated in specific tissues and contributes to organogenesis. To elucidate the molecular events induced by such a marked increase in Pez expression, we overexpressed Pez in MDCK epithelial cells and selected for clonal lines (referred to as Pez-MDCK cells). Pez overexpression transformed the normal epithelial morphology of MDCK cells into a fibroblast-like morphology (), whereas cells transfected with the empty expression vector (vector-MDCK) remained indistinguishable from parental MDCK cells. The transformation to fibroblast-like morphology appeared to occur in two steps: an initial scattering phenotype (, middle) that was evident ∼2–3 wks after transfection and clonal selection was followed by a more spindly and elongated morphology (, right) typical of epithelial cells that have undergone EMT. The Pez-MDCK cells appeared to have lost cell–cell contact, consistent with loss of junctional E-cadherin visualized by indirect immunofluorescence (). Total E-cadherin protein expression was examined in four representative Pez-MDCK clones and found to be highly down-regulated compared with a pool of vector-MDCK cells (), suggesting that the loss of junctional E-cadherin was not merely due to a translocation of the protein away from the cell junctions. Similar results were obtained in a repeat of the experiment in which three individual vector-MDCK clones and another four representative Pez-MDCK clones were analyzed (unpublished data). Analysis of E-cadherin mRNA expression in these clones by qRT-PCR indicated that E-cadherin mRNA was also markedly down-regulated (). Averaging the data from four independent Pez-MDCK clones and three vector-MDCK clones, a >30-fold reduction in E-cadherin mRNA was observed in Pez-MDCK clones. The Pez-MDCK clones also had markedly up-regulated fibronectin (an extracellular matrix protein secreted by mesenchymal cells) compared with the vector-MDCK clones (). In addition, using a Boyden chamber assay to measure motility, two representative Pez-MDCK clones were found to be highly motile compared with vector-MDCK cells (). The alteration of cellular morphology accompanying Pez overexpression in MDCK cells, together with the loss of the epithelial marker E-cadherin and gain of mesenchymal marker (fibronectin) and function (motility), suggests that Pez had induced an EMT.
Many studies have shown that the loss of E-cadherin expression associated with EMT is a consequence of transcriptional repression of the E-cadherin gene, particularly by the zinc finger transcriptional repressors such as Snail and Slug (for review see ). We compared the mRNA levels of the E-cadherin repressors Snail, Slug, ZEB1, and ZEB2 between Pez-MDCK and vector-MDCK cells. Snail and Slug expression were elevated in all four Pez-MDCK clones analyzed, averaging ∼5- and ∼8-fold higher, respectively, than expression in the vector-MDCK clones (), while ZEB1 and ZEB2 were induced ∼9- and ∼30-fold, respectively (). The marked loss of E-cadherin mRNA along with the induction of expression of the E-cadherin repressors Snail, Slug, ZEB1, and ZEB2 suggest that the Pez-induced reduction in E-cadherin is due to transcriptional repression.
We noted that ZEB1 levels were relatively constant between the Pez-MDCK clones whereas Snail, Slug, and ZEB2 mRNA levels were highest in clone D, intermediate in clone A, and lower in clones G and K (). Upon further passaging, Snail expression in both clones A and D declined to the lower levels expressed by clones G and K (unpublished data), whereas ZEB1 expression remained relatively unchanged. These observations are consistent with the transient nature of Snail and ZEB2 expression noted previously (), and suggest that at the time of the analysis shown in , down-regulation of Snail and ZEB2 expression in individual Pez-MDCK clones had proceeded to different extents.
TGFβ has previously been shown to induce an EMT in MDCK and other epithelial cells through induction of all or a subset of the transcription repressors Snail, Slug, ZEB1, and ZEB2 (), all of which were induced in Pez-MDCK cells. We therefore investigated the possibility that Pez may be mediating its effects through the action of TGFβ. By qRT-PCR, we found that the mRNAs for TGFβ-1, -2, and -3 were 5, 3, and 15× higher, respectively, in the Pez-MDCK cells than in the vector-MDCK cells (). To determine whether increased TGFβ mRNA expression in the Pez-MDCK cells was accompanied by increased secretion of active TGFβ capable of inducing an EMT, conditioned medium from the Pez-MDCK cells cultured for 72 h was collected and used to culture parental MDCK cells. After 1–2 d, parental MDCK cells cultured in Pez-MDCK–conditioned medium began to lose cell–cell contact and scatter, whereas cells cultured in control parental MDCK cell conditioned medium remained unaltered (unpublished data). After 4–6 d in culture in Pez-MDCK conditioned medium (, Pez CM), the parental MDCK cells underwent a transformation to a spindly fibroblast-like morphology resembling that of Pez-MDCK cells () and cells treated with rhTGFβ (, rhTGFβ) that have undergone an EMT. In contrast, parental MDCK cells cultured in conditioned medium from parental (, parental MDCK CM) or vector-MDCK cells (unpublished data) remained epithelial and indistinguishable from cells cultured in normal medium (unpublished data). These observations suggest that Pez-MDCK conditioned medium contained a soluble factor that was able to promote EMT in a similar fashion to rhTGFβ, and that this factor was not present at levels sufficient to induce EMT in conditioned medium obtained from parental MDCK or vector-MDCK cells. The transition to a mesenchymal morphology induced by Pez-MDCK conditioned medium was inhibited by the addition of the inhibitor of TGFβ receptor kinase activity, SB-431542, and by a blocking pan-TGFβ Ab to the Pez-MDCK–conditioned medium, but not by an anti-HGF Ab (), suggesting that the EMT-promoting factor secreted by Pez-MDCK cells was active TGFβ.
In Pez-MDCK cell lines, Pez expression has persisted over a long period of time, raising the possibility that TGFβ secretion and subsequent signaling may be an indirect consequence of Pez overexpression. In mammalian cells, ligand binding to the TGFβ receptor triggers downstream signaling by phosphorylation of the R-Smads, which then complex with the coSmad, Smad4, and translocate into the nucleus (for review see ). To address the kinetics of Pez-induced TGFβ signaling, we initially examined the effect of Pez expression on Smad4 nuclear translocation in transient transfection assays at 24 h post-transfection, when at least 50% of the Pez-transfected cells showed elevated Pez expression compared with empty vector-transfected cells (). In control empty vector-transfected MDCK cells, Smad4 remained predominantly cytoplasmic (). In contrast, in MDCK cells transfected with Pez, Smad4 was translocated into the nucleus (), suggesting that Pez overexpression induces Smad-dependent TGFβ signaling. Addition of a neutralising TGFβ Ab following Pez transfection blocked nuclear translocation of Smad4 (), confirming that the nuclear localization of Smad4 was TGFβ specific. The ablation of Smad nuclear translocation by the exogenous addition of TGFβ Ab also indicates that the activation of TGFβ signaling by Pez involves the secretion of active TGFβ.
Examination of Pez expression and TGFβ signaling at earlier time points after Pez or vector transfection showed no detectable increase in Pez expression in the Pez-transfected cells compared with vector-transfected cells at 6 h post-transfection, but by 12 h post-transfection a small proportion of cells with elevated Pez expression was evident (), indicating that elevated Pez expression occurs between 6 to 12 h after transfection. Examination of Smad4 subcellular localization at 12 h post-transfection showed an increase in nuclear Smad in cells with elevated Pez expression, suggesting that canonical TGFβ signaling is induced shortly after Pez expression (). The expression of TGFβ mRNAs was also assayed by qRT-PCR up to 24 h post-transfection, but no induction was detected (unpublished data), suggesting that TGFβ production induced by Pez initially occurs by post-transcriptional mechanisms. No analyses were performed beyond 24 h post-transfection, as Pez expression began to decline after this point (unpublished data), making the interpretation of data obtained beyond 24 h post-transfection equivocal.
The induction of TGFβ signaling and mRNA synthesis by Pez in vitro prompted us to investigate if this is paralleled in vivo. Published data on the pattern of TGFβ3 expression () revealed that its expression closely resembled that seen for Pez in zebrafish embryos. We therefore compared the profiles of Pez and TGFβ3 expression during zebrafish development and found that they coexpressed in the same subset of cells in the ventricular zone of the brain at 24 hpf () and in the endocardial cells of the heart at 44 hpf (). We also noted that TGFβ3 was expressed in some tissues where Pez was not expressed, such as in the lens of the developing eye at 24 hpf ( and ). In addition to the co-temporal expression of Pez and TGFβ3, the high degree of spatial correlation, particularly in the ventricular zone of the brain () and in the heart (), is highly consistent with the notion that in these developing tissues Pez may up-regulate TGFβ3 expression. To explore this further, we examined the expression of TGFβ3 in Pez morphants at 24 hpf. Unlike embryos at later stages of development, the morphology and architecture of the developing brain and heart tube of the Pez morphants at 24 hpf are not overtly different from control MO-injected or wild-type embryos. Although morphologically similar at 24 hpf, expression of TGFβ3 was undetectable in the ventricular zone of the brain and the heart tube in the Pez morphants, in contrast to control embryos where TGFβ3 was clearly expressed in these tissues (). It is also worth noting that in the lens of the developing eye where Pez is not expressed, TGFβ3 expression was unaffected by knocking down Pez expression ().
EMT is a process shown to be essential for cell movement to facilitate the formation of new tissues and organs during development. Collectively, several lines of evidence generated from this study indicate that Pez is a novel regulator of TGFβ-mediated EMT in vitro and of organogenesis during embryonal development. Epithelial MDCK cells overexpressing Pez underwent morphological and functional changes, including the acquisition of cell motility, characteristic of an EMT. The morphological and functional changes were accompanied by and are likely to be a consequence of changes in gene expression typical of EMT, including loss of E-cadherin and up-regulation of the EMT-inducing transcription factors, Snail, Slug, ZEB1, and ZEB2. The transient nature of Pez expression during zebrafish development which correlated with specific stages of organ development suggests that Pez plays a role in regulating organogenesis. This is further supported by data obtained from zebrafish hypomorphic for Pez expression where we observed developmental defects in all organs that expressed Pez during development. Given that Pez can promote EMT in vitro, some or all of the defects observed in these morphant embryos may be due to an inability to induce EMT during organ development.
The role of TGFβ as a driver of developmental EMT is well recognized although little is known of the mechanisms that initiate TGFβ activation. In these studies we made several observations suggesting that Pez is a novel inducer of TGFβ signaling. In Pez-MDCK cell lines, TGFβ secretion is up-regulated to a level capable of promoting EMT. In addition, transient Pez overexpression induces TGFβ secretion and canonical Smad-dependent TGFβ signaling within 6 h of its expression, suggesting that TGFβ signaling is activated as an immediate-early response to Pez expression. The early activation of TGFβ by Pez was, however, not accompanied by increases in TGFβ mRNAs, suggesting that this initial activation occurs by post-transcriptional mechanisms which may be due to activation of latent TGFβ (for review see ) or de-repression of translation (). At later stages following Pez expression, TGFβ mRNAs were up-regulated as seen in Pez-MDCK cell lines that had undergone an EMT. It is well documented that TGFβ induces its own transcription (; ) and we suggest that the initial activation of canonical TGFβ signaling by Pez acts to induce subsequent de novo synthesis of TGFβ mRNAs by an autoinductive mechanism. This is supported by studies in our laboratory showing that exogenous addition of rhTGFβ to MDCK cells induced TGFβ and TGFβ mRNAs after several days of stimulation (unpublished data). Furthermore, in other studies we have also found that the exogenous addition of rhTGFβ to MDCK cells induced the downstream effectors of EMT, Snail1 (Fig. S3), ZEB1, and ZEB2 (unpublished data), albeit this also took 3 d or more of continuous stimulation to achieve substantial levels of induction. Together, these observations suggest that Pez initially induces TGFβ activation by post-transcriptional mechanisms and the resultant activation of canonical TGFβ signaling then acts in an autocrine feed-forward mechanism to activate subsequent transcription of TGFβs and the downstream effectors of EMT. Finally, the high degree of correlation between Pez and TGFβ3 expression in the developing brain and heart of zebrafish embryos and the loss of TGFβ3 expression in these tissues when Pez was knocked down strongly supports our in vitro data that Pez regulates TGFβ production.
In the developing heart, rhythmic contractions signifying the beginning of cardiac development are evident at ∼22 hpf and development continues for several days (). Although Pez expression was evident at 24 hpf, strong expression was not observed until 36 hpf (unpublished data), suggesting that it is not essential for the early stages of cardiac development, a notion supported by the development of atrium, ventricle, and rhythmic contractions in the Pez morphants. Instead, strong Pez expression in the heart coincides with looping of the heart tube and specification of the AV canal, which begin around 36 and 37 hpf, respectively (; ). In the early steps of A-V valve development, the designated endocardial cells of the A-V canal undergo a TGFβ-dependent EMT to form the endocardial cushions, which precede formation of the A-V valves (; ). In zebrafish development, this begins at ∼45–48 hpf when Notch1b expression, which is required in conjunction with TGFβ signaling to induce EMT (), is restricted to the A-V canal (; ). Our studies show that Pez expression persists until ∼48 hpf and in 44-hpf embryos both Pez and TGFβ3 colocalize to the endocardium, coinciding precisely with the timing of restricted Notch1b expression at the A-V canal and the onset of EMT. The defects in cardiac valve development and loss of TGFβ3 expression when Pez is knocked down strongly suggest that Pez plays a role in initiating TGFβ3- induced EMT in cardiac valve development. However, TGFβ isoforms have also been found to be expressed in newly formed mesenchyme as a consequence of EMT (; ) and that they play subsequent roles in further valve development and remodelling. If Pez expression results in activation of latent TGFβ, Pez may have a more far-reaching role in cardiac valve formation beyond the initial induction of EMT.
Zebrafish embryos hypomorphic for Pez expression showed a loss of TGFβ3 expression and increased cellularity in the ventricular zone of the developing brain at the expense of the cortex, indicative of increased proliferation of the progenitors and/or lack of migration into the cortical region. It is therefore of interest to note that TGFβ plays a dual role in neurogenesis of the developing cerebral cortex. It encourages the progenitor cells of the ventricular zone to exit cell cycle, thereby maintaining an appropriate number of cycling cells (), and it promotes migration of the post-mitotic cells into the emerging cortex (), where they can undergo differentiation into mature neurons. The increase in cellular density and expansion of the ventricular zone observed in the Pez morphants is therefore consistent with a loss of TGFβ3 signaling, leading to an inability of the neuronal precursors to exit cell cycle and migrate out of the proliferation zone.
Although our in vitro and in vivo data suggest that Pez is a crucial regulator of TGFβ signaling, the lack of correlation between Pez and TGFβ3 expression in some tissues such as the somites suggest that Pez may also mediate its effects by TGFβ-independent pathways or through TGFβ1- or TGFβ2-mediated pathways. Pez expression in the developing somites is restricted to the central mesenchymal cell compartment surrounded by the epithelium (see Fig. S1) and occurs in a very narrow window after all the somites have formed and when EMT and re-epithelialization by mesenchymal–epithelial transition is complete. Consistent with the expression profile of Pez, somite boundaries were formed and maintained in Pez morphants, albeit ragged and uneven in appearance. At this stage, it is unclear what the role of Pez is in somite development and further work is underway to elucidate this.
Finally, loss of E-cadherin accompanied by induction of its repressors, Snail, Slug, ZEB1, and ZEB2 in various combinations have been observed in a number of epithelial cancer cell lines that have undergone an EMT (; ). Furthermore, TGFβ signaling is emerging as a dual regulator of oncogenesis and metastasis (). The data reported here suggests that Pez has the potential to play a role in oncogenic EMT. It is therefore of interest to note that () reported a number of somatic mutations in Pez associated with colorectal cancers, although how these mutations affect Pez expression or function is unknown. Based on our findings here, we would suggest that the mutations may cause changes either in Pez expression or activity that could lead to dysregulation of TGFβ signaling.
Anti–E-cadherin and anti-fibronectin mAbs were from Transduction Laboratories, anti-tubulin mAb was from Abcam, goat anti-actin pAb was from Santa Cruz Biotechnology, Inc. Anti-TGF-β1,-β2, -β3, and anti-HGF mAbs were from R&D systems. The TGFβ receptor kinase inhibitor, SB-431542, was obtained from Tocris Bioscience, UK.
MDCK cells were maintained in DMEM (JRH Biosciences) supplemented with 10% fetal bovine serum (JRH Biosciences), or 5% serum for conditioned medium experiments. MDCK cells were transfected with pcDNA3 eukaryotic expression vector (Invitrogen) or N-terminal Flag-tagged human Pez cDNA cloned into pcDNA3 () and stable transfectants selected by resistance to 500 μg/ml Geneticin (G418; Promega). Single clones of Pez-MDCK cells and either single clones or a pool of vector- MDCK clones were isolated for analysis.
Cells were lysed in Triton X-100 lysis buffer (50 mM Hepes, pH 7.5, 150 mM sodium chloride, 10 mM sodium pyrophosphate, 5 mM EDTA, 50 mM sodium fluoride, 1 mM sodium orthovanadate, and 1% Triton X-100 with protease inhibitor cocktails) and whole-cell lysates analyzed by Western blotting with the indicated primary antibodies and appropriate HRP-conjugated secondary antibodies detected using enhanced chemiluminescence (ECL Plus; GE Healthcare).
Total RNA was extracted from cells using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. 1–2 μg total RNA was primed with 10 μM pd(N)6 random hexamer (GE Healthcare) and reverse transcribed using Omniscript reverse transcriptase (QIAGEN) according to the manufacturer's instructions. 2 μl of each 20-μl reverse transcription reaction was used as cDNA template for qRT-PCR, using SYBR Green 2x RT-PCR mix (QIAGEN) and 1 μM of each primer. Primer sequences and annealing temperatures for primers are shown in . Reactions were performed on a Rotor-gene RG-3000 (Corbett Research), with the following cycle parameters: 95°C, 15 min; 94°C, 30 s; 50–65°C, 20 s, 72°C, 20 s × 30–35 cycles. Melt curves were analyzed from 72–99°C and products run on 2% agarose gels to verify homogeneity and correct size of all products.
Migration assays were performed using a modified Boyden chamber assay (Transwells, 6.5-mm diameter, 8-μm pore size; CoStar). 5 × 10 cells were plated in the upper chamber in serum-free medium (DMEM; JRH Biosciences), and migration toward 10% FBS quantified 24 h post-plating using an MTS/PMS-based assay (CellTiter 96 AQueous Non-Radioactive Cell Proliferation Assay; Promega) according to the manufacturer's instructions. Percentage of migrated cells was derived by comparison to a standard curve generated using known cell numbers. Standards were assayed in duplicate, and MDCK clones in quadruplicate.
Whole-mount in situ hybridization probe templates for zebrafish Pez (zfPez) and zfTGF-β (zfTGF-β3) were generated by PCR amplification from adult (1 yr) zebrafish cDNA, and cloned into pBluescript SK (+) vector (Stratagene). For zfPez, a 600-bp region of the zebrafish Pez 3′ UTR was amplified using forward primer: 5′-GTGTCCGCGGTCCAGCACCTGTCAGGATTT-3′ and reverse primer: 5′-GTGTGAATTCGATCAGTCTGGAGTTTTTCAGGA-3′, based on EST sequence BQ285767.1. For zfTGF-β3, the 1.2-kb coding region was amplified using forward primer: 5′-GTGTGAATTCATGCATTTGGGCAAAGGACTG-3′ and reverse primer: 5′-GTGTGAATTCCCGCGGTCAGCTGCACTTGCAGGATTTG-3′, Based on GenBank sequence . DIG-11-UTP–labeled (Roche) sense and antisense probes were generated by in vitro transcription (MAXIscript in vitro transcription kit; Ambion) according to the manufacturer's instructions.
Wild-type zebrafish embryos were collected 0–48 hpf, chorions removed, and embryos fixed in 4% formaldehyde/PBS for a minimum of 24 h. In situ hybridization was performed as described previously () but using 0.2–0.5ng/μl sense or antisense RNA probe for hybridization, 1:4,000 dilution of 0.75 U/μl alkaline phosphatase–conjugated anti-DIG antibody (Roche) at 4°C for 16–18 h for antibody binding, and 340 μg/ml 4-nitroblue tetrazolium chloride (NBT)/175μg/ml 5-bromo-4-chloro-3-indolyl-phosphate toluidine salt (BCIP) stain solution (Roche) for signal detection. Where indicated, yolk proteins were cleared with benzylbenzoate/benzyl alcohol (2:1) post-staining, as described in before image capture.
The human Pez mRNA sequence (GenBank accession no. ) was used to blast the zebrafish EST database for homologous sequences. ESTs spanning the 5′ (GenBank accession no. ) and 3′ (GenBank accession no. ) ends of the zebrafish Pez coding sequence were identified. Forward (5′-TCATGTGTTTGTCTTGTGGAG-3′) and reverse (5′-TGCTGGACACCGTGTATCTC-3′) PCR primers were designed to amplify a 172-bp region spanning the ATG start site. Based on sequence amplified from adult (1 yr) zebrafish cDNA, three independent antisense MOs (GeneTools) (PezMO1-3) were designed to block translation by binding upstream of or spanning the start site. Wild-type or fli1-GFP transgenic zebrafish embryos were collected and injected at one-cell stage with Pez-specific MO or an irrelevant control using standard procedures. 96-hpf embryos were fixed in 4% formaldehyde/phosphate-buffered saline (PBS) for 16–24 h and either resuspended in 80% glycerol/PBS or embedded in paraffin, and 3-μm sections were cut and stained with hemotoxylin and eosin (H&E). For live imaging, embryos were anaesthetised by the addition of 0.003% tricaine methanesulfonate to embryo medium and time-lapse fluorescent images were collected using a microscope (MVX 10; Olympus) illuminated with a mercury lamp. MO sequences were as follows: PezMO1 (nt −40 to −16): 5′-AGCTTTCCTCTGCGCTTAATCCATG-3′; PezMO2 (nt −70 to −46): 5′-CGACCTTCCTCCAACAGAACGATTC-3′; PezMO3 (nt −19 to +5): 5′-GGCATGTTGAACCCGCGCCGCGAC-3′ where +1 is the A of the ATG start site. GeneTools standard control MO: 5′-CCTCTTACCTCAGTTACAATTTATA-3′. All zebrafish manipulations were performed according to guidelines set out and approved by the University of Adelaide Animal Ethics Committee.
Phase-contrast images were captured using a microscope (IX70; Olympus), UPlanFl 10× NA 0.3 Ph1 and LCPlanFl 20× NA 0.4 Ph1 objectives, and a camera (DP12; Olympus). Color and fluorescent images were captured using a microscope (IX81; Olympus) with 10× UPLSApo NA 0.4, 20× UPLSApo NA 0.75, and 40× UAPO/340 NA 1.15w objectives and CC12 (Soft Imaging System, color) and Hamamatsu Orca-ER (fluorescent) cameras, respectively. Videos were generated from a microscope (MVX10; Olympus) and F-view camera (Soft Imaging System). Fluorescence images were acquired using Cell software (Olympus Soft Imaging System), all other images were acquired and processed using the Analysis software (Olympus Soft Imaging System). For immunofluorescence images where comparisons of staining intensities were made, the images were acquired using the same attenuator and exposure settings.
Two videos are included displaying cardiac morphology and pumping in live control (Video 1) and Pez morphant (Video 2) fli-1-GFP transgenic zebrafish at 53 hpf. In addition, there are three supplemental figures showing a high magnification view of Pez mRNA localization in somite of zebrafish (Fig. S1), pericardial edema in Pez morphants (Fig. S2), and time course of Snail mRNA induction by TGFβ in MDCK cells (Fig. S3), and a table summarizing the defects in Pez morphants (Table S1). Online supplemental material is available at . |
Commissural axons cross the midline by elongating between the floor plate cells and the basement membrane. Dye tracing and ultrastructural studies have shown that commissural axons are restricted to the basal floor plate, growing between the floor plate cells and the underlying ECM/basement membrane (). The avoidance of penetrating floor plate cells, the source of the chemoattractants netrin and sonic hedgehog (; ; ), is puzzling. Two possible mechanisms may account for this pattern of trajectory. The floor plate cells may produce a short-range repellent signal that prevents axonal growth into the floor plate cells or, alternatively, the basement membrane may positively attract axons.
The basement membrane of the floor plate is a site of deposition of various adhesion molecules (; ), among them F-spondin (). , a gene expressed at the ventral midline of the embryonic spinal cord (i.e., the floor plate) encodes a secreted, ECM-attached protein (). It plays a dual role in patterning axonal trajectory in the embryonic spinal cord by promoting outgrowth of commissural axons and inhibiting outgrowth of motor axons and migration of neural crest cells (; ; ). The carboxyl half of F-spondin contains six thrombospondin type-one repeats (TSRs). The TSRs of F-spondin protein are typical of class-two TSRs (). Vertebrate class-two TSR proteins are represented in the nervous system by F-spondin, mindin, subcommissural organ (SCO)–spondin and two remotely related proteins, heparin-binding growth-associated molecule and midkine ().
The TSR domain of F-spondin is proteolytically processed. The cleaved products of F-spondin have different adhesive properties: the fifth and sixth TSRs (TSR5 and TSR6, respectively) bind to the ECM whereas the TSR1-4 fragment does not bind to the ECM (). In this paper we demonstrate that the two fragments of F-spondin, which are generated in vivo by a proteolytic cleavage, are spatially restricted to different extracellular compartments at the midline. An outgrowth-promoting fragment, TSR6, is deposited at the basement membrane that underlies the floor plate, whereas an outgrowth-inhibiting fragment, TSR1-4, is bound to the floor plate cell surface via binding to lipoprotein receptor–related protein (LRP) receptors apolipoprotein E receptor 2 (ApoER2), LRP2/megalin, and LRP4. The spatial localization of F-spondin fragments forces commissural axons to deflect from the repulsive TSR1-4 and extend on the permissive TSR6. Consequentially, commissural axons elongate between the basement membrane and the floor plate cells, rather than into the floor plate cells, which are the source of the chemoattractants netrin and sonic hedgehog. Hence, two novel posttranslational modifications elicit F-spondin activity: the coordinated generation of two functionally opposing polypeptides from a single protein by proteolysis and immobilization by membranal receptors.
The outgrowth-promoting activity of F-spondin resides in its carboxyl half, the TSR domains. F-spondin is proteolytically processed in vivo between the amino-terminal reelin/spondin and the carboxyl-terminal TSR domain. The TSR domain of F-spondin is further cleaved into three fragments. Plasmin cleavage between repeats 1–4, 5, and 6 generates a nonadhesive TSR1-4 fragment and adhesive TSR5 and 6 fragments (). To test whether the diverse adhesive properties of the TSRs of F-spondin are also reflected in their outgrowth-promoting potential, we tested the outgrowth of E6 chick dorsal spinal cord neurons on cell surface–immobilized F-spondin substrates. Glycosylphosphatidylinositol (GPI)-anchored forms of TSR1-4, 5, and 6 were transiently expressed in COS cells. A confluent culture of mixed F-spondin–expressing cells marked by EGFP and nonexpressing cells served as a substrate for the dissociated neurons. Chick E6 dorsal spinal cord neurons preferably elongate on TSR5 and 6 (), and avoid growing on TSR1-4 (). The ratio between neurites growing on F-spondin fragments versus neurites growing on the nonexpressing cells was calculated. Cultured EGFP-expressing COS cells served as a control (). The outgrowth on TSR5 and 6 was significantly higher than the control whereas the outgrowth on TSR1-4 was significantly lower than the control (). Similar results were obtained in an outgrowth assay with rat E13 spinal cord neurons cultured on purified TSR1-4 and 6 proteins (Fig. S1, available at ). Thus, the dual activity of F-spondin in either promoting or inhibiting neurite outgrowth may result from the activity of different domains; the TSR1-4 fragment inhibits outgrowth whereas the TSR5 and 6 fragments support outgrowth.
The pattern of expression of the TSR fragments of F-spondin protein in chick embryos was examined with R3 antibody raised against TSRs 3–6 of chick F-spondin (). Epitope mapping with the isolated TSRs indicates that R3 recognizes TSR3 and 5 (Fig. S2, available at ). Hence, R3 recognizes the repulsive (TSR3) and the outgrowth-promoting, ECM-attached domain (TSR5) of F-spondin that flank the plasmin cleavage site P2 (). F-spondin immunoreactivity is first detected at stage 21. As the pioneer commissural axons invade the midline, expression is detected on the apical floor plate cells and the pia that underlies the floor plate cells (). Costaining with axonal marker (antineurofilament 3A10 mAb) reveals that F-spondin accumulates beneath and along the pathway of the crossing axons (). After stage 24, expression was also detected on the crossing fibers of the commissural axons (stage 26; ). The segregation between the apical floor plate cells and the axonal and pial staining persists through development (stage 27, the last stage analyzed was stage 29; ). The pial staining is not restricted to the basement membrane underlying the floor plate. It also spreads lateral to the floor plate (, arrowheads). This suggests that F-spondin binding to the pia is governed by a ubiquitous, not midline-specific, basement membrane component.
To test the subcellular deposition site of F-spondin fragments, the nonadhesive TSR1-4 and the adhesive TSR6 were expressed at the chick embryonic floor plate. Cell-specific expression was achieved by electroporation of DNA using the floor plate–specific enhancer III of the gene (Fig. S3 A, available at ; ). The TSR1-4 and 6 fragments were expressed at the floor plate at stages 12–14 along with cytoplasmic EGFP. F-spondin fragments were tagged with the myc epitope at the amino terminus. The localization of the ectopically expressed fragments was analyzed at stages 22–24. The TSR1-4 fragment labeled the floor plate cell membrane (, compare myc staining with the cytoplasmic EGFP staining). The staining obtained for the TSR1-4 fragment resembles the apical floor plate staining obtained with the R3 antibody (). The TSR1-4 fragment is likely to accumulate on the cell membranes of the expressing cells. In contrast, the TSR6 fragment labeled the basement membrane underlying the floor plate (), similar to the basal staining acquired with the R3 antibody (). Thus, the different TSR domains of F-spondin accumulate in different floor plate compartments. The nonadhesive TSR1-4 fragment binds to the surface of the floor plate cells, whereas the adhesive TSR6 fragment is deposited in the ECM that underlies the floor plate cells.
Two plasmin-sensitive sites were predicted based on amino acid sequence and were found, in vitro, to be plasmin sensitive: between the first and second antiparallel β strands of TSR5 and between TSR5 and 6 (). To test whether the predicted cleavage sites are indeed targets for floor plate–derived proteases, a double-tagged F-spondin protein was used. EGFP and myc epitope were fused to the amino and carboxyl ends of the TSR domain, respectively, and expressed in the floor plate at stages 12–14. At stages 22–24, the intact TSR domain EGFP–TSR1-6–myc gave rise to two distinct labeling patterns. TSR6 accumulated at the ECM that underlies the floor plate cells whereas the TSR1-4 domain bound to the floor plate cells (). A protein that lacks the adhesive fifth and sixth repeats, EGFP–TSR1-4–myc, was deposited on the floor plate cells, as revealed by identical staining with either myc or EGFP (). Thus, F-spondin is processed in vivo and gives rise to two proteolytic fragments that either bind to the membrane of floor plate cells or are deposited at the ECM.
Corroboration of the predicted cleavage sites was assessed with two mutant forms of F-spondin: deletion of the putative cleavage sites by deleting TSR5 and the region between TSR5 and 6 (TSR1-6Δ5) and point mutations in the plasmin cleavage sites (TSR1-6m; ). The mutated proteins were flanked with EGFP and myc tags. On electroporation of the mutated proteins into the floor plate, the amino and carboxyl tags labeled the floor plate cells (). There was no accumulation of TSR6 at the basement membrane. The complete overlapping of the amino and carboxyl tags suggests that deleting or mutating the plasmin cleavage sites generates a plasmin-resistant F-spondin protein. The unprocessed TSR domain of F-spondin, generated by the various mutations, accumulates on the membrane of floor plate cells rather than at the ECM that underlies the floor plate.
To test whether plasmin is involved in the processing of F-spondin, the serine protease inhibitor aprotinin was used. Aprotinin was injected repeatedly into the lumen of the spinal cord in ovo. The pattern of the endogenous F-spondin and exogenous double-tagged F-spondin EGFP–TSR1-6–myc was studied. After aprotinin application, the endogenous protein, as revealed by the R3 antibody, binds uniformly to the floor plate cells (). The segregation in the deposition of F-spondin between the apical floor plate and the pia that is evident in the floor plate of untreated embryos () was eradicated. In addition, the pial staining was reduced (), suggesting that the release of TSR5 and 6 from TSR1-4 is inhibited. To test directly whether aprotinin inhibits F-spondin processing, EGFP–TSR1-6–myc was expressed at the floor plate, followed by aprotinin treatment. The amino and carboxyl tags labeled the floor plate cells in an overlapping manner (), resembling the pattern of the mutated proteins (). The resistance of F-spondin to cleavage after aprotinin injection provides evidence that F-spondin processing occurs extracellularly. Thus, the cleavage of F-spondin by serine protease is required for the diffusion and accumulation of the adhesive TSRs at the basement membrane.
The localization of the TSR1-4 fragment, together with the finding that it does not support neurite outgrowth in vitro, implies that it may prevent the growth of commissural axons into the floor plate cells. To test the potential inhibitory effect of the TSR1-4 fragment on commissural axons, the fragment was ectopically expressed in the lateral chick neural tube. Expression of a guidance molecule in the presumed responding neurons can hinder the ability of the neurons to respond to the exogenous, target-derived guidance molecule. Thus, an alternate reporter/guidance molecule approach was used. This method generates two different expression patterns: EGFP in the dI1 neuronal subpopulation using the enhancer () and a guidance molecule in nondI1 cells using a cytomegalovirus (CMV) enhancer (). dI1 interneurons give rise to commissural and ipsilaterally projecting axons ( and Fig. S3, B–D). Embryos were electroporated at stages 17 and 18 and analyzed at stage 26. The expression of TSR1-4 in nondI1 cells was barely detectable (), suggesting that the ectopically expressed TSR1-4 does not bind avidly to the cell surface of the lateral spinal cord neurons. The presence of TSR1-4 along the axonal trajectory pathway of dI1 axons did not alter their projection. dI1 cells projected axons in commissural and ipsilateral patterns (). The projection pattern is similar to the axonal pattern of embryos electroporated with Math1-Cre and EGFP-Cre–dependent plasmids ().
The inability of TSR1-4 to inhibit axonal outgrowth in vivo contradicts the inhibitory activity obtained in vitro with a substrate-bound protein ( and Fig. S1; ), thus suggesting that the repulsive activity of TSR1-4 is context dependent and may require immobilization of the protein. It is conceivable that the TSR1-4 protein is anchored to the floor plate cells by binding to a floor plate–specific receptor, whereas the ectopically expressed protein is soluble. We hypothesized that a membrane-tethered form of F-spondin may mimic the immobilized form of the protein, yielding its inhibitory effect. Therefore, a GPI signal was added to the carboxyl end of the TSR1-4 fragment.
A TSR1-4/EGFP alternate plasmid with a Math1-Cre plasmid were expressed in the chick neural tube. In contrast to TSR1-4, the staining of ectopic TSR1-4 is intense, and cell bodies as well as the axons express it, as revealed by the myc- positive axons at the contralateral side of the spinal cord (). The expression of TSR1-4 along the trajectory pathway of dI1 axons resulted in substantial elimination of the diagonally crossing commissural axons (). The number of diagonally crossing axons toward the floor plate was evaluated. A mean of 5.45 ± 2.2 axons per section was scored in the TSR1-4–treated embryos ( = 20) whereas 1.48 ± 1.35 axons were scored in the TSR1-4–treated embryos ( = 25; ). The repulsion of dI1 axons for TSR1-4, demonstrated in this protocol, is specific for commissural axons because motor neurons expressing either TSR1-4 or EGFP projected laterally in a normal manner (unpublished data). These results support the hypothesis that the endogenous TSR1-4 fragment of F-spondin should be associated with the floor plate cell surface to exert its repulsive effect. TSR1-4 and TSR1-4 have no effect on cell fate, as determined by cell fate markers (Fig. S4, A and B, available at ).
The TSR1-4 fragment of F-spondin binds to the ApoER2 (). Fluorescent in situ hybridization with the chick probe demonstrates that is expressed in the lateral floor plate and in the ventral ventricular zone. A low level of expression is also evident at the midline (). Thus, is expressed at the floor plate, mostly at the lateral floor plate cells. Therefore, ApoER2 may bind to F-spondin, immobilize it, and present it to the commissural axons.
The possible immobilization of TSR1-4 by ApoER2 was tested in vitro. COS cells were transfected with a myc-tagged TSR1-4 or cotransfected with TSR1-4 and ApoER2. Surface staining (unfixed cells) with 9E10 mAb was used to reveal immobilization of TSR1-4 to the cell membranes. Cells expressing () or coexpressing TSR1-4 and a secreted form of ApoER2, ApoER2 (), did not present TSR1-4 on their cell surface. However, coexpression of ApoER2 and TSR1-4 resulted in the immobilization and presentation of TSR1-4 on the cell surface of the expressing cells (). TSR1-4 did not bind to DCC when coexpressed in COS cells (Fig. S5, A–C, available at ). Thus, ApoER2 is required for the cell surface immobilization of the TSR1-4 fragment of F-spondin. Consistent with the in vitro experiment, when expressed in the chick hemitube, the combination of ApoER2 and TSR1-4 yielded intense cell surface staining in vivo (). The staining, however, is confined to cell bodies and is not detected on axons.
The possible binding of F-spondin to other LRP receptors was tested using the aforementioned immobilization assay. The TSR1-4 fragment of F-spondin binds to very-low-density lipoprotein receptor (VLDLR), LRP4, and megalin (Fig. S5, D–L). The expression pattern of genes was analyzed by mRNA in situ hybridization. is expressed in the floor plate and in the ventricular zone (). is expressed at embryonic day (E) 4 in the dorsal third neural tube (). Its expression overlaps the region of commissural neurons. At E5 the expression of spreads to the ventricular zone and the floor plate (). is expressed in the dI1 subpopulation of commissural neurons (). Thus, LRP receptors that bind F-spondin are expressed at the floor plate and in commissural axons.
To test whether ApoER2 may collaborate with the TSR1-4 fragment of F-spondin in inhibiting commissural axon outgrowth, TSR1-4 and ApoER2 were jointly expressed unilaterally at the chick neural tube using TSR1-4/EGFP and ApoER2/EGFP alternating plasmids with a Math1-Cre plasmid. Ectopic expression of ApoER2 at stage 18 had no effect on cell fate as determined by cell fate markers (Fig. S4 C). Coexpression of TSR1-4 and ApoER2 along the dI1 axonal pathway resulted in an axonal failure to turn diagonally toward the floor plate (). Axons seem to circumvent the ApoER2/TRS1-4 domains. The ratio between the accurate, diagonally crossing axons and the erroneous axon growing in the ventral lateral neural tube is 1.09 ± 1.07 ( = 76; ). Electroporation of ApoER2 did not cause any dI1 axonal errors and axons projected diagonally toward the floor plate (). The ratio between normal and erroneous axonal projection in ApoER2 + TSR1-4 ectopic expression is significantly different from ectopic ApoER2 (4 ± 2.41, = 18) and ectopic TSR1-4 (5.58 ± 4.39, = 25; ).
In ectopic TSR1-4/ApoER2 expression, EGFP axonal labeling is evident in the white matter on the ipsi- and contralateral sides of the floor plate, indicating that dI1 axons elongated in the white matter rather than in the neuroepithelium expressing ApoER2/TSR1-4. Thus, the attraction to the floor plate is retained. The extent of commissural axonal lateral deflection and the failure to cross the midline to the contralateral side in ApoER2/TSR1-4 is smaller than the erroneous phenotype obtained with TSR1-4. This may result from the larger spinal cord domain that is occupied by TSR1-4 neuronal cell bodies and axonal processes as opposed to ApoER2-immobilized TSR1-4 that is restricted to cell bodies.
ApoER2 binds to the nonadhesive TSR of F-spondin TSR1-4. The adhesive TSR5 and 6 do not bind to ApoER2 (). This provides the means to specifically block the immobilization of the endogenous TSR1-4 fragment of F-spondin using a secreted (ecto) form of ApoER2. To test whether ApoER2 can block the binding of TSR1-4 to the membranal ApoER2, COS cells were transfected with myc-tagged TSR1-4, ApoER2, and various concentrations of ApoER2. At an ApoER2/ApoER2 ratio of 1:10, no binding of TSR1-4 to the cell surface was detected (). Coexpression experiments with ApoER2 and other LRP demonstrated that it blocks the binding of F-spondin to LRP4, megalin, and VLDLR (unpublished data).
Thus, ApoER2 can serve as a dominant-negative form of ApoER2 that blocks the binding and immobilization of the endogenous TSR1-4 to the endogenous floor plate–derived LRPs ApoER2, megalin, and LRP4. ApoER2 was expressed ectopically at the ventral spinal cord. The ventral ectopic expression of ApoER2 yielded erroneous axonal projection at the midline. Axons are detached from the ventral bundle that underlies the floor plate and turn dorsally within the floor plate cells (64%, = 115; and M). ApoER2 has no patterning effect when expressed in the lateral and ventral spinal cord and no axonal guidance effects when expressed in the lateral spinal cord (unpublished data). However, it cannot be excluded that ApoER2 may directly affect axon guidance specifically at the floor plate.
F-spondin activity at the midline might be shared by other class-two TSR proteins that are expressed at the central nervous system. Mindin and SCO-spondin (SCO-spondin contains 13 class-one and 12 class-two TSR repeats) are expressed at the floor plate (; ; ). The TSR of mindin and 19 out of the 25 SCO-spondin TSRs are not adhesive (as predicted by the absence of basic amino acids at their third antiparallel β strand) and thus resemble TSR1-4 of F-spondin. The TSRs of SCO-spondin bind with ApoER2 (Fig. S5, M–R) and thus may also serve as a repellent cue for commissural axons. The binding of the endogenous LRPs and class-two TSR proteins can be blocked by secreted LRP and also by the dominant-negative form of TSR1-4. In vitro TSR1-4 does not promote outgrowth of spinal cord neurons. However, the inclusion of an adhesive TSR together with TSR1-4 (as in the TSR1-5 protein) converts in vitro the outgrowth activity of F-spondin to a promoting one (). The mutated noncleavable forms of F-spondin bind to the surface of floor plate cells () via binding to the floor plate LRPs and probably compete with the endogenous TSR1-4. The recruitment of the adhesive TSR5 and 6 (TSR1-6m) or the TSR6 (TSR1-6Δ5) to the cell surface of floor plate cells may either serve as a gain of function (positioning of the outgrowth promoting modules on the surface of floor plate cells) or a loss of function (competing away the repulsive TSR1-4 with a nonrepulsive TSR mutant proteins).
The mutant forms of the TSR domain were electroporated into the ventral spinal cord. Commissural axons encountering the TSR1-6Δ5 (82%, = 100; and M) and TSR1-6m (81%, = 147; and M) turned within the floor plate dorsally into the floor plate cells. Electroporation of EGFP (6%, = 100) or the nonadhesive fragment TSR1-4 (18%, = 100) did not alter axonal trajectory at the floor plate (). The erroneous projection of the commissural axons was always confined to the ventral midline, though the mutated proteins were expressed nonspecifically throughout the entire ventral or lateral/ventral neural tube. This suggests that the adhesive TSR5 and 6 are not sufficient to alter commissural outgrowth but rather function as dominant-negative TSR1-4 that inhibits the binding of the endogenous TSR1-4 to the endogenous floor plate LRPs. In support of this, expression of a membrane-tethered form of TSR6 (TSR6) at the ventral neural tube did not cause pathfinding errors at the floor plate (19%, = 100; ). These results support the hypothesis that the TSR1-4 fragment of F-spondin serves as a repellent that prevents the invasion of commissural axons into the floor plate.
#text
Fertilized white leghorn chicken eggs were incubated at 38.5–39°C. A DNA solution of 5 mg/ml was injected into the lumen of the neural tube at either Hamburger and Hamilton stages 12–14 (CMV enhancer in pCAGG plasmid) or stages 17 and 18 (Math1 and HoxA1 enhancers, provided by T. Lufkin, Mount Sinai School of Medicine, New York, NY). Electroporation was performed using 3 × 50 ms pulses at 25 V applied across the embryo using a 0.5-mm tungsten wire and an electroporator (ECM 830; BTX). Embryos were incubated for 2–3 d before analysis.
Aprotinin (Protosol [500,000 kallikrein inactivator units/ml]; Kamada) was injected into the lumen of the spinal cord at 10-h intervals from stages 16–23 (five injections). Embryos were analyzed at stage 25.
For the R3 antibody, embryos were fixed in 4% paraformaldehyde/0.1 M phosphate buffer and processed for embedding in paraffin. 8-μm sections were cut and collected on Superfrost Plus slides. Paraffin was removed by immersion in xylene and the sections were rehydrated using a graded ethanol/HO series. For antigen retrieval, slides were submerged completely in 10 mM sodium citrate, pH 6.0, in a glass histology box. The buffer was then brought to boiling in a pressure cooker (PickCell Laboratories), and the slides were boiled in buffer for 1 h. Slides immersed in buffer were allowed to cool for 15 min and washed twice for 5 min in PBS at room temperature.
E13 rat and E6 chick dorsal spinal cord neurons were obtained and plated on F-spondin fragments as described previously (). F-spondin proteins TSR1-4_HIS, and TSR6_HIS were affinity purified on an affinity column (Talon; CLONTECH Laboratories, Inc.).
The floor plate–specific enhancer enhancer III was provided by T. Lufkin. The human was obtained from the I.M.A.G.E. Consortium. The chick (ChEST392o8) was obtained from the Chick EST Project. To generate the ApoER2 the sequences downstream to amino acid 472 were deleted using a BglII site. To generate TSR1-4, TSR5 , and TSR6, a synthetic DNA sequence encoding the prion GPI signal (TCTAGATCCAGCGCGGTGCTGTTCTCCTCCCCTCCTGTGATCCTCCTCATTTCCTTTCTCATCTTCCTGATGGTGGGATGA) was cloned in frame downstream from the TSR fragments of F-spondin.
A minigene was constructed from two partially overlapping EST clones (ChEST548d5 and ChEST437i20) that encompass the third complement-type repeats of chick megalin gene (amino acids 2704–3102), obtained from the Chick Est sequencing project. The ESTs were annealed and “filled in” by PCR using a 5′ (5′-CGAAGCTTTGGAAATGTGACAACGACAATG-3′) and 3′ primer (5′-TAGGATCCCACTGCATTCATTTATACCAC-3′). The PCR product was subcloned into a pSecTagB vector (Invitrogene). A membranal form of megalin was generated by fusing the secreted chick minigene of to the carboxyl end of (transmembrane and cytoplasmatic domains).
Fluorescence in situ hybridization of chick spinal cords was performed as described in the TSA plus protocol (PerkinElmer).
Fig. S1 depicts the outgrowth of rat E13 dorsal spinal cord neurons on a substrate of purified TSR1-4 and 6. Neurons extend neurites only when cultured on TSR6. Fig. S2 depicts epitope mapping of the R3 antibody. R3 recognizes TSR3 and 5. Fig. S3 describes the expression pattern of the floor plate–specific enhancer and the commissural-specific enhancer . Fig. S4 depicts the effect of TSR1-4, TSR1-4, and ApoER2 on the patterning of the neural tube. None of these proteins changes the patterning of neural tube when expressed ectopically in the spinal cord. Fig. S5 illustrates the binding of F-spondin's and SCO-spondin's TSR to LRP receptors. The TSR1-4 fragment of F-spondin binds to LRP2, LRP4, and VLDLR. TSR12 and 26 of SCO-spondin bind to ApoER2. Online supplemental material is available at ). |
Actin filament assembly is essential for numerous cellular events, including the formation of endocytic vesicles at the plasma membrane (; ; ; ). However, the role of actin filament disassembly (a key step in actin dynamics) during actin-mediated endocytosis is less clear. A major modulator of actin filament disassembly is the actin-depolymerization factor/cofilin family of proteins (hereafter referred to as cofilin). These small highly conserved actin-binding proteins are essential regulators of actin dynamics in living cells.
Cofilin and its role in actin dynamics have been studied extensively in vitro. Cofilin binds preferentially to ADP-actin subunits within actin filaments. This binding induces a twist in the filament, accelerates the release of Pi from ADP-Pi subunits, and severs actin filaments (). Severing increases the number of actin filament pointed ends and, in conjunction with capping of barbed ends by capping proteins, stimulates filament disassembly ().
The molecular determinants of cofilin recruitment to actin filaments in vivo have not been fully investigated. demonstrated the restriction of cofilin to zones slightly displaced from the leading edge in keratocytes. This localization was found to be unaffected by manipulations that were expected to change polymerization rates at the leading edge, which should alter the relative abundance of ATP-, ADP-Pi-, and ADP-actin subunits within the filaments. These observations led to the hypothesis that cofilin localization is largely dependent on accessory factors or other regulatory inputs, such as phosphorylation or pH.
In the budding yeast , fixed cell imaging analysis demonstrated that cofilin localizes with actin at sites of endocytosis (; ). Genetic and biochemical analysis of a conditional allele of cofilin showed that the actin filament disassembly activity of cofilin is essential for fluid-phase endocytosis (). A later study indicated that at elevated temperatures, cofilin is important in the initial internalization step of endocytosis (). However, this study did not address the molecular basis for a drastic block in the endocytic delivery of a fluid-phase marker to the vacuole observed in the cofilin mutant at permissive temperatures. Here, we sought to understand cofilin dynamics in live cells, to address the molecular determinants of its recruitment to actin structures in vivo, and to determine how cofilin influences actin dynamics, endocytic internalization, and delivery of cargo to the vacuole.
To investigate cofilin dynamics in live cells, we attempted to tag cofilin with GFP. However, both N- and C-terminal fusion proteins were nonfunctional. Therefore, we attempted to make a functional construct by creating internal in-frame fusions of GFP to cofilin. We inserted GFP flanked on both sides by either 12–, 7–, 4– or 0–amino acid linkers after amino acids D34, P48, P58, and N74. One construct, in which GFP was inserted between amino acids N74 and G75 and was flanked on both sides by 12–amino acid linkers (see Materials and methods section Media, plasmids, and strains), complemented a mutation when expressed from a high-copy plasmid (Fig. S1, available at ). The localization and dynamics of GFP-cofilin expressed from high-copy or low-copy plasmids were identical. However, the expression levels of GFP-cofilin from cell to cell were less variable using the low-copy plasmid. Therefore, we used the low-copy plasmid in wild-type cells to express GFP-cofilin for our microscopy studies. These cells exhibited normal cortical actin patch dynamics (unpublished data).
Actin assembly at cortical patches in budding yeast powers the invagination and vesicle scission steps of endocytosis (). To determine the timing of cofilin recruitment to actin patches, we used simultaneous two-color imaging of Abp1–monomeric RFP (mRFP) as a marker for actin at endocytic sites and GFP-cofilin. GFP-cofilin localized to actin patches in live cells ( and Video 1, available at ), which is in agreement with previous studies in fixed cells (; ). We then examined cofilin and Abp1 dynamics at the actin patch. Strikingly, GFP-cofilin associated with patches 3.6 s (±0.6 s) after the initial assembly of actin/Abp1p (, bottom). Analysis of cofilin spatiotemporal dynamics with kymographs () indicated that cofilin associates with patches during the internalization/disassembly phase of the endocytic patch lifetime.
To further document and analyze the temporal delay in cofilin recruitment to actin structures, we examined cofilin localization in mutant cells in which endocytic coat internalization is defective but actin still assembles at cortical patches. In this mutant background, actin comet tails stably associate with the endocytic machinery at the cell cortex (). Actin subunits actively treadmill through these elongated Arp2/3-dependent filamentous actin (F-actin) structures as assessed by fluorescence recovery after photobleaching, with new actin filament assembly taking place at the cell cortex (). Consistent with a temporal delay in cofilin association with actin patches, there is a significant (P < 0.01), spatially resolvable exclusion of cofilin from regions of actin comet tails adjacent to the cell cortex (). The spatial exclusion of untagged cofilin from regions of active actin assembly in cells was verified by immunolocalization of cofilin and actin (Fig. S2, available at ). The localization of these two proteins was quantified as a function of fluorescence intensity along the length of the actin tails, moving from the cell surface to the cell interior (). These observations further strengthened our conclusion that the low-copy GFP-cofilin construct is a faithful reporter of cofilin dynamics in vivo.
We next focused on the molecular mechanisms responsible for the delay in cofilin recruitment to actin filaments. This delay could reflect regulation by phosphorylation, pH, accessory proteins, or the nucleotide-bound state of actin subunits within the filaments. In , cofilin is not phosphorylated, and its severing activity is not regulated by pH (; ). Therefore, we examined the contributions of two proteins that are important for cofilin function in vivo, Srv2p and Aip1p, to cofilin localization within actin tails in cells (; ). Previously, we showed that in the absence of Aip1p, cofilin, which is normally restricted to cortical actin patches, becomes distributed on both patches and cytoplasmic cables (). When either Aip1p or Srv2p was eliminated by gene deletion in cells, no detectable effect on cofilin localization within actin comet tails was observed (unpublished data). In vitro experiments have shown that cofilin has preferential affinity for ADP-actin, but the in vivo relevance of this observation has never been established (). To test this relevance, we used an allele of actin, , which forms filaments that structurally mimic ADP-Pi-actin and that have reduced F-actin disassembly rates in vitro and in vivo (; ). Strikingly, cofilin localization to F-actin tails is quantifiably reduced in cells with as the sole source of actin ().
double mutants, GFP-cofilin in cells also showed an increase in cytoplasmic localization and, in addition, in a subset of cells, was assembled into aberrant cablelike structures (unpublished data).
To further test whether cofilin binds preferentially to ADP-actin filaments in vivo, we used the actin filament–stabilizing molecule jasplakinolide. Jasplakinolide blocks F-actin disassembly, which is expected to deplete the assembly-competent actin monomer pool and stop actin subunit flux through filament networks. Actin nucleotide hydrolysis would be expected to convert ATP-actin into ADP-actin subunits within jasplakinolide-stabilized filaments, which should eliminate the zone of the actin tails that lacks associated cofilin. We generated an strain, which expressed Abp1-mRFP and GFP-cofilin and was sensitized to jasplakinolide by elimination of two multidrug transporters, Snq2 and Pdr5 (). When these cells were treated with jasplakinolide, actin tails began to lengthen, indicating that F-actin disassembly was blocked. After 5 min, no actin subunit flux was detectable by following fiduciary marks in the elongated actin tails using kymographs (unpublished data). Strikingly, cofilin was found to be uniformly distributed along the elongated actin tails ( and Video 2, available at ). This localization effect is quantitatively documented in . As a complimentary analysis, treatment of cells with the actin filament assembly inhibitor latrunculin A (lat A) is also predicted to eliminate ATP and ADP-Pi from actin filaments because new assembly is blocked. Treatment of cells with lat A caused the tails to begin to shorten, and, over time, cofilin was no longer restricted to regions distal to the cell cortex (Fig. S3 and Video 3A). These independent pharmacological approaches to modify actin dynamics through the monomer pool with lat A and through the filamentous population with jasplakinolide implicate the actin filament nucleotide state in controlling cofilin localization in vivo.
Having established the molecular determinants of cofilin recruitment to F-actin in vivo, we wanted to better understand how cofilin contributes to actin subunit flux through filament networks. To do this, we examined the effects of a mutant of cofilin, Cof1-22p, on actin flux in actin comet tails and on tail morphology in mutant cells (). In vitro studies suggested that Cof1-22p has weakened interactions with actin filaments and is defective in actin filament disassembly (; ).
cells expressing actin-GFP had drastic alterations to actin tail morphology (), and the tails were significantly elongated (∼3 μm; P < 0.01) compared with tails in cells with wild-type cofilin function (∼700 nm; ).
cells (P < 0.01; and Video 4, available at ) and averaged ∼55 nm/s in cells with wild-type cofilin function compared with ∼10 nm/s in cells expressing Cof1-22p (). Interestingly, despite the previously reported contribution of Srv2p and Aip1p to cofilin-dependent actin filament turnover (; ), these proteins did not contribute to actin comet tail morphology or subunit flux in cells (unpublished data). These studies provide evidence that cofilin-dependent actin filament disassembly contributes to actin subunit flux in vivo and suggest that cofilin is the major mediator of this process in .
Having established that cells are defective in actin network dynamics in vivo, we wanted to determine how these defects affect endocytosis. In wild-type and cells, we simultaneously imaged the dynamics of Sla1p, an endocytic coat protein, and Abp1p, a marker for actin in endocytic patches, respectively. In wild-type cells, Sla1p assembles at the cell cortex before actin. Subsequent actin assembly, which is marked by Abp1p, drives internalization and triggers disassembly of the endocytic coat (, left; and Video 5A, available at ). Maximum intensity projections of 2-min medial focal plane videos show that assembly and disassembly of the endocytic and actin complexes are restricted to the vicinity of the cell cortex in wild-type cells. In contrast, cells have a dramatic redistribution of F-actin to the cell interior (, right; and Video 5 B). However, the endocytic coat marker Sla1p was not abnormally redistributed to the cell interior.
Examination of the dynamics of the endocytic coat and actin markers at the cell cortex in cells by kymograph revealed unexpected behaviors. First, as in wild-type cells, endocytic coat components assembled at the cell cortex and moved inward concurrent with the burst of actin assembly (, right kymograph). This observation suggests that endocytic internalization occurs in this mutant, which was not expected because of previously reported defects in fluid-phase endocytosis for this mutant (). These behaviors were quantified, and 90% ( = 71) of Sla1-GFP patches internalized in cells versus 93% ( = 76) of Sla1-GFP patches in wild-type cells. Quantification of Abp1p-mRFP and Sla1p-GFP dynamics in cells indicates that these proteins still assemble at endocytic sites, but with slowed kinetics (, panels 2 and 4). After reaching a peak in fluorescence intensity, Sla1p internalized and rapidly disassembled, whereas Abp1p internalized but did not efficiently disassemble, indicating that the F-actin clumps in the cell interior are derived from structures assembled at the cell cortex. The observation that Sla1p is internalized in mutant cells suggests that perturbations to cofilin-mediated actin filament turnover do not block endocytic internalization from the plasma membrane. In addition, the delay in actin assembly at endocytic sites suggests that cofilin function is important for rapid actin assembly.
The observation that endocytic coat complexes internalize concomitant with actin assembly in cells () was intriguing given the observed defects in the accumulation of fluid-phase endocytic markers in the vacuole (). We tested whether the endocytic coat internalization events reflect receptor-mediated endocytic internalization by quantifying the internalization of radiolabeled yeast mating pheromone. Indeed, in cells, radiolabeled α factor is internalized, albeit with delayed kinetics (). The delay in internalization could be reflected by a decrease in the steady-state number of endocytic events in cells. We tested this possibility by quantifying the total number of Sla1-GFP patches per unit area in wild-type and mutant cells and found that there is a slight increase in the number of steady-state endocytic sites in mutant cells (1.7 ± 0.6 endocytic sites per 10 μm for wild-type cells [ = 25] vs. 2.5 ± 0.5 endocytic sites per 10 μm for mutants [ = 25]). Therefore, we suspect that the delay in α-factor internalization reflects the longer lifetime of the endocytic machinery at the plasma membrane seen in cells ().
The observation that actin patches have elongated postinternalization lifetimes spurred us to investigate the association of these actin patches with internalized plasma membrane. We used simultaneous imaging of the lipophilic dye FM4-64 and Abp1-GFP to analyze the dynamic interrelationship between plasma membrane internalization and actin in cells. Consistent with the observation that radiolabeled α factor is internalized in cells, 5 min after the addition of FM4-64, labeled internal membranes were observed, indicating that plasma membrane internalization occurs in cells (). Strikingly, we observed partial colocalization between cytoplasmic Abp1-GFP and internalized FM4-64 in cells ( and Video 8, available at ) but not in wild-type cells ( and Video 7). The internalized membranes moved rapidly within the cytoplasm. The duration of the colocalization of FM4-64 and Abp1-GFP is evident in kymographs, which demonstrate prolonged colocalization compared with wild-type cells ().
We biochemically compared the association of Abp1p with membranes in and wild-type cells by fractionating cell extracts and examining the subcellular distribution of Abp1p. We found that Abp1p and actin are enriched in the heavy membrane fraction (P13,000) from cells (). Further fractionation experiments were unsuccessful in determining the exact identity of these Abp1p/actin-associated membranes. These data coupled with the in vivo live cell imaging results suggest that cofilin-mediated actin filament turnover is required for the timely disassembly of actin from endocytic membranes after internalization from the plasma membrane.
Two observations in cells, the internalization of plasma membrane and endocytic proteins and the lack of accumulation of endocytic cargo in the vacuole, seemed contradictory. We sought to determine the basis for this contradiction, reasoning that the membrane marker FM4-64 would provide better resolution of membrane-trafficking events between the plasma membrane and the vacuole than fluid-phase endocytosis markers. In wild-type cells, FM4-64 initially labeled the plasma membrane (). After 15 min, it labeled small, intensely stained endocytic intermediates. After ∼45 min, vacuolar labeling was observed, and, at 60 min, the vacuole was strongly stained. In cells, a dramatic alteration in the morphology of endocytic compartments was observed (). After 15 min, large membranous structures were labeled, which persisted throughout the course of the analysis (60 min). These membranous structures were observed in 75% of cells ( = 20) and were never seen in wild-type cells ( = 15). To better identify these structures, we used the vacuolar endopolyphosphatase Phm5p, which was tagged with GFP, as a steady-state marker of vacuolar morphology and followed the internalization of FM4-64 over time (Fig. S4 A, available at ). In mutant cells, FM4-64 colocalized with Phm5-GFP in the multilobed and fragmented vacuoles after 30 min of internalization of the dye. Multilobed and fragmented vacuoles were observed in 68% of cells ( = 65) versus 7% of wild-type cells ( = 25). These observations indicate that productive cofilin-mediated actin filament turnover is required for the normal morphology of endocytic compartments.
One hallmark of a subset of postinternalization endocytosis mutants, particularly () mutants, is the missorting and subsequent secretion of soluble vacuolar hydrolases such as carboxypeptidase Y (CPY). We tested for the secretion of CPY in cells using a colony blot assay (). The cells showed markedly elevated CPY secretion compared with wild-type cells. Secretion of CPY from mutant cells was similar to levels observed for two mutants, and . To control for the deposition of CPY by cell lysis rather than secretion, log-phase liquid cultures of wild-type and mutant cells were compared for release of the soluble cytosolic marker 3-phosphoglycerate kinase (Pgk1p). No difference in Pgk1p levels was observed in the culture media under these conditions (unpublished data). To further understand the nature of the trafficking defects in mutant cells, we examined the distribution of alkaline phosphatase (Pho8p), a protein that is trafficked to the vacuole via a different biosynthetic route from the one used by CPY (). We found that Pho8p localizes properly to fragmented/multilobed vacuoles (Fig. S4 B). This result establishes that although certain aspects of vacuole sorting and morphology are abnormal in mutant cells, others remain intact. Collectively, these results indicate that cofilin-mediated actin filament turnover is required for the timely delivery of endocytic material to the vacuole after internalization from the plasma membrane, for certain vacuole sorting processes, and for normal vacuole morphology.
i
n
v
e
s
t
i
g
a
t
e
d
c
o
f
i
l
i
n
'
s
d
y
n
a
m
i
c
s
a
n
d
r
o
l
e
i
n
a
c
t
i
n
s
u
b
u
n
i
t
f
l
u
x
t
h
r
o
u
g
h
f
i
l
a
m
e
n
t
n
e
t
w
o
r
k
s
i
n
l
i
v
i
n
g
y
e
a
s
t
c
e
l
l
s
a
n
d
o
b
t
a
i
n
e
d
i
n
v
i
v
o
e
v
i
d
e
n
c
e
f
o
r
t
h
e
d
e
p
e
n
d
e
n
c
e
o
f
c
o
f
i
l
i
n
r
e
c
r
u
i
t
m
e
n
t
o
n
t
h
e
a
c
t
i
n
f
i
l
a
m
e
n
t
n
u
c
l
e
o
t
i
d
e
s
t
a
t
e
.
T
h
i
s
i
n
s
i
g
h
t
i
n
t
o
c
o
f
i
l
i
n
-
d
e
p
e
n
d
e
n
t
a
c
t
i
n
f
i
l
a
m
e
n
t
t
u
r
n
o
v
e
r
i
n
y
e
a
s
t
a
l
l
o
w
e
d
u
s
t
o
a
s
k
h
o
w
a
c
t
i
n
s
u
b
u
n
i
t
f
l
u
x
c
o
n
t
r
i
b
u
t
e
s
t
o
e
n
d
o
c
y
t
i
c
i
n
t
e
r
n
a
l
i
z
a
t
i
o
n
a
n
d
t
o
s
u
b
s
e
q
u
e
n
t
s
t
e
p
s
i
n
t
r
a
n
s
p
o
r
t
t
o
t
h
e
v
a
c
u
o
l
e
.
A
c
o
f
i
l
i
n
a
l
l
e
l
e
w
i
t
h
d
e
f
e
c
t
s
i
n
c
o
f
i
l
i
n
-
d
e
p
e
n
d
e
n
t
a
c
t
i
n
f
i
l
a
m
e
n
t
t
u
r
n
o
v
e
r
s
t
r
o
n
g
l
y
b
l
o
c
k
e
d
a
c
c
u
m
u
l
a
t
i
o
n
i
n
t
h
e
v
a
c
u
o
l
e
o
f
a
m
a
r
k
e
r
f
o
r
f
l
u
i
d
-
p
h
a
s
e
e
n
d
o
c
y
t
o
s
i
s
b
u
t
o
n
l
y
p
a
r
t
i
a
l
l
y
b
l
o
c
k
e
d
p
l
a
s
m
a
m
e
m
b
r
a
n
e
a
n
d
r
e
c
e
p
t
o
r
-
m
e
d
i
a
t
e
d
c
a
r
g
o
i
n
t
e
r
n
a
l
i
z
a
t
i
o
n
.
C
o
n
s
i
s
t
e
n
t
w
i
t
h
t
h
e
s
e
o
b
s
e
r
v
a
t
i
o
n
s
,
e
n
d
o
c
y
t
i
c
p
a
t
c
h
e
s
w
e
r
e
a
b
l
e
t
o
i
n
t
e
r
n
a
l
i
z
e
f
r
o
m
t
h
e
p
l
a
s
m
a
m
e
m
b
r
a
n
e
c
o
n
c
o
m
i
t
a
n
t
l
y
w
i
t
h
a
s
l
o
w
e
d
b
u
r
s
t
o
f
a
c
t
i
n
p
o
l
y
m
e
r
i
z
a
t
i
o
n
.
D
e
f
e
c
t
s
i
n
a
c
t
i
n
d
i
s
a
s
s
e
m
b
l
y
d
i
d
n
o
t
b
l
o
c
k
d
i
s
a
s
s
e
m
b
l
y
o
f
t
h
e
o
t
h
e
r
c
o
m
p
o
n
e
n
t
s
o
f
t
h
e
e
n
d
o
c
y
t
i
c
m
a
c
h
i
n
e
r
y
,
i
n
c
l
u
d
i
n
g
c
o
a
t
p
r
o
t
e
i
n
s
s
u
b
s
e
q
u
e
n
t
t
o
i
n
t
e
r
n
a
l
i
z
a
t
i
o
n
,
b
u
t
r
e
s
u
l
t
e
d
i
n
t
h
e
p
r
o
l
o
n
g
e
d
a
s
s
o
c
i
a
t
i
o
n
o
f
a
c
t
i
n
w
i
t
h
i
n
t
e
r
n
a
l
m
e
m
b
r
a
n
o
u
s
c
o
m
p
a
r
t
m
e
n
t
s
.
F
-
a
c
t
i
n
d
i
s
a
s
s
e
m
b
l
y
d
e
f
e
c
t
s
r
e
s
u
l
t
e
d
i
n
v
a
c
u
o
l
a
r
m
o
r
p
h
o
l
o
g
y
d
e
f
e
c
t
s
a
n
d
t
h
e
i
m
p
r
o
p
e
r
s
o
r
t
i
n
g
o
f
a
t
l
e
a
s
t
o
n
e
t
y
p
e
o
f
s
o
l
u
b
l
e
c
e
l
l
u
l
a
r
c
a
r
g
o
d
e
s
t
i
n
e
d
f
o
r
t
h
e
v
a
c
u
o
l
e
.
Yeast strains used in this study are listed in Table S1 (available at ). Yeast strains were grown in standard rich media (YPD) or synthetic dextrose media (SD) supplemented with the appropriate amino acids. CasAA media is selective for uracil prototrophy and is composed of 1% casamino acids (Difco), 0.7% yeast nitrogen base (Difco), and 2% dextrose. All strains were cultured at 25°C unless otherwise noted. T. Doyle (Stanford University, Stanford, CA) provided the ACT1-GFP plasmid, and H. Pelham (Medical Research Council Laboratory of Molecular Biology, Cambridge, UK) provided the PHM5-GFP plasmid. The GFP-cofilin construct, which has S65T-GFP and a 12–amino acid linker inserted in between amino acids N74 and G75 of cofilin, was made as follows. Genomic DNA from DDY426 was used as a template for PCR to generate a DNA fragment corresponding to 750 bp upstream of the translational start site and 400 bp downstream of the translational start flanked by BamHI and SpeI restriction sites at the 5′ and 3′ ends, respectively, which were introduced into the primers. A second fragment was generated corresponding to the region 401 bp downstream from the translational start and 350 bp downstream of the translational stop flanked by NotI and SacII sites at the 5′ and 3′ ends, respectively, which were introduced into the primers. pFA6a-GFP(S65T)-kanMX6 was used as a template to generate by PCR a fragment of DNA corresponding to GFP(S65T) flanked by a linker sequence encoding the amino acids GHGTGSTGSGSS and flanked by SpeI and NotI sites at the 5′ and 3′ ends, respectively. These constructs were cloned into pRS426 to generate pDD2090, which was used in Fig. S1, and pRS316 to generate pDD2091, which was used in the rest of this study. The entire construct was sequenced, and the only error introduced into the relevant portion of the coding sequence was a missense mutation converting the third glycine in the 5′ GFP linker into an arginine. Primer sequences used for generating this construct are provided in Table S2.
All cells were grown at 25°C to early log phase in SD media lacking tryptophan (TRP [SD−TRP]) to minimize background fluorescence. Simultaneous two-color imaging and fluorescence recovery after photobleaching experiments were performed as described previously (). All images were acquired with MetaMorph software (Molecular Devices). P-values were determined using a two-tailed test in Excel (Microsoft).
For simultaneous imaging of Abp1-GFP and FM4-64 internalization, FM4-64 (Invitrogen) at 8 μM in SD–TRP was perfused over cells in flow chambers as described above for visualization of cofilin redistribution upon treatment with jasplakinolide and lat A. Labeling of endocytic intermediates and the vacuole was performed essentially as described previously () with the following modifications. Cells were grown in SD–TRP at 25°C to early log phase. A final concentration of 25 μM FM4-64 was added to the cells, and aliquots were harvested at 0, 15, 30, 45, and 60 min after dye addition. The aliquots of cells were washed into ice-cold SD–TRP containing 15 mM sodium azide and 15 mM sodium fluoride and were then imaged on a microscope (IX71; Olympus) equipped as described in the previous paragraph.
For immunofluorescence of alkaline phosphatase (Pho8p), wild-type and cells were grown to early log phase in YPD. Cells were fixed with 4% formaldehyde for 30 min at 25°C in YPD and overnight in 4% formaldehyde, 50 mM potassium phosphate, pH 7.4, and 1 mM MgCl. The fixed cells were treated with TEB (200 mM Tris-HCl, pH 8.0, 20 mM EDTA, and 144 mM β-mercaptoethanol) for 30 min at 30°C. Cells were spheroplasted with 0.75 mg/ml Zymolyase 20T (Seikagaku Corp.) in 1.2 M sorbitol, 50 mM potassium phosphate, pH 7.4, and 1 mM MgCl for 45 min at 30°C. Cells were washed twice with 1.2 M sorbitol and were treated for 2 min with 1.2 M sorbitol with 2% SDS. Cells were washed twice with 1.2 M sorbitol, allowed to settle on a chambered slide, and processed for immunofluorescence using a 1:10 dilution of monoclonal α-ALP antibodies (Invitrogen) and a 1:50 dilution of FITC-conjugated donkey α-mouse antibodies (Jackson ImmunoResearch Laboratories).
S-labeled α factor was prepared as described previously (). A continuous presence protocol was used. Cells were grown in YPD media at 25°C, harvested by centrifugation at 1,500 for 4 min, and resuspended in internalization media (YPD media with 0.5% casamino acids and 1% BSA). Aliquots were taken at the indicated time points and diluted in ice-cold 50 mM potassium phosphate buffer at pH 6.0 (total bound α factor) or pH 1.1 (internalized α factor). The cells were collected by filtration, and counts per minute were determined in a scintillation counter (LS 6500; Beckman Coulter). Percent uptake is represented by multiplying 100 times the ratio of counts per minute at pH 1.1/pH 6.0 for each time point.
1.5 liters of cells were grown at 25°C to log phase (A = 0.5/ml) in YPD. Cells were pelleted by centrifugation at 1,500 for 5 min and washed once with water. The cell wall was partially destabilized by treating cells at 20 OD/ml with 100 mM Tris-HCl, pH 9.4, and 50 mM β-mercaptoethanol for 15 min at 30°C. Cells were washed once with 100 ml spheroplast buffer (1.2 M sorbitol, 20 mM potassium phosphate, pH 7.5, and 5 mM MgCl) and resuspended to 50 OD/ml in spheroplast buffer with 3 mg/ml Zymolyase 20T (Seikagaku Corp.). Cell walls were digested for 1–1.5 h at 30°C, whereas spheroplast formation was assayed using phase-contrast light microscopy. Cells were pelleted at 1,500 for 5 min at 4°C, washed once with 100 ml of ice-cold spheroplast buffer, and resuspended in ice-cold membrane isolation buffer (0.6 M sorbitol, 20 mM Hepes-KOH, pH 7.5, 1 mM PMSF, and 3 mM benzamidine) to 50 OD/ml. Cells were lysed in a Dounce homogenizer (Wheaton Scientific) with 15 strokes using the tight pestle. The resulting cell lysate was centrifuged at 300 for 5 min at 4°C to pellet unlysed cells and other dense cell debris. Half of the supernatant (S300) was harvested and spun at 13,000 for 10 min at 4°C, yielding the S13,000 and P13,000 fractions. Three quarters of the S13,000 was spun at 100,000 for 1 h at 4°C to yield the S100,000 and P100,00 fractions. Fractions were solubilized in Tris urea buffer (50 mM Tris-Cl, pH 6.8, 3 M urea, 1% SDS, and 5% β-mercaptoethanol) and analyzed by SDS-PAGE followed by immunoblotting with polyclonal Abp1p and Act1p antibodies.
Sterile 0.45-μm nitrocellulose filter disks were placed on YPD plates, and cells were streaked on top and allowed to grow for 10 h. The nitrocellulose disks were then washed and subjected to standard immunoblotting techniques using polyclonal CPY antibodies (provided by R. Schekman, University of California, Berkeley, Berkeley, CA).
Fig. S1 shows complementation of a strain by a high-copy GFP- cofilin plasmid, the inability of this strain to grow in the absence of the plasmid, and the relative amounts of GFP-cofilin in vivo when it is expressed from low or high-copy plasmids. Fig. S2 shows that native untagged cofilin is restricted from regions of actin tails engaged in active actin filament assembly. Fig. S3 shows GFP-cofilin relocalization in actin tails in cells upon treatment with lat A. Fig. S4 shows vacuolar morphology defects in cells using Phm5-GFP and FM4-64 and shows that alkaline phosphatase is properly trafficked to fragmented vacuoles in mutant cells. Fig. S5 shows low levels of lat A slow actin assembly at endocytic sites. Video 1 shows GFP-cofilin and Abp1-mRFP dynamics in vivo using two-color real-time image acquisition. Videos 2 and 3 show GFP-cofilin relocalization in Abp1-mRFP–labeled actin tails in cells treated with jasplakinolide (Video 2) and lat A or DMSO (Video 3) using two-color real-time image acquisition.
cells using real-time acquisition. Video 5 shows Sla1-GFP and Abp1-mRFP dynamics in wild-type cells and mutant cells using two-color real-time image acquisition. Video 6 shows Abp1-GFP dynamics in wild-type cells treated with low levels of lat A using real-time image acquisition. Videos 7 and 8 show FM4-64 and Abp1-GFP dynamics in wild-type (Video 7) and mutant (Video 8) cells using two-color real-time image acquisition. Table S1 presents the yeast strains used in this study. Table S2 presents the primer sequences used to generate GFP-cofilin. Online supplemental material is available at . |
Virulence associated with several Gram-negative bacterial pathogens requires the translocation of “effector” proteins from bacteria into host cells through a dedicated protein translocation apparatus termed the type III secretion system (TTSS) (; ). Many bacterial effector proteins possess a specialized activity required to limit anti-microbial immune response and promote bacterial growth and dissemination during pathogenesis. We are particularly interested in the mechanisms of type III effector proteins found in the attaching and effacing (A/E) pathogen group including (EPEC) and its close relatives, (EHEC 0157:H7) and that cause severe gastrointestinal disease. Although the coordinated actions of several type III effector proteins including Tir, Map, EspF, EspG, EspH, EspI, EspJ, EspK, EspZ, and NleA-F are required for virulence associated with A/E pathogens, the biochemical activities of most effectors are poorly defined.
Microbial pathogens may hijack the actin cytoskeleton machinery to perform a variety of functions that include actin-based motility, cellular invasion, and intracellular trafficking through the endocytic pathway (). A key regulator of the actin cytoskeleton is neuronal Wiskott-Aldrich syndrome protein (N-WASP), a eukaryotic protein that initiates actin filament branching and assembly through the direct activation of the Arp2/3 complex. Due to its critical role in this process, N-WASP is tightly regulated by upstream signals including phospholipids (PIP), the small G-protein Cdc42, and Src homology-3 (SH3) adaptor proteins (). Recent work also indicates that N-WASP is intimately associated with mechanisms of endocytosis (; ; ), a cellular process required for the uptake of extracellular components including microbial pathogens and viruses ().
Membrane remodeling during endocytosis requires the spatiotemporal coordination of several phospholipids and F-actin binding proteins (; ). As a specific example, sorting nexin 9 (SNX9) is dynamically recruited to clathrin-coated pits (CCPs) at the late stages of vesicle formation (; ). Although the exact role of SNX9 is not well understood, its domain architecture suggests that it links the plasma membrane to proteins associated with the cellular cortex. SNX9 possesses two lipid interaction domains, a phospholipid-binding module termed the phox (PX) domain followed by a putative Bin/Amphiphysin/Rvs (BAR) lipid-binding domain. Interestingly, the BAR domain is a banana-shaped helical dimer that senses membrane curvature and can reconfigure lipid vesicles or sheets into membrane tubules (). In addition to its lipid-binding properties, SNX9 also possesses an N-terminal Src homology-3 (SH3) protein interaction module that was recently shown to bind WASP () and to functionally activate dynamin at CCPs (). Thus, SNX9 is uniquely suited to regulate the membrane/cytoskeletal interface during clathrin-mediated endocytosis.
In this paper, we found that the type III effector EspF coordinates membrane remodeling and F-actin polymerization during EPEC pathogenesis. Similar to previous results (), we found that EspF binds the SH3 domain of SNX9 and we further demonstrated that this interaction induces membrane remodeling, a phenotype that is functionally coupled to N-WASP–dependent actin polymerization in eukaryotic cells. EspF activated both SNX9 and N-WASP in a coordinated spatiotemporal pattern at CCPs. Importantly, these data provide a molecular mechanism for EspF function in host cells and further suggest that the dynamic interplay between bacterial type III effector proteins and eukaryotic signaling pathways is a critical aspect of host–pathogen interactions.
Database searches reveal that EspF orthologues are found in an A/E pathogen group that includes EPEC, EHEC 0157:H7, and (). EspF is composed of 47 amino acid proline-rich repeats (PRRs) that are repeated several times throughout the coding sequence. Although the repeat numbers vary amongst EspF orthologues in A/E pathogens, there are no cases where the PRR sequence is present as a single copy. In addition, EHEC 0157:H7 possesses a second PRR domain protein called EspFu/TccP that shares 35% homology with EspF, but has a distinct biological function (; ). To define eukaryotic binding partners we screened EspF against a mouse embryo cDNA library using the yeast two-hybrid system. Screening of 8 million yeast transformants yielded five positives. Two cDNAs encoded the SH3 domain of SNX9 (, residues 1–111) and the Cdc42/rac interactive binding (CRIB) domain of N-WASP (, residues 200–258).
Our yeast two-hybrid clone spanned residues 1–111 of SNX9, a region encompassing its SH3 domain. This interaction was confirmed using [S]-methionine labeled SNX9 1–111 (SH3 domain) and recombinant GST-EspF fusion protein, which selectively interacted in pull-down assays (). Grb2, Nck1, and Nck2, three SH3 domain–containing proteins with broad ligand specificity, could not be copurified with EspF (). Similar binding specificity was also observed in vivo by coimmunoprecipitation assays from transfected Hek293A cells. These data demonstrate that EspF interacts with full-length SNX9 but not Grb2, Nck1, or Nck2 in cells ().
In general the ∼300 SH3 domains in the human genome bind ligands with a canonical PxxP motif. It was surprising that only one SH3 protein was identified in our high coverage yeast two-hybrid screen. To better understand this apparent specificity, we used phage-displayed peptide libraries to explore the binding specificity profiles of the SH3 domain of SNX9. The SH3 domain was screened against a library composed of 10 completely random dodecapeptides. After four rounds of affinity selection, 150 individual binding-phage were sequenced. Remarkably, only 13 unique peptides were identified from the 150 sequences, suggesting that SNX9 recognizes a highly conserved sequence motif (). Homology alignment was used to derive a SNX9 specificity profile in the form of a preferred binding motif (). Notably, the specificity profile confirmed that the SH3 domain recognizes multiple features of the ligand sequence; an arginine was invariant at the −3 position, alanine was preferred at the −1, and prolines were selected at the 0 and +3 positions of the PxxP motif. These data determine the preferred consensus SNX9 SH3 ligand motif as RxAPxxP.
Next, an in vitro–translated S-SNX9 was used to screen a solid phase library containing peptides of 15-mer residues (offset every three residues) spanning amino acids 1–166 in EspF (). By overlaying S-SNX9 onto the EspF peptide library, we found that the SH3 binding site in EspF was confined to two distinct regions within residues 75–81 and 122–128 (). Remarkably, these regions in EspF both possess a common RxAPxxP motif that conforms to the SNX9 SH3 consensus-binding motif identified in our unbiased phage display. Because residues 75–81 and 122–128 are found at the N-terminal portion of PRR domain 1 and 2, respectively, and because homologous regions are found in all EspF orthologues, we can conclude that each PRR domain of EspF possesses a single SNX9 binding site (see the alignment in ). In support of this conclusion, a dissociation constant of 2.2 μM was experimentally derived from an EspF peptide (PPP) bound to the SNX9-SH3 domain (), suggesting that these residues constitute the minimal sequence required for SNX9 interaction. In addition, we substituted aspartic acids for the arginines in position 75, 122, and 169 (Arg→Asp or EspF-D3) in all three PRR domains of EPEC EspF. Mutant EspF-D3 failed to bind SNX9 in both in vitro pull-down assays (unpublished data) and coimmunoprecipitation experiments ().
Next, we turned to the interaction between EspF and N-WASP. Full-length N-WASP interacted with EspF in coimmunoprecipitation assays, but not with control EGFP (). This interaction was comparable to the interaction between N-WASP and prophage-expressed EspFU/TccP gene that displays 35% identity to EspF () (). These data suggest that the functionally distinct EspF and EspFU/TccP proteins both interact with N-WASP in eukaryotic cells.
We tested the ability of EspF to directly activate N-WASP by assaying Arp2/3 complex-dependent actin polymerization kinetics using a pyrene-actin assembly assay in vitro. For these assays, we used Arp2/3-complex purified from bovine brain extracts and purified N-WASP ΔEVH1 (residues 138–501), an N-WASP truncation mutant lacking the WH1/EVH1 domain that exhibits strong autoinhibition () (). In addition, a purified 6×-His tagged EspF protein possessing three intact PRR domains was used in these studies (EspFΔ47, residues 48–206) (). In the absence of either N-WASP or the Arp2/3 complex, EspF did not alter F-actin assembly dynamics (). However, in the presence of both Arp2/3 and N-WASP ΔEVH1, EspF stimulated actin nucleation (). These data indicate that EspF directly activates N-WASP in vitro.
As demonstrated in , residues 200–258 encompassing the Cdc42 binding CRIB domain of N-WASP constituted the minimal interaction site with EspF, suggesting that EspF could functionally mimic this GTPase. We used an engineered “mini N-WASP” protein possessing the minimal Basic and CRIB regulatory domains linked to the Arp2/3 activating VCA domain to test this hypothesis () (). EspF potently stimulated mini-N-WASP in a dose-dependent manner and the half-maximal activity of mini-N-WASP occurred at an EspF concentration of 11.8 nM as determined by saturation binding experiments (). In total, these data indicate that EspF can relieve N-WASP autoinhibition through a direct binding interaction with the CRIB regulatory domain.
Because both N-WASP and SNX9 bind to residues within the PRR domains of EspF, it was important to determine if their binding sites were mutually exclusive. We found that the SNX9 binding deficient mutant EspF-D3 induced mini-N-WASP–dependent actin polymerization to a similar extent as EspF, suggesting that the RxAPxxP motif did not play a role in N-WASP activation (). Moreover, addition of saturating concentrations of GST-SNX9-SH3 (10 μM) to EspF had a minor affect on N-WASP activation (), suggesting the possibility that EspF can bind SNX9 and N-WASP through two distinct interaction sites.
To directly test the ability of EspF to nucleate a functional SNX9 and N-WASP complex, we first incubated recombinant EspF with GST-SNX9-SH3 immobilized to glutathione-Sepharose beads (). The formation of a stable protein complex between the SH3 domain (34 kD) and EspF (26 kD) was confirmed by SDS-PAGE analysis (). Next, we added the stable SH3/EspF complex to mini-N-WASP and measured actin polymerization rates (). The SH3/EspF complex directly activated mini-N-WASP–mediated actin assembly kinetics (, blue). In control experiments, actin polymerization kinetics were not increased with SH3 beads alone (, gray), with mutant EspF-D3 that could not form a complex with the SH3 beads (, purple), or with GST control proteins (, black). In total, these data provide strong evidence that EspF coordinates the binding of SNX9 and the activation of N-WASP through two independent binding motifs found within its highly conserved PRR domains (see ).
SNX9 and N-WASP participate in several membrane trafficking events and localize to cellular membranes at CCPs and trafficking organelles (, ; ; ; ). We used total internal reflection-fluorescence microscopy (TIR-FM), a technique that allows fluorophore-conjugated proteins to be visualized within 100 nm of the cell surface (), to determine if EspF also localized to the plasma membrane in living cells. EGFP-EspF was transfected into Swiss-3T3 cells stably expressing DsRed-Clathrin light chain- a (Clc-DsRed), a cellular marker of CCPs and the plasma membrane (). Several 10–12-min videos were recorded in which EGFP-EspF and Clc-DsRed fluorescence was simultaneously captured by dual-color TIR-FM (Video 1, available at ).
Analysis of living cells indicated that EspF was highly dynamic at the plasma membrane and it transiently accumulated into patches at the cell surface (Video 1). Interestingly, EspF partially colocalized with clathrin in still images () and examination of time-resolved videos indicated that 92% (357 of 388 events from 8 cells) of clathrin structures transiently colocalized with EspF ( and Video 2, available at ). In addition, larger EspF tubules could be found extending near the cell surface (Video 3). Expression of EspF did not affect normal clathrin dynamics, as the lifetime of clathrin at the plasma membrane was nearly identical between untransfected cells (62 ± 4 s, = 60 events) and EGFP-EspF transfected cells (64 ± 5 s, = 58 events) (). These dynamics were similar to those previously reported for clathrin in Swiss 3T3 cells ().
We noticed that the spatial and temporal dynamics of EspF were exquisitely coordinated with clathrin-mediated endocytic events. To quantify these observations, the surface kinetics of EspF was directly compared with clathrin dynamics. As a reference index, the Clc-DsRed fluorescence intensity values of 25 clathrin structures were measured from 8 cells expressing EGFP-EspF. We analyzed events in which clathrin appeared at the plasma membrane, was stable for ∼60 s in the TIR-FM imaging field, and exponentially decayed from the plasma membrane (, red trace). These measurements were directly compared with the EGFP-EspF fluorescence intensities at each CCP. By aligning all 25 EspF traces relative to clathrin membrane departure, it was clear that EspF displayed a tightly coordinated spatiotemporal localization pattern at the plasma membrane. First, EspF was recruited to the plasma membrane at preexisting CCPs (, green trace). The peak EspF fluorescence signal coincided with the exponential decay of clathrin signal from the plasma membrane (, arrow). Because these clathrin dynamics are indicative of clathrin coated-vesicle movement away from the plasma membrane and vesicle scission (), EspF is likely to associate with actively budding membrane domains. Second, EspF remained at the cell surface beyond the time of clathrin departure, suggesting that EspF was associated with the plasma membrane and not in CCPs (). Finally, EspF signal decayed in the TIR-FM image, indicating its retreat from the plasma membrane.
Next, we examined if EspF dynamics at the plasma membrane correlated with the protein interactions defined by our in vitro data. The triple-mutant EspF-D3 protein did not interact with SNX9, but maintained N-WASP binding activity ( and ). Therefore, this mutant was used to determine the potential contributions of SNX9 binding on EspF dynamics. We recorded several TIR-FM videos from cells expressing EGFP-EspF-D3 () and compared its dynamics to wild-type EspF (). EspF-D3 was recruited to 86% (112 of 130 events from 5 cells) of cell surface clathrin structures. Unlike EspF, however, EspF-D3 departed from the plasma membrane coincident with or slightly before clathrin-mediated endocytosis (). The average lifetime of EspF-D3 associated with endocytic sites (12.3 ± 5 s, = 108 events) was approximately fivefold shorter than wild-type EspF (68 ± 14.5 s, = 98 events) (). It was also clear that EspF-D3 did not exhibit the prolonged plasma membrane localization that wild-type EspF displayed after clathrin departure. These data are consistent with the ability of EspF to interact with SNX9 at the plasma membrane (); however, additional experimental approaches are necessary to directly establish this link.
Using wide-field fluorescence microscopy, we found that endogenous SNX9 localized to CCPs at the surface of HeLa cells and accumulated in perinuclear regions near the Golgi apparatus () (; ). Transient transfection of EspF induced a redistribution of SNX9 from CCPs and perinuclear regions to “worm-like” tubular structures found near the cell periphery (; Fig. S1 A, available at ). In these cells SNX9 and EspF colocalized at tubular structures, suggesting that SNX9 redistribution was directly associated with EspF binding ( and Fig. S1 B). Consistent with this notion, mutant EspF-D3 had no affect on SNX9 localization. To test these observations in a more controlled cellular environment, we artificially increased the SNX9 expression levels by transfecting mCherry-SNX9 into HeLa cells. Exogenous SNX9 had a predominantly perinuclear localization and a small proportion could also be found in cell surface puncta reminiscent of CCPs (). We found that coexpression of EspF with SNX9 induced an array of tubular structures that extended into dense tubular networks (). Similar tubular networks were identified in >90% of EspF and SNX9 cotransfected cells and both proteins colocalized at these sites (). As expected, mutant EspF-D3 failed to induce SNX9 tubules indicating that SH3 binding is required for SNX9 localization. Moreover, wild-type EspF did not induce tubules upon cotransfection of SNX9ΔBAR, a C-terminal deletion mutant of SNX9 in which the putative membrane deforming BAR domain was removed (). These data reveal a novel activity of SNX9 to induce cellular tubulation, a phenomenon that is directly linked to EspF binding.
To determine the nature of the abnormal tubules formed by coexpression of EspF and SNX9, EspF was first immuno-localized by electron microscopy (EM). EspF decorated tubular and vesicular membrane structures but was not found in nuclear areas, inside organelles, or randomly distributed throughout the cytoplasm (). Using thin-section EM we identified a striking array of membrane structures in EspF transfected () but not untransfected cells (). The EspF-induced tubules were abnormally curved and in many cases budding profiles could be observed along both sides of the membrane tubules (). In addition, unusually long (∼0.5 μm) cell surface invaginations appeared to be continuous with the plasma membrane () and several membrane projections extended from multivesicular organelles (). A dense negative stain was also observed around the new membrane structures, most likely representing the EspF/SNX9 protein coat. We now propose a model whereby EspF activates SNX9; engagement of the SNX9-SH3 domain by the highly specific EspF-ligand motif exposes the C-terminal BAR domain to induce membrane remodeling in eukaryotic cells.
N-WASP was originally shown to induce cell surface filopodia projections during cell migration (); however, recent findings indicate that it also regulates membrane trafficking events including actin-based vesicular transport and endocytosis (, ; ; ; ). We did not detect the formation of cell surface filopodia in EspF-transfected cells. However, de novo F-actin nucleation occurred near membrane tubules induced by EspF and SNX9 coexpression (; Fig. S2 A, available at ). EspF activation of SNX9 was prerequisite for these downstream signaling events, as F-actin appeared normal in cells expressing mutant EspF-D3 (). These data suggest that membrane tubulation occurs before actin polymerization in the EspF/SNX9/N-WASP signaling cascade. In support of this conclusion, depolymerization of actin had no affect on the formation of membrane tubules; however F-actin was perturbed at these sites (Fig. S2, C and D). Moreover, N-WASP redistributed from the cellular cytoplasm to membrane tubules in EspF-transfected cells () but not in cells expressing mutant EspF-D3 (). In total, these data indicate that membrane targeting and subsequent remodeling by EspF is required for its activation of downstream signaling events.
Several additional lines of evidence also suggest that EspF activates N-WASP at membrane sites. Careful examination of the cell surface dynamics of EspF by TIR-FM indicated that a small proportion of EspF puncta (∼5%) formed “comet tails” and seemed to propel the clathrin structure over a very short distance just before clathrin internalization (; Video 4, available at ). This EspF behavior is reminiscent of actin filament dynamics observed in comet tails () and the projectile motion of endocytic vesicles at the tips of actin tails (). EspF was also found at mobile clathrin structures that moved parallel to the plasma membrane, a phenotype that may occur due to changes in the actin cytoskeletal architecture (; Video 5) (). In total, these cellular observations indicate that EspF functions as a bacterial signaling node, integrating the activation of two signaling cascades at membranes of eukaryotic cells ().
Next, we sought to determine the potential role of EspF signaling in epithelial models of EPEC infection. Genetic studies have implicated EspF in several pathogenic phenotypes, including the deregulation of the tight-junction ion barrier function in EPEC-infected polarized epithelial cells (; ; ). We found a dramatic loss of trans-epithelial electrical resistance (TER, a measure of tight junction integrity) in polarized T84 colonic epithelial cells infected with wild-type EPEC that was dependent on the gene () (). Interestingly, both plasmid-expressed EPEC (p) as well as EHEC (p) trans-complemented the mutant strain () (). Similarly, we found that mutant strain expressing the SNX9 binding–deficient mutant EspF-D3 () also induced a loss of TER (). These confounding data suggest that multiple genes in EPEC are responsible for tight junction breakdown and suggest that the complex phenotype cannot be ascribed to EspF function alone.
Next, we wanted to directly test the cytoplasmic role of EspF in polarized epithelial cells. Tight junction breakdown was monitored by the redistribution of occludin from a uniform band outlining the cell junctions in uninfected cells () to a discontinuous beaded pattern in EPEC infected cells (). To directly test the role of EspF in this process, stable cell lines expressing a tandem affinity protein tag (TAP; protein A and flag) fused to the N-terminus of EPEC EspF were created (). As shown in , TAP-EspF was expressed in every cell and localized to small puncta near the apical cell surface. SNX9 interaction seemed to be required for this localization as the SH3-binding mutant EspF-D3 was distributed throughout the cytoplasm and accumulated in nuclear regions (). Importantly, MDCK cells expressing either EspF or EspF-D3 formed normal tight junctions (), suggesting that EspF activity alone is not sufficient to break down tight junctions. Next, we biochemically purified EspF from MDCK cells as an unbiased approach to determine protein complexes associated with EspF in a polarized cell type (). TAP-purified EspF or control parental cell lines were separated by SDS-PAGE and Coomassie stain indicated that the major EspF-binding protein had a molecular weight of ∼70 kD (). This protein was unambiguously identified as canine SNX9 (MW = 71 kD) by tandem mass spectrometry from two independent experiments (see Materials and methods). Thus, SNX9 is the major target of cytoplasmic EspF in polarized epithelial cells.
Finally, we sought to determine if the localization and function of bacterial delivered EspF relies on interaction with SNX9 in polarized cell types. Human CaCo intestinal epithelial cells were infected with wild-type EPEC or the mutant strain. Whereas wild-type EPEC induced SNX9 membrane tubules, this phenotype did not occur in mutant EPEC-infected cells (). The wild-type phenotype was rescued by introducing a plasmid encoding the allele (p) into the mutant strain (). However, there was no phenotype associated with mutant strains complemented with SNX9 binding–deficient (p) (). Similar results were also found in EPEC-infected HeLa cells (unpublished data). Western blot analysis confirmed the expression and secretion of the EspF proteins by wild-type and trans-complemented strains (). These data further indicate that the biochemical activities of EspF that we have described in vitro are likely to occur during EPEC pathogenesis of intestinal epithelial cells.
#text
The gene from EHEC O157:H7 (GenBank accession no.) and the EPEC E2348/69 were PCR cloned in-frame into pEGFP-C2 (CLONTECH Laboratories, Inc.). The gene was cloned from a clinical isolate of EHEC 0157:H7 and has the sequence identical to GI: 15831969. The yeast two-hybrid clone of mouse SNX9 (NM_025664) residues 1–111, as well as full-length human SNX9, N-WASP, Grb2, NckI, and Nck2 were Topo cloned into pcDNA3.1-V5-His (Invitrogen). To produce mCherry-SNX9, the EGFP from pEGFP-C1 was replaced with mCherry by PCR subcloning. For bacterial expression, 47 amino acid N-terminal deletions (ΔN47) of EspF were PCR subcloned into pGEX-4T1 (GST-tag) (GE Healthcare) or pet30A (6xHis tag) (Novagen). EspF-D3 was generated with Multi-site QuikChange Site-Directed Mutagenesis (Stratagene) following the manufacturer's instructions. For bacterial complementation, the EPEC or mutant was Topo cloned into pTrcHis2 with a epitope tag (Invitrogen). GST and His-tagged mini-N-WASP were obtained from Wendall Lim (University of California, San Francisco, San Francisco, CA; ). All constructs were verified by DNA sequencing.
The yeast expression vector pLexA encoded a gene with N-terminal LexA binding domain and residues 1–248 of EHEC EspF. 250 μg of a d 9.5 and 10.5 mouse embryo library in VP16 were screened using the yeast two-hybrid system as previously described ().
Cell culture, transfections, and wide-field fluorescence microscopy was performed as previously described (). Images were acquired on a microscope (Axiovert; Carl Zeiss Microimaging, Inc.) using a digital camera (MicroMax; Roper-Princeton Instruments) controlled by MetaFluor software (Universal Imaging, Corp.). Optical filters were obtained from Chroma Technologies and 40 or 63× objectives were used for image acquisition.
Concentrations of anti-GFP polyclonal and monoclonal (CLONTECH Laboratories, Inc.), anti-His (QIAGEN), anti-GST, anti-flag M2, anti-flag polyclonal (Sigma-Aldrich), anti-HA (Covance), and anti-occludin (Zymed Laboratories) were used for immunoprecipitation, immunocytochemistry, and immunoblotting as recommended by the manufacturer. Rabbit polyclonal SNX9 antibody was used at 1:200 (). Rabbit polyclonal EspF was used at 1:10,000 (). Anti-tubulin antibody was used at 1:200 (Sigma-Aldrich), and Latrunculin-A at 1 μM (Calbiochem).
Recombinant 6×His EspFΔ47, mini-N-WASP (from Wendall Lim), N-WASP ΔEVH1 (from Jack Taunton; University of California, San Francisco, San Francisco, CA) and SNX9 were produced in BL-21/DE3 strain following standard methods. Cells were lysed in either His buffer (50 mM Tris, pH 7.4, 300 mM NaCl, 10 mM imidazole, and 0.5% Triton-X 100) or GST buffer (phosphate-buffered saline) supplemented with protease cocktail (Roche). Proteins were purified with Nickel agarose (QIAGEN) or glutathione-Sepharose (GE Healthcare) following the manufacturer's instructions. Proteins were dialyzed overnight in 2 l of lysis buffer at 4°C. Glycerol was added to 30% and aliquots were stored at −80°C. For GST pull-down assays, 10 μg of recombinant GST proteins immobilized to glutathione-Sepharose was incubated with 50 μl of [S]-methionine (GE Healthcare) produced with TNT reaction (Promega) for 4 h at 4°C. Proteins were separated by SDS-PAGE, dried, and exposed by autoradiography.
Immobilized peptides were synthesized on cellulose paper using a Multipep Autospot synthesis robot following the manufacturer's directions (Intavis AG). Membranes were blocked in 5% milk in TBS-Tween. 50 μl of [S]-methionine SNX9 produced by TNT (Promega) was added to 5 ml of block solution and overlayed onto peptide membranes for 4 h at room temperature. Membranes were washed 3× 10 min in TBS-T and exposed by autoradiography.
FITC-labeled peptides (Cell Essentials) used for fluorescence polarization include EspF residues 69–86 (FITC-ATSFTPSRPAPPPTSGQA) or control peptide (FITC-QIAKRRRLSSLRA). Peptides (5 nM) were suspended to working dilutions in PBS with 5 mg/ml BSA. Increasing concentrations of purified GST-SNX9-SH3 or control GST proteins were mixed with peptides in 100 μl and incubated at room temperature for 10 min. Fluorescence polarization (FP) was measured on a GENios Pro (TECAN) fitted with FP excitation and emission filters, 485/535 nm. Polarization values (mP) were determined at equilibrium and normalized to the highest value of saturation. Saturation binding curve was generated with PRISM software (Graph Pad) and dissociation constants () were calculated from the nonlinear regression curve from averages of three independent experiments.
A library of random dodecapeptides fused to the N terminus of the M13 gene-8 major coat protein was constructed and cycled through rounds of binding selections with the bacterially expressed SH3 domain immobilized on 96-well Maxisorp immunoplates (NUNC), as described previously (; ). After four rounds of selection, individual phage were isolated and analyzed in a phage ELISA. Phage that bound to the SH3 domain was subjected to DNA sequence analysis.
1:1 mix of pyrene actin (∼40% labeled, final is ∼20% labeled) to cold G-actin (Cytoskeleton, Inc.) were mixed in G-buffer (5 mM Tris, pH 8.0, 0.2 mM CaCl, 0.2 mM ATP, and 1 mM DTT) and centrifuged at 100,000 for 2 h. A final of 2.5 μM pyrene actin mix was added to 270 μl of Arp2/3 buffer (20 mM Tris, pH 7.5, 25 mM KCl, 1 mM MgCl, 0.1 mM ATP, and 1 mM ATP). 40 nm bovine Arp2/3 (a gift from Tom Pollard; Yale University, New Haven, CT) and 100 nM mini-N-WASP or N-WASP ΔEVH1 was mixed with various concentrations of GST, GST-EspF, or SNX9 constructs in 30 μl of 10× actin polymerization buffer (100 mM Tris, pH 7.5, 500 mM KCl, 20 mM MgCl, and 10 mM ATP). 270 μl pyrene actin (2.5 μM) was mixed with 30 μl mini-N-WASP-Arp2/3 and actin-assembly kinetics was monitored by pyrene fluorescence over a 10–20-min time interval. Peak fluorescence values were normalized to 2.4 μM actin consumption and graphed with PRISM Software.
TIR-FM imaging procedures and microscopic manipulations were conducted as previously reported (). Swiss 3T3 cells stably expressing clathrin-light chain DsRed () were grown to 70% confluency in a 100-mm tissue culture dish in DME + 10% fetal calf serum and transfected with 20 μg EGFP-EspF or EGFP-EspF-D3 DNA using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were allowed to recover in fresh medium for 2 h, replated onto 22 × 22-mm, no. 1.5 glass coverslips (Corning; = 1.523), and filmed the following morning.
Total internal reflection fluorescence (TIRF) microscopy was performed using an inverted microscope (TE2000U; Nikon) custom-modified to allow for through-the-objective multispectral total-internal reflection fluorescence microscopy using a 100×, 1.45 NA objective (Nikon), as previously described (). TIRF images were taken for 10 min at 2–3-s intervals, with a 100–300-ms exposure time, depending on the intensity of the signal.
Low expressing EGFP-EspF or EGFP-EspF-D3 cells stably expressing Clc-DsRed were chosen for ∼8–12 min time-lapse movies. Stacks of red (Clc-DsRed) and green (GFP-EspF) images were de-interleaved into separate stacks. Clc-DsRed stacks were analyzed for candidate clathrin structures that were stable for 30 s before disappearance and seemed likely to represent single CCPs. Clathrin structures were excluded similar to the criteria described previously (). We marked the position of each Clc-DsRed candidate with a ∼1–2-μm diameter circle and observed GFP-EspF dynamics at these sites. This procedure yielded the coordinates of the CCS in both the red and green channels so that red and green regions precisely centered on the CCS could be excised and stored as ministacks for analysis. All time and fluorescence intensity measurements were based on these criteria. Image analysis was done with Metamorph (Universal Imaging Corp), Excel, and graphed using PRISM (Graph Pad).
To measure 25 CCPs, we first chose 25 candidate CCPs from 8 different cells. Each CCP appeared during the 10-min video series was stable for at least 30 s, disappeared from the evanescent field, and did not reappear for the remainder of the video series. The peak fluorescence intensity measurements were collected for each individual EspF and clathrin image and averaged background values were subtracted. Clathrin results were normalized to the mean fluorescence during the 21 s (7 frames) before the peak clathrin fluorescence (). The time of clathrin departure (time 0) was defined as the last frame in which the normalized fluorescence value was within 90% of peak fluorescence and the signal reached <5% over the following 45 s of imaging. Each clathrin event ( = 25) was aligned to the time of departure and averaged. For EspF, the results of each event were normalized to the mean of three contiguous intensity values, including the peak value, then aligned to the time of clathrin departure and finally averaged (). Kymograph analysis was performed using MetaMorph software.
T84 colonic epithelial cells were cultured and TER experiments were performed as described previously (). EPEC and mutant strains () were complemented with plasmid EspF subcloned into pTrcHis2 (Invitrogen). Plasmid EspF was expressed in EPEC by incubation with 100 μM IPTG for 3 h before EPEC infections. T84 cells were infected with an MOI of 100 for the indicated time points. CaCo experiments were performed by infecting cells with EPEC or the indicated mutants for 4 h before fixation and immunofluorescence processing.
TAP-tagged EspF or mutant EspF-D3 were subcloned into pCDNA4T/O and MDCK stable cell lines were generated following the manufacturer's instructions (Invitrogen). Clonal lines expressing TAP-EspF or TAP-EspF-D3 were selected with 500 μg/ml zeocin and confirmed by Western blot and immunofluorescence microscopy. Purification of TAP-EspF was performed as follows: MDCK cells were lysed in 50 mM Hepes, pH 7.5, 150 mM NaCl, 5% glycerol, 0.5% TX-100, 1.5 mM MgCl, and 1 mM EGTA and incubated with anti-flag M2 beads for 2 h. The beads were washed 3× 10 min with lysis buffer and eluted with LDS-PAGE buffer (Invitrogen). Immunocomplexes were run down SDS-PAGE and gels were stained with Coomassie. Proteins were in-gel digested and analyzed with nano liquid chromatography tandem mass spectrometry. Acquired data was searched with Inspect against a dog database (National Center for Biotechnology Information, Canis familiaris fasta protein database). Identified peptides were manually verified to be canine SNX9: aa 218–237, SAAPYFKDSESAEAGGAQR; aa 313–327, SYIEYQLTPTNTNR; aa 346–359, FGSAIPIPSLPDK; aa 515–529, TYEEIAGLVAEQPK; aa 543–557, K.GFLGCFPDIIGAHK.G; and aa 605–613, IYDYNSVIR.
Fig. S1 shows membrane tubule formation induced by EspF, and colocalization of endogenous SNX9 at these sites (Fig. S1, A and B). Fig. S2 demonstrates that actin is polymerized de novo at sites of EspF/SNX9 membrane tubules. Video 1 shows dynamics of EGFP-EspF and Clc-DsRed in living Swiss 3T3 cells captured by TIR-FM. Video 2 shows EspF and clathrin dynamics at a single CCP. Video 3 shows a dynamic membrane tubule induced by EspF expressing at the cell surface. Video 4 demonstrates a “rocketing” clathrin structures seemingly being propelled by EspF. Video 5 demonstrates a splitting clathrin structures with EspF localizing to the center of this event. |
Leukocyte trans-endothelial migration (TEM) is a key event in host defense. The passage of leukocytes across the vascular wall into the underlying tissues can be divided into distinct phases, including firm adhesion of leukocytes to the endothelium and subsequent diapedesis (; ; ; for review see ). Leukocyte adhesion to the endothelium initiates the formation of dynamic dorsal membrane protrusions, assembling a cuplike structure, which surrounds adherent leukocytes and contains the cell adhesion molecules intercellular adhesion molecule-1 (ICAM1) and VCAM1 (; ; ). They have been referred to as docking structures () or trans-migratory cups (). Little is known about the mechanisms that regulate their assembly, and their role in TEM remains uncertain.
During TEM, leukocytes adhere to ICAM1 on the endothelial cell surface, and this triggers diverse intracellular signals (; ). Engagement of ICAM1 can be mimicked by cross-linking ICAM1 with ICAM1-specific antibodies (; ; ) or by beads coated with antibodies against ICAM1 (). Actin dynamics in endothelial cells are important for leukocyte TEM, which is prevented by inhibiting endothelial actin polymerization by cytochalasin D (; ). Cross-linking of ICAM1 stimulates the assembly of actin stress fibers (; ). In addition, actin polymerization is involved in assembly of the cups ().
Actin membrane dynamics are controlled by small Rho-like GTPases. These proteins function as molecular switches and cycle between an inactive GDP-bound state and an active GTP-bound state. Blocking RhoA activity using C3 transferase prevents the adhesion or migration of leukocytes across endothelial cell monolayers (; ). However, the role of RhoA in the assembly of the cups is unclear. reported that assembly of these structures induced by VCAM1 is inhibited by Y27632, an inhibitor of Rho-associated coil-containing protein kinase (ROCK)/Rho kinase, which is a downstream effector of RhoA. In contrast, found that treatment with Y27632 or C3 was unable to prevent cup formation downstream from ICAM1 engagement. The similarity of these apical cups to phagocytic cups (; ) together with the role of RhoG in the phagocytosis of apoptotic cells () has led us to examine whether RhoG may contribute to the formation of endothelial cups and participate in TEM.
In this study, we demonstrate that RhoG is a critical mediator of leukocyte TEM. RhoG and a guanine-nucleotide exchange factor (GEF) for RhoG, SH3-containing GEF (SGEF), are recruited to sites of ICAM1 engagement, where RhoG becomes activated. We find that ICAM1 interacts with SGEF through its SH3 domain. Finally, reduction of RhoG or SGEF expression in endothelial cells using siRNA decreases the assembly of the cups as well as the migration of leukocytes across endothelial cell monolayers.
Adhesion of myeloid leukemia HL60 cells to TNF-α–activated endothelial cells induced not only the recruitment of ICAM1 to sites of adhesion () but also ICAM1-positive membrane protrusions that surrounded the adhered leukocyte (), which is consistent with previously reported findings (; ). Also, GFP-actin, which is transiently expressed in endothelial cells, distributed to sites of leukocyte binding and colocalized with ICAM1 (). Of note, the endothelial cell–cell junctional marker vascular endothelial (VE) cadherin did not localize to these membrane protrusions (). Three-dimensional projections showed that ICAM1-positive protrusions arose from the apical plane of the endothelial cells but did not fully cover the leukocyte (). These protrusions resembled cuplike structures that extended ∼6–7 μm above the baso-lateral membrane (, d). To determine whether these ICAM1-rich cups formed around cells that were transmigrating, HL60 cells were plated on endothelial monolayers growing on transwell filters. Confocal analysis of fixed and stained preparations revealed rings of ICAM1 staining at the apical surface (i.e., cups) surrounding cells that were traversing the monolayer (Fig. S1, available at ). Scanning EM confirmed the presence of endothelial cuplike protrusions surrounding but not fully covering leukocytes 30 min after leukocyte adhesion ().
The small GTPase RhoG and its specific GEF, SGEF, are known to induce dorsal ruffles (). RhoG and SGEF are endogenously expressed in endothelial cells as well as in COS7 and HeLa cells (). Overexpression of the constitutively active mutant RhoG-Q61L or SGEF in endothelial cells induced ruffles on the apical surface ().
To study the involvement of ICAM1 in the regulation of dorsal ruffles, COS7 cells that lack endogenous ICAM1 were used. The expression of ICAM1 tagged with GFP or the V5 epitope in COS7 cells showed distributions similar to ICAM1 in endothelial cells (). Interestingly, cotransfection of RhoG-Q61L or SGEF not only induced dorsal ruffles but also induced a redistribution of ICAM1 to these ruffles (, A and B; and Videos 1 and 2, available at ). ICAM1 colocalized with RhoG-Q61L or SGEF (). The localization of ICAM1 to ruffles required active RhoG because neither wild-type (wt) RhoG nor a dominant-negative mutant, T17N, colocalized with ICAM1 (unpublished data). As a control, transmembrane protein PECAM-1 was expressed together with RhoG-Q61L or SGEF and showed no colocalization (unpublished data). These data suggested a role for RhoG and SGEF in the formation of endothelial apical cup structures; therefore, we next tested the involvement of RhoG and SGEF in ICAM1 signaling and cup formation.
COS7 cells lacking endogenous ICAM1 were used to express ICAM1-GFP. Incubation of these COS7 cells with HL60 cells resulted in the majority of HL60 cells adhering to the ICAM1-GFP–transfected cells (). Three-dimensional projections showed that ICAM1-positive protrusions surrounded the adhered HL60 cells (, d), similar to those observed with endothelial cells (). To specifically study ICAM1 engagement and downstream signaling that would mimic leukocyte binding to ICAM1, beads coated with antibodies against ICAM1 were used as described in Materials and methods (see Bead adhesion assay section; ). These beads, which are hereafter referred to as αICAM1 beads, specifically adhered to ICAM1 and recruited ICAM1-GFP within 30 min ( and Video 3, available at ). X-Z projections showed that ICAM1-GFP protruded around adhered αICAM1 beads (, d). Additionally, scanning EM images revealed that adhesion of αICAM1 beads induced dorsal ruffles comparable with those induced by leukocytes (). The αICAM1 beads did not bind to VCAM1-GFP–transfected cells or to nontransfected cells (unpublished data). In addition, blocking antibodies to ICAM1 completely inhibited binding of the αICAM1 beads to ICAM1 (unpublished data).
To show that ICAM1-GFP was recruited specifically to the beads, cotransfections with ICAM1-V5 and GFP as a control were performed and revealed that GFP alone was not recruited to sites of adhesion (). Also, neither β-catenin–GFP nor VE-cadherin–GFP was recruited to sites of adhesion (). In contrast, GFP-SGEF and GFP–RhoG-Q61L were recruited to sites of ICAM1 engagement (). Additionally, as a control, beads coated with major histocompatibility complex (MHC) antibodies were incubated on human umbilical vein endothelial cells (HUVECs), and z-stack analysis was performed to measure actin-rich protrusions around adhered beads. The results revealed that αICAM1 beads induced substantially more F-actin–rich protrusions than the αMHC class I beads, whereas the total number of beads that adhered to the endothelium was equivalent (Fig. S2 A, available at ). Expression of GFP–RhoG-Q61L in HUVECs showed that RhoG is recruited by αICAM1 beads but not by αMHC class I beads (Fig. S2 B). Previous work has indicated that actin is a major component of the ICAM1- positive cup structures (; ; ). Using GFP-actin, which is transiently expressed in endothelial cells, we confirmed that αICAM1 beads efficiently recruited actin to sites of adhesion (Fig. S2 C). These data indicate that ICAM1 specifically induces these protrusions and recruits RhoG to sites of adhesion.
We next performed RhoG activation assays to determine RhoG activity downstream from ICAM1 engagement. We made use of the RhoG downstream effector ELMO (engulfment and cell motility), which specifically binds GTP-bound RhoG (; ). In our initial experiments, we used an adenoviral vector to deliver myc-tagged RhoG to HUVECs and found that engagement of ICAM1 with αICAM1 beads induced RhoG activation (). Examining the activation of endogenous RhoG using a monoclonal antibody revealed that ICAM1 engagement showed a similar response (). It should be noted that TNF-α pretreatment did not change the activity of RhoG in endothelial cells, although overnight treatment slightly diminished RhoG expression (unpublished data). To delineate the pathway downstream from ICAM1, myc-tagged RhoG-wt together with ICAM1-GFP were expressed in COS7 cells as described in Materials and methods (see RhoG, RhoA, and Rac1 activation assay section). Treatment with αICAM1 beads induced RhoG activation after 10 and 30 min (). This activation was transient because the induced activity of RhoG declined after 60 min ( and Fig. S3 A, available at ). Beads coated with MHC class I antibodies did not induce any RhoG activation (Fig. S3 B). To examine whether leukocytes could activate RhoG through ICAM1, we added HL60 cells to myc–RhoG-wt and ICAM1-GFP–expressing COS7 cells. RhoG activation was stimulated by the adhesion of HL60 cells (). To study whether closely related GTPases Rac1 and Cdc42 are activated downstream from ICAM1 engagement, pull-down assays using the p21-activated kinase–binding domain (PBD) as bait were performed. Interestingly, Rac1 and Cdc42 were transiently activated downstream from ICAM1 engagement as well, although Rac1 activation peaked at 10 min (Fig. S3 C). RhoA activity measurements confirmed that RhoA became activated after ICAM1 engagement (; ), and this was maximal after 10 min (Fig. S3 E).
Previously, it has been shown that the intracellular domain of ICAM1 is required for leukocyte passage across the endothelium but is dispensable for the initial adhesion (; ). To investigate whether the intracellular domain of ICAM1 is required to transmit the signal that triggers RhoG activation, a C-terminal deletion mutant of ICAM1 lacking the intracellular domain and tagged to a V5 epitope (ICAM1-ΔC-V5) was generated and expressed in COS7 cells. The overexpression of ICAM1-ΔC-V5 together with GFP–RhoG-Q61L showed that ICAM1 required its intracellular domain to localize to RhoG-induced dorsal ruffles (). No difference in the adhesion of αICAM1 beads to either full-length or ICAM1-ΔC was observed (unpublished data). However, the αICAM1 beads were unable to activate RhoG in cells expressing ICAM1-ΔC (). Additionally, cells that expressed ICAM1-ΔC induced substantially less ICAM1-positive protrusions around adhered leukocytes than ICAM1-wt (). Together, these data show that ICAM1 engagement induces RhoG activation and subsequent membrane protrusions in a pathway that is dependent on its intracellular domain.
The finding that RhoG is activated downstream from ICAM1 engagement coupled with the observation that SGEF and RhoG colocalized with ICAM1 led us to investigate whether ICAM1 and SGEF physically interact. Immunoprecipitation experiments showed that endogenous ICAM1 was precipitated with endogenous SGEF from TNF-α–treated endothelial cells (). To study this interaction in more detail, pull-down experiments were performed using biotinylated peptides. A peptide corresponding to the cytoplasmic domain of ICAM1 bound myc-tagged SGEF as well as endogenous SGEF (, B and C, respectively). Interestingly, the intracellular domain of ICAM1 comprises only 28 amino acids, and its C terminus contains four prolines in close proximity. We examined whether the SH3 domain of SGEF could directly associate with the cytoplasmic domain of ICAM1. Biotinylated ICAM1–intracellular domain peptide sedimented the SH3 domain of SGEF, which was fused to GST (GST-SH3) in vitro (, a). To further explore the interaction of SGEF with ICAM1, we used a myc-tagged mutant of SGEF lacking the SH3 domain (SGEF-ΔSH3). This mutant SGEF failed to coimmunoprecipitate with ICAM1-GFP (, b). Interestingly, the association between SGEF and ICAM1 did not depend on the GEF activity of SGEF; ICAM1 still associated with a catalytically dead mutant of SGEF (SGEF-ΔDH) that contained the SH3 domain (, b). An inactivating point mutant in the SH3 domain of SGEF (myc– SGEF-W826R) was previously generated in which the catalytic activity of SGEF remained intact (). This construct and SGEF-wt were overexpressed in COS7 cells together with ICAM1-GFP. Immunoprecipitation assays confirmed that SGEF-wt interacted with ICAM1, but SGEF-W826R revealed decreased binding (). These data indicated that the ICAM1–SGEF interaction requires an intact SGEF-SH3 domain. To test whether ICAM1 associates through its proline-rich sequence to SGEF, we deleted this proline-rich sequence from the cytoplasmic domain of ICAM1. Immunoprecipitation studies revealed that ICAM1 lacking the proline-rich sequence failed to bind to SGEF ().
To study RhoG involvement in TEM, siRNA was used to reduce RhoG expression in primary endothelial cells. Western blot analysis revealed that the relevant siRNA reduced RhoG protein expression in endothelial cells but did not affect other proteins known to be present in cup structures or involved in transmigration, such as moesin and ICAM1 (; ; ). Also, the expression levels of other closely related small GTPases such as Rac1, Cdc42, and RhoA were unaffected (). Adhesion of leukocytes to endothelial monolayers that showed reduced RhoG expression was not affected. Similarly, expression of dominant-negative RhoG did not affect leukocyte adhesion (unpublished data). However, the formation of cup structures, which was quantified as ICAM1-positive ringlike structures that surrounded adhered leukocytes, was decreased compared with control cells (). Transmigration of HL60 cells across endothelial cell monolayers was also substantially attenuated by the knockdown of RhoG expression ().
Several previous studies have addressed the role of RhoA in endothelial cells during leukocyte TEM, demonstrating that it is required for TEM (; ) and showing that it becomes activated downstream from ICAM1 cross-linking (; ; ). We were interested to relate our RhoG results to this previous body of work on RhoA. Reducing RhoG expression by siRNA did not affect RhoA activation downstream of ICAM1 engagement (Fig. S4 A, available at ), which is consistent with the activation of RhoA occurring faster than the activation of RhoG (Fig. S4, A and E). Interestingly, reducing RhoA expression by siRNA depressed ICAM1-induced RhoG activation (Fig. S4 B). This suggested that RhoA acts upstream of RhoG activation in the pathway from ICAM1 engagement. Whether RhoA has a role in cup formation has been controversial. found that inhibiting the RhoA effector ROCK/Rho kinase with Y27632 diminished cup formation. However, this was not found by , who also were unable to block cup formation by treating endothelial cells with C3 or Y27632 (). Our finding that RhoA is required upstream of RhoG activation suggested that RhoA may be necessary for cup formation. Consequently, we investigated this directly using micro-RNA (miRNA) of RhoA to depress its expression. We have found that the depletion of RhoA reduced the formation of cups induced by αICAM1 beads (Fig. S4 C).
We wished to explore whether SGEF has a role in leukocyte TEM and, thus, have used siRNA to knockdown SGEF expression in endothelial cells. We confirmed that the siRNA decreased SGEF expression and that it did not affect the expression of RhoG, Rac1, or other proteins involved in cups, such as ICAM1 or moesin (). Importantly, SGEF knockdown did impair the activation of RhoG downstream from ICAM1 engagement (), and, consistent with this, it also resulted in decreased cup formation, as judged by the number of ICAM1-positive rings surrounding adherent leukocytes (). Together, these data indicate a pathway from ICAM1 clustering to SGEF to RhoG activation resulting in the formation of cups. Finally, we examined the effect of SGEF knockdown on TEM and found that it caused a decrease in the migration of HL60 cells across endothelial monolayers by up to 50% ().
During the last decade, it has become increasingly clear that endothelial cells, rather than being a passive barrier, actively participate in the process of leukocyte TEM. This study focuses on a recently discovered phenomenon that occurs during TEM in which the endothelial cell extends sheets of membrane to form a cuplike structure that surrounds adherent leukocytes (; ; ; ). Although their precise function is unclear, evidence has been presented that these structures assist leukocytes on their way through the endothelium ().
Our data reveal a new signaling pathway downstream from leukocyte adhesion that involves the small GTPase RhoG. We show here that RhoG activation is triggered through the engagement of ICAM1 and is critical for formation of the apical cups. Additionally, RhoG expression is needed for optimal leukocyte passage across the endothelium. Our data show a strong correlation between formation of the cups and TEM. The endothelial apical cups resemble phagocytic cups, and it is notable that RhoG has been implicated previously in the phagocytosis of apoptotic cells in (). Recent work has also implicated RhoG as well as its exchange factor SGEF in the uptake of by epithelial cells (). Engulfment of is promoted by several bacterial proteins that function to activate multiple Rho family GTPases. Interestingly, the protein SopB was found to activate SGEF and RhoG, thereby stimulating the formation of phagocytic cups on the surfaces of epithelial cells (). Together, these results suggest that SGEF and RhoG may function in a variety of physiological and pathological situations in which phagocytosis or the uptake of particulate material is involved.
The route by which leukocytes pass through the endothelium, whether it is paracellular or transcellular, has generated considerable debate for many years. In tissue culture models, it has been estimated that only 10–25% of all leukocytes use the transcellular route, with the majority migrating through cell–cell junctions (). have shown that the redistribution of ICAM1 to caveolin-rich membrane domains in response to engagement is followed by transcytosis to the baso-lateral side of the endothelium. The induction of apical cups by RhoG as well as the similarity of these structures to phagocytic cups might lead to the idea that RhoG would function primarily in transcellular rather than paracellular migration. However, our data show that silencing RhoG results in >70% inhibition of leukocyte TEM. Although our work does not discriminate between the para- and trans-cellular migration routes, this decrease in TEM cannot be explained by blocking the trans-cellular pathway only. Consistent with this, the work of suggests that trans-migratory cups are not restricted to the trans-cellular route but may function to facilitate and guide leukocyte TEM in general. Alternatively, RhoG may have additional functions in TEM besides mediating cup formation.
The role of Rho family GTPases in formation of the cup structures has begun to be investigated. found that Y27632, which inhibits ROCK/Rho kinase downstream of RhoA, decreased the assembly of these structures induced by VCAM1 engagement. However, Carman and Springer reported that neither C3 nor Y27632 inhibited the assembly of the cups induced by ICAM1 cross-linking (; ). In our hands, we have observed the partial inhibition of cup formation by Y27632 (our unpublished data) and have found that knockdown of RhoA also inhibits cup formation (Fig. S4 C). The depression of cup formation may, in part, be caused by the inhibition of RhoG activation in cells in which RhoA has been knocked down (Fig. S4 B). How RhoA regulates RhoG activation remains to be determined. In addition, RhoA may play other roles in the assembly of endothelial apical cups.
In this study, we have focused on RhoG, a close relative of Rac1 (), because it induces dorsal membrane ruffles and has been implicated in phagocytosis (). However, we have observed that ICAM1 engagement leads to the activation of not only RhoG and RhoA but also Rac1 and Cdc42 ( and S3). It is notable that RhoG can activate Rac1 through the DOCK180-binding protein ELMO (), raising the possibility that the activation of RhoG we observe stimulates Rac1 activation. However, the time course of the activation of Rac1 and RhoG is not consistent with this idea. In future work, it will be interesting to identify the pathways leading to the activation of these other Rho family members.
The intracellular domain of ICAM1 is a prerequisite for optimal TEM of leukocytes (; ). ICAM1 lacking its intracellular domain (ICAM1-ΔC) fails to promote leukocyte TEM, although leukocyte adhesion to ICAM1-ΔC is unaffected. Engagement of ICAM1-ΔC by αICAM1 beads also fails to activate RhoG. The fact that ICAM1-ΔC cannot activate RhoG is likely the result of its inability to bind SGEF. We found that the proline-rich region of the intracellular domain of ICAM1 binds the SH3 domain of SGEF. This interaction is independent of SGEF activation because catalytically inactive mutants of SGEF that express the SH3 domain still bind ICAM1. Engagement of ICAM1 does not promote the association between SGEF and ICAM1 but does increase the activation of SGEF, as judged by the increased binding of SGEF to nucleotide-free RhoG (unpublished data). Thus, SGEF and ICAM1 likely form a stable interacting pair.
Additional signals such as tyrosine phosphorylation may be necessary to trigger SGEF activation, as has been shown for other GEFs (). One such signal may depend on Src-kinase activity. Src-kinase is rapidly activated after ICAM1 engagement and is required for optimal leukocyte TEM but also does not affect leukocyte adhesion (; ; ; ). Our preliminary results show that inhibiting Src family kinases using PP2 prevented RhoG activation downstream from ICAM1 engagement (unpublished data). These data support the idea that additional signals such as tyrosine phosphorylation are needed to activate SGEF. It is likely that there are multiple targets for Src downstream from ICAM1. One Src substrate that has been implicated in TEM is cortactin (). Cortactin is a regulator of the actin cytoskeleton that is notably prominent in structures like membrane ruffles and phagocytic cups ().
The passage of leukocytes across the endothelium is a critical event in immune surveillance and in inflammation. Although inflammation is physiologically important, it also underlies many pathological conditions. Consequently, there is considerable interest in understanding the pathways by which leukocytes cross the endothelial barrier so that inappropriate inflammation can be controlled. Much remains to be learned about TEM, including the role of the cups that are formed in response to ICAM1 engagement. Different leukocyte types may induce different effects on the kinetics of ICAM1 signaling and subsequent apical cup formation. In this study, we have identified a pathway downstream from ICAM1 involving RhoG and its exchange factor SGEF that leads to endothelial apical cup formation. Inhibition of either RhoG or SGEF not only inhibits apical cup formation but also depresses TEM, which is consistent with, although does not prove, a role for the cups in TEM.
pAbs against ICAM1 (for Western blotting) and mAb against RhoA were obtained from Santa Cruz Biotechnology, Inc. mAbs against Rac1, Cdc42, and MHC class I (MHC-A, -B, and -C) were purchased from BD Biosciences. Recombinant TNF-α and a mAb against ICAM1 were purchased from R&D Systems. The GFP and myc (clone 9E10) mAbs were purchased from Invitrogen. Polyclonal rabbit antibody against VE-cadherin was purchased from Cayman Chemical. The SGEF rabbit pAb was generated in our laboratory as described previously (). The mAb against RhoG (clone IF-3-B3-E5) was raised in the laboratory of M.A. Schwartz (Robert M. Berne Cardiovascular Research Center, University of Virginia, Charlottesville, VA) against the C-terminal RhoG peptide (AA162-180) of the sequence QQDGVKEVFAEAVRAVLNPT. Dot blots showed that the mAb did not cross react with bacterially expressed Rac1, Cdc42, and RhoA. Western blotting analysis showed that the RhoG antibody did recognize GFP–RhoG-wt but not GFP–Rac1-wt expressed in COS7 cells.
SGEF cDNA was subcloned using BamHI–EcoRI restriction sites into pCMV6M, an N-terminal myc epitope–tagged eukaryotic expression vector, as described previously (). SGEF deletion mutants were generated using the QuikChange Site-Directed Mutagenesis kit (Stratagene) and were subcloned into pCMV6M. pGEX-4T2-ELMO was a gift from K. Ravichandran (University of Virginia, Charlottesville, VA). Generation of eukaryotic expression vectors pCMV-myc-Rac(Q61L), pCMV-myc–Rac-wt, pCMV-myc-Rac(T17N), pCMV-myc-RhoG(Q61L), pCMV-myc–RhoG-wt, and pCMV-myc-RhoG(T17N) was described previously by our laboratory (). wt and mutant Rac1 and RhoG constructs were subsequently subcloned into pEGFP-C3 (CLONTECH Laboratories, Inc.) as described previously (). SGEF was subcloned into pEGFP-C2. ICAM1-GFP was a gift from F. Sanchez-Madrid (Hospital de la Princesa, Universidad Autónoma de Madrid, Madrid, Spain). For ICAM1-ΔPro-GFP, the last 11 amino acids of the intracellular tail of ICAM1 were deleted. ICAM1-wt and C-terminal deletion mutant (lacking the last 28 amino acids) cDNA was subcloned into the pAdCMV-V5-DEST vector using the Gateway expression system (Invitrogen).
HUVECs were obtained from Cambrex and cultured as described previously (). Endothelial cells were activated with 10 ng/ml TNF-α overnight as indicated to mimic inflammation. All cell lines were cultured or incubated at 37°C at 10% CO. The HL60 promyelocytic cell line was obtained from the University of North Carolina's Lineberger Comprehensive Cancer Center Tissue Culture Facility and grown in Optimem plus 5% FBS. In all experiments described, differentiated HL60 cells were used. Differentiation to a neutrophil-like lineage was achieved by adding 1.3% DMSO for 3–5 d (). COS7 cells were maintained in growth medium (Iscove's modified Dulbecco's medium with 10% FCS; Sigma-Aldrich). Cells were transiently transfected with the expression vectors indicated in each experiment according to the manufacturer's protocol using LipofectAMINE PLUS (Invitrogen) or Fugene 6 (Roche). Myc–RhoG-wt cDNA was transferred to an AdV expression vector and transfected into 293 cells, and high titer virus stocks were produced. Subsequently, myc–RhoG-wt was transiently delivered into HUVECs by adenovirus transduction.
Cells were cultured on glass coverslips, fixed, and immunostained with the indicated primary antibodies for 60 min at RT as described previously (). Subsequent visualization was performed with AlexaFluor-conjugated secondary antibodies for 30 min (Invitrogen). F-actin was visualized with fluorescently labeled phalloidin (Invitrogen). Glass coverslips were mounted in MOWIOL at RT. Images were collected with a confocal microscope (LSM510; Carl Zeiss MicroImaging, Inc.) equipped with a microscope (Axiovert 100M; Carl Zeiss MicroImaging, Inc.) and an oil immersion plan-Neofluar 63× NA 1.3 oil lens (Carl Zeiss MicroImaging, Inc.). Cross talk between the different channels was avoided by the use of sequential scanning. Images were processed using imaging examiner software (Carl Zeiss MicroImaging, Inc.) and Photoshop CS (Adobe).
Transfected cells were grown on glass coverslips, fixed in 2.5% glutaraldehyde/PBS for 30 min at room temperature, and processed for scanning EM as described previously (). In brief, samples were incubated with 2% aqueous osmium tetroxide for 45 min, dehydrated in a graded ethanol series, and critical point dried in liquid CO using a drying apparatus (CPD 010; Balzers Instruments). Samples were mounted on aluminum stubs (Ted Pella, Inc.) and sputter coated with gold/palladium using Polaron scanning EM. Cells were examined on a scanning electron microscope (model 820; JEOL) at 15 kV.
Migration assays were performed in transwell plates (Corning) of 6.5-mm diameter with 8-μm pore filters. Approximately 20,000 endothelial cells were plated on matrigel-coated transwell filters, which were treated the next day with siRNA as indicated. The following day, endothelial cells were treated with siRNA again and with 10 ng/ml TNF-α overnight at 37°C and 10% CO. 100,000 differentiated HL60 cells were added to the upper compartment, and HL60 cells were allowed to migrate to 50 ng/ml stromal cell–derived factor-1 (SDF-1; placed in the lower chamber to generate a chemotactic gradient; R&D Systems) for 4 h at 37°C and 10% CO. An input control (i.e., 100,000 HL60 cells) was set as 100%. After collecting the migrated HL60 cells, filters were inspected by confocal laser-scanning microscopy using fluorescently labeled phalloidin to stain F-actin; coating of matrigel on the transwell filter did not affect the formation of a confluent endothelial monolayer. Migrated HL60 cells were counted and compared with 100% input, and the percent migration of HL60 cells was calculated. To confirm efficient knockdown of the protein by siRNA, cells were simultaneously grown in six-well plates and equally treated with siRNA constructs and were analyzed by Western blotting.
Cells were grown to confluency, washed twice gently with ice-cold Ca- and Mg-containing PBS, and lysed in 300 μl lysis buffer (25 mM Tris, 150 mM NaCl, 10 mM MgCl, and 1% Triton X-100 with the addition of fresh protease inhibitors, pH 7.4). Immunoprecipitation was performed as previously described () and analyzed by Western blotting using an enhanced ECL detection system (GE Healthcare). The intensity of the bands was quantified by using ImageJ version 1.36 (National Institutes of Health, Bethesda, MD).
Using confocal laser-scanning microscopy, z-stacks were taken to confirm the formation of a cup around an adhered leukocyte. The length of the protrusion was ∼6–7μm above the baso-lateral plane of the substrate (, d). The apical plane was set to 4 μm from the baso-lateral plane (). ICAM1-positive rings in the apical plane were counted as positive cups.
For RhoG activation assays, a transient coexpression of myc-tagged RhoG was used because of the lack of a high affinity antibody that is appropriate for these assays (according to ). Transfected cells were lysed in 300 μl of 50 mM Tris, pH 7.4, 10 mM MgCl, 150 mM NaCl, 1% Triton X-100, 1 mM PMSF, and 10 μg/ml each of aprotinin and leupeptin. Lysates were cleared at 14,000 for 10 min. Supernatants were rotated for 30 min with 60–90 μg GST-ELMO (GST fusion protein containing the full-length RhoG effector ELMO) conjugated to glutathione–Sepharose beads (GE Healthcare). Beads were washed in 50 mM Tris, pH 7.4, 10 mM MgCl, 150 mM NaCl, 1% Triton X-100, and protease inhibitors. Pull-downs and lysates were then immunoblotted for the myc epitope tag. For RhoA and Rac1, GST-Rhotekin and GST-PBD were used as baits, respectively, and used as described for GST-ELMO.
GST-ELMO, GST-SH3 (SGEF), GST-Rhotekin, and GST-PBD fusion proteins were purified from BL21 cells (Stratagene) using glutathione–Sepharose 4B as previously described (). GST fusion proteins were eluted with free, reduced glutathione in TBS medium (50 mM Tris, 150 mM NaCl, 5 mM MgCl, pH 7.4, and 1 mM DTT) and stored in 30% glycerol at −80°C.
3 μm polystyrene beads (Polysciences, Inc.) were pretreated with 8% glutaraldehyde overnight, washed five times with PBS, and were incubated with 300 μg/ml ICAM1/MHC mAb according to the manufacturer's protocol.
For immunofluorescence or scanning EM, 1 μg/ml of antibody-containing beads was washed and resuspended in culture medium. 1 μg/ml of antibody-coated beads was incubated in wells of 24-well dishes containing glass coverslips, on which TNF-α–pretreated HUVECs or COS7 cells were cultured. After the appropriate time, unbound beads were removed, and coverslips were put on ice, gently washed three times with ice-cold PBS containing 1 mM Ca/Mg, and subsequently processed for immunofluorescence. For biochemistry, 10 μg/ml of antibody-coated beads were incubated on the cells, after which cells were washed as described above (see Bead adhesion assay section) and subsequently lysed and processed as described (see Immunoprecipitation and Western blotting and RhoG, RhoA, and Rac1 activation assay sections).
siRNA duplexes against human RhoG (sense, GCAACAGGAUGGUGUCAAGUU; antisense, 5′-P-UCGUCCAAGAUCGACAUCC UU) and SGEF mRNA (sense, CAAAUGGCCUUGCCGCUAAUU; antisense, 5′-P-UUAGCGGCAAGGCCAUUUGUU) and siControl nontargeting siRNA were obtained from the Dharmacon siRNA collection. HUVECs were transfected twice with 50 nmol/l siRNA using RNAifect transfection reagent (QIAGEN). After 48 h, cells were processed as described in the previous paragraph.
miRNA adenoviral constructs were engineered according to the manufacturer's protocol (Invitrogen). In brief, two sets of DNA oligonucleotides were designed to target human RhoA mRNA and were named RhoA#1 and RhoA#2: TGCTGAAGACTATGAGCAAGCATGTCGTTTTGGCCACTGACTGACGACATGCTCTCATAGTCTT and CCTGAAGACTATGAGAGCATGTCGTCAGTCAGTGGCCAAAACGACATGCTTGCTCATAGTCTTC (RhoA#1) and GCTGTTTCCATCCACCTCGATATCTGTTTTGGCCACTGACTGACAGATATCGGTGGATGGAAA and CCTGTTTCCATCCACCGATATCTGTCAGTCAGTGGCCAAAACAGATATCGAGGTGATGGAAAC (RhoA#2). The oligonucleotides were annealed and ligated into pcDNA6 EmGFP. The EmGFP MiR RNA cassette was subsequently transferred to pDONR221 and finally to pAd by two sequential Gateway BP and LR recombinations. Each construct was sequence verified, and viral particles were produced by transfection in 293A cells.
Peptides were synthesized with the following sequence: ICAM1–intracellular tail peptide; biotin-GRQRKIKKYRLQQAQKGTPMKPNTQATPP-OH; αv peptide; and biotin-GHENGEGNSET-OH.
COS7 cells were cultured on glass coverslips and transfected with cDNA as indicated. After 24 h, cells were placed in a heating chamber at 37°C and recorded with a confocal microscope (LSM510; Carl Zeiss MicroImaging, Inc.) equipped with a microscope (Axiovert 100M; Carl Zeiss MicroImaging, Inc.) and an oil immersion plan-Neofluar 63× NA 1.3 oil lens (Carl Zeiss MicroImaging, Inc.).
Fig. S1 shows the recruitment of endogenous ICAM1 around a migrating HL60 cell. Fig. S2 shows the recruitment of GFP–RhoG-Q61L and F-actin to αICAM1 beads but not to αMHC class I beads. Fig. S3 shows activation of the small GTPases RhoG, Rac1, Cdc42, and RhoA downstream of ICAM1 engagement. Fig. S4 shows that reduced RhoG expression does not affect ICAM1-mediated RhoA activation but that the reduced expression of RhoA does influence RhoG activity downstream from ICAM1 engagement, which is induced by αICAM1 beads. Video 1 shows a real-time recording of 10 min of GFP–RhoG-Q61L expression in COS7 cells. Video 2 shows a real-time recording of 10 min of GFP-SGEF expression in COS7 cells. Video 3 shows a real-time recording of ICAM1-GFP expressed in COS7 cells and incubated with αICAM1 beads for 30 min. Online supplemental material is available at . |
Dendritic spines are highly specialized microscopic protrusions on dendrites that typically have an enlarged head connected to the dendritic shaft by a narrow neck (; ). Dendritic spines are the primary site where the postsynaptic components of excitatory synapses are located in the mammalian central nervous system. The spine neck serves to compartmentalize Ca in the spine head away from the dendritic shaft, thus modulating excitatory synaptic transmission (; ; ).
The formation and maintenance of dendritic spines require the coordinated activity of structural and signaling molecules (; ). Particularly important are proteins involved in the regulation of actin filaments, which are the main cytoskeletal component of spines that help define their shape (). Time-lapse imaging has revealed that dendritic spines are dynamic structures that can undergo actin-based shape modifications and, in some cases, can form and disassemble within minutes (; ; ; ). Spine remodeling also occurs during the physiological stimulation of synapses and is correlated with changes in the strength of synaptic transmission. Structural reorganization of spines could be important for cognitive processes such as learning and memory (; ; ). Indeed, several cognitive disorders are associated with spine malformations and changes in spine density (; ). Thus, molecular dissection of the mechanisms involved in regulating dendritic spine morphology could be critical for understanding cognitive function and pathological conditions of the brain.
The Eph family of receptor tyrosine kinases and their ligands, the ephrins, play critical roles in both the developing and mature nervous system (; ; ; ). Eph receptors and ephrins interact at sites of cell–cell contact, where both are anchored to the cell surface. Ligand–receptor engagement elicits bidirectional signals: forward signals are generated through the Eph cytoplasmic region, including the tyrosine kinase domain, and reverse signals are communicated by the ephrin through cytoplasmic signaling molecules. Eph receptors and ephrins modify cell shape and movement by reorganizing the actin cytoskeleton. Several recent studies have shown that signaling by EphB receptors regulates dendritic spine morphogenesis, a process whereby the long and thin dendritic filopodial protrusions of immature neurons are replaced by dendritic spines in more mature neurons (; ; ; ). Among the Eph receptors of the A class, EphA4 has been shown to regulate dendritic spine morphological plasticity in the adult hippocampus ().
Signaling by EphA4, which is expressed on dendritic spines of pyramidal neurons, reduces spine length and density in acute hippocampal slices. A ligand for EphA4 in the hippocampus is ephrin-A3, which is glycosyl-phosphatidylinositol linked and localized on the surface of the astrocytic processes that surround spines. The molecular interplay between EphA4 and its ligand ephrin-A3 likely mediates a form of cell contact– dependent communication between astrocytes and neurons that helps to maintain the organization and architecture of dendritic spines by restricting their ability to grow and change shape (). Indeed, EphA4 knockout mice have disorganized, longer, and more numerous spines than wild-type mice. Spine defects also accompany the introduction of a kinase-inactive form of EphA4 in pyramidal neurons, suggesting that kinase-dependent signals are involved in spine regulation. Therefore, identifying the signaling pathways that are activated downstream of EphA4 is important for understanding how dendritic spines can be structurally modified by cell surface receptors as well as for elucidating aspects of neuron–glia communication. In this study, we show that EphA4 regulates the morphological plasticity of dendritic spines by modulating integrin activity and downstream signaling molecules.
To investigate EphA4 signaling pathways involved in the regulation of dendritic spine morphology, we stimulated acute slices from postnatal mouse hippocampus with the ephrin-A3 ligand. Ephrin-A3 Fc, a soluble dimeric form of ephrin-A3, increased the tyrosine phosphorylation of EphA4 () and several other proteins ().
Src and Fyn are nonreceptor tyrosine kinases of the Src family that function downstream of EphA receptors and regulate the actin cytoskeleton (; ; ; ). Fyn in particular is present in mature synapses, plays a role in synaptic plasticity, is required for spatial learning and memory, and is needed for maintenance of dendritic spines (; ). Interestingly, we found that an ∼120-kD tyrosine-phosphorylated protein (or proteins) coprecipitated with Fyn but not Src from hippocampal slices (), and ephrin-A3 Fc stimulation decreased the phosphotyrosine signal (). This suggests that one or more Fyn-associated proteins are dephosphorylated or become uncoupled from Fyn upon EphA receptor activation by ephrin in hippocampal slices. On the other hand, we did not detect changes in Fyn activation after ephrin-A3 Fc stimulation ().
To identify the tyrosine-phosphorylated proteins associated with Fyn, we examined the interaction of Fyn with ∼120-kD proteins known to bind Src family kinases, including the two closely related nonreceptor tyrosine kinases focal adhesion kinase (FAK) and proline-rich tyrosine kinase 2 (Pyk2), and the scaffolding protein Crk-associated substrate (Cas). Probing Fyn immunoprecipitates from hippocampal lysates for FAK or Pyk2 did not reveal a reproducible association (), although in some experiments, weak signals for FAK and Pyk2 were detectable ( and not depicted). In contrast, substantial amounts of Cas coprecipitated with Fyn (). Both the SH2 and SH3 domains of Src family kinases can associate with the Src-binding domain of Cas (; ). Treatment of hippocampal slices with ephrin-A3 Fc decreased Cas binding to the SH2 domain of Fyn alone or fused with the SH3 domain in pull-down experiments () and inhibited Cas association with full-length Fyn in coimmunoprecipitation experiments ().
As expected from the reduced Cas–Fyn association (; ), ephrin-A3 Fc also decreased Cas tyrosine phosphorylation (). This inhibition of Cas phosphorylation was not observed in hippocampal slices from EphA4 knockout mice ( and Fig. S2, available at ), suggesting that EphA4 is the main EphA receptor responsible for the effects of ephrin-A3 Fc in hippocampal slices. The reason for the lower basal levels of Cas phosphorylation in the EphA4 knockout slices remains to be determined.
Using a phosphospecific antibody, we found that ephrin-A3 Fc stimulation decreased phosphorylation of the substrate domain of Cas (), which contains multiple YxxP motifs. When phosphorylated, these motifs mediate binding of the SH2 domain of the adaptor protein Crk, which is an important effector of Cas (; ). Indeed, ephrin-A3 Fc treatment inhibited Cas binding to Crk in pull-down () and coimmunoprecipitation experiments (). Phosphorylation of Crk on tyrosine 221 is also known to negatively regulate Crk interaction with Cas (), but ephrin stimulation did not affect the phosphorylation of this site (). Thus, EphA4-induced dephosphorylation of Cas in the hippocampus disrupts Cas complexes.
Cas is a well-known regulator of cell shape and motility (; ) and, therefore, may play a role in dendritic spine regulation downstream of EphA4. Indeed, both EphA4 and Cas are present in brain synaptosomal preparations, including the postsynaptic density (PSD) fraction (). Furthermore, Cas expression in primary hippocampal neurons but not in glial cells () supports the idea that Cas is involved in EphA4 forward signals that control dendritic spine morphology.
Biolistic transfection of a pool of four Cas siRNAs or individual Cas siRNAs in cultured hippocampal slices caused a substantial decrease in dendritic spine density (30–45%; ) and a small but significant decrease in spine length (13–16%; P < 0.01) compared with transfection with control siRNAs that do not knock down Cas expression ( and Fig. S3, available at ). Cas siRNA transfection also decreased the number of mushroom-shaped spines, which represent the most common category of spines in hippocampal pyramidal neurons, and increased the number of stubby spines (). These effects were caused by the down-regulation of Cas expression because they were abolished when human Cas was cotransfected with a mouse-specific Cas siRNA (Fig. S4). Expression of the different Cas siRNAs did not detectably affect the overall morphology of the pyramidal neurons (unpublished data). These results implicate Cas as an important regulator of dendritic spines.
To identify the functional domains of Cas that play a role in dendritic spines, we transfected Cas deletion mutants (). Spine density was reduced by 48% in pyramidal neurons expressing the CasΔSB mutant (which lacks the Src-binding domain) and by 47% in neurons expressing the CasΔSH3 mutant (). In contrast, transfection of a Cas mutant that lacks the substrate domain (CasΔSD) and, therefore, cannot bind Crk did not significantly affect spine density (P > 0.05; ). Spine length and width were not significantly affected by any of the Cas mutants (P > 0.05), with the exception of a small increase in spine width in the neurons transfected with CasΔSH3 and CasΔSB. With regard to spine shape, the CasΔSH3 and CasΔSB mutants caused a decrease in mushroom spines and a prominent increase in stubby spines, whereas CasΔSD decreased mushroom spines while increasing thin spines (). These results suggest that protein interactions mediated by the Cas SH3 and Src-binding domains contribute to maintaining spine density and play a role in defining spine shape.
The Cas SH3 domain is known to be a docking site for the FAK and Pyk2 kinases (; ), and we found that treatment of hippocampal slices with ephrin-A3 Fc decreases both FAK and Pyk2 tyrosine phosphorylation (, a and b). Because FAK, Pyk2, and Cas all become tyrosine phosphorylated upon integrin-mediated adhesion (; ), the coordinated decrease in their phosphorylation suggested that EphA4 activation may suppress integrin activity. Indeed, ephrin-A3 Fc stimulation of the hippocampal HT22 neuronal cell line, which expresses endogenous EphA4, caused a 35% decrease in cell attachment to surfaces coated with the integrin ligand fibronectin, whereas the integrin-independent attachment to poly--lysine was not affected (). Similar results were obtained with the neuronal-like B35 cell line, which stably expresses EphA4 (). Concurrent with the reduction in cell attachment, ephrin-A3 Fc stimulation reduced the level of Cas tyrosine phosphorylation in both cell lines (), which is similar to the effect observed in hippocampal slices.
Integrins are heterodimers containing an α and β subunit, and integrins containing the β1 subunit are highly expressed in the brain and influence hippocampal synaptic plasticity (; ; ; ). To determine whether EphA4 may inactivate β1 integrins, we used the 9EG7 antibody, which, at low concentrations, detects the ligand-occupied (active) conformation of β1 integrins without interfering with the binding of extracellular matrix ligands (). Ephrin-A3 Fc treatment of HT22 cells caused a strong decrease in 9EG7 antibody binding () without affecting the cell surface levels of β1 integrins, as determined by surface protein biotinylation (). Thus, EphA4 activation by ephrin-A3 inhibits β1-integrin activity and adhesion to fibronectin in neurons.
To examine whether integrins regulate dendritic spine morphology, we treated cultured hippocampal slices with an integrin function-blocking peptide (RGDFV) or a control peptide (RADFV; ). Peptides containing the RGD (Arg-Gly-Asp) motif competitively antagonize integrin binding to extracellular matrix ligands (; ; ). Treatment of slices with the RGDFV peptide significantly decreased Cas tyrosine phosphorylation (P < 0.05; ). Although RGD peptides have also been shown to promote integrin activation depending on the experimental conditions (; ; ), this result is consistent with the expected inhibition of integrin downstream signaling pathways. The dendritic spines in slices treated with RGDFV peptide were 28% shorter than those treated with the control peptide (), whereas spine width was reduced only slightly, and spine density was not significantly altered (P > 0.05; ). The shape of spines treated with the RGDFV peptide shifted from a mushroom-shaped to a stubby appearance (). These results show that integrin activity is required to maintain proper spine length and suggest that suppression of integrin function could play an important role in the effects of EphA4 on dendritic spine morphology.
To prevent β1-integrin inactivation during ephrin-A3 Fc stimulation, we treated acute hippocampal slices with high concentrations of the 9EG7 antibody. The antibody alone did not affect dendritic spine length, density, or width (; compare Fc with β1-Ab/Fc). As previously shown (), treatment of hippocampal slices with ephrin-A3 Fc caused a significant decrease in spine length (P < 0.001; ) and density (), with minor effects on spine width (, compare Fc with A3-Fc). In addition, as we show here, ephrin-A3 Fc decreased the proportion of mushroom-shaped spines while increasing stubby spines (). Remarkably, the 9EG7 antibody completely blocked all of the effects of ephrin-A3 Fc on the spines (; compare β1-Ab/Fc with β1-Ab/ephrin-A3 Fc). This is consistent with the lack of detectable Cas dephosphorylation when ephrin-A3 Fc was used together with the 9EG7 antibody (), although the antibody itself decreased Cas phosphorylation. We obtained very similar results when we treated hippocampal slices with the divalent cation Mn, which is known to promote the activation of many integrins, including β1 integrins (Fig. S5, available at ; ; ). Thus, preventing β1-integrin inactivation counteracts the effects of EphA4 on dendritic spines.
This study provides the first direct evidence that EphA receptors regulate β1-integrin activity and downstream signaling pathways in neurons, which is crucial for governing the morphological plasticity of dendritic spines. Activation of EphA4 by ephrin-A3 may disrupt critical integrin-mediated attachment to the extracellular matrix and cell surface integrin ligands, thus inhibiting integrin signaling to Cas and other downstream signaling proteins. Inhibition of integrin function with an RGD antagonistic peptide and inhibition of Cas function with siRNAs decreased spine length and the proportion of mushroom-shaped spines while increasing stubby spines in hippocampal slices. These spine morphological changes are strikingly similar to those elicited by EphA4 activation with ephrin-A3 Fc, which is consistent with the idea that the different manipulations affect components of a common signaling pathway. However, the RGD peptide did not decrease spine density, suggesting that integrin inhibition by the peptide down-regulates Cas function less drastically than Cas siRNAs or that it may affect additional Cas-independent pathways.
Several studies have shown that Eph receptor forward signaling modulates the adhesion and migration of nonneuronal cells by affecting integrin function (Table S1 in ). Both positive and negative effects of Eph receptors on integrin-mediated adhesion have been reported, depending on the receptor involved and the cellular context. The interplay between Eph receptors and integrins may involve a direct physical association, as reported for EphA4 and αIIbβ3 integrin in platelets ().
Many integrin α and β subunits are expressed in the brain (; ). Among them, the β1 subunit, which can associate with various α subunits, is found in brain synaptosomal preparations and in the dendritic spines of cultured hippocampal neurons (; ; ). Signals mediated by integrins, including β1, have been shown to increase the length and density of dendritic protrusions and promote synapse formation and remodeling in cultured hippocampal neurons (; ). Furthermore, siRNA knockdown of the α5- or β1-integrin subunits in hippocampal neurons has been shown to decrease the number of dendritic protrusions and synapses (; D.J. Webb, H. Zhang, and A.F. Horwitz. 2005. Society for Neuroscience. Abstr. 501.5). These findings are consistent with the idea that integrin inactivation contributes to the effects of EphA4 on dendritic spine morphology and density. The ability of a β1 integrin–activating antibody to block spine retraction downstream of EphA4 and the observed inhibition of β1-integrin activity by EphA4 suggest that integrins containing the β1 subunit play a particularly important role downstream of EphA4.
It is noteworthy that inhibiting integrin function with RGD peptides through the use of integrin function–blocking antibodies or by targeted gene deletion has been shown to cause defects in long-term potentiation (LTP), a form of synaptic plasticity (; ; ; ). The effects of integrin inactivation on synaptic plasticity have been associated with decreased -methyl--aspartate (NMDA) receptor function and defects in actin polymerization (; ). It will also be interesting to examine whether the changes in spine shape induced by integrin inactivation may have a functional impact on synaptic transmission and plasticity. Recent findings suggest that the induction of LTP, which is dependent on an increase in cytosolic Ca concentration, may be more difficult to achieve in dendritic spines with thick neck regions such as stubby spines as a result of the faster dispersion of Ca ions into the dendritic shaft (; ; ). Inhibition of integrin activity may therefore moderate LTP induction and stabilization by EphA4 ().
Consistent with a role of integrins in dendritic spine regulation and synaptic physiology, the extracellular matrix has also been shown to influence spine morphological plasticity. For example, mutant mice that have reduced levels of reelin, a large secreted glycoprotein that binds to the α3β1 integrin, have fewer and shorter spines compared with wild-type mice (). Furthermore, degradation of extracellular matrix proteins by tissue plasminogen activator, which likely reduces integrin activity, has been implicated in the structural remodeling of spines that occurs in the visual cortex during experience-dependent plasticity (; ). Whether EphA4 may inhibit integrin activity by promoting proteolytic degradation of the extracellular matrix remains to be determined.
We have found that integrin inactivation by EphA4 results in the disassembly and inactivation of integrin signaling complexes, including the dissociation of Cas–Fyn complexes and the dephosphorylation of Cas, FAK, and Pyk2. The phosphatase mediating these dephosphorylation events remains unknown, although the tyrosine phosphatase SHP2 has been previously implicated in FAK dephosphorylation downstream of an EphA receptor in nonneuronal cells ().
Even though Cas is most highly expressed in the brain and has been detected in complex with the NMDA receptor (; ), its role in the nervous system is poorly characterized. Recent data have implicated Cas in axon elongation in cerebellar granule neurons (), and our data show that Cas is preferentially expressed in hippocampal neurons rather than glial cells and is enriched in the PSD fraction of synaptosomal preparations. Furthermore, Cas knockdown in hippocampal slices decreased dendritic spine density and caused changes in spine morphology. The observed involvement of the Src-binding and SH3 domains of Cas in spine regulation further suggests that Cas may cooperate with Fyn and kinases of the FAK family, which interact with these two domains, to maintain proper spine morphology.
Interestingly, we detected Cas association with Fyn but not Src in hippocampal tissue. Although both Src and Fyn are components of the PSD and of NMDA receptor complexes (), differences in Src and Fyn protein interactions in the brain have been reported (; ). Thus, the synaptic functions of these two closely related kinases are not completely redundant. Accordingly, Fyn but not Src knockout mice demonstrate defects in the induction of LTP in adult CA1 hippocampal synapses (; ).
The dephosphorylation of FAK and Pyk2 that we observed in hippocampal slices stimulated with ephrin-A3 is consistent with reduced integrin function and suggests that inactivation of these two kinases may also contribute to spine morphological plasticity downstream of EphA4. However, the role of FAK in dendritic spines seems to be complex because FAK activation has been recently shown to be important for spine shortening and morphogenetic changes downstream of EphB receptors in cultured hippocampal neurons ().
Surprisingly, we did not detect significant changes (P > 0.05) in dendritic spines expressing the CasΔSD mutant, which does not bind Crk, except for a decrease in mushroom-shaped spines. Crk couples phosphorylated Cas to the guanine nucleotide exchange factor DOCK180, which activates Rac1, thus promoting actin polymerization and cell migration (). Therefore, either the Cas–Crk pathway does not play a major role in spine regulation or another signaling pathway can compensate for the disruption of Cas–Crk complexes. Indeed, EphA4 activation does not affect Crk phosphorylation on tyrosine 221, suggesting that the adaptor function of Crk is not impaired by EphA4 and that molecules other than Cas could still recruit Crk to promote Rac1 activation ().
Our results demonstrate that EphA4 function in dendritic spines is closely tied to the regulation of β1-integrin signaling pathways, but additional pathways can also influence spine structure downstream of EphA4. For example, in neurons, EphA4 promotes the activity of at least two exchange factors that activate Rho family GTPases, Ephexin-1, and Vav2 (; ; ). Recently, a pathway involving the serine-threonine kinase Cdk5 and Ephexin-1 has been reported to decrease dendritic spine density downstream of EphA4 (). This pathway likely affects dendritic spines through the activation of RhoA, a GTPase that promotes actomyosin contraction (; ). Nevertheless, β1-integrin inactivation appears to be a crucial event because preventing it counteracts all EphA4-dependent spine modifications. Further studies will be necessary to fully elucidate the Eph forward and ephrin reverse signaling pathways that modulate the morphological plasticity of excitatory synapses.
Monoclonal antibodies to phosphotyrosine (clone 4G10), Src, PSD95, and polyclonal antibody to Pyk2 were obtained from Upstate Biotechnology. Monoclonal antibody to FAK, Cas, Crk, and β1 integrin were purchased from BD Biosciences. The phosphospecific antibodies to Crk tyrosine 221, Cas tyrosine 165, and Src tyrosine 416 were obtained from Cell Signaling. The monoclonal 9EG7 antibody to activated β1 integrin was obtained from BD Biosciences. Monoclonal antibodies to β-actin and syntaxin were purchased from Sigma-Aldrich. Antibodies to Fyn and SHP2 and a Cas polyclonal antibody were purchased from Santa Cruz Biotechnology, Inc. The monoclonal antibody to EphA4 was obtained from Zymed Laboratories.
Tissues or cells were lysed in Hepes buffer (50 mM Hepes, pH 7.5, 150 mM NaCl, 5 mM KCl, 1 mM EDTA, 10% glycerol, and 1% Triton X-100) or modified radioimmunoprecipitation assay buffer with 1 mM sodium orthovanadate, 1 mM NaF, 1 mM PMSF, and other protease inhibitors (Sigma-Aldrich). GST-Fyn SH2 and GST-Crk SH2 fusion proteins have been previously described (, ). Recombinant GST fusion proteins were affinity purified from BL21 (Pfizer) by adsorption to glutathione–Sepharose 4B beads (Pfizer). GST-Fyn SH3/SH2 fusion protein was purchased from Santa Cruz Biotechnology, Inc. For GST pull-down assays, cell lysates were incubated with 10 μg GST fusion proteins immobilized on glutathione beads. For immunoprecipitations, cell lysates were incubated with 2–5 μg of antibodies and GammaBind Plus Sepharose beads (GE Healthcare) for 3 h at 4°C. Protein extracts were separated by SDS-PAGE and probed by immunoblotting. Detection of HRP-conjugated secondary antibodies (GE Healthcare) was performed with enhanced chemiluminescence detection systems (GE Healthcare or Pierce Chemical Co.).
All steps were performed at 4°C as described previously (). Freshly dissected adult mouse brains were homogenized in 10 vol of ice-cold homogenization buffer (buffer H [50 mM Tris-HCl, pH 7.4, 1 mM CaCl, 1 mM MgCl, and 1 mM NaHCO] containing 320 mM sucrose) using a motor-driven glass-fluorinated ethylene polymer homogenizer. The homogenate was centrifuged at 700 for 10 min to produce supernatant S1. S1 was further centrifuged at 17,000 for 15 min to produce pellet P2. P2 was resuspended in homogenization buffer, layered on top of sucrose gradient (equal volumes of 1.2, 1.0, 0.8, and 0.65 M sucrose in buffer H), and centrifuged at 100,000 for 2 h. A turbid fraction was recovered from the 1.0 /1.2 M sucrose interface and resuspended in at least 2 vol of homogenization buffer. The suspension was centrifuged for 30 min at 100,000 to produce the pellet P2′ (synaptosome fraction). P2′ was resuspended in Tris buffer (50 mM Tris-HCl, pH 7.4, and 1 mM EDTA with protease inhibitors), incubated for 45 min on ice, and centrifuged at 100,000 for 30 min. The supernatant containing soluble synaptic proteins was collected (synaptic cytosol fraction). The membrane pellet was resuspended in homogenization buffer, applied on top of a five-step discontinuous sucrose gradient (1.2, 1.0, 0.8, and 0.65 M sucrose), and centrifuged at 100,000 for 2 h. Pure synaptic membranes were recovered from the 1.2/1.0 M sucrose interface and diluted with at least 2 vol of homogenization buffer (synaptic membrane fraction). A pure membrane fraction was pelleted at 100,000 for 30 min, resuspended in homogenization buffer, and mixed with an equal volume of 3% Triton X-100 in homogenization buffer. Membranes were incubated for 30 min on ice and centrifuged at 100,000 for 2 h to produce a supernatant (soluble membrane fraction) and a Triton X-100–insoluble material pellet. Triton X-100–insoluble proteins were resuspended in homogenization buffer containing 1.5% Triton X-100 and applied on top of a 0.83-M sucrose cushion. After centrifugation at 105,000 for 1 h, PSDs were collected as the resulting pellet and resuspended in Tris buffer containing 0.25% Triton X-100 (PSD fraction). Equal amounts of protein from synaptosome, synaptic cytosol, synaptic membrane, soluble membrane, and PSD fractions were analyzed by immunoblotting.
Primary hippocampal neurons were prepared from embryonic day 17 rat embryos and cultured as described previously () with minor modifications. In brief, after trypsinization and mechanical dissociation, hippocampal cells were suspended in DME supplemented with 10% FCS and preattached on uncoated culture plates for 2 h to remove glial cells. Neurons recovered as nonadherent cells were plated on poly--lysine– (50 μg/ml) and laminin (5 μg/ml)-coated plates at a density of 3 × 10 cells/cm in neurobasal medium supplemented with B27, -glutamine, and penicillin/streptomycin. Neurons were cultured for 14 d in the presence of 2 μM AraC to prevent the growth of contaminating glial cells. The glial cells separated from the neurons by attachment on uncoated culture plates were cultured for 14 d in DME supplemented with 10% FCS, -glutamine, and penicillin/streptomycin.
Hippocampal slices (300 μm in thickness) were prepared and stimulated for 45 min as previously described using the middle one third of the hippocampus (; ). Slices were kept for 2–5 min in MEM (Invitrogen) supplemented with 25 or 5% horse serum (Invitrogen), 25% HBSS (Irvine Scientific), and 0.65% dextrose before stimulation. In some experiments, 9.5 μg/ml ephrin-A3 Fc (R&D Systems) was preclustered with 1 μg/ml anti-Fc antibodies, which produced similar EphA4 phosphorylation and Cas dephosphorylation as unclustered ephrin-A3 Fc. To promote β1-integrin activation, slices were pretreated with 20 μg/ml 9EG7 antibody for 25 min or 500 μM Mn for 20 min before stimulation with Fc proteins. Slices were either lysed and analyzed by immunoblotting or fixed in 4% PFA in 0.1 M PBS for 30 min, labeled with DiI crystals (Invitrogen), and analyzed by confocal z-series imaging as previously described (). EphA4 knockout mice were provided by A. Boyd and M. Dottori (Queensland Institute of Medical Research, Brisbane, Australia).
A gene gun (Helios; Bio-Rad Laboratories) was used to transfect plasmids into pyramidal neurons. Hippocampal slices prepared as described in the previous paragraph were cultured for 6 d on 0.4-μm Millicell-CM organotypic inserts (Millipore) in MEM containing 25% horse serum, 25% HBSS, and 0.65% dextrose. Slices were then transfected (80 pounds per square inch) with pSSRα/farnesylated enhanced GFP (EGFP-F; control), Cas/EGFP-F, CasΔSD/EGFP-F, CasΔSH3/EGFP-F, or CasΔSB/EGFP-F plasmids ((25 μg Cas plasmid mixed with 20 μg EGFP-F plasmid coated onto 25 mg of 1.6-μm gold particles; ). Cas plasmids were provided by K. Vuori and A. Motoyama (Burnham Institute for Medical Research, La Jolla, CA). After 48 h, the slices were fixed in 4% PFA in 0.1 M PBS for 30 min before z-series imaging of dendrites from transfected CA1 pyramidal neurons. For integrin inhibition, 170 μM RGDFV or RADFV (control) peptide (BIOMOL Research Laboratories, Inc.) was added to the media of EGFP-F–transfected cultured hippocampal slices. After an additional 24 h in culture, the slices were analyzed either by confocal z-series imaging or immunoblotting.
Four siRNAs for mouse Cas were obtained from Dharmacon, including siCas #1 (GCATAGGGCATGACATCTA; nucleotides 678–696; GenBank/EMBL/DDBJ accession no. ), siCas #2 (AGACGGAGCAGGATGAGTA; nucleotides 804–822), siCas #3 (CCGAGGAAGTCCAGGATTC, nucleotides 1,816–1,834), and siCas #4 (GATTCTGGTTGGCATGTA, nucleotides 292–310). Two scrambled siRNAs (nontargeting siRNA #1 and #2; Dharmacon) were used as controls. To obtain the pEGFP-hCas vector, human Cas cDNA (GenBank/EMBL/DDBJ accession no. ; Open Biosystems) was amplified by PCR with the forward primer 5′-GCTAGCCTCGAGCATGAACCACCTGAACGT-3′ and the reverse primer 5′-GCTAGCGAATTCTCAGGCGGCTGCCAGACA-3′. The PCR product was digested and inserted into the XhoI and EcoRI sites of pEGFP-C2 (BD Biosciences) and verified by sequencing.
Hippocampal slices were cotransfected with the siRNAs and the indicated plasmid using a gene gun as described previously (; ). 24 μg of duplex siRNA and 12 μg of plasmid were mixed and coated onto 10 mg of 1.6-μm gold particles. Transfected (green fluorescent) dendrites were analyzed 2 d after transfection by confocal z-series imaging. The effectiveness of siRNA transfection into hippocampal pyramidal neurons was verified by using a rhodamine-labeled siRNA transfected together with EGFP-F. The effectiveness of Cas knockdown by the siRNAs and the rescue of Cas expression by EGFP-hCas was verified in NIH3T3 cells. Cells were plated at 80% confluency, cultured overnight in DME and 10% bovine calf serum, and transfected in Opti-MEM with 78 nM siRNAs (with or without plasmids) using LipofectAMINE 2000 (Invitrogen). After 4 h, the medium was replaced by the DME culture medium, and Cas expression was assessed after 48 h by immunoblotting of cell lysates.
Images were collected using a confocal microscope (MRC 1024; Bio-Rad Laboratories) and analyzed using ImageJ 1.62 software (National Institutes of Health [NIH]), and the analysis of spines was confined to secondary and tertiary dendrites from the stratum radiatum. Quantification of spine lengths and widths was performed as described previously (). In brief, the length was measured from the tip of spine head to the interface with the dendritic shaft. The width was taken as a diameter of the spine head perpendicular to the length of the spine. Spines were assigned to morphological categories as described previously () by an investigator blind to the conditions. Mushroom spines are <2 μm in length, >0.5 μm in width, and are connected to the dendritic shaft by a narrower portion (neck). Stubby spines are <2 μm in length, >0.5 μm in width, and lack a defined neck. Thin spines are <2 μm in length, <0.5 μm in width, and have a neck. Filopodia spines are >2 μm in length, <0.5 μm in width, and do not have a distinct spine head. Irregular spines are branched spines (two heads and two necks merged into a single neck at the base) and multilobed spines (many heads fused together on top of a single neck). All measurements were obtained by counting 478–2,182 spines per condition (from 6–13 neurons) from independent experiments. Spine cumulative distributions were compared using the Kolmogorov-Smirnov test, and means were compared using the test or analysis of variance (ANOVA) test followed by Dunnett's, Tukey's, or Bonferroni's posthoc tests.
To obtain the EphA4-B35 cells, the rat B35 neuronal-like cell line () was transfected with the EphA4 cDNA in pcDNA3 using SuperFect (QIAGEN) and selected for 2 wk with 500 μg/ml G418. To perform the cell attachment assays, the HT22 hippocampal neuron cell line () or EphA4-B35 cells were starved overnight in DME with 1% BSA, detached, and kept in suspension for 30 min at 37°C in DME–1% BSA. Cells (7 × 10 cells per well of a 12-well plate) were then allowed to attach to poly--lysine– (20 μg/ml; Sigma-Aldrich) or fibronectin-coated coverslips (20 μg/ml fibronectin for EphA4-B35 and 10 μg/ml for HT22; Chemicon) for 10 min at 37°C in the presence of 10 μg/ml Fc or 10 μg/ml ephrin-A3 Fc. Attached cells were fixed in 4% PFA, stained with DAPI (Invitrogen), and counted blind (seven to nine 20× microscope fields were counted per condition). To determine Cas phosphorylation, both adherent and nonadherent cells were pooled, lysed, and probed by immunoblotting. HT22 and B35 cell lines were provided by W. Stallcup and B. Moosmann (Burnham Institute for Medical Research, La Jolla, CA), respectively.
HT22 was starved and plated on tissue culture plates coated with 10 μg/ml fibronectin for 1 h at 37°C in the presence of 10 μg/ml Fc, 10 μg/ml ephrin-A3 Fc, or 1 mM MnCl. To determine the levels of activated β1 integrin, live cells were rinsed with PBS and incubated with 2 μg/ml 9EG7 rat monoclonal antibody or control rat IgG for 15 min at 37°C. Cells were rinsed with PBS and lysed in SDS sample buffer, and bound antibodies were detected by immunoblotting with biotin-conjugated anti–rat IgG antibody (Vector Laboratories) followed by HRP-streptavidin (Zymed Laboratories). To determine the levels of cell surface β1 integrin, cells were washed with PBS, and surface proteins were labeled with 1 mM sulfosuccinimidyl-6-(biotin-amido)hexanoate (Pierce Chemical Co.) for 30 min at 4°C. Cells were washed with PBS and lysed in Hepes buffer with protease inhibitors, and biotin-labeled proteins were incubated with 50 μl streptavidin-conjugated beads (Biochemika) for 1 h at 4°C. Bound proteins were eluted and analyzed by immunoblotting with anti–β1-integrin antibodies.
Protocols for the animal studies were approved by the Burnham Institute's Animal Research Committee.
Fig. S1 summarizes the quantification and statistical analysis of the data from immunoblotting experiments. Fig. S2 confirms that ephrin-A3 Fc decreases Cas tyrosine phosphorylation in hippocampal slices from EphA4 mice but not in slices from EphA4 mice. Fig. S3 confirms that Cas knockdown decreases dendritic spine density and length in hippocampal slices by using a second siRNA control. Fig. S4 shows that the expression of human Cas abrogates the effects of mouse Cas siRNA on dendritic spine morphology. Fig. S5 demonstrates that integrin activation with Mn blocks ephrin-A3 Fc–induced spine changes. Online supplemental material is available at . |
As biologists, we are all aware that ionic and hydrogen bonds, plus van der Waals and hydrophobic forces, act within and between macromolecules to shape the final structure. However, a distinct interaction, known as the “depletion attraction,” may also play a substantial role (; ). This force is only seen in crowded environments like those found in cells, where 20–30% of the volume is occupied by soluble proteins and other macromolecules (; , ). Crowding increases effective concentrations, which has important consequences (Box 1), but it also creates a force apparently out of nothing. We argue that this force drives the assembly of many large structures in cells.
Consider , where many small and a few large spheres are contained in a box, representing the many small, crowding macromolecules and the fewer, larger complexes in a cell. In physicists' terminology, both types of sphere are “hard” and “noninteracting,” so that none of the forces familiar to biologists act between them. The small spheres bombard the large ones from all sides (arrows). When two large spheres approach one another, the small ones are excluded from the volume between the two. Therefore, the small ones exert an unopposed force equivalent to their osmotic pressure on opposite sides of the two large ones to keep them together. This osmotic effect depends on the volume that is inaccessible to the small spheres; if the small spheres could gain access to this (depleted) volume, they would force the two large ones apart. gives an alternative view. The centers of mass of the small spheres can access the yellow volume, but not the gray volumes, around each large sphere or abutting the wall. When one large sphere approaches another, these excluded volumes overlap; as a result, the small spheres can now access a greater volume. The resulting increase in entropy of the many small spheres generates a depletion attraction between the large spheres. At first glance, this seems like an oxymoron; entropy usually destroys the order that an attraction creates. But if we consider the whole system (not just the large spheres), the excluded volume is minimized and thus entropy is maximized (because there are so many small spheres).
The Asakura–Oosawa theory (“AO theory”; ), allows us to estimate the scale of this depletion attraction (Box 1). In cells, the diameters of the large spheres are the major determinants, as the other variables in the equation in Box 1 are constant; larger spheres tend to cluster more than smaller ones (, compare i with ii). The attraction can easily be recognized in vitro; adding an inert crowding agent like a dextran or polyethylene glycol (PEG) promotes aggregation (by increasing the volume fraction, , of the small spheres). However, the force has a maximum range of only ∼5 nm, which is the diameter of a typical crowding protein; it will be larger if the two large objects fit snugly together (or are “soft” enough to fuse into one with conservation of volume) and smaller if surface irregularities limit close contact ().
units; 1
is ∼0.7 kcal/mol, which is roughly comparable to the energy associated with one hydrogen bond in a protein ().
are within the range that biologists know can stabilize a structure.
sub
xref
#text
xref
italic
sub
#text
xref
italic
sub
#text
xref
#text
xref
italic
#text
We have argued that an osmotic depletion attraction drives the organization of many cellular structures. Unlike other noncovalent interactions (i.e., ionic and hydrogen bonds, van der Waals and hydrophobic forces), this one only becomes significant in crowded environments like those in cells. It is nonspecific in the sense that it can bring spheres together without orienting them. It also depends on size and shape; the larger the overlap volume, the larger the attraction. Just as the entropy of the solvent (water) mainly underlies the hydrophobic effect, that of the solute (the crowding macromolecules) creates the attraction. These generalizations come with caveats because the underlying physics is complicated, and AO theory involves several simplifications (e.g., it becomes less accurate when is >0.3, and it takes no account of kinetics). Nevertheless, the concept of a hydrophobic force is useful to biologists despite the underlying complexity, and we believe the concept discussed in this work will be similarly useful, especially because its scale can be calculated so simply.
Many questions remain. On the theoretical side, what happens when increases above 0.3, and the AO equation becomes less precise and the theory much more complicated ()? What are the relative advantages and disadvantages of the different theories of crowded solutions (Box 1)? On the experimental side, what exactly is the volume fraction within a cell, and how closely can typical proteins approach each other? Could the attraction help nucleosomes strung along DNA pack into the chromatin fiber (, ii). Can clumps of heterochromatin be treated as spheres that are subject to the attraction? If so, the attraction could underpin the condensation of an (interphase) string of such clumps into the mitotic chromosome (, ii; ). Could it also underpin the pairing of chromosomes seen during meiosis and polytenization, where a string bearing a unique array of factories and heterochromatic clumps aligns in perfect register with a homologue, but not with others carrying different arrays (, iii; )? Could it drive end-to-end pairing of chromosomes? For example, diploid human lymphocytes contain 10 chromosomes encoding nucleolar organizing regions (NORs), but only ∼6 NORs are transcribed, and only these aggregate to form nucleoli (). Does the attraction act through the thousands of active polymerizing complexes associated with each active NOR to drive nucleolar assembly (, iv)? Could it similarly drive the clustering of heterochromatic centromeres into chromocenters (, iv)? We have also seen how the attraction contributes to protein folding, but what of the special case where a protein is so confined that the overlap volume resulting from contact with the surrounding wall becomes significant ()? Do pores, and the barrels formed by chaperonins, proteasomes, and exosomes (), all exploit the attraction to promote ingress of their target proteins (; Cheung et al., 2005; see , for a review of how crowding affects protein folding in confined spaces)? Clearly, we need to extend the experimental studies on the simple model systems reviewed in this study to complex subcellular assemblies, much as describes.
As soon as cellular structures become larger than ∼75 nm, the overlap volume can generate an attraction of ∼5
; this is probably sufficient to promote irreversible aggregation when cooperative effects are included (, i, inset; ). This begs the obvious question: why don't all large structures in the cell end up in one aggregate (just as overexpressed bacterial proteins form inclusion bodies)? We suggest they will tend to do so unless energy is spent to stop aggregation and/or inert mechanisms prevent it. For example, anchorage to a larger structure (e.g., the cytoskeleton), surface irregularities (), or charge interactions could all prevent close contact, and thus reduce the attraction. All seem to operate; for example, >70% of proteins in and (and >90% of the most abundant ones) are anionic at cellular pH, and thus would be expected to repel each other (; ). We also note that structures like the cytoskeleton and membrane-bound vesicles are not rigid and permanent; rather, they continually turn over, to reduce their effective size and ensure that a large structure does not persist long enough to aggregate (; ). Nature, although constrained by the second law of thermodynamics, finds ways around it. |
Cell polarization studies have unveiled many of the molecular pathways by which cells can break symmetry in response to asymmetric stimuli. The stimuli can be either intracellular, like during cytokinesis, in which the mitotic spindle induces the position of the cleavage furrow (), or extracellular, such as chemical gradients during chemotaxis (). Interestingly, cells conserve the ability to polarize even in the absence of an asymmetric signal (). Such spontaneous polarization could be caused by a biochemical instability generated by the amplification of small stochastic variations in polarity protein concentrations (; ). In many cases, however, symmetry breaking and polarization seem to be driven by a mechanical instability of the actomyosin cytoskeleton.
The cell membrane is supported by a thin cortical layer between 100 nm and 1 μm thick that consists of cross-linked actin filaments, myosin motors, and actin-binding proteins, the spatial organization and dynamics of which are only beginning to be resolved (; ; ). The motors present in the cortex generate a contractile tension in the actin network () that can be relaxed if the cortex ruptures (). Local relaxation of the cortical tension can trigger polarization events such as global cortex flows (; ) or the growth of membrane protrusions called blebs (; ; ). Similarly, during early neuronal differentiation, breaking of the neuronal sphere and sprouting of neurites seem to require local relaxations of the cortical actin meshwork, although, in this case, the role of myosin motors is unclear (). Polarization induced by a release of mechanical tension is also observed in simpler systems, such as in actin networks growing on beads that mimic motility. In this paper, we compare the biochemistry and the mechanics of polarization in cells and around beads, and we argue that the bead system can serve as a simple model system to study mechanically driven polarization in cells. Furthermore, we argue that both actin gels around beads and the actomyosin cortex in cells are close to an instability threshold. Instability can be triggered by an intracellular or extracellular signal or can occur spontaneously when a fluctuation exceeds the mechanical threshold. Finally, we discuss the likelihood that polarization, by locally overcoming a mechanical threshold, could apply more generally to a variety of biological systems.
The mechanism by which cortical tension relaxes in cells is difficult to characterize because of the complexity of the cell. Mimicking the phenomenon under simplified conditions provides an alternative experimental way to study the mechanism of cortex breakage. An actin network that is mechanically comparable with the cell cortex is the actin gel that grows from the surface of a bead coated with an activator of actin polymerization (). Such beads have been used widely in the last 10 yr as a model system for studying actin-based movement of intracellular objects and lamellipodium extension (for review see ; ).
Beads (radii of 1–10 μm) are first covered with an activator of actin polymerization and are placed in cell extracts or in a mixture of purified proteins that reconstitutes the dynamics of actin-based movement observed for the bacterium (). Actin polymerization is activated at the surface of the bead, and an actin gel grows outward in spherical geometry. During gel growth, new monomers are incorporated at the bead surface underneath the preexisting gel, which is thus pushed outward and stretched as a result of the curved surface (). As a consequence, stresses build up, and the actin shell is under tension (). When this tension exceeds the maximum tension that the actin network can bear, the actin shell breaks, and the actin gel develops into a comet tail (; ; ). The gel rupture is most likely to take place in a region where the actin network is locally weaker. Interestingly, symmetry breaking does not necessarily occur at a single point in the actin gel; for large beads, under conditions in which gel growth is slow (e.g., at low gelsolin concentration), the gel may rupture at multiple locations, giving rise to several comets (; and unpublished data).
In some cases, the gel stops growing before the rupture threshold is reached. The stress in the gel is then below the critical value, and symmetry breaking is delayed. However, symmetry breaking may still occur if a local perturbation is induced in the gel or, for example, if a spontaneous fluctuation in the cross-linker density is large enough to bring the system over the threshold ().
Like the actin layer that grows around beads, the cell cortex is a cross-linked actin meshwork under tension. Indeed, myosin motors exert contractile forces on the actin network (). At a microscopic scale, the actin gels around the bead and the cell cortex appear to differ in several ways: the origin of the tension is different in the two systems, and the orientation of the actin filaments and the direction of network growth differ as well. However, at a mesoscopic scale, the two networks are very similar: both are cross-linked actin meshworks in which stresses develop tangentially to the actin layer (). Several aspects of the behavior of the gels growing around beads may therefore be reproduced in the cell cortex. Indeed, it has been proposed that just like the gel around beads, the cell cortex can rupture to relax the tension (). The relaxed region then expands as a result of pulling forces from the adjacent regions, which may lead to large cortical reorganizations and cell polarization (). By adding a bias with intracellular or extracellular cues, cells can use this cortical instability and the associated cortical flows in several ways.
For example, flows of the actomyosin cortex have been observed in various cell lines at the onset of cytokinesis, where they presumably contribute to cleavage furrow formation (; ). One mechanism that has been proposed to cause these cortical flows is a local relaxation of the cortex at the cell poles by astral microtubules (; ). However, this hypothesis remains controversial, as several experiments have shown that myosin can be recruited and activated in the equatorial zone even in the absence of cortex flows (; ; ). It is well possible that the cell uses several redundant mechanisms and that direct myosin recruitment mediated by the spindle midzone and aster-triggered cortex flows both contribute to furrow positioning (; ). Another process that is thought to depend on local cortex relaxation is the polarization of the one-cell embryo. Here, the sperm provides the external cue: after fertilization, the sperm centrosome moves toward the point of sperm entry, where it locally relaxes cortical contractility (). As during cytokinesis, the cortex flows away from the relaxed region, transporting polarity proteins and shaping the pseudocleavage furrow (; ). Polarization by cortex relaxation may also precede cell migration in some cells (; ).
In the aforementioned examples, cortex instabilities and polarization are triggered by a spatial cue that presumably relaxes the cortex locally. However, like the tension in actin gels grown around beads, the cortical tension can also relax spontaneously. For example, this is observed in cell blebbing (). Blebs are spherical bare membrane protrusions that are commonly observed during apoptosis (), cell division (), cell migration (; ), and spreading (). Bleb formation is driven by the pressure generated by contraction of the actomyosin cortex and occurs in regions where the actin cortex is weakened. Blebs are thought to be initiated by rupture of the actomyosin cortex (; ; ) or by detachment of the membrane from the cortex (; ; ). Interestingly, blebbing cells can form one single large bleb (; ; ) or multiple smaller blebs over the cell surface (; ; ).
Symmetry breaking in gels around beads and cell polarization caused by cortex breakage or relaxation are both driven by a release of mechanical tension in the actin gel. Spontaneous rupture of the actin network occurs when the tension in the gel exceeds a threshold value that is determined by the strength of the network. If the tension is just below the threshold, symmetry breaking may still occur if a spontaneous fluctuation in the density of actin, myosin, or cross-linkers, for example, is able to bring the system locally over the threshold. This implies that symmetry breaking can be enhanced either by lowering the threshold (the strength of the network) or by increasing the global tension (the driving force). Observations of symmetry breaking in both the bead system and the cell cortex support this idea.
In both systems, the instability threshold can be lowered by lowering the density of cross-linkers in the actin gel, like filamin or α-actinin, which leads to a softer and weaker network. Indeed, the depletion of filamin or degradation of α-actinin in cells enhances blebbing, probably as a result of cortical breakage, or at least a local release in the cortical tension (; ). Conversely, shell breakage in the bead system is slowed down by the presence of filamin or α-actinin (). In both systems, actin gel rupture is thus facilitated by the depletion of cross-linkers.
The driving force for cortex breakage in cells can be enhanced by increasing the activity of myosin II, leading to an increased contractility of the cortex and a larger cortical tension. Indeed, blebbing in cells is enhanced when the global contractility of the cortex is increased (), and, conversely, blebbing is reduced when contractility is decreased (). In the bead system, the tension is related to the thickness of the gel layer. Thus, the analogous effect of decreased contractility (leading to a lower tension) in the bead system is a decrease in gel thickness. For example, this can be achieved by adding actin-depolymerizing factor/cofilin, which enhances the depolymerization of filaments in the outer regions of the actin gel. Indeed, at high actin-depolymerizing factor/ cofilin concentrations, the gel thickness remains small, and no symmetry breaking is observed, indicating that the threshold tension for gel rupture can never be reached ().
A growing actin shell in spherical geometry can break spontaneously and form a propelling comet at the opposite side of the breakage point, although the original breakage and, thus, direction of the comet is random. If gel growth stops before the instability threshold is reached, symmetry breaking can still be triggered by an external perturbation (for example, by a local disruption of the actin network by photodamage; ). Likewise, a local alteration of the actin cortex in cells, either by locally applying drugs that affect actin or by increasing the local stress mechanically, induces cortex rupture and bleb formation ().
We can compare the forces necessary for shell breakage around beads and for cortex breakage in cells. The stresses in the gel around beads can be estimated from the elastic modulus of the actin gel and the thickness of the gel (). This produces a value of 10–10 Pa for the critical tensile stress for gel rupture (). The cell cortical tension has been estimated in different cell types and is on the order of 10 N/m for (; ), lymphocytes (), or fibroblasts (), whereas it is ∼20–30 times smaller for neutrophils (). With a cortical thickness of a few hundred nm, this provides a value of 10–10 Pa for the tensile stress in the cortex, which is very similar to the stress in the bead system. Interestingly, in , the deletion of either myosin II or of two myosins I leads to a decrease of the tension by ∼50%, suggesting that most of the cortex tension is caused by myosin motors (). Note that the cortical tension is very close to the threshold for cortex breakage, as breakage can be induced by applying pressures as small as 100 Pa, which is only 10% of the cortical stress ().
xref
#text
The concept of polarization driven by a global driving force that can locally exceed a mechanical threshold is not restricted to actin gels under tension but can be applied more generally. For example, in plant cells, fungi, or bacteria, the force that drives cell deformation and growth comes from the internal osmotic pressure, whereas the mechanical strength that resists deformation is provided by the cell wall. Because the pressure in the cell is homogeneous, the polarized growth of walled cells requires an inhomogeneous extensibility of the cell wall (). For example, root hairs and pollen tubes in plants and buds in budding yeast are all initiated as small bulges growing at the cell periphery in regions where the cell wall is locally softened (). To achieve such a local wall softening, a cell needs to direct vesicles that contain cell wall–loosening enzymes to specific sites at the cell periphery. This directed transport requires a polarized cytoskeleton, which may, in turn, be achieved by a biochemical instability (). Similarly, neuritogenesis starts by the growth of small buds at the initially spherical neuron surface. Buds are thought to result from pushing forces exerted by microtubules at spots where the actin network underlying the membrane is locally relaxed (). This relaxation could be tension driven because activation of the Rho–ROCK pathway, which activates myosin II, has been reported previously (). It could also result from some other kind of instability triggered by external signals ().
On a larger scale, a mechanical instability has been proposed to explain the shape and size of oscillations observed during the regeneration of fresh water polyp . At the initial stages, cells form a hollow sphere consisting of a cell bilayer. This sphere inflates by the uptake of fluid and builds up pressure as a result of stretching of the cells, which is analogous to the accumulation of stress in the actin gel growing around a bead. It has been proposed that this stress is released by rupture of the cell layer followed by rapid shrinkage of the cell ball (). Repeated cycles of growth followed by rupture and rapid shrinkage might be important for the first polarization step in morphogenesis.
We have argued that the mechanical states of the actomyosin cell cortex and of actin gels growing from beads are comparable. Both actin networks are under tension, which can be released by breaking symmetry in answer to a cue or spontaneously. Both systems appear to be operating close to a mechanical threshold, which would increase their sensitivity to small stimuli but would also make the system sensitive to fluctuations.
Highly reactive systems operating close to instability thresholds may be frequently found in biology. A similar, although not mechanical, threshold mechanism is observed in budding yeast, for example, where Cdc42, a small GTPase, is required for bud formation. The expression of a constitutively active Cdc42 results in spontaneous polarization with random orientation (; ). It is possible that the enhanced activity of Cdc42 brings the system closer to a chemical threshold, where it becomes sensitive to random fluctuations ().
Comparing the forces necessary for rupture and the effects of various proteins on symmetry breaking suggests that the mechanisms of polarization of the cell cortex and of the rupture of gels growing around beads are very similar. As a consequence, understanding symmetry breaking in biomimetic systems may provide essential insight into spontaneous cortex rupture in cells. There are many open questions as to how exactly polarizing signals trigger the mechanical instability leading to cortex rupture. The centrosome–microtubule system plays an essential role here, but, to a large extent, the pathways by which it controls the cortex mechanics are still unknown. |
Mutations in the tumor suppressor breast cancer–associated protein 1 (BRCA1) are associated with a high risk of breast and ovarian cancer. BRCA1 is a nuclear protein implicated in multiple processes, including genomic stability, transcription regulation, chromatin remodeling, and cell-cycle control (; for reviews see ; ). In normal S-phase cells, BRCA1 shows a punctate distribution with typically ∼10–20 prominent accumulations (foci), but upon induced DNA damage, it relocalizes to sites of DNA repair (; ; ). Although many studies have investigated BRCA1 foci in relation to DNA repair, little is known about the BRCA1 foci in nonirradiated cells. These have been suggested to be storage sites or, possibly, sites of endogenous damage. They are not thought to be sites of DNA replication because they distribute in a pattern distinct from that of replicating DNA (). However, it remains an important consideration that normal S-phase BRCA1 foci may reflect an unrecognized, but fundamental, function of BRCA1.
A key to understanding whether the BRCA1 foci in nonirradiated cells have biological significance is whether they form at specific genomic loci. The spatial association of BRCA1 at sites of DNA damage provided key evidence for its role in DNA repair. We investigate whether BRCA1 foci in normal cells form at specific nuclear or chromosomal sites, or distribute more randomly, as might be expected for storage sites or endogenous damage. BRCA1 localizes to the unpaired X and Y chromosomes in spermatocytes, implicating BRCA1 in recombination and meiotic silencing (; ). However, in normal somatic nuclei there is no evidence that BRCA1 spots associate with specific sites of chromatin, other than a reported association of BRCA1 with XIST RNA on the inactive X chromosome (Xi; ). Findings in this study demonstrate that BRCA1 foci form at particular classes of heterochromatin, linked to their replication, and suggest a novel role of BRCA1 with implications in the maintenance of genomic stability.
The report that BRCA1 colocalizes with XIST RNA on the inactive X chromosome (Xi) in a subset (5–10%) of cells () led us to further investigate the spatial relationship between XIST RNA and BRCA1. In extensive investigation of multiple cell lines, using several BRCA1 antibodies (see Materials and methods), we did not find that BRCA1 substantially overlaps XIST RNA on Xi (; this study). However, using methods optimized for simultaneous detection of nuclear RNA and protein (see Materials and methods), BRCA1 partially overlapped or closely abutted XIST RNA in 3–5% of hundreds of cells viewed in 2D. 3D analysis of deconvolved optical sections () shows that even in cases where BRCA1 and XIST RNA appear to overlap in 2D, they largely occupy distinct spatial territories, typically with BRCA1 tightly abutting the XIST signal (Video 1, available at ). BRCA1 also did not colocalize substantially with other hallmarks of Xi-facultative heterochromatin (H3mK27 or ubiquitin; Fig. S1 A), which colocalize throughout the XIST RNA territory (; ). We also recorded a fraction of cells (∼13% in TIG1 fibroblasts, with similar results for multiple cell lines) in which a BRCA1 spot was directly adjacent to, but clearly not overlapping (even by 2D analysis), XIST RNA ( and not depicted; see Materials and methods for definition of scoring terms). The significance of these more limited associations is investigated in this study. However, overall, these findings are consistent with other evidence that BRCA1 does not have a direct role in localizing XIST RNA (); if BRCA1 has a spatial relationship to the Xi, it is not via an association with XIST RNA.
To address whether the ∼10–20 prominent BRCA1 foci associate with a particular category of chromatin, we investigated whether they preferentially localize to the euchromatic or heterochromatic compartments. To delineate these compartments, we used hybridization to heterogeneous nuclear RNA (hnRNA) and labeling of splicing-factor–rich domains. Hybridization to hnRNA with a Cot-1 DNA probe delineates the inactive X chromosome () and heterochromatin abutting the nuclear envelope and nucleolus (). Analysis, in two different cell lines, revealed a strong propensity for BRCA1 foci to localize in hnRNA-depleted regions; only ∼19% overlapped the Cot-1 RNA signal, which fills most of the nucleoplasm (). A surprisingly large fraction of BRCA1 foci (∼32%) localized to the Cot-1–depleted region abutting or within the nucleolus (). Another 14% localized to the peripheral heterochromatin, and 35% precisely colocalized with small discrete “holes” in the hnRNA signal (). Although not our focus in this study, an association with the centrosome () was not noted with cells and antibodies used here. The association of BRCA1 with the nucleolus is interesting because many centromeres localize there.
The preference for heterochromatic regions contrasted to the paucity of BRCA1 foci with SC-35 and SRM300, which are splicing components that label 20–30 large domains linked to RNA metabolism. These regions are surrounded by active genes in the euchromatic compartment (). These BRCA1 foci only rarely overlap (<1%) or contact (3%) SC-35 () or SRM300 in mouse cells, suggesting they are largely excluded from these euchromatic “neighborhoods.”
BRCA1 in normal S phase has not been thought to reflect routine DNA replication because BRCA1 distribution does not mirror that of replicating DNA (). We reexamined the relationship of BRCA1 foci to mid-to-late replicating DNA, which comprises largely heterochromatic DNA. Unlike the dispersed particulate pattern of early replication, the mid-to-late pattern comprises a smaller number of larger spots (; ). Examination of whether BRCA1 foci overlapped BrdU spots confirmed the earlier conclusion that, in general, the two patterns are not the same (). However, close scrutiny suggested a substantial, but incomplete, relationship. Approximately 3% of the discrete BRCA1 foci overlapped a BrdU spot, but an additional 18% were abutting or adjacent to (contacting) BrdU spots. Although these mid-to-late S-phase BrdU spots occupy a much smaller area of the nucleus than SC-35 domains (), BRCA1 shows greater spatial association with them. Many BRCA1 spots (an additional ∼27%) seemed to position consistently close (∼0.3 μm) to a BrdU spot, the significance of which was initially unclear.
We also reexamined the relationship to the replicating Xi. As shown in , the subset of cells that showed BRCA1 association (either abutting or adjacent; ) during replication of Xi increased two- to threefold over asynchronous cultures, which is consistent with an increased association in late S phase ().
The aforementioned findings led us to investigate whether BRCA1 has a relationship to heterochromatin associated with centromeres. Using an antibody to centromere protein C (CENP-C), which is a constitutive component of the interphase centromere–kinetochore complex, the patterns of CENP-C and BRCA1 in human fibroblasts (TIG1) were again distinct, yet exhibited a substantial spatial association (). We categorized these associations into three types, as follows: 3% of BRCA1 spots were completely coincident with CENP-C spots, another 12% directly abutted or contacted (no separation visible by light microscopy), and an additional 24% were suggestive of a close pairing. Very similar observations (3% coincident, 14% abutting/adjacent, and 16% close) were made when we hybridized to α-satellite DNA () or used CENP-B, which binds α-satellite, as a marker ().
We next asked whether BRCA1 foci that abut Xi reflect a relationship to centromeres. The frequency with which we found BRCA1 partially overlaps (∼2%) or resides adjacent to (8%) the X centromere may largely account for BRCA1–XIST RNA association (3% partial overlap and 13% adjacent). Using the Barr body to distinguish the active and inactive X (), there was not a major difference in BRCA1 association with Xi versus active X chromosomes (Xa) centromeres (10% vs. 7%, respectively). Thus, the relationship of BRCA1 to Xi primarily reflects a relationship to centromere-associated constitutive heterochromatin, rather than specifically Xi-facultative heterochromatin. However, we do not exclude the possibility that the slightly higher association with Xi is caused by its more heterochromatic nature.
The frequency with which BRCA1 signals either overlap or directly abut interphase centromere markers indicates a substantial, albeit incomplete, association. Although the “close, but not contacting” category is less clear, this could reflect a spatial linkage to some component of the centromere–kinetochore complex (for review see ), which has many components that do not all completely coincide in nuclei () and are not all known. Therefore, the relatively consistent gap between BRCA1 and CENP-C could contain some other component of this structure (see the following section).
Because BRCA1 most often “neighbors” (rather than overlaps) these centromere components, we next investigated BRCA1's relationship to pericentric heterochromatin (PCH), which would also lie adjacent to centromeric DNA. Mouse cells have a more well-defined organization of centric and pericentric DNA than do human cells (); in mouse cells, centromeres cluster into 5–10 chromocenters that are easily visualized with DAPI stain. Fig. S1 B confirms a recent report () demonstrating that the DAPI-dense chromocenters comprise pericentric heterochromatin (mouse major satellite) and the centromeric DNA (minor satellite) is smaller and positions at the periphery of the larger blocks of PCH.
BRCA1 staining revealed a clear structural relationship with chromocenters (). Although not all chromocenters have associated BRCA1, and vice versa, in all of the several different lines examined (mouse 3T3, 3X mouse, mouse embryo fibroblasts (MEFs), and mouse ES cells), 26–38% of BRCA1 spots in an asynchronous population directly associated with a chromocenter. In addition, a subpopulation of cells showed higher association; in some cells, almost all BRCA1 spots were with a chromocenter (). Typically one or two BRCA1 foci were at the immediate periphery of each chromocenter, but, not infrequently, several foci or elongated BRCA1 accumulations “hugged” the contour of the chromocenter (). Occasionally, a “paint” of nearly all the DAPI-bright PCH was apparent (, top middle). Interestingly, this association is present even in very early (1-d differentiated) embryonic stem cells (, bottom right). This is potentially important because BRCA1 knockout is early embryonic lethal ().
We next examined the relationship between BRCA1 and the minor satellite (equivalent of human α satellite) of the centromere proper. When visualized together (), their relationship mirrored that seen (see previous section) between human centromeres and BRCA1, as follows: 6% coincident, 10% adjacent/abutting, and 27% close. However, when viewed with DAPI in three colors, it became apparent that, often, minor satellite and BRCA1 signals that had no direct contact were in fact associated with a common chromocenter. These observations bolster the significance of close/“paired” signals in the human; even when the BRCA1 foci are not coincident with a centromeric marker, they are spatially linked by their common association with the PCH of the chromocenter ( and Video 2, available at ). The link between BRCA1 and centric DNA may be through the PCH, but, in either case, results indicate a connection between the discrete BRCA1 S-phase foci and centromeres, which are structures key to the proper segregation of chromosomes and maintenance of genomic integrity.
In both human and mouse, some cells show greater BRCA1 association with chromocenters than others, as illustrated in (middle). We addressed whether this difference might relate to replication, using proliferating cell nuclear antigen (PCNA) as a marker of the replication machinery (). Chromocenters replicate roughly synchronously in a given cell in mid-to-late S phase (; ; ). In cells in which most chromocenters had prominent PCNA label, a higher association of BRCA1 was clearly evident. Of chromocenters that label with PCNA, 55% have BRCA1 associated (15% overlap and 40% abutting), in contrast to <7% with no PCNA label (<1% overlap and 6% abutting; ). Cells with the most striking BRCA1 painting of chromocenters also labeled for replication of the chromocenter. This demonstrates a temporal relationship between widespread, largely synchronous BRCA1 association and replication of PCH.
Previous work has shown that mouse chromocenters have a defined architecture such that DNA replication (and likely chromatin assembly) occurs at the periphery of the large major satellite block, and the newly replicated DNA then moves into the central region of the chromocenter (). This fits well with the distribution of BRCA1, which is mostly concentrated at the chromocenter periphery. Because BRCA1 did not always localize to PCNA-labeled chromocenters, it may transiently associate close to the time of replication. The fact that BRCA1 is more juxtaposed to PCNA than overlapping it is consistent with other evidence that BRCA1 may have a post-replicative role. Similar observations were made with a 15-min terminal pulse of BrdU (Fig. S1 C).
Because one recent study reports that BRCA1 regulates topoisomerase IIα (topoIIα) during routine DNA replication (), we briefly addressed whether topoIIα associates with mouse chromocenters. Although topoIIα is enriched at mitotic centromeres, and there is one report of its association with late S-phase BrdU (), it is not known to be enriched at chromocenters/centromeres during S phase. As shown in , we found topoIIα commonly enriched on mouse chromocenters; in ∼40% of interphase cells, topoIIα concentrates on essentially all chromocenters. Many of these cells are in S phase, with PCNA on their chromocenters ().
Our findings suggest that BRCA1 may have a role in replication-linked maintenance of peri/centromeric heterochromatin. As the study of X inactivation has demonstrated, the epigenetic state of heterochromatin is controlled at numerous levels that work synergistically and provide redundancy; for example, heterochromatic features of the Xi are compromised only very slightly over the long term if XIST RNA is lost from somatic cells (). Similarly, reintroduction of XIST RNA in somatic cells would not simply correct a deficit in heterochromatin (; ). Thus, short-term loss or gain of BRCA1 could have no immediate effect on heterochromatin but still be important for its long-term maintenance and stability in an organism. We found that short-term acute loss of BRCA1 by RNAi in HeLa cells impacts proliferation and reduces mitotic figures by >60% (Fig. S2, available at ), which is consistent with other reports. Although this could reflect an impact on the complex epigenetic state or repair of pericentric heterochromatin, it could also reflect other short-term effects of BRCA1 loss on centrosome function (), DNA decatenation (), or cell-cycle checkpoints (; ).
Breast tumor cells such as HCC1937 are exposed to longer-term BRCA1 loss. As an initial effort to investigate some properties of centric heterochromatin, we examined CENP-A, a constitutive interphase kinetochore component directly linked to specifying a centromeric property, and CENP-F, the first transient kinetochore protein bound in G2 (). Localization of these centromeric components appeared normal in these BRCA1 mutant cells (). However, given the essential role of CENP-A in kinetochore assembly, this may not be surprising.
reported that a high fraction (∼10%) of HCC1937 cells had lagging chromosomes or DNA bridges after mitosis. We attempted to confirm these findings, but extended our analysis to MCF7 (BRCA1+) breast cancer cells and normal diploid fibroblasts. It was obvious in DAPI-stained slides of HCC1937 that many early G1 daughter pairs contain a “bridge” of DNA extending between them (); in contrast, this was almost never seen in BRCA1+ MCF7 or in normal fibroblasts. Many mitotic figures showed lagging chromosomes, and early G1 pairs showed thin bridges of DNA. For example, in 100 G1 daughter pairs, visible DNA bridges were seen 31 times in HCC1937 cells, in contrast to 3 times in diploid fibroblasts (TIG1) and 4 times in MCF7 cells. Most mitotic figures with lagging chromosomes showed a normal bipolar configuration; thus, in most cases, a multipolar spindle (i.e., centrosome) defect was not apparent. Although we did not observe an appreciable difference in mitotic defects in a BRCA1-reconstituted HCC1937 cell line (), as noted earlier, once any defects in PCH or aneuploid cells are generated, reversion to normal mitotic figures would be difficult. Finally, we addressed whether the thin bridges connecting G1 daughters contained satellite DNA. Although many DNA bridges were just thin threads, a large fraction (35/40) contained α-satellite DNA (). These results are consistent with the possibility that a defect in centric/pericentric heterochromatin is present.
Although BRCA1 nuclear distribution has been studied for some time, this is the first study to identify a preferential relationship with centric and pericentric heterochromatin, and link this temporally to replication of these structures (summarized in ). This may have escaped earlier detection because BRCA1 distribution does not simply mirror that of replicating DNA, but we show there is indeed a meaningful relationship suggesting a novel biological role for BRCA1. Because most (∼80%) of the bright BRCA1 foci localize to hnRNA-depleted sites, those not with peri/centric DNA may be mostly with some other heterochromatin (e.g., telomeres, etc). The widespread, largely synchronous localization of BRCA1 foci to mouse chromocenters suggests a link to routine replication rather than just repair, but the highly repetitive (or condensed) nature of this DNA poses specific requirements that may involve repair-related or other BRCA1 functions (e.g., chromatin remodeling/assembly, transcriptional regulation, or topoII-mediated roles).
This work points to a new direction of BRCA1 research involving routine replication and maintenance of peri/centric heterochromatin. Although a specific role of BRCA1 requires further investigation, any involvement of BRCA1 in maintaining centric/pericentric heterochromatin would have profound significance for understanding how BRCA1 mutations contribute to genomic instability and cancer. This study also clarifies an earlier report that BRCA1 appeared to have an extensive and specific relationship to XIST RNA and Xi facultative heterochromatin (). Recent studies from the Livingston laboratory, and other laboratories, indicate that mitotic loss of the Xi and gain of an Xa is the most common means whereby X chromosome dosage is increased in certain types of breast cancers (; ). Therefore, the demonstration in this study that BRCA1 associates with constitutive more than facultative heterochromatin fits well with recent evidence that the most prevalent mechanism of Xi loss in BRCA−/− cancer may involve increased errors in chromosome segregation. Additionally, loss of BRCA1 may contribute to a generalized failure of heterochromatin maintenance.
Human diploid fibroblast lines WI38 (CCL-75) and IMR-90 (CCL-186) were obtained from the American Type Culture Collection, and TIG-1 (AG06173) were obtained from Coriell Cell Repositories. In addition, 3X mouse cells () and MCF7 cells were used. WI38, TIG-1, MCF7, and 3X mouse cells were grown in MEM supplemented with 10% FBS, 2 mM -glutamine, 1% penicillin/streptomycin, NIH 3T3 cells, and MEFs and were grown in DME supplemented with 10% FBS, 2 mM -glutamine, and 1% penicillin/streptomycin. HCC1937 cells were obtained from American Type Culture Collection (CRL-2336) and were grown in RPMI with Hepes medium supplemented with 10% FBS, 2 mM -glutamine, and 1% penicillin/streptomycin. For BrdU labeling, cells were plated for 48 h, followed by a 15-min treatment with 30 μM BrdU just before fixation. HCC1937 cells reconstituted with BRCA1 were obtained from J. Chen (Mayo Clinic, Rochester, MN; ).
Monoclonal antibodies to mouse (GH118) and human (MS110) BRCA1 were a gift from D. Livingston and S. Ganesan (The Dana Farber Cancer Institute, Boston, MA; ). A monoclonal antibody to polyubiquitinated proteins (UbFk2) was obtained from Affiniti BioReagents. Polyclonal antibodies to BRCA1 (KAPST0201) were obtained from Assay Designs and C. Deng (National Institutes of Health, Bethesda, MD). Antibodies to BrdU (rat monoclonal) were obtained from Harlan. Polyclonal antibodies to CENP-C (rabbit) were obtained from W. Earnshaw (University of Edinburgh, Edinburgh, UK), and CENP-B antibodies (rabbit; H-65) were obtained from Santa Cruz Biotechnology, Inc. An antibody to CENP-A was obtained from M. Valdivia (Universidad de Cadiz, Cadiz, Spain), and an antibody to CENP-F was obtained from D. Cleveland (University of California, San Diego, La Jolla, CA). Antibodies to PCNA were obtained from Immunovision (Human). An antibody to topoIIα was obtained from Lab Vision (rabbit).
Initial studies involved testing two fixation methods; paraformaldehyde fixation followed by Triton X-100 extraction, as described by , or brief Triton X-100 extraction before fixation, as described previously (; ). For extraction, cells were extracted in CSK buffer with 0.5% Triton X-100 and 2 mM vanadyl adenosine for 5 min. Cells were then fixed in 4% paraformaldehyde in 1×PBS for 10 min, incubated with primary antibodies in 1×PBS/1% BSA for 1 h at 37°C, and rinsed successively in 1×PBS, 1×PBS + 0.1% Triton X-100, and 1×PBS for 10 min. Detection was carried out with secondary antibodies tagged with fluorescein, rhodamine, or Texas red (Jackson ImmunoResearch Laboratories).
RNA FISH used previously established protocols (; for review see ). XIST RNA was detected with a 10-kb plasmid (pG1A) spanning intron 4 to the 3′ end of , or with plasmid (pXISTHb-B) containing intron 1 (). Probes were nick translated using biotin-11-dUTP or digoxigenin-6-dUTP (Boehringer Mannheim). Hybridization was detected with either antidigoxigenin antibody (Boehringer Mannheim) coupled with rhodamine or fluorescein or, for biotin detection, avidin conjugated to Alexa Fluor–streptavidin 594 (red) or fluorescein (Boehringer Mannheim). RNAsin was added for simultaneous RNA FISH and antibody staining. After detection and washing, cells were re-fixed in 4% paraformaldehyde in PBS for 10 min at 25°C and processed for RNA FISH as described in this section.
Digital imaging analysis was performed using an Axiovert 200 or an Axiophot microscope (Carl Zeiss MicroImaging, Inc.) equipped with a 100× PlanApo objective (NA 1.4; Carl Zeiss MicroImaging, Inc.) and 83,000 multibandpass dichroic and emission filter sets (Chroma Technology Corp.) set up in a wheel to prevent optical shift. Images were captured with a camera (Orca-ER; Hamamatsu) or a cooled charge-coupled device camera (200 series; Photometrics). Where rhodamine was used for detection in red, a narrow band-pass fluorescein filter was inserted to correct for any bleed-through of rhodamine fluorescence into the fluorescein channel. Optical sections and 3D images were created using Axiovision 4.4 (Carl Zeiss MicroImaging, Inc.). Images were captured at 0.1-μm intervals, and stacks were deconvolved with a constrained iterative algorithm. Rendered images are maximum value projections.
For scoring purposes, the following definitions are used for scoring terms: association indicates any relationship that appears to involve “contact” (no physical separation visible by 2D light microscopy). These were further divided into three categories of association. Painting, which typically indicates almost complete overlap of two signals, but in this analysis any overlap >50% would have been included. Abutting/partial overlap, which indicates a signal which very closely pressed against another, such that as viewed in two dimensions there appears a slight overlap of the two signals. 3D analysis may show the two signals are actually not overlapping. Adjacent, which indicates two signals that are juxtaposed and appear in contact, but for which even 2D analysis indicates no overlap. “Closely paired” signals are distinct in that this category indicates two signals that do not contact, but are separated by ∼0.2–0.4 μm.
Fig. S1 shows hallmarks of X inactivation on the Xi, localization of major and minor satellite DNA relative to chromocenters, and localization of BRCA1 and BrdU to mouse chromocenters. Fig. S2 shows decreased proliferation in BRCA1 siRNA-treated versus control siRNA-treated cells and contains RNAi methods. Video 1 is a 3D movie of BRCA1 and XIST. Video 2 is a 3D movie of BRCA1 and a mouse chromocenter. Online supplemental material is available at . |
DNA damage blocks the progression of replicative DNA polymerases and causes stalled replication forks at S phase. Persistent stalled replication forks collapse and cause genomic instability or cell death (). In , stalled replication forks are resolved either by bypassing DNA damage with translesion synthesis (TLS) polymerases or by template switching to the nascent strand of sister chromatid (; ). The selection of these pathways is regulated through the modification of proliferating cell nuclear antigen (PCNA), a homotrimeric complex that encircles DNA strands and functions as a loading dock for DNA polymerases as well as various DNA repair machinery.
PCNA is monoubiquitinated at the lysine 164 (K164) by the ubiquitin ligase (E3) yRad18 with the ubiquitin-conjugating enzyme (E2) yRad6/Ubc2 after the cells are treated with a low concentration (0.02%) of the alkylating agent methyl methanesulfonate (MMS; ). Monoubiquitinated PCNA switches the replicative DNA polymerase δ to nonessential polymerases specialized for TLS (; ). After the treatment of yeast cells with MMS, the same monoubiquitinated lysine residue of PCNA is further modified with a noncanonical lysine 63 (K63)–linked polyubiquitin chain by yRad5 (E3) along with the yUbc13–yMms2 (E2 and E2 variant) heterodimer complex (). This modification of PCNA presumably promotes the error-free mode of bypass, which is thought to use a template-switch type of recombination through reversal of stalled forks (; ). However, nothing is known about molecular mechanisms downstream from PCNA polyubiquitination.
Mammalian PCNA also undergoes monoubiquitination after a low dose of MMS (0.02%) and UV irradiation, and monoubiquitinated PCNA preferentially binds to TLS polymerases (; ). So far, no evident PCNA polyubiquitination has been observed in mammals. Furthermore, even though homologues of yRad18 (RAD18), yRad6 (HHR6A and HHR6B), yUbc13 (UBC13), yMms2 (MMS2/ UEV2), and downstream TLS polymerases have been identified (; ; ; ; ), no apparent homologues have been discovered. Therefore, it has been controversial whether the mammalian error-free mode of bypass exists, and PCNA regulation through differential ubiquitinations is a conserved and fundamental mechanism in mammals.
is a putative tumor suppressor gene encoding a large protein of 1,683 amino acids with various predicted functional domains, including SWI2/SNF2 and RING domains (; ). The human gene was mapped to chromosome 6q24, which has been suggested to contain tumor suppressor genes. Four point mutations in the gene were identified in melanoma and ovarian cancer–derived cell lines, although roles of SHPRH in cancer development have been largely unexplored ().
In an effort to investigate whether mammals have a yeast Rad5–like pathway that prevents genomic instability, we identified as a human orthologue. We demonstrate that SHPRH has conserved biochemical properties with yeast Rad5 and suppresses genomic instability by promoting K63-linked polyubiquitination of PCNA.
Yeast Rad5 is a member of the SWI2/SNF2 family of helicases with the E3 activity (). A unique structural feature of yRad5 is that the RING domain for its E3 activity is embedded between the conserved motifs IV and V of the SWI2/SNF2 domain (). To identify human proteins with this unique domain structure, we performed a SMART search () and identified a putative tumor suppressor gene, . SHPRH and yRad5 show 45.5 and 36.4% identities and 62.1 and 47.3% similarities in the SWI2/SNF2 and RING domains, respectively, but have little homology in other sequences ( and Fig. S1, available at ). SHPRH also contains predicted linker-histone and PHD-finger domains, which are not found in yRad5. Just like mammalian , which shares even higher sequence similarity with (), expression could not rescue the UV sensitivity of the strain (Fig. S2).
To test whether SHPRH is a functional orthologue of yRad5, we examined whether SHPRH could polyubiquitinate PCNA. The overexpression of HA-tagged ubiquitin (HA-Ub), FLAG-tagged PCNA (FLAG-PCNA), and myc-His–tagged SHPRH (SHPRH-myc-His) in human embryonic kidney (HEK) 293T cells induced mono- and polyubiquitinations of PCNA (, lane 5). PCNA polyubiquitination by SHPRH was further enhanced by UBC13–MMS2 (, lane 7) but not by RAD6 (lane 6) or UBC13 with the C87A mutation, which inactivates the E2 activity of UBC13 (lane 8). In contrast, RAD18 exclusively induced PCNA monoubiquitination with RAD6 (Fig. S3 A, lanes 6 and 7, available at ). Enhancement of PCNA monoubiquitination by SHPRH may be caused by competition between SHPRH and the deubiquitin enzyme USP1, which removes monoubiquitin from PCNA (). To clarify the role of the RING domain in SHPRH, we created two proteins: one with a mutation at the conserved cysteine 1432 in the RING domain (C1432A) and the other with a deletion of the entire RING domain at the C terminus (ΔRING). Both mutants showed reduced levels of PCNA polyubiquitinations (, lane 4; and Fig. S3 B), suggesting a role of the RING domain of SHPRH. The remaining levels of PCNA ubiquitinations observed in these mutants imply that SHPRH mutants may be able to “recruit” or “nucleate” endogenous SHPRH or other E3 ligases. Supporting the specificity of overexpressed SHPRH to PCNA, SHPRH expression did not affect the basal polyubiquitination level of His-c-JUN, a nuclear protein modified with a lysine 48–linked polyubiquitin chain by several ubiquitin ligases, such as SCF (; Fig. S3 C).
To determine whether SHPRH polyubiquitinates the conserved K164 of PCNA, which is also targeted by yRad5, PCNA (either wild type or K164R mutant) was immunoprecipitated and the level of mono- and polyubiquitinations was analyzed. PCNA with the K164R mutation was defective in mono- and polyubiquitinations (). Furthermore, the SHPRH-promoted polyubiquitination of PCNA was reduced by the overexpression of the ubiquitin(K63R) mutant but was enhanced by the ubiquitin(K48R) mutant (). We therefore concluded that SHPRH functions with the UBC13–MMS2 complex to modify the K164 of PCNA with a noncanonical K63-linked polyubiquitin chain in vivo.
Previous yeast studies suggested that the monoubiquitination of PCNA by yRad18 precedes polyubiquitination by yRad5 (). It is consistent with genetic observations that y is epistatic to y after exposure to various DNA-damaging agents (; ; ). Notably, stable knockdown of RAD18 by short hairpin RNA (shRNA) substantially reduced SHPRH-mediated PCNA polyubiquitination (). Furthermore, coexpression of RAD18 and RAD6 with SHPRH and UBC13–MMS2 synergistically promoted the polyubiquitination of endogenous PCNA (). These results clearly indicate that the PCNA monoubiquitination by RAD18–RAD6 and polyubiquitination by SHPRH–UBC13–MMS2 are sequential, rather than competitive, events.
Yeast Rad5 interacts with PCNA, yUbc13, and yRad18 and self-multimerizes (). We observed that SHPRH physically interacted with GST-fused PCNA (wild type or K164R mutant) and GST-UBC13, but not with GST-MMS2, GST-RAD6, or GST alone (). The interactions between the PCNA(K164R) mutant and SHPRH or RAD18 suggest that the ubiquitination of PCNA is not essential for these interactions. SHPRH interacted with UBC13 and weakly with UBC13(C87A) in vivo (). Although structural studies predicted that the interaction site on UBC13 with RING domains is distinct from the cysteine C87 (), C87 may affect this interaction in vivo. In addition, we observed an increased level of SHPRH protein in cells coexpressing wild-type UBC13 and, to a lesser extent, UBC13(C87A), suggesting the stabilization of SHRPH through the complex formation with UBC13. We also identified the self-association of SHPRH and the interaction between SHPRH and RAD18 by coimmunoprecipitation experiments (). These observations demonstrate that SHPRH has interaction features similar to yRad5.
To investigate the biological significance of PCNA polyubiquitination in mammalian cells, we transfected HEK 293T cells with low amounts of plasmids that express HA-ubiquitin and FLAG-PCNA, with or without SHPRH-myc-His, and treated with various DNA-damaging agents. No apparent changes in PCNA polyubiquitination were detected after treating cells with 30 J/m UV or 0.3 mM of the DNA cross-linking agent mitomycin C (MMC; ). In contrast, PCNA polyubiquitination was induced after treating cells with 0.01% MMS (, lane 7). MMS-induced PCNA polyubiquitination was strongly enhanced by the additional expression of SHPRH. We also noticed that the signals of SHPRH in the cell extracts were reduced only when cells were treated with MMS (, lane 8), implying that SHPRH was redistributed to the insoluble (chromatin bound) fraction. MMS treatment of HCT116 human colon carcinoma cells (without any transfections) also showed endogenous PCNA polyubiquitination in a dose-dependent manner (). To examine the cell cycle specificity of PCNA polyubiquitination, we treated cells arrested in G1 phase or 4 h after release from the G1 arrest with MMS. We observed a somewhat substantial level of PCNA polyubiquitination in untreated cells in G1 phase, but not in S phase (, lanes 1 and 3). Importantly, PCNA polyubiquitination was most efficiently induced in mid–S phase (, lane 4).
Mutations in genes in the yRad5 pathway cause genomic instability and increased cell sensitivity to various DNA-damaging agents (; ; ; ). To determine whether reduced expression of could cause similar cellular phenotypes, we transduced HCT116 cells with lentiviral vectors that express two different -interfering shRNAs (constructs B or C; ). silenced cells (by construct C) showed a substantial reduction in MMS-induced PCNA polyubiquitination compared with control (, lane 4). All three clones with reduced expression showed higher sensitivity to MMS than wild type or the two control clones infected with an empty lentivirus (). In contrast, -silenced cells showed no substantial sensitivity to UV mimetic 4-nitroquinoline 1-oxide (4-NQO), MMC, or γ-irradiation (unpublished data). Moreover, -silenced clones (B2 and C4) showed a greater number of chromosome breaks than did wild type after 0.01% MMS treatment (). Notably, the levels of reduction of expression () were well correlated with their levels of sensitivity to MMS and the frequencies of chromosome breaks (). These observations suggest that SHPRH is involved in MMS-induced DNA-damage responses.
Accumulating evidence suggests that DNA damage bypass followed by PCNA modifications is important for suppressing genomic instability and cancers. For example, the overexpression of HHR6B was implicated in the chromosomal instability phenotypes of human breast cancer cells (). The targeted disruptions of the or gene in either mouse embryonic stem cells or mouse embryonic fibroblasts, respectively, increased genomic instability, including sister chromatid exchange, homologous recombination, and illegitimate recombination (; ). Mutations of , which encodes a TLS polymerase η, were found in cancer-prone xeroderma pigmentosum variant syndrome (; ). Our observations prove that the regulatory mechanisms of PCNA through differential modifications by RAD18 and yRad5/ SHPRH are fully conserved and constitute a fundamental mechanism to prevent genomic instability throughout evolution.
HEK 293T and HCT116 cells were cultured in DME and McCoy's media supplemented with 10% fetal bovine serum, respectively. Commercially available anti-PCNA (PC10), anti-ubiquitin (P4D1), anti-polyubiquitin (FK2), anti-myc (9E10), anti-HA (12CA5), anti-FLAG (M2), anti-V5 antibodies, and anti-RAD18 (K-15) were used. Polyclonal anti-SHPRH antibody was previously described (). MMS, MMC, and mimosine were purchased from Sigma-Aldrich.
Full-length cDNA of human was obtained by PCR with IMAGE clones (available from GenBank/EMBL/DDBJ under accession nos. and ) and RACE-PCR products as templates. cDNA encoding human , , , and were obtained from Mammalian Genome Collection (, MHS1010-52750; , MHS1011-58526; , MHS1011-62471; , MHS1011-62750). Expression plasmids were constructed by subcloning each cDNA into pcDNA3.1-myc-His, p3XFLAG-CMV, or pGEX-6P-1. UBC13(wild type or C87A)-HA–expressing plasmids were gifts from D. Bohmann (European Molecular Biology Laboratory, Heidelberg, Germany) and J. Kehrl (National Institute of Allergy and Infectious Diseases, Bethesda, MD), respectively. Point mutations in the RING finger domain of (C1432A) and in PCNA(K164R) were introduced by using an in vitro mutagenesis method (QuikChange; Stratagene).
For coimmunoprecipitation assay, HEK 293T cells were transfected with various combinations of expression plasmids using FuGENE 6 transfection reagent (Roche) and lysed in the TNE buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.5% NP-40, 1 mM EDTA, 8% glycerol, 0.5 mM DTT, 50 mM NaF, 1 mM PMSF, 1 μg/ml aprotinin, and 1 μg/ml leupeptin). For the GST pull-down assay, various GST fusion proteins were expressed in strain BL21 (DE3)-RIL (Stratagene) and purified using glutathione–Sepharose beads (GE Healthcare). 20 μg of GST fusion proteins were used for pulling down SHPRH-FLAG or RAD18-FLAG expressed in HEK 293T cells.
-silencing vectors were constructed by cloning the target sequences of 5′-TTCAATGCCCTCCTACAC-3′ (construct B) and 5′-AGTGTCCATCCTTTCCAT-3′ (construct C) into the lentivirus-based expression vector pLL3.7 (a gift from V. Parijs, Massachusetts Institute of Technology, Cambridge, MA). -silencing lentivirus vector was purchased from Open Biosystems. Lentivirus packaging plasmids were gifts from F. Candotti (National Human Genome Research Institute, Bethesda, MD). Lentivirus-infected cells were selected by the expression of GFP () or by puromycin (). The expression level of was examined by Western blot or by RT-PCR with primers 5′-GAGCAACTCTGATCATCTCTCCAAG-3′ and 5′-GATAGAGAAGTCGAACCCACCAGTG-3′. Primers used for amplifying as a control were 5′-GCTCGTCGTCGACAACGGCTC-3′ and 5′-CAAACATGATCTGGGTCATCTTCTC-3′. SHPRH-silenced clones B2, B11 (construct B), and C4 (construct C) were used in this study.
HCT116 cells were treated with MMS for 2 h, washed with PBS, and lysed in RIPA buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.1% Triton X-100, 0.5 mM DTT, 50 mM NaF, 1 mM PMSF, 1 μg/ml aprotinin, and 1 μg/ml leupeptin). In some experiments, cells were treated with 0.5 mM mimosine for 20 h. For measuring MMS sensitivity, -knockdown cells were treated with increasing concentrations of MMS for 1 h, washed with PBS three times, replated onto 6-well plates at defined cell densities, and cultured for 10 d. Colonies grown from the surviving cells were counted and expressed as a survival fraction (%) compared with untreated cells as 100%. For analyzing MMS-induced chromosomal breaks, -knockdown cells were treated with 0.01% MMS for 1 h, washed with PBS three times, and cultured for 24 h. At least 100 metaphase spreads from each clone were analyzed.
Fig. S1 shows moderate homology between the SWI2/SNF2 domains in SHPRH and yeast Rad5. Fig. S2 shows that expression in the yeast strain could not complement UV sensitivity. Fig. S3 shows that RAD18 and RAD6B induced exclusive monoubiquitination of PCNA, the reduced activity of SHPRH(ΔRING) for PCNA polyubiquitination, and that SHPRH did not promote polyubiquitination of c-JUN. Online supplemental material is available at . |
Fas receptor apoptotic signaling is critical for normal development and tissue homeostasis (). At the molecular level, Fas receptor is activated by the binding of either membrane-bound or -soluble Fas ligand (FasL), which induces receptor oligomerization (; ). Fas receptor oligomerization stimulates formation of the death-induced signaling complex (DISC), which is minimally composed of the adaptor Fas-associated death domain protein and caspase 8/10 (). Caspase 8/10 activation can lead to direct activation of the effector caspases, such as caspase 3. In some cells, the activation of effector caspases is amplified by engaging the mitochondrial pathway via the Bcl-2 family member Bid (). Caspase-dependent Bid truncation leads to translocation of the protein to the mitochondria, where it promotes the release of proapoptotic factors such as cytochrome . Proapoptotic proteins released from mitochondria stimulate formation of the apoptosome and efficient caspase 3 activation ().
In addition to activating the apoptosome, cytochrome released from mitochondria can bind and modulate the activity of the inositol 1,4,5-trisphosphate receptor (IPR), which is an ER-resident calcium channel (; ). As such, Fas-mediated apoptosis requiring engagement of the mitochondrial pathway may be associated with IPR-dependent signaling. However, although it has been suggested that elevated cytosolic calcium may contribute to the progression of Fas apoptosis (; ; ; ), the molecular mechanisms are unknown.
We show that efficient Fas signaling requires calcium release from the ER in two separate phases mediated by distinct but interdependent mechanisms. The first phase of calcium elevation is rapid and associated with activation of PLC-γ1, subsequent IP generation, and calcium release via IPR channels. The second phase of calcium elevation occurs over the course of hours and is mediated by cytochrome binding to IPR. Blocking either phase of calcium elevation inhibits FasL-mediated apoptosis, highlighting possible therapeutic targets for disorders associated with this apoptotic pathway.
Jurkat T-lymphoma cells are sensitive to FasL-induced apoptosis in a manner dependent on engagement of the mitochondrial pathway (). To determine whether IPR activation contributes to FasL-mediated cell death in Jurkat cells, we first examined whether IP, which is the obligate ligand of IPR, is produced after treatment with FasL. Jurkat cells were transfected with the phosphoinositide 4,5-bisphophate biosensor comprised of the pleckstrin homology domain of PLC-δ1 coupled to GFP (PLCδ1PH-GFP; ). This protein localizes to the plasma membrane under basal conditions, but rapidly redistributes to the cytosol after treatment with agonists that stimulate IP production (; ). FasL stimulation of Jurkat cells resulted in rapid dissipation of PLCδ1PH-GFP fluorescence from the plasma membrane, indicating that FasL stimulates IP production (). The ratio of plasma membrane to cytosol fluorescence oscillates after an initial sharp decrease, which is consistent with other agonists coupled to PLC activation and oscillatory calcium release events (; ). This redistribution is not seen in Jurkat cells lacking PLC-γ1 expression (PLC-γ1 null; ; ). A marker of PLC-γ1 activation is phosphorylation of tyrosine residue 783 (; ). FasL stimulation resulted in a time-dependent increase in phosphorylation of tyrosine 783 (), indicating that PLC-γ1 is activated in response to FasL. There was no PLC-γ1 immunoreactivity in lysates prepared from PLC-γ1–null cells (not depicted).
To examine whether PLC-γ1 activation is associated with calcium release, we monitored alterations in intracellular calcium in response to FasL stimulation in wild-type and PLC-γ1–null Jurkat cells. FasL stimulated immediate oscillations in cytosolic calcium (). Importantly, PLC-γ1–null cells did not release calcium in response to FasL treatment (). Stable expression of wild-type PLC-γ1 in PLC-γ1–null cells () rescued calcium release in response to FasL application ().
FasL stimulation of Jurkat cells causes release from mitochondria in cytochrome , which subsequently binds IPR channels (). In subcellular fractionation experiments, this is observed as a translocation of cytochrome immunoreactivity from a 10,000 mitochondrial-enriched pellet to a 100,000 ER-enriched and mitochondria-free pellet (; Jurkat WT; ; ). Cytochrome release was not observed in PLC-γ1–null cells upon FasL treatment, suggesting that PLC-γ1 activation and calcium release are upstream of mitochondrial permeabilization (; PLC-γ1 null). Consistent with this observation, PLC-γ1–null cells were defective in caspase-3 activation () and were resistant to cell death induced by FasL (). Stable expression of wild-type PLC-γ1 in PLC-γ1–null cells restored cytochrome release, caspase 3 activation, and cell death induced by FasL (, H to J).
To investigate if IPR channels are responsible for calcium mobilization downstream of PLC-dependent IP production, we transiently knocked down the expression of IPR-1 by RNAi. This isoform is the predominant isoform expressed in Jurkat cells (). Because of the low (∼10%) transfection efficiency of Jurkat cells, biochemical determination of RNAi efficacy was first tested in HeLa cells (). RNAi transfection resulted in a substantial reduction in IPR-1 levels, which could be rescued by transfection of the rat IPR-1 gene, which has several nucleotide changes within the RNAi-targeted sequence (see Materials and methods). Expression of another closely related IPR isoform (IPR-3) abundantly expressed in HeLa cells was unaffected, demonstrating specificity. To test whether knockdown of IPR-1 in Jurkat cells affected calcium release induced by FasL stimulation, we cotransfected RNAi and YFP and examined single-cell calcium responses simultaneously in RNAi-transfected and nontransfected cells. IPR-1 knockdown cells were resistant to calcium release induced by FasL (). Control RNAi-transfected cells responded similarly to untransfected controls (). Overexpression of rat IPR-1 rescued calcium release induced by FasL in IPR-1 knockdown Jurkat cells (). IPR-1 knockdown cells were also resistant to cell death induced by exposure to FasL, an effect rescued by overexpression of rat IPR-1 (). Longer incubations of IPR knockdown cells with FasL resulted in increased cell death, suggesting delayed activation of alternative pathways, such as mitochondrial-independent caspase activation (), residual IPR activities, or RNAi turnover and loss of down-regulation. PLC-γ1–null cells, with or without treatment with IPR RNAi, are refractory to FasL-induced cell death at all time points, suggesting that recovery of cell death in IPR RNAi-treated cells is caused by loss of down-regulation (). These results indicate that activation of IPR channels and calcium release from internal stores are downstream effectors of PLC-γ1 activation after FasL stimulation.
We hypothesized that cytochrome binding to IPR contributes to changes in calcium levels during FasL-mediated apoptosis, as we have previously shown for staurosporine-induced apoptosis (). It can be predicted that cytochrome binding to IPR would occur downstream of PLC-γ1 activation and mitochondrial permeabilization with kinetics similar to those observed biochemically for cytochrome release (12–24 h; ). Blocking cytochrome binding to IPR with a dominant-negative peptide (B-IP3RCYT) inhibits staurosporine-induced caspase activation and cell death (). We predicted B-IP3RCYT would not affect rapid PLC-γ1–dependent calcium release events, but would block potential sustained elevations in cytosolic calcium (). Consistent with these hypotheses, B-IP3RCYT or a peptide with two point-mutations eliminating cytochrome binding (B-IP3RCYTmut) had no effect on initial calcium release events induced by FasL (). Treatment of Jurkat cells for 24 h with FasL resulted in increased basal cytosolic calcium (), which is consistent with previous studies (). Pretreatment with B-IP3RCYT, but not B-IP3RCYTmut, blocked this late elevation in cytosolic calcium (). FasL treatment of Jurkat cells is also associated with rapid alterations in mitochondrial calcium immediately after FasL treatment, and sustained increases in mitochondrial calcium 24 h after FasL treatment (Fig. S1 and Video 1, available at ). B-IP3RCYT selectively inhibited the sustained increases in mitochondrial calcium at 24 h. Finally, preincubation with B-IP3RCYT, but not B-IP3RCYTmut, attenuated cytochrome release and cell death in response to FasL stimulation (). Although not detectable in , we predict that small amounts of cytochrome are released in the presence of B-IP3RCYT, even though cell-wide cytochrome release is blocked. These results suggest that PLC-γ1 activation is critical to the initial oscillatory phase of calcium elevation, and cytochrome binding to IPR is necessary for a second, temporally delayed increase in cytosolic and mitochondrial calcium. It appears that both phases are required for optimal Fas signaling, as specific blockage of either pathway is cytoprotective. It has been shown that caspase-3 cleavage of the IPR is associated with apoptotic calcium release (; ). Treatment of Jurkat cells with the caspase-3 inhibitor z-DEVD-fmk had no effect on calcium mobilization in response to FasL, suggesting this pathway is not a required component of FasL-induced calcium release (Fig. S2).
To determine if calcium is required for other models of FasL-mediated apoptosis, we tested whether primary mouse hepatocytes released calcium in response to FasL stimulation. FasL treatment of hepatocytes was associated with calcium release that could be inhibited with the PLC inhibitor U73122, but not the inactive analogue U73343 (). The kinetics of release were distinct from those observed in Jurkat cells, with a nonoscillatory rise which diminished after ∼50 min. Interestingly, we observed a spike in cytosolic and in particular nuclear calcium in individual cells after the initial rise in cytosolic calcium (, arrowheads), which would likely influence FasL-dependent nuclear events such as gene transcription. FasL treatment of hepatocytes for 24 h was associated with cytochrome release and translocation to both the ER and cytosol, and these effects were reversed by B-IP3RCYT treatment (; compare 24 and 24P, where P indicates peptide treatment). The presence of cytochrome in the cytosol in hepatocytes may reflect saturation of IPR channels with cytochrome at 24 h caused by the low density of IPRs in this cell type (). We also observed a partial loss of IPR-1 immunoreactivity after FasL treatment (), which may be caused by caspase-3 cleavage. Cell death induced by FasL was inhibited by U73122 or B-IP3RCYT, suggesting that the PLC activation and cytochrome binding to IPR are both required for FasL-mediated apoptosis in hepatocytes ().
The molecular basis for calcium mobilization during apoptosis has remained enigmatic. We suggest a linear sequence of events, which lead to calcium-dependent mitochondrial permeabilization, caspase activation, and cell death during FasL-mediated apoptosis (). FasL binding to the Fas receptor recruits the canonical components of the DISC complex, and, concurrently, PLC-γ1 is activated by an unknown mechanism. These two events occur on the order of seconds to minutes, leading to calcium release from internal stores, elevated mitochondrial calcium, and Bid-mediated increases in mitochondrial permeability (). FasL-mediated apoptosis might be reversible at this stage, providing a critical control point. The second step occurs over the course of minutes to hours, and is characterized by limited cytochrome release from mitochondria and binding to IPR. This would sensitize the channel to increased calcium release (), ultimately resulting in depletion of ER calcium, mitochondrial calcium overload, and global cytochrome release from all mitochondria. Consistent with the critical role of calcium in regulating FasL-mediated apoptosis, the calcium chelator BAPTA-AM suppresses cytochrome release, caspase activation, and cell death in response to FasL (Fig. S3, A–C, available at ). Furthermore, if ER-derived calcium was critical for cytochrome release from mitochondria and cell death in Jurkat cells, it would be expected that passive depletion of ER stores by thapsigargin would also cause cytochrome release and cell death. We find that thapsigargin dose-dependently results in cytochrome release, caspase activation, and cell death in Jurkat cells (Fig. S3, D–F). Finally, reducing the functional expression of PLC-γ1 and IPR or blocking cytochrome to IPR is cytoprotective, suggesting that these proteins may be useful drug targets for treating disorders in Fas or other calcium-dependent apoptotic pathways.
Jurkat T cell leukemia and Jurkat derivatives j.gamma1 and j.gammaWT were obtained from the American Type Culture Collection. Primary mouse hepatocytes were prepared as described elsewhere () and plated on collagen-coated plates or coverslips. Primary hepatocytes were used within 24 h to limit dedifferentiation.
Caspase activity was determined as previously described () using z-DEVD-R110 as the protease substrate at a final concentration of 50 μM (American Peptide Company).
Cells were transiently transfected with PLCδ-PH-GFP (), plated on poly--lysine–coated coverslips, and imaged at 25°C. PLCδ-PH-GFP was a gift from T. Balla (National Institutes of Health, Bethesda, MD). Images were acquired every 5 s using MetaFluor imaging software (Molecular Devices). Fas ligand-bearing vesicles (Millipore) were administered at 1 ng/ml. Quantification of PLC activity was calculated as a ratio of plasma membrane to cytosol fluorescence over time, as described elsewhere (). For immunoblot analysis, Jurkat cells were treated with 1 ng/ml FasL for the indicated times, and lysates were probed using phosphotyrosine 783-PLC-γ1 and PLC-γ1 antibodies (Cell Signaling Technologies).
Calcium measurements were performed as previously described (). Fura-2–loaded cells were allowed to attach to poly--lysine–coated coverslips and imaged at 25°C. Cells were treated with 10 ng/ml FasL vesicles, and images were acquired every 3 s for 2,000 s. Jurkat cells respond heterogeneously to FasL (), thus, we displayed representative single-cell traces. Cells with spontaneous release activity in the absence of FasL were identified by imaging at least 100 s before FasL addition and were eliminated from analysis. In experiments where DNA or RNAi were transfected into Jurkat cells, expressing cells were identified by cotransfecting YFP. Nonexpressing cells were imaged simultaneously with expressing cells as internal controls. Each experiment was repeated a minimum of five times, comprising hundreds of single-cell traces. In and S2 B, cells treated for 24 h with FasL were loaded with Fura-2 and cytoplasmic calcium imaged in three separate fields, each comprising 50–100 cells. The mean of the three fields comprised one data point. The experiment was repeated an additional two times and presented as the SEM of three experiments. Our previous study indicated that B-IP3RCYT loading into HeLa cells may augment staurosporine-induced calcium elevations (). Testing B-IP3RCYT in multiple apoptosis model systems, including Jurkat cells in this study, indicated that this effect is unique to that particular model system.
Fura-2 and GFP images were acquired on an inverted microscope (TE2000; Nikon) using a 60× oil immersion objective (SuperFluor; Nikon) with a 1.3 NA. All imaging was performed at 25°C in 107 mM NaCl, 7.2 mM KCl, 1.2 mM MgCl, 1 mM CaCl, 11.5 mM glucose, 0.1% bovine serum albumin, and 20 mM Hepes 7.2. Images were captured with a camera (CoolSNAP HQ; Roper Scientific). Rapid filter changes for ratiometric imaging were computer controlled via a 10–2 filter wheel controller (Lambda; Sutter) and MetaFluor data acquisition and analysis software. Raw data was acquired with MetaFluor and graphed in Sigma Plot (SPSS Scientific). Fluorescent images were pseudocolored using the IMD display mode in MetaFluor for display purposes in , and assembled without further manipulation in Photoshop (Adobe).
Subcellular fractionation was performed as described previously (). Fractionation purity in each experiment was determined by blotting with cytochrome oxidase (mitochondria), heme oxygenase (ER), and lactate dehydrogenase (cytosol).
Cell death was quantified as previously described (), either by propidium iodide staining, or, in the case of BODIPY-labeled peptide-treated samples, by trypan blue staining. The number of dead cells was determined either by manual counting in a light microscope or by flow cytometry. Cell death in transfected cells was determined by cotransfecting YFP. YFP was retained in dying cells, including propidium iodide–positive cells, allowing the determination of the effects of transfection on cell death. As such, there was no significant difference in the relative amount of cell death between nontransfected and YFP only–transfected cells (compare , YFP, with , Jurkat WT, or , FasL).
Stealth-modified (Invitrogen) double-stranded RNA against the human IPR-1 gene (sense sequence, 5′-GAGGGAUCGACAAAUGGAUUUAUUA-3′) targeting ORF nucleotides 314–338 was purchased from Invitrogen. Control RNA was similar, but with several deletions and insertions (sequence, 5′-GAGUAGCCAAAUAGGUAUUAGGUUA-3′). Transfection with Lipofectamine 2000 (Invitrogen) of various doses of RNA was used to determine that 100 pmol of double-stranded RNA complex per well of a six-well dish gave maximum knockdown. Knockdown was readily evident at 24 and 48 h. The blot in is 24 h after transfection. Rescue of RNAi knockdown was accomplished by overexpression of the rat IP3R-1 gene, which has three substitutions within the targeted sequence (rat sequence, with sequence changes in bold, 5′-GAGGGAUCACAAUGGAUUUAUA-3′).
The IP3RCYT sequence is DNKTVTFEEHIKEEHN, comprising amino acids 2,567–2,582 of human IPR-1. IP3RCYTmut replaces two glutamic acid residues critical for binding () with glutamine, DNKTVTFHIKEEHN. A C-terminal cysteine was added during synthesis to facilitate coupling to BODIPY 577/618 via a maleimide linkage, as previously described (). IP3RCYT and IP3RCYTmut were synthesized by the Protein Chemistry Laboratory core facility at the University of Texas Medical Branch.
All data is presented as the mean ± the SEM. Statistical significance was examined with a test. P < 0.05 was determined to be significant. Actual P values are listed in each figure.
Fig. S1 shows mitochondrial calcium levels early and late after FasL stimulation, and the effect of B-IP3RCYT. Fig. S2 shows the effect of z-DEVD-fmk on FasL-induced calcium signals. Fig. S3 shows the effects of BAPTA and thapsigargin on cytochrome release and cell death. Video 1 depicts mitochondrial calcium dynamics in response to FasL. There is also a referenced Supplemental materials and methods. Online supplemental material is available at . |
Transmembrane proteins monoubiquitinated on their cytosolic domains are sorted into the lumenal vesicles of late endosomal multivesicular bodies (MVBs; for review see ). MVB vesicles and their cargoes are exposed to the hydrolytic interior of the lysosome upon fusion of the limiting endosomal membrane with the lysosomal membrane. The mechanism of MVB cargo sorting is conserved and mediated by class E Vps proteins originally identified in . Most class E genes encode soluble cytosolic proteins recruited transiently to endosomes. Genetic and biochemical data suggest a sequence that begins with recruitment of the Vps27–Hse1 complex, which recognizes monoubiquitinated cargoes, followed by recruitment of three distinct endosomal sorting complexes required for transport (ESCRTs; for review see ). Like the Vps27–Hse1 complex, ESCRT-I and -II bind monoubiquitinated cargoes, whereas ESCRT-III lacks ubiquitin-binding subunits and functions downstream of cargo recognition.
ESCRT-III is comprised of the Vps20–Snf7 and Vps2–Vps24 subcomplexes (). Although its molecular function is not fully understood, one role for ESCRT-III is the recruitment of late-acting components of the sorting machinery. Snf7 recruits Bro1 (), and Bro1 recruits Doa4, which deubiquitinates cargoes before their enclosure within MVB vesicles (). Vps4 is an ATPase that catalyzes the dissociation of class E Vps proteins from endosomal membranes, and, in the absence of Vps4 activity, ESCRT complexes accumulate on endosomes (; ,).
A central question of Vps4 function concerns how its activity is coordinated to dissociate multiple protein complexes. We report that Did2, a protein related to ESCRT-III subunits (), directs Vps4 activity to the dissociation of ESCRT-III. In the absence of Did2, ESCRT-I and -II dissociation occurs, whereas ESCRT-III and downstream components accumulate on endosomes. Surprisingly, MVB vesicle budding proceeds in the absence of Did2 despite the requirement for Did2 in sorting cargoes, demonstrating that vesicle formation and MVB cargo sorting can be uncoupled.
The N terminus of Did2 is predominantly comprised of basic amino acids, whereas its C terminus predominantly contains acidic residues (). As shown in , bacterially expressed His-Vps4 copurified with GST-Did2 but not GST alone. This interaction occurred regardless of whether Vps4 was locked in its ATP-bound state (His-Vps4) or was disabled from binding ATP (His-Vps4; ). In contrast, GST-Vta1 showed a strong preference for binding His-Vps4 (), which is consistent with Vta1 interacting with ATP-bound Vps4 to stimulate its oligomerization ().
Studies of CHMP1b and Vps4a, the mammalian orthologues of Did2 and Vps4, respectively, demonstrated that the C terminus of CHMP1b binds Vps4a and that this interaction is disrupted by the mutation of leucine-64 in the microtubule interaction and trafficking (MIT) domain of Vps4a (). This leucine is conserved in the MIT domain of yeast Vps4 (), suggesting that it is important for the interaction between Did2 and Vps4. Indeed, His-Vps4 failed to bind GST-Did2 but still bound GST-Vta1 (), which is in agreement with Vta1 binding the AAA domain rather than the MIT domain of Vps4 (). We further observed that His-Vps4 interacted with the C terminus of Did2 (GST-Did2) but not its N terminus (GST-Did2; ). Thus, the binding mechanism between Vps4 and Did2 appears conserved.
Because the MIT domain of Vps4 is essential for its localization to endosomes (), we addressed whether its binding to Did2 mediates the endosomal recruitment of Vps4. Locked in the ATP-bound state, Vps4 is unable to catalyze the dissociation of itself and its substrate proteins from endosomes (). Thus, GFP fused to Vps4 appeared concentrated at class E compartments stained with the lipophilic dye FM 4-64 (, arrowheads). GFP-Vps4 also localized to endosomes in cells (, arrowheads), indicating that the recruitment of Vps4 does not require Did2. Localization of Vps4 to the endosomal membrane in the absence of Did2 was also observed by subcellular fractionation (Fig. S1 A, available at ), which was not surprising given that Vps4 binds multiple distinct ESCRT-III components. Indeed, the deletion of did not affect the ability of GST-Vps4 to pull down Snf7 from yeast lysates (Fig. S1 B), which is consistent with our observation that GST-Vps4 interacts directly with His-Snf7 (Fig. S1 C) and with a previous study showing that Vps4 interacts directly with Vps20 ().
Did2 binds Vta1 and is required for the interaction of Vta1 with the ESCRT-III component Snf7 (). Therefore, we examined whether Vta1 or ESCRT-III proteins mediate the recruitment of Did2 to endosomes. Genetic and biochemical studies suggest that ESCRT-III consists of a Snf7–Vps20 subcomplex and a Vps2–Vps24 subcomplex (). Did2 accumulated at the class E compartment in cells regardless of whether , (), or (not depicted) had been deleted. In contrast, the deletion of either () or (not depicted) caused Did2-GFP to remain cytosolic.
and
extracts ().
Fluorescence microscopy () and subcellular fractionation () indicated that the N terminus of Did2 is necessary and sufficient for endosomal localization. Therefore, the simplest explanation for the mislocalization of Did2 to the cytosol in cells lacking either Vps2 or Vps24 is that the N terminus of Did2 interacts directly with the Vps2–Vps24 subcomplex. Indeed, recombinant His-Vps24 bound GST-Did2 but not GST-Did2 or GST alone (). In contrast, recombinant His-Snf7 failed to bind GST-Did2 (unpublished data).
To address the functional significance of the interaction between Did2 and Vps4, we examined the ability of Vps4 to mediate the dissociation of ESCRT complexes in the absence of Did2. As shown previously (; ), Vps23-GFP of ESCRT-I () and Vps36-GFP of ESCRT-II () in wild-type cells were predominantly cytosolic in addition to being localized weakly at punctate structures. As expected, Vps23- and Vps36-GFP in cells accumulated at class E compartments (, A and B; arrowheads). However, the distributions of both proteins in cells appeared to be similar to their distributions in wild-type cells (). Thus, Did2 is not required for the dissociation of either ESCRT-I or -II.
cells.
Fusion of GFP to ESCRT-III proteins disrupts their function in MVB sorting (unpublished data). Thus, we assessed the distributions of endogenous Snf7 and Vps24 by subcellular fractionation and Western blotting. As shown previously (), Snf7 was predominantly soluble, and Vps24 was evenly distributed between membrane and soluble fractions in wild-type cells, whereas both proteins shifted entirely to the membrane pellet in cells (). Snf7 and Vps24 were similarly concentrated in the pellet fraction of cells (), indicating that Did2 is essential for the membrane dissociation of both proteins. Likewise, Did2 was required for the endosomal dissociation of Bro1 and Doa4 (), components that function downstream in the MVB pathway but depend on ESCRT-III for recruitment to endosomes (; ). The ability of Bro1 and Doa4 to dissociate from endosomes may require Did2 to coordinate the Vps4-mediated dissociation of ESCRT-III. Although Vta1 binds Did2 and requires Vps4 to dissociate from endosomes (), Vta1 appeared predominantly cytosolic in the absence of Did2 (), indicating that it does not need Did2 for dissociation. This quality makes Vta1 unique among ESCRT-III–associated Vps4 substrates acting late in the MVB pathway.
Class E compartments stained by FM 4-64 are a hallmark phenotype caused by mutations in and other class E genes. By EM, these abnormal late endosomes appear as flattened stacks of cisterna-like structures devoid of lumenal vesicles (; ). We examined cells versus wild-type and cells using high resolution EM and tomographic modeling. An example of a typical wild-type MVB is shown in the tomogram in and is modeled in (B and C; and see Videos 1 and 2, available at ). The limiting membrane of this MVB is approximately spherical and surrounds numerous lumenal vesicles. As expected, no multivesicular endosomes were detected in cells, which instead contained class E compartments similar to the structures described previously in cells lacking Vps4 function (). An example of a class E compartment in cells is shown in the tomogram in and is modeled in (E and F; and see Videos 3 and 4). Three-dimensional analysis indicated that its elongated cisternae-like elements did not connect with one another. Similar characteristics were observed in serial sections that included entire class E compartments (unpublished data), and lumenal vesicles were not observed in >300 class E compartments of cells examined by EM (including three structures modeled by tomography), which is consistent with an essential role for Vps4 function in the biogenesis of MVB vesicles.
Surprisingly, multivesicular endosomes were readily apparent in cells, an example of which is shown in the tomogram in and modeled in (H and I; and Videos 5 and 6, available at ). The limiting membrane, rather than being spherical as seen in wild-type cells, was typically elongated, which is similar to the cisternae-like elements of class E compartments. In >200 sections examined by EM, these vesicular tubular endosomes (VTEs) were most often observed crowded together, which is reminiscent of the compact organization displayed by class E compartments in cells (). Three-dimensional analysis of tomograms and serial sections that encompassed entire VTEs (unpublished data) indicated that the lumenal vesicles were not interconnected, nor were they connected to the limiting membrane. The interior of lumenal vesicles in VTEs of cells appeared electron dense and were uniformly larger (by 38%) than lumenal vesicles of MVBs in wild-type cells (P < 0.0001; 23.98 ± 0.23 vs. 33.01 ± 0.56 nm, respectively; ), raising the possibility that ESCRT-III, which is unable to dissociate from the membrane, is mistakenly packaged as cargo. However, ESCRT-III was only detected at the limiting membrane of VTEs in thin sections of cells examined by immunogold labeling using antibodies against Vps24 (Fig. S2 D, available at ).
The similarity of the class E compartment and VTE when viewed by fluorescence microscopy () underscores the need for EM when reaching any conclusion regarding endosome morphology. Moreover, the ultrastructural differences between class E compartments and VTEs suggest that the class E phenotype warrants subdivision based on endosome morphology. The absence of lumenal vesicles in class E compartments as a result of the loss of function of ESCRTs, Vps4, or Bro1 has been thought to signify that these components comprise the core class E Vps machinery required for vesicle budding (for review see ). However, the VTEs observed in cells contradict the view that the dynamic cycling of ESCRT-III, a subset of this core machinery, is either a pre- or corequisite for MVB vesicle formation, although it remains likely that the assembly of ESCRT-III on endosomes is critical to the budding event.
Like Vps4, Did2 is required for efficient sorting of MVB cargoes, as indicated by the failure of GFP–carboxypeptidase S (CPS), a biosynthetic protein, to be sorted into the vacuole lumen in cells (). Sna3-GFP (Fig. S2 A), another biosynthetic protein, as well as Ste3-GFP (Fig. S2 B), an endocytic protein, are also mislocalized, demonstrating that the loss of Did2 function causes a broad cargo-sorting defect. Intriguingly, the sorting of GFP-CPS in cells was partially rescued upon in-frame fusion of ubiquitin to its cytosolic domain (Fig. S3 C, available at ). However, the mislocalization of Sna3- GFP suggests that the molecular basis for the cargo-sorting defect in cells is not directly related to ubiquitination because Sna3 does not require ubiquitin to be sorted into the MVB pathway (). The ubiquitin-independent localization of Sna3-GFP to the vacuole lumen indicates, albeit indirectly, that cargo sorting can be uncoupled from lumenal vesicle formation in yeast. Indeed, MVB vesicles are observed in cells lacking functional Doa4 or Rsp5, the primary E3 ubiquitin ligase for MVB cargoes in yeast (unpublished data). Similarly, MVB vesicles are observed by EM despite deletion of the gene (unpublished data), which blocks MVB sorting of CPS but not Ste2, an endocytic cargo protein (). In mammalian cells, the overexpression of a mutant form of Hrs that is defective in ubiquitin binding has no effect on MVB vesicle formation but reduces the efficiency of cargo sorting, perhaps because of a failure in the concentration of cargoes at the site of vesicle budding (). Although the nature of the sorting defect in is not clear, it might be caused by the trapping of cargoes within an ESCRT-III network that is unable to release from endosomes, in which case the in-frame fusion of ubiquitin could promote sorting by enhancing cargo interactions with ESCRT-I and -II to the exclusion of ESCRT-III.
Our findings suggest that Did2 functions to coordinate Vps4 activity to ESCRT-III dissociation (). The C terminus of Did2 binds the MIT domain of Vps4, whereas the N terminus of Did2 binds Vps24 of ESCRT-III. Did2 has a position downstream of the Vps2–Vps24 subcomplex in order of assembly because Vps24 can be recruited to the membrane in the absence of Did2 but not vice versa. The significance of Did2 recruitment is that Vps4 requires Did2 to catalyze the endosomal dissociation of ESCRT-III as well as factors that function downstream. Therefore, this set of Vps4 substrates is Did2 dependent, which is in contrast with ESCRT-I and -II, which are Did2 independent (). The selective role Did2 plays in coordinating Vps4 with ESCRT-III dissociation implies that additional factors couple Vps4 function to the dissociation of ESCRT-I and -II.
Yeast strains and plasmids used in this study are listed in Table S1 (available at ). Yeast manipulations were performed using standard protocols. Gene deletions and introduction of epitopes in yeast were constructed by homologous recombination of PCR products (). Genes PCR amplified from genomic DNA were TOPO cloned into pCR2.1 (Invitrogen) and subcloned using T4 DNA ligase into an expression vector.
BL21(DE3) cells transformed with pGEX4T1 or pET-His PL plasmids were grown at 37°C to logarithmic phase and were induced to express recombinant genes by the addition of 0.5 mM isopropyl-β--thiogalactoside. Cells were harvested 1–2 h later, lysed under native conditions, and cleared of cell debris. GST-tagged proteins were purified using GSTrap FF columns (GE Healthcare), and His-tagged proteins were purified using Ni-agarose beads (QIAGEN). ∼1.2 μg of purified recombinant proteins were tested for stable interactions using glutathione–Sepharose beads (GE Healthcare) essentially as described previously ().
Fractionation of proteins into membrane-associated pellet and soluble cytosolic fractions was performed as described previously (). One-half of OD unit equivalent of each sample was resolved by SDS-PAGE and analyzed by Western blotting. Yeast 3-phosphoglycerate kinase and the mitochondrial porin (Por1) were examined as soluble and membrane-bound controls, respectively.
Strains were grown to logarithmic phase at 30°C in synthetic medium before observation at room temperature in synthetic medium at 100× using a microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.) equipped with an NA 1.40 oil immersion objective (Carl Zeiss MicroImaging, Inc.). Differential interference contrast (DIC) and fluorescence microscopy images were acquired with a digital camera (Cooke SensiCam; Applied Scientific Instruments) and processed using Slidebook (Intelligent Imaging Innovations) and Photoshop 7.0 software (Adobe). GFP-CPS was introduced by transforming cells with pGO45 (). Pulse-chase staining of cells with FM 4-64 has been described previously ().
Cells were high-pressure frozen, freeze substituted with 0.1% uranyl acetate, 0.25% glutaraldehyde, and anhydrous acetone at −90°C, embedded in Lowicryl HM20, and polymerized under UV light at −50°C (). 200-nm semithick sections were placed on Rhodium-plated Formvar-coated copper slot grids and mapped on an electron microscope (CM10 TEM; Phillips) at 80 kV. Dual tilt series images were collected from 60 to −60° with 1° increments at 200 kV using an electron microscope (Tecnai 20 FEG; FEI). Tomograms were imaged at 29,000× with a 0.77-nm pixel (binning 2). Sections were coated on both sides with 15-nm fiducial gold for the reconstruction of back projections using IMOD software (). 3dmod software was used for mapping structure surface areas. Mean z-scale values for wild-type and sections were within 3%. Best fit sphere models were used to measure vesicle diameters to the outer leaflet of membrane bilayers. IMOD calculated limiting membrane surface areas using three-dimensional mesh structures derived from closed contours that were drawn each 3.85 nm using imodmesh software.
Table SI describes strains and plasmids used in this study. Fig. S1 shows that Did2 is not required for Vps4 to interact with ESCRT-III. Fig. S2 shows MVB cargo localization in . Videos 1–6 depict the tomograms and three-dimensional models of wild-type, , and endosomes shown in . Online supplemental material is available at . |
The autoimmune blistering skin diseases pemphigus vulgaris (PV) and pemphigus foliaceus (PF) are caused by autoantibodies (IgG) against desmosomal cadherins and different antigens, including cholinergic receptors (; , ; ; ). We have previously shown that PF-IgG caused cellular dissociation without directly inhibiting desmoglein (Dsg) 1 binding when probed by laser tweezers and atomic force microscopy (). These data favor the contribution of cellular signaling rather than a direct inhibitory action on Dsg binding.
To investigate signaling events in pemphigus blistering, we used an ex vivo human skin model (; ) and the human epidermal cell line HaCaT. We focused on Rho GTPases in this study because they are involved in the regulation of adhesion mediated by several members of the cadherin family (), and we found that the inhibition of Rho GTPases by toxin B induced epidermal splitting similar to pemphigus IgG (unpublished data). However, the role of Rho GTPases in the control of desmosomal adhesion is unclear at present (). A previous study reported that Rho proteins regulate epithelial cadherin (E-cadherin)–mediated adhesion but not the maintenance of desmosomes in keratinocytes because the microinjection of C3 toxin to inhibit Rho A and a constitutively inactive mutant of Rac 1 altered the localization of E-cadherin but not of desmoplakin, a component of the desmosomal plaques (). However, it is possible that the short incubation period of 25 min after microinjection used in the study was sufficient to destabilize adherens junctions but not desmosomes. Thus, a definitive conclusion concerning the role of Rho GTPases in the regulation of desmosomal adhesion cannot be drawn from these studies. Our data demonstrate for the first time that Rho A is involved in the maintenance of desmosomes and that interference with Rho A signaling considerably contributes to pemphigus pathogenesis.
Incubation of human skin for 24 h in the presence of PV- or PF-IgG caused epidermal splitting, whereas no epidermal splitting was found after incubation in the absence of patients' IgG or with IgG from a healthy volunteer (). PV-IgG1–induced splitting occurred suprabasally, whereas in the PF-IgG1–treated epidermis, the cleavage plane was found to be both deep (not depicted) and within the spinous layer (). In control skin, Dsg 3 was localized along cell junctions throughout the entire epidermis except for the granular layer, which displayed weak staining (, a). We used two different bacterial toxins specific for Rho family GTPases: cytotoxic necrotizing factor 1 (CNF-1), a toxin that activates Rho A, Rac 1, and Cdc42 by deamidation (, ), as well as CNFy from because it selectively activates Rho A (). Selective activation of Rho A by CNFy was equally effective as activation of Rho A, Rac 1, and Cdc42 by CNF-1 to block pemphigus IgG–induced skin splitting (). From these data, we concluded that Rho A is the primary Rho GTPase targeted by pemphigus IgG–triggered signaling mechanisms.
To further test this hypothesis, we performed parallel experiments with cultured human keratinocytes (HaCaT) that, in contrast to the skin model, allow biophysical and biochemical studies. To detect pemphigus IgG–induced cell dissociation, we labeled the peripheral filamentous actin (F-actin) ring in HaCaT monolayers by Alexa-phalloidin staining. F-actin was enriched in a beltlike marginal zone along cell junctions (). After 24 h of incubation, PV-IgG1 (, arrows) as well as PF-IgG1 (, arrows) caused intercellular gap formation accompanied by the reorganization of F-actin from cellular junctions to a cytoplasmic meshwork. CNFy completely abolished pemphigus IgG–induced cell dissociation and actin reorganization and increased stress fiber formation along junctions (; c, d, and f). Similar results were obtained using PV- and PF-IgG2 (unpublished data). We studied the effects of pemphigus IgG on desmosome integrity by immunostaining for Dsg 3. In control cultures and monolayers incubated for 24 h in the presence of control IgG or CNFy, immunostaining for Dsg 3 remained localized to cell junctions (, a–c). Under the same conditions, PV-IgG2 and PF-IgG1 induced profound alterations of Dsg 3 immunolocalization (, e and i). In experiments using PF-IgG1, Dsg 3 staining remained continuously localized to cell junctions in areas where no gaps were present and was abolished at gap margins only (, i; arrows). Overall, changes induced by PV-IgG were stronger compared with effects induced by PF-IgG and ranged from the complete disappearance of Dsg 3 staining in some areas of the monolayer to substantial fragmentation of the continuous junction-associated Dsg 3 immunostaining (, e). Similar results were obtained using PV-IgG1 and PF-IgG2 (unpublished data).
Cell dissociation was observed with all patients' IgG fractions as well as after the inhibition of Rho A by a cell-permeable C3 fusion toxin for 24 h (, d). The latter finding is in contrast to an earlier study from . In this study, the microinjection of C3 toxin to inhibit Rho A was found to displace E-cadherin from intercellular junctions in human keratinocytes, whereas the distribution of desmoplakin was not affected. From this result, it was concluded that Rho GTPases are not involved in the regulation of desmosomal adhesion. As outlined above, the failure of Rho A inhibition might be caused by the technical design of the study. When incubation with PV- and PF-IgG was performed in the presence of CNFy to selectively activate Rho A or with CNF-1, cellular dissociation and alteration of Dsg 3 staining in monolayers treated with PV-IgG2 (, f and g) or PF-IgG1 (, j and k) were almost completely abrogated. Similarly, the simultaneous inhibition of p38 MAPK by SB 202190 (, h and l) but not the inhibition of PKC by chelerythrine (not depicted) inhibited effects of PV-IgG2 and PF-IgG1, supporting recent findings from , ) both in vivo and in vitro. Collectively, these results indicate that autoantibody-induced keratinocyte dissociation in situ and in vitro as well as the loss of desmosomes in vitro can be suppressed by the activation of Rho A. Moreover, together with the observation that specific inhibition of Rho A by C3 fusion toxin also induced keratinocyte dissociation and the loss of desmosomes, these data demonstrate for the first time that the GTPase Rho A is critically involved in the maintenance of desmosomes.
Next, we tested the effect of pemphigus IgG and Rho A activation on the localization of desmoplakin and E-cadherin as well as on the keratin filament cytoskeleton. Compared with controls (, a–c), PV-IgG1 induced strong alterations of desmoplakin localization paralleled by keratin retraction, whereas the effects on the localization of E-cadherin and plakoglobin (not depicted) were less severe (, d–f; arrows pointing to the sides of cell dissociation). This is in line with previous studies describing desmoplakin fragmentation and keratin retraction (; ). However, when Rho A was activated by CNFy in the presence of PV-IgG1, the staining of desmoplakin and E-cadherin along cell junctions was enhanced compared with controls (, g and j), and the fragmentation of desmoplakin was only rarely observed (, g; arrow). Under these conditions, keratin retraction was abolished, and many cells displayed dense keratin filament bundles running perpendicular to cell junctions (, h; arrowheads). Similar results were obtained using PV-IgG2 and PF-IgG1/2 (not depicted).
Inhibition of Rho kinase by Y27632 in addition to treatment with PV-IgG1 and CNFy partially blocked the protective effect of Rho activation on desmoplakin localization (, a and c), whereas the CNFy-induced inhibition of cell dissociation was not affected (, b and c), suggesting that Rho kinase is involved in some of the mechanisms regulated by Rho A. In summary, these data indicate that Rho A activation abrogates the pemphigus IgG–mediated disruption of desmosome anchorage to the keratin filament cytoskeleton.
On the basis of these results, we conclude that the activation of Rho A is sufficient to abrogate central events involved in pemphigus blistering, like keratinocyte dissociation and epidermal splitting. To determine whether pemphigus IgG reduced the activity of Rho A in cultured HaCaT cells, we assayed Rho A activity by the ability of the GTPase to bind to the specific effector protein rhotekin coupled to agarose beads (). After incubation for 120 min, both PV- and PF-IgG1 reduced the amount of active Rho A to ∼50% of control levels ( = 3; ). Comparable results were obtained using PV- and PF-IgG2 (unpublished data). Next, we investigated whether the activation of Rho A was modified in response to pemphigus IgG because constant activation is required for Rho proteins to cycle between active and inactive states. When CNF-1 was applied together with PV- and PF-IgG1 for 120 min, the ability of CNF-1 to cause the activation of Rho A was decreased to 30 and 36% ( = 3) compared with cells treated with CNF-1 alone ().
Because the inhibition of p38 MAPK was equally efficient as Rho A activation to block pemphigus IgG–triggered keratinocyte dissociation, we investigated whether p38 MAPK was involved in Rho A inactivation in response to pemphigus IgG. When HaCaT monolayers were incubated with PV- and PF-IgG1 in the presence of the p38 MAPK inhibitor SB 202190, the IgG-triggered inactivation of Rho A was abolished (), indicating that p38 MAPK is part of the signaling cascade leading to Rho A inactivation. Finally, we proved that CNF-1 activated all three Rho proteins in human keratinocytes (i.e., Rho A [], Rac 1, and Cdc42 []), whereas CNFy selectively activated Rho A (). Collectively, the results indicate that pemphigus IgG causes the inhibition of Rho A in a p38 MAPK–dependent manner, presumably by interference with GTPase activation.
We further characterized the mechanisms underlying Rho A–mediated maintenance of desmosomes and used the laser tweezer technique as a functional assay to study Dsg 1–mediated adhesion. In this assay, microbeads coated with recombinant Dsg 1 are allowed to settle on the surface of HaCaT cells for 30 min, where many beads undergo formation of desmosome-like cell to bead contacts as described previously (). During this time course, typically ∼65% of Dsg 1–coated beads had formed tight adhesive contacts and could no longer be displaced by the laser beam (). These control values were set to 100%. Adhesion of Dsg 1–coated beads was not significantly changed by incubation with control IgG for 120 min, whereas incubation with CNF-1 or CNFy increased the number of bound beads to 118 ± 5 and 122 ± 6%, respectively (P < 0.05). However, both the inhibition of Rho A by C3 fusion toxin (180 min) and the inhibition of Rho kinase by Y27632 for 180 min resulted in a significant reduction of bound beads to 52 ± 5 and 58 ± 4% of controls, respectively (P < 0.05). PV-IgG1 and 2 (120 min) also reduced bead binding to 82 ± 7 and 52 ± 5% of control levels. PV-IgG1– and 2–induced reduction of Dsg 1 binding was completely inhibited by simultaneous treatment of cultures with either CNF-1 or CNFy. Similarly, both CNF-1 and CNFy completely inhibited the weakening of bead binding caused by PF-IgG1 and 2. Thus, these experiments support our findings on cell dissociation in that the activation of Rho A alone is sufficient to prevent autoantibody-triggered weakening of Dsg 1–mediated adhesion. Because the specific inhibition of Rho A by C3 fusion toxin and the inhibition of Rho kinase also reduced Dsg 1 binding, we conclude that Rho A and Rho kinase are required for strong Dsg 1–mediated adhesion.
Our findings indicate that Rho A regulates the anchorage of desmosomes to the keratin filament cytoskeleton, which is required for strong Dsg 1 binding and maintenance of desmosomes. To test the cytoskeletal anchorage of desmosomal proteins, triton extraction can be used (). Both PV- and PF-IgG1 increased Dsg 3 in the triton-soluble fraction of cell lysates, indicating that the anchorage of Dsg 3 to the cytoskeleton is reduced by pemphigus IgG (). Activation of Rho A by CNFy reversed the effects of PV- and PF-IgG1. Collectively, these data are consistent with our aforementioned results that the activation of Rho A blocked pemphigus IgG–induced keratin retraction and fragmentation of Dsg 3 and desmoplakin. Dsg 3 levels were not substantially decreased in the cytoskeletal fractions after 24 h of incubation with pemphigus IgG. This was similarly observed using higher concentrations (1 mg/ml) of patients' IgG (unpublished data). However, a recent study showed that Dsg 3 was substantially reduced after incubation with PV-IgG (), indicating that the mechanisms underlying pemphigus skin blistering are rather complex and do not seem to be the same in different patients at all stages of the disease.
Collectively, we provide evidence that Rho A is critically involved in pemphigus pathogenesis and in the maintenance of desmosomes. However, at present, it is not clear how the inactivation of Rho proteins is induced by antibody binding. It is possible that signaling pathways caused by autoantibodies inhibit dissociation of the negative regulator Rho guanine nucleotide dissociation inhibitor from Rho A, thereby preventing GTP loading and making it ineffective to bind to its effector molecules. We found that the inactivation of Rho A was dependent on p38 MAPK, which is in line with a recent study (). Plakoglobin and PKC may also be involved in pemphigus pathogenesis (; ; ). Moreover, the precise mechanisms by which Rho A regulates desmosomal adhesion remain to be elucidated. We demonstrate that Rho A regulates the anchorage of desmosomal proteins to the keratin filament cytoskeleton. Finally, Rho A and Rho kinase seem to be required to provide strong Dsg 1 binding. Because Rho A activation inhibits the blistering effects of pemphigus IgG in human epidermis, this might be used as a new therapeutic approach with the possibility of topical epidermal application. Future studies are required to test whether Rho A activation by different approaches is effective in vivo.
HaCaT cells were cultured as described previously and were used for all experiments when grown to confluent monolayers (). CNF-1 and CNFy were prepared and characterized as described previously (, ; ). Preliminary experiments indicated that CNF-1 at a dose of 300 ng/ml for 120 min and CNFy at a dose of 600 ng/ml for 6 h were necessary to induce the activation of Rho proteins (). For experiments using human epidermis, 1,200 ng/ml of both toxins were required. The cell-permeable C3 fusion toxin was a gift from H. Barth (University of Ulm, Ulm, Germany) and was used at 300 ng/ml for 180 min. Y27632 (Calbiochem) was used at 30 μM, and p38 MAPK inhibitor SB 202190 (Calbiochem) was applied at 10 μM. The specific inhibitor of PKC, chelerythrine (Sigma-Aldrich), was used at 10 μM.
Sera from two PF patients and two patients suffering from a mucocutaneous form of PV whose diagnosis was confirmed clinically, histologically, and serologically and from a volunteer without any skin disease (control) were used for this study. Patients' sera were tested by ELISA for reactivity against Dsg 1 and 3, respectively. All sera contained Dsg 1 antibodies, whereas Dsg 3 antibodies were only present in PV-IgG fractions. Purification was performed as described previously (). Concentrations of all IgG fractions were adjusted to a 150-μg/ml final concentration for all experiments.
The model is similar to the technique established previously (; ). Skin pieces were taken from fresh cadavers of individuals not suffering from any skin disease who had donated their bodies to the Institute of Anatomy and Cell Biology of Würzburg. Viability of the epidermis after incubation in DME at 37°C in the absence and presence of PV-IgG was confirmed by incubation with 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl-tetrazolim bromide (MTT; Sigma-Aldrich) for 30 min, which was converted to a colored formazan by mitochondrial dehydrogenases using heat-inactivated epidermis (65°C for 30 min) as a negative control (). Specimens were incubated with DME containing 10% FCS and 1.8 mM Ca for 24 h in the presence or absence of PV- or PF-IgG alone or in combination with 1,200 ng/ml CNF-1 or CNFy. After brief rinsing with PBS (consisting of 137 mM NaCl, 2.7 mM KCl, 8.1 mM NaHPO, and 1.5 mM KHPO, pH 7.4), skin specimens were mounted on copper plates using Reichert-Jung mounting medium (Cambridge Instruments) and frozen in liquid nitrogen. 5-μm–thick cryosections were obtained using a cryostat (Reichert-Jung 2800 Frigocut; Cambridge Instruments). Cryosections were stained with hematoxylin and eosin for histomorphological evaluation. For each condition, three to five pieces of skin (2 × 2 mm) from at least two different cadavers were used and incubated in the presence or absence of patients' IgG separately. From three to five skin pieces, 24–111 different sections were evaluated under the various conditions (numbers of sections evaluated for each condition are given in the figure legends). After each section harvested, at least 50 μm of tissue were discarded. In the next section, it was verified by microscopic evaluation that no epidermal splitting was found to ensure that each blister measured was counted not more than once. For each blister, the total split length was measured and normalized to the length of the specific section given in millimeters.
After incubation with autoantibodies, immunostaining was performed as described previously (). After fixation with ice-cold acetone for 2 min and incubation with normal goat serum and 1% BSA at RT, monolayers were incubated for 16 h at 4°C with mouse monoclonal antibody directed to Dsg 3 (Zytomed), rabbit polyclonal to desmoplakin (Biozol), mouse monoclonal to cytokeratin 14 (Chemicon), or mouse monoclonal to E-cadherin (BD Transduction Laboratories; each dilution was 1:100 in PBS). After several rinses with PBS (three times for 5 min each), monolayers were incubated for 60 min at RT with Cy3-labeled secondary antibodies (Dianova). For the visualization of F-actin, Alexa-phalloidin (Invitrogen) diluted 1:60 in PBS was used (incubation for 1 h at RT). Cells incubated with antibodies or Alexa-phalloidin were rinsed with PBS (three times for 5 min each).
The expression and purification of recombinant Dsg 1, coating of polystyrene beads, and the laser tweezer setup were described previously (). Throughout all experiments, the laser intensity was 42 mW in the focal plane. Coated beads (10 μl of stock solution) were suspended in 200 μl of culture medium and allowed to interact with HaCaT monolayers for 30 min at 37°C before measuring the number of bound beads (equal to control values). Beads were considered tightly bound when resisting laser displacement at a 42-mW setting. For every condition, 100 beads were counted. Afterward, control IgG, PV-IgG, or PF-IgG were applied under various conditions for 120 min, and the number of bound beads was counted again. The percentage of beads resisting laser displacement under various experimental conditions was normalized to control values.
To test the effect of pemphigus IgG on the activation of Rho proteins, HaCaT cells were treated with either PF-IgG, PV-IgG, CNF-1, CNFy, or SB 202190 alone or in combinations as indicated before cells were subjected to activation assay kits for Rho A (Upstate Biotechnology), which were applied according to the manufacturer's instructions. The technique was used as described previously ().
Extraction, Western blotting, and immunodetection were essentially performed as described previously () with the following exceptions. HaCaT cells were grown to full confluence in six-well dishes. After incubation with pemphigus IgG in the absence or presence of CNFy, monolayers were incubated with extraction buffer (1% Triton X-100, 10 mM Tris, 140 mM NaCl, 5 mM EDTA, and 2 mM EGTA) containing protease inhibitor solution (20 μg/ml each of leupetin, pepstatin, and aprotinin and 1 mM PMSF; all were obtained from Sigma-Aldrich) for 10 min on ice. Cells were scraped, vortexed briefly, and centrifuged at 14,000 for 30 min. The supernatant was collected and defined as the Triton-soluble fraction. The pellet was defined as the cytoskeletal fraction and dissolved immediately in 10% SDS-containing sample buffer at 95°C for 5 min. Finally, immunodetection was performed using mouse monoclonal Dsg 3 antibody (Zytomed) diluted 1: 1,000 in PBS.
The histology and immunofluorescent images were taken with a microscope (LSM 510; Carl Zeiss MicroImaging, Inc.) using 40× NA 1.3 oil plan Neofluar and 63× NA 1.4 oil plan Apochromat objectives (Carl Zeiss MicroImaging, Inc.). Images were taken by RT in antifade as an imaging medium for immunofluorescence and 80% glycerol for hematoxylin and eosin staining. The fluorochromes used were Cy3 and Alexa488. Densitometric quantification of Western blot bands was performed using Photoshop 7.0 software (Adobe). Photoshop 7.0 software was also used for the adjustment of contrast and brightness.
Differences in bead adhesion and epidermal splitting were assessed using the test. In text and bar diagrams, values were expressed as the mean ± SEM. Statistical significance was assumed for P < 0.05. |
Tight coordination of the cellular events associated with every phase of the cell cycle is essential for orderly progression through the cell cycle (). Cells arrest the cell cycle when confronted with stresses that may compromise cellular integrity. The mechanisms that halt the cell cycle are referred to as checkpoints (). Checkpoints result in key dependency relationships; mutants that fail to complete DNA replication also block nuclear division and cytokinesis, and mutants that fail to complete nuclear division also block cytokinesis (). Cells arrest the cell cycle in response to DNA damage. The DNA damage checkpoint was originally defined as the pathway that promotes cell cycle delay in response to DNA damage (), although it is now generally accepted that checkpoint responses involve additional processes, such as the activation and recruitment of DNA repair factors () and the stabilization of replication forks (). At least two checkpoints operate during S phase in ; the replication checkpoint, which was originally defined as causing hydroxyurea (HU)-induced cell cycle arrest and inhibition of late-firing origins (), and the intra–S phase checkpoint, which reduces the rate of DNA replication and slows cell cycle progression in response to DNA-damaging agents (). A series of checkpoints also act to monitor assembly of the mitotic spindle and regulate progression through mitosis (). More recently, additional checkpoints have been identified, including a cell wall morphology checkpoint and a morphogenesis checkpoint, which monitor cell wall synthesis, bud formation, cell size, actin perturbation and, possibly, septin organization (; ; ).
DNA replication stress and DNA damage induce activation of two phosphoinositide 3-kinase–related kinases, Tel1 and Mec1, which are similar to mammalian ataxia telangiectasia mutated and ataxia telangiectasia mutated and Rad3-related. These function in the activation of downstream protein kinases, including Chk1 and Rad53. Activation of Rad53 ( Chk2) is mediated by two partially redundant adaptor proteins Mrc1 and Rad9 (). Although in and higher eukaryotes cell cycle progression is blocked in response to replication stress, mainly by stimulating inhibitory phosphorylation of cyclin-dependent kinases (CDKs), in inhibition of Cdk1 ( CDK) activity does not appear to play a role in S phase checkpoint-induced cell cycle arrest (; ). Rather, blocks cell cycle progression by directly inhibiting late origin firing and chromosome segregation (; ). Cdk1 in appears to have taken on a different function, and it synchronizes bud morphogenesis with the cell cycle (). In , Cdk1 activity is determined by various factors, including Swe1 ( equivalent of mammalian and Wee1), which phosphorylates the conserved Y19 residue on Cdk1, resulting in inhibition of Cdk1. This phosphorylation is removed by the phosphatase Mih1. Swe1 levels are controlled by a pathway often referred to as the morphogenesis checkpoint (), which responds to the state of the actin cytoskeleton, bud formation, and cell size, resulting in accumulation of Swe1, thereby delaying entry into mitosis (; ). Degradation of Swe1 depends on various factors, including septins. Septins serve as a scaffold to recruit Swe1 and various kinases that phosphorylate Swe1, ultimately targeting it for destruction by an as yet unknown Skp1–Cul1–F-box complex and, possibly, the anaphase-promoting complex (; ). A defect in Swe1 degradation results in prolonged inhibition of Cdk1. As a consequence, Cdk1 cannot induce the switch from polar to isotropic bud growth, resulting in the formation of elongated buds (,).
Thus far, studies on the S-phase checkpoints have mainly focused on regulation of DNA replication, replication fork stabilization, and chromosome segregation. We show a novel role for components of the replication stress checkpoints in control of morphogenesis during replication stress. Our results are consistent with a model in which checkpoint proteins promote timely degradation of Swe1, thereby restricting bud growth during replication stress.
We observed that checkpoint mutants frequently have morphological aberrations. For instance, double mutant cells are commonly misshapen, have somewhat elongated buds, and often fail to complete cytokinesis (). and Δ single mutants did not have morphological aberrations (unpublished data), whereas mutants lacking both Mrc1 and Rad9, which function upstream of Rad53 (), as well as Δ mutants had a phenotype similar to that of mutants (). Deletion of , which acts downstream of Rad53, resulted in a milder phenotype, whereas and mutants did not have a morphology defect (unpublished data). In addition, and Δ double mutants, and Δ single mutants, frequently deposited abnormally large amounts of chitin (), not only at the bud neck but often at other apparently random sites of the cell wall (unpublished data), suggesting a possible defect in orchestrating cell wall architecture. Because cells with defective cell walls lyse in the presence of SDS or Calcofluor white (; ), we tested the sensitivity of various mutants to these chemicals. , , and single mutants were no more sensitive to Calcofluor white and SDS than wild-type cells, whereas Δ mutants were one to two orders of magnitude more sensitive (), which is indicative of a defective cell wall architecture. double mutants showed severely reduced viability on yeast extract/ peptone/dextrose (YPD), probably because these cells suffer major endogenous DNA damage from lack of DNA repair and checkpoint functions (), and these mutants were sensitive to Calcofluor white (Fig. S1 A, available at ). However, because of the dramatic growth defects of double mutants, we decided to focus on Δ and Δ mutants instead. Δ mutants were as sensitive to Calcofluor white as Δ mutants (Fig. S1 A), whereas and mutants were only weakly sensitive to Calcofluor white (Fig. S1 A). Therefore, Rad53 seems to be a critical mediator of resistance to Calcofluor white, whereas there appear to be redundancies between Mec1 and Tel1, between Rad9 and Mrc1, and, potentially, between Chk1 and Dun1 (not tested), paralleling the roles of these proteins in checkpoints (). Cells with cell wall defects are usually more sensitive to zymolase (). Untreated wild-type and Δ mutants did not lyse when incubated with a hypotonic buffer containing zymolase (). However, pretreatment with HU increased sensitivity of Δ mutants to zymolase, whereas wild-type cells remained unaffected. Cell lysis typically occurred at the bud tips (unpublished data), indicating that the replication checkpoint promotes formation of a healthy bud during replication stress (see below). Cell wall stress activates the Pkc1–Slt2 pathway (), and, consistent with our findings that checkpoint mutants have cell wall defects, Slt2 is hyperphosphorylated in and Δ mutants (). These results suggest that and Δ double mutants, and Δ single mutants, have considerable defects in control of cell morphology and cell wall structure. Finally, we noticed that in Δ mutants bud scars are frequently positioned at distal poles or at random positions on the cell (), indicating that Δ mutants may have a bud site–selection defect because wild-type haploid cells have an axial budding pattern, forming buds at just one pole (; ). As shown in , the majority of wild-type cells had an axial budding pattern. However, ∼45% of log phase Δ mutants had bud scars that deviated from that pole. Cells expressing a kinase-dead mutant of (; K227A, D319A, and D339A; see Smolka et al. on p. ▪▪▪ of this issue), which results in a checkpoint deficiency similar to that of Δ mutants, had a bud site selection defect similar to that of a Δ mutant (), indicating that the kinase function of Rad53 is essential for correct bud site selection. Furthermore, whereas , , and mutants had very mild or no bud site selection defects, double mutants had a bud site selection defect similar to that of Δ mutants, whereas and mutants had a more intermediate phenotype. mutants were examined as a control and were found to have a 100% random distribution of bud scars (unpublished data). In conclusion, we found that various checkpoint proteins contribute to cell wall architecture and maintenance of cell polarity.
Both bud site selection and cell wall synthesis are controlled by factors that regulate the actin cytoskeleton (). To test whether S-phase checkpoint proteins function in regulation of the actin cytoskeleton, we treated log-phase wild-type cells with HU for 4 h and visualized F-actin using rhodamine-phalloidin. Untreated wild-type, , Δ, and Δ mutants had a polarized actin cytoskeleton, with actin cables extending from the mother cell into the bud (). Treatment of wild-type cells with HU arrested the cells with large buds and a depolarized actin cytoskeleton (). In contrast, the actin cytoskeleton of both Δ and Δ mutant cells remained polarized upon HU treatment (). Incubation of Δ mutants at 42°C for 5 min resulted in complete actin depolarization, showing that Δ mutants did not have a general defect in stress responses (unpublished data). HU also failed to induce actin depolarization in mutants. mutants, as well as mutants (a DNA helicase that is thought to play roles in both the replication and intra–S phase checkpoints; ), also failed to fully depolarize the actin cytoskeleton. and mutants were similar to wild-type cells, and in mutants, HU treatment only partially depolarized the actin cytoskeleton (, Fig. S1 B, and not depicted). Together, these results suggest that checkpoint proteins like Sgs1, Dun1, Rad53, and a combination of Mrc1 and Rad9 affect the polarity state of the actin cytoskeleton upon induction of replication stress.
Cytoskeletal polarity is guided by the Cdc24–Cdc42 pathway (). We found that in wild-type cells, treatment with HU resulted in the disappearance of Cdc24 from the bud tip membrane (), as might be expected from cells that have arrested with large buds. However, in Δ and single mutants, and in Δ double mutants, Cdc24 remained at the membrane (), whereas mutants were similar to wild-type cells. Similar results were obtained with cells expressing Sec4-GFP, a Rab GTPase that is an essential component of the secretory machinery found primarily at sites of polarized growth (; Fig. S1 C). Therefore, these results indicate that checkpoint proteins, including Rad53, Dun1, and Mrc1, in combination with Rad9, promote removal of the bud growth machinery from the bud tip upon treatment with HU.
Mutations in genes involved in regulation of the actin cytoskeleton often render cells sensitive to pharmacological actin inhibitors like latrunculin A (). Therefore, we speculated that Δ mutants might also be sensitive to latrunculin A. As shown in , Δ mutants were more sensitive to latrunculin A than wild-type cells, whereas , , and single mutants were no more sensitive to latrunculin A than wild-type cells ( and not depicted). Interestingly, a deletion suppressed the sensitivity of Δ mutants to latrunculin A (), and we found that Rad53 may control Swe1 (see the following section).
In addition to failure to depolarize the actin cytoskeleton, treatment of Δ, but not wild-type, cells with HU for extended periods of time (16–20 h, although visible after 6 h [see Smolka et al. on p. ▪▪▪ of this issue], after which >80% of the cells are still alive, as indicated by staining with vital dyes; Fig. S2 A, available at ) resulted in formation of elongated buds (). Analysis of W303 and W303 Δ strains revealed a similar effect of a Δ mutation (unpublished data). Formation of elongated buds is a common feature of mutants that fail to properly control levels of Swe1 (). This, and the finding that deletion of suppressed the sensitivity of Δ mutants to latrunculin A, raised the possibility that the elongated bud phenotype of HU-treated Δ mutants is caused by failure to down-regulate Swe1. Indeed, deletion of completely rescued the HU-induced elongated bud phenotype of the Δ mutant (). Similar results were obtained with cells harboring the allele ( and Fig. S2 B). Deletion of also caused cells to form elongated buds after treatment with HU (), and deletion of in a Δ background resulted in an augmented phenotype ().
We next tested the effect of a range of mutations in genes encoding checkpoint functions on bud morphogenesis during replication stress. , , , Δ, and single mutations all caused no, or a very small, increase in HU-induced elongated bud growth. In contrast, Δ and double mutants, the latter of which express an allele of Mrc1 that is proficient in DNA replication, but unable to activate the replication checkpoint (), showed elongated bud growth upon HU treatment. Furthermore, double mutants also showed an increase in elongated bud growth; however, we noticed that cells that attempted to elongate their buds frequently lysed, possibly because of their severe cell growth and cell wall defects, and this may obscure the phenotype. The , , , , , and mutations alone did not increase HU-induced elongated bud growth (). In contrast, cells harboring the allele displayed considerable HU-induced bud elongation, which is consistent with the function of the replication protein A complex in the DNA replication stress response. mutants also showed HU-induced elongated bud growth, which was -dependent, and the extent of elongation of these buds was often severe (Fig. S2 B). The double mutant had a more severe HU-induced elongated bud phenotype compared with the respective single mutants, similar to a previous report on the effects of and mutations on the sensitivity to HU (). Other combinations of mutations did not reveal significant genetic interactions (). Finally, mutations in , , and , which are the three components of the MRX complex, resulted in HU-induced elongated bud growth. These results identify Sgs1, the MRX complex, replication protein A, Mec1 in combination with Tel1, Mrc1 in combination with Rad9, Rad53, and Dun1 as important mediators in controlling bud morphology during replication stress.
The aforementioned results indicate that a pathway requiring Rad53 may control Cdk1 activity during S phase, when most bud growth takes place. Because different checkpoint mutants undergo HU-induced, Swe1-dependent elongated bud growth, we analyzed lysates of HU-treated cells by Western blotting using Swe1 antibodies (). Treatment of wild-type cells with HU for 1 or 2 h resulted in accumulation of moderately phosphorylated Swe1, which then became hyperphosphorylated after 3 h, ultimately resulting in its destruction (4 h), which is consistent with the results of of Liu and Wang (). In contrast, in HU-treated Δ mutants, only moderately phosphorylated species of Swe1 accumulated, and Swe1 largely failed to get degraded (); this was not caused by cell cycle effects, as the budding index of Δ mutants was similar to that of wild-type cells and cells did not reenter the cell cycle (Fig. S3 A, available at ). Importantly, this resulted in hyperphosphorylation of Y19 of Cdk1 (), which, as expected, was Swe1-dependent (unpublished data). A similar effect of a Δ mutation was seen with W303 and W303 Δ strains (unpublished data). When analyzed as a control, Y19 of Cdk1 also became hyperphosphorylated in and mutants, which are known to be defective in degradation of Swe1 (Fig. S3 B). Furthermore, high levels of Swe1 accumulated in Δ mutants (), which is consistent with the fact that Rad9 and Mrc1 are upstream regulators of Rad53. Swe1 levels were also elevated in mutants (), which is consistent with the known role of Cla4 in phosphorylating Swe1 to target it for destruction.
For Swe1 to be hyperphosphorylated it needs to localize to the bud neck (), which is dependent on various factors, including septins. We could not detect endogenous GFP-tagged Swe1 (unpublished data), but, when overexpressed, we could detect bud neck–localized Swe1-GFP in 50–60% of the cells (), in accordance with a previous study (). Importantly, we found that Swe1 localization in Δ mutants was similar to that of wild-type cells, even after treatment with HU (unpublished data), showing that the defect in Swe1 degradation does not result from failure to localize Swe1 to the bud neck. Altogether, these data show that a Rad53-dependent pathway restricts bud growth when cells are confronted with DNA replication stress by controlling Swe1-Cdk1.
Various pathways are known to control the actin cytoskeleton and bud morphogenesis (,). Because we found that checkpoint proteins control bud morphology during replication stress, one would predict that defects in known pathways that regulate bud morphology might also result in HU-induced elongated bud growth. Based on knowledge from the literature (,), we tested several candidates in an attempt to identify the pathways that may be used by the replication stress checkpoints to ensure proper bud growth during replication stress. The results of this analysis, described in this section, show that regulation of bud morphology is an essential part of the response to DNA replication stress, and that cells use a network of diverse pathways to ensure proper bud morphology when confronted with replication stress.
The p21-activated kinase–like kinase Cla4 is a key factor in regulation of the actin cytoskeleton, and is a central component of the Cdc42 pathway. We not only found that the Cdc24–Cdc42 pathway may not be properly down-regulated in several checkpoint mutants after HU treatment () but also that mutants accumulate Swe1 during replication stress (). Therefore, we tested whether mutants might be defective in HU-induced actin depolarization. Indeed, mutants failed to depolarize the actin cytoskeleton and dramatically elongated their buds upon HU treatment (; quantified in ), indicating that Cla4 is necessary for regulation of the actin cytoskeleton when cells are confronted with replication stress. Interestingly, also genetically interacts with (see below). Bem1 is another component of the Cdc24–Cdc42 pathway. We found that mutants, like mutants, failed to depolarize actin upon HU treatment, resulting in elongated bud growth. We also tested strains lacking GTPase-activating proteins for the small Ras-like GTPases Cdc42 and Rho, which function in septin organization and bud growth (). Cells lacking the Bem2 and Bem3 failed to depolarize the actin cytoskeleton and elongated their buds after HU treatment, whereas and mutants responded to HU like wild-type cells (). This supports our finding that proper regulation of the Cdc42–Cla4 pathway is important for control of the actin cytoskeleton and cell morphology in response to HU. These results also suggest a possible involvement of specific -controlled Rho pathways.
The kinases Elm1 and Gin4 function in septin ring assembly (). The septin ring itself serves as a scaffold to bind various proteins, including Hsl1, Gin4, Elm1, and Hsl7. Hsl7 is an adaptor protein that helps recruit Swe1, bringing it into proximity of several kinases, including Hsl1, Clb2-Cdk1, Cla4, and Cdc5, which phosphorylate Swe1 to target it for degradation (). Consistent with a role of Swe1 in regulating bud growth in response to HU, , , , and mutants failed to depolarize their actin cytoskeleton and formed hyperpolarized buds after treatment with HU (, and not depicted).
We also found that cells lacking a variety of cell cycle regulators, such as the transcription factors Swi4, Swi6, and Fkh2, the F-box protein Grr1 ( and not depicted), and the mitotic exit–regulating proteins Kel2 and Dbf2, hyperpolarized bud growth upon HU treatment. Swi4 and Swi6 are involved in transcriptional control of various genes, including Hsl1, Kcc4, and Gin4 (; ; ), which might explain why and mutants elongate their buds upon HU treatment. It is not clear why mutants elongate their buds after treatment with HU, but it was previously found that a deletion synthetically interacts with , indicating a role for Fkh2 in the Cdc42–Cla4 pathway (). Kel2, Dbf2, and Grr1 regulate mitotic exit and actomyosin ring contraction. We are currently investigating why HU induces elongated bud growth in cells lacking these factors.
Bni1 is a member of the polarisome and functions together with Cla4 in septin organization. mutants did not develop elongated buds when challenged with HU, but instead appeared to undergo an extra round of cell growth, resulting in formation of “strings” of cells (graphically displayed in and quantified in ). We also noticed that , , and mutants form such strings of cells, whereas deletion of any of the Rho–GTPase-activating proteins or the septin-regulating kinases tested in our study had no such effect. Thus, Bni1, Dbf2, Swi4, and Swi6 also appear to assure that cells maintain proper morphology when challenged with prolonged replication stress.
If regulation of actin, septins, and cell morphology is part of the response to replication stress, one would predict that cells defective in these pathways are more sensitive to HU. Indeed, chronic treatment of strains lacking Cla4, Elm1, Hsl1, Hsl7, and Gin4 with HU caused reduced growth compared with wild-type cells (). Analysis of the colonies that formed on HU plates showed that these mutants formed elongated buds and clusters of cells with multiple unseparated cell bodies (unpublished data), indicating that cell proliferation may have slowed because of morphological aberrations rather than an inability to activate the replication checkpoint. All of these mutants, including , recovered normally from a 4-h HU arrest, indicating that all mutants are checkpoint proficient, whereas additional deletion of in these mutants resulted in a complete loss of viability (Fig. S4 A, available at ). mutants also recovered normally from HU arrest, whereas the Δ double mutant did not recover, indicating that our finding that a Δ mutation suppresses some of the morphological defects of a Δ mutant is unrelated to an effect on checkpoint regulation of cell cycle arrest. These results suggest that septin/Swe1-regulating factors are unlikely to be directly involved in checkpoint control of cell cycle arrest, but instead have a different function in the cellular response to replication stress, such as ensuring proper bud growth.
has been found to genetically interact with a variety of DNA replication, repair, and checkpoint factors (Fig. S4 B; ; ). In line with this, we found that a Δ mutation caused synthetic fitness defects when combined with mutations in , , and () and caused temperature-sensitive growth when combined with a mutation in (). These results indicate that Rad53 may cooperate with the pathway involving Cla4, Elm1, Gin4, and Shs1. Therefore, we tested whether Rad53 is involved in septin organization. We studied septin ring organization by immunofluorescence rather than expression of Cdc12-GFP because we found that Cdc12-GFP weakly interfered with septin organization (unpublished data). Interestingly, whereas Δ mutants had no noticeable septin defect, deletion of augmented the septin defects of , , and mutants (; see Fig. S4 C for examples of septin phenotypes). These data indicate that Rad53 has a redundant function in the organization of septin rings.
Finally, we observed that a Δ deletion, as well as mutations in upstream regulators of Rad53, often lead to formation of small aggregates of cells (), indicating that Rad53 may also be involved in regulation of cytokinesis. Therefore, we treated cells with zymolase to digest the cell wall, which results in removal of the buds of cells that have successfully completed cytokinesis, but preserves the buds of cells that have not completed cytokinesis, as these cells contain buds that are still connected through the plasma membrane. We found that log-phase cultures of wild-type cells contained ∼40% of cells with a single bud and 7% of cells with more than one bud after zymolase treatment (). In contrast, zymolase treatment of Δ mutants, as well as Δ mutants, left 36 and 41% of cells with two or more buds, respectively, indicating that these cells failed to complete cytokinesis (). Treatment of cells with HU modestly increased the number of multibudded cells in Δ mutants and did not increase the number of multibudded cells in wild-type and Δ cultures. In addition, mutants also had cytokinesis defects (unpublished data). Based on these results, we conclude that Rad53 kinase activity is required for the successful completion of cytokinesis and separation of mother and daughter cells.
During replication stress, cells of most wild-type strains (see second to last paragraph of this section) arrest with spherically shaped buds approximately the same size as the mother cell, indicating that mechanisms exist to monitor and control cell growth and morphology when DNA replication has halted. We observed that checkpoint mutants were frequently misshapen, had defective cell walls, and displayed hyperphosphorylation of Slt2. This is in line with a recent report showing genetic interactions between Rad9 and Slt2 (). Checkpoint proteins were also found to support proper bud site selection, indicating an involvement in maintenance of cell polarity and control of the actin cytoskeleton (,). Indeed, we found that checkpoint proteins, including Rad53, are involved in the removal of Cdc24 and Sec4 from the bud tip during replication stress, resulting in depolarization of the actin cytoskeleton, and that failure to do so resulted in the formation of elongated buds. We identified several additional pathways involved in preventing elongated bud growth during replication stress, and at least some of these pathways (e.g., the Cla4 pathway) genetically interact with the Rad53 pathway. Our results indicate that checkpoints and additional pathways cooperate to support cell morphology and cytokinesis during chronic replication stress.
To identify the checkpoint pathways that restrict bud growth during replication stress, we tested mutants lacking a wide range of checkpoint factors. Mutations in and , but not in other genes (i.e., , or ) thought to encode upstream-acting checkpoint factors, caused a significant defect in control of cell morphology during replication stress (). Although a single mutant had no defects, double mutants had a more severe phenotype than single mutants, in line with previous reports showing that double mutants have increased sensitivity to HU and methylmethanesulfonate (). Consistent with this, Rad53 and its downstream-acting factor Dun1 were involved in controlling cell morphology during replication stress. Mec1 is generally thought to function upstream of Rad53 in checkpoint signaling. Although and single mutants did not have a strong morphological phenotype, double mutants had morphological aberrations that were comparable to those of Δ and mutants. This fits with the view that Mec1 and Tel1 can have redundant functions, and is reminiscent of the Mre11 complex–mediated S phase checkpoint response that is Tel1-dependent, but is predominantly seen in the absence of Mec1 (). Indeed, we found that the MRX complex is important for maintaining proper morphology in response to HU treatment. In addition, we found that Mrc1 and Rad9 together were required for the morphological response to replication stress, which is consistent with the finding that these proteins are redundant activators of Rad53 (). Defects in Chk1 did not cause morphological defects, which is not surprising because mutations do not cause sensitivity to HU (). Finally, mutations in and did not cause morphology defects, suggesting that there is redundancy among these genes or that they act in different aspects of checkpoint responses. Overall, the results presented in this study link replication stress checkpoint functions to the control of cell morphology.
A recent study has shown that and Δ mutants show premature spindle elongation during HU-induced replication stress because of failure to inhibit the microtubule-associated proteins Cin8 and Stu2, resulting in precocious chromosome segregation (). This raises the possibility that the defects in regulation of morphology and spindle dynamics in checkpoint mutants may be related. Premature spindle elongation is not sufficient for HU-induced bud elongation in checkpoint mutants because mutants, which elongate their spindles prematurely (), do not show HU-induced elongated bud growth. Furthermore, nocodazole treatment, which inhibits spindle elongation, did not block HU-induced elongated bud growth in Δ mutants (Fig. S4 D), indicating that premature spindle elongation is not required for elongated bud growth induced by HU treatment of checkpoint mutants.
In contrast to and higher eukaryotes, does not target Cdk1 to block cell cycle progression during replication stress, but instead directly targets processes such as late origin firing and chromosome segregation (; ). Rather, Cdk1 has taken on a different role; it is essential for the morphogenetic switch from polar to isotropic bud growth, and mutants that fail to degrade Swe1 have a delayed morphogenetic switch, resulting in elongated bud growth. We found that a Rad53-dependent pathway is important for timely degradation of Swe1 during replication stress, and showed genetically that it is this failure of checkpoint mutants to degrade Swe1 that results in elongated bud growth. However, it is currently unknown how Rad53 controls Swe1. Swe1 recruitment to septin rings was unaffected in Δ mutants, which is consistent with our finding that septin organization is normal in these mutants. In , the Rad53 homologue Cds1 may directly phosphorylate Wee1 to control its activity (), raising the possibility that in Rad53 also directly phosphorylates Swe1. However, it is also possible that Rad53 indirectly controls Swe1 levels. For instance, Rad53 may target kinases that in turn directly phosphorylate Swe1, like Cdc5 and Cla4 (). Indeed, Rad53 has been suggested to control Cdc5 activity ().
We found that genetically interacts with a variety of upstream regulators of septins, as well as with the nonessential septin . Furthermore, replication stress checkpoint mutants showed a cytokinesis defect, which is often linked to septin malfunction. Therefore, we investigated the possibility that Rad53 may affect septin organization. We found that Rad53 by itself is not essential for the organization of septin rings, but has a redundant role with upstream septin regulators like Cla4, Gin4, and Elm1 in septin organization and cell growth. Consistent with this, defects in septin-regulating factors caused sensitivity to HU and HU-induced bud elongation. Although the molecular nature of the Rad53 interactions with septins is not yet clear, interestingly, Rad53 can directly phosphorylate Shs1 in vitro, and a pool of Rad53 may localize to the septin ring in vivo (see Smolka et al. on p. ▪▪▪ of this issue). It is possible that defects in this cooperation could explain the deregulation of Swe1 that occurs in checkpoint mutants in response to HU. Further studies are needed to unravel the function of Rad53 in control of septin organization.
Bud morphogenesis during replication stress has been studied previously. Several studies have shown that replication stress induced filamentous differentiation strongly in one wild-type strain (Σ1278b), weakly in two wild-type strains (W303 and A374), and not at all in another wild-type strain (S288c; ; ). Induction of this phenotype in Σ1278b cells appeared to require Mec1, Rad53, and Swe1, but not Sgs1 or Dun1. The HU-induced bud elongation phenotype we have studied does not occur in our wild-type strain (S288c) and is induced in various checkpoint mutants, and thus appears to be a different phenotype. To better understand the differences between wild-type strains, we tested seven different laboratory wild-type strains collected over the years and identified one strain showing a strong bud elongation phenotype (DBY745), two strains showing a weaker phenotype (W303 and W303 ), and five strains (MGD, BY4741, L2955, Y55, and JKM139) that did not show HU-induced bud elongation. Because DBY745 was derived by crossing markers into S288c (unpublished data, Botstein, D., personal communication) and JKM139 was derived by crossing markers from Y55 into DBY745 (unpublished data, Haber, J., personal communication), it seems likely that the bud elongation resulted from a mutation introduced during the construction of DBY745. Similarly, the bud elongation phenotype of the W303 strains may have resulted from the introduction of a mutation during the intercrossing of the S288c derivatives used to construct W303 (for detailed information on W303 see the Genome Database; ). Further analysis will be required to understand the exact genetic basis for the replication stress–induced bud elongation and filamentous differentiation that some wild-type strains show.
In conclusion, we found that replication stress checkpoint proteins like Rad53 function together with additional pathways to promote the timely degradation of Swe1, thereby relieving inhibition of Cdk1. Cdk1 then induces the switch from polar to isotropic bud growth, thus preventing elongated bud growth and contributing to cell viability because such elongated buds are susceptible to cell wall stress. Therefore, replication stress checkpoint–mediated control of bud morphology is part of the response to replication stress and contributes to cell survival.
strains were grown in standard YPD medium or synthetic complete medium lacking the appropriate amino acid. Strains were directly derived from the S288c strain RDKY3615 using either standard gene replacement methods or intercrossing (Table S1, available at ). To construct cells harboring the allele, a cassette was inserted into a PacI site 250 bp downstream of the allele in plasmid pAO138 (gift from S. Elledge, Harvard Medical School, Boston, MA). Subsequently, was PCR amplified and used to replace a cassette previously inserted at the locus. All other strains were obtained from the systematic deletion project (BY4741; derived from the same parental strains as RDKY3615, i.e., S288c) and were only used when specifically stated in the figure legends. Plasmids pRS414- and pRS415- were provided by M. Peter (Swiss Federal Institute of Technology, Zurich, Switzerland), and pRS414- was obtained from S. Emr (University of California, San Diego, La Jolla, CA). The YepT- plasmid was obtained from K. Lee (National Institutes of Health, Bethesda, MD).
Live cells (expressing either Cdc24-GFP, Sec4-GFP, or Swe1-GFP) were imaged at room temperature in synthetic complete medium with an inverted microscope (Eclipse TE300; Nikon) equipped with a 100×/1.40 NA Plan Apo objective lens (Nikon), using a charge-coupled device camera (Orca-ER; Hamamatsu) and MetaMorph software (Universal Imaging Corp.), and images were processed using Photoshop and Illustrator software (both Adobe). Alternatively, cells were fixed with 3.7% formaldehyde and either stained with 50 μg/ml Calcofluor white or rhodamine-phalloidin according to manufacturer's instructions (Invitrogen) or processed for septin immunofluorescence, as previously described (). Rabbit anti-Cdc11p antibodies (Santa Cruz Biotechnology, Inc.) were used at 1:10 dilution, followed by TRITC–conjugated goat anti–rabbit secondary antibody at a 1:50 dilution (Jackson ImmunoResearch Laboratories). Fixed cells were embedded in Vectashield HardSet mounting medium (Vector Laboratories) to reduce photobleaching. At least 100 cells were counted per strain and per treatment. For analysis of bud scar positioning, only cells with at least three bud scars were counted. Scars were scored as “normal” when all scars on a cell were located at the same pole. When one or more scars deviated from that pole, bud scar positioning was scored as “abnormal.”
Log-phase cells were treated with HU as indicated, pelleted, and resuspended in hot (95°C) Laemmli sample buffer supplemented with protease inhibitors and boiled for 5 min, after which glass beads were added and cells were vortexed for 15 min at 4°C. After centrifugation, soluble proteins were analyzed by SDS-PAGE and Western blot using phospho-Cdc2 antibodies (New England Biolabs) to analyze Cdk1 phosphorylation. Total Cdk1 levels were determined using PSTAIRE antibodies (Millipore). Antibodies against Swe1 were provided by D. Kellogg (University of California, Santa Cruz, Santa Cruz, CA). Phosphorylated Slt2 was analyzed with phospho-specific p42/44 MAPK antibodies (Cell Signaling Technology), and pan-Slt2 was detected using Slt2 antibodies (Santa Cruz Biotechnology, Inc.). Detection was performed using the SuperSignal West Femto Detection kit according to the manufacturer's instructions (Pierce Chemical Co.).
10 ml of cells were grown to OD of 1 and quickly washed with 15 ml HO. Cells were then resuspended in 10 ml hypotonic buffer (10 mM Hepes, pH 7.5) supplemented with 20 U/ml lytic enzyme (Puregene) and incubated at 30°C with occasional agitation. OD was measured every 15 min.
100 μl of 1–2 M HU in HO was added to 1 ml log-phase cells (in YPD), followed by 8–16 h at 30°C while shaking. The fraction of cells with elongated buds was determined as the percentage of large-budded cells only (unbudded and small-budded cells were ignored). A bud was scored as “elongated” when the length of the bud was at least twice its width.
1 ml of log-phase cells was washed with PBS and incubated at 30°C in 25 mM Hepes, pH 7.5, containing 1 M sorbitol (for osmotic support), and 5 mg/ml zymolase. After 30 min, cells with no buds, one bud, or two or more buds were counted. At least 100 cells were counted per treatment.
Halo assays were performed as previously described (). In brief, 250 μl of a log-phase YPD culture was diluted in 3.5 ml YPD, after which 3.5 ml molten Agarose (1% wt/vol) in YPD (cooled to ∼50°C) was added. The cell suspension was then poured onto a YPD plate. Paper disks (6 mm diam; Becton Dickinson) were placed on top of the plates, and either 10 μl latrunculin A or 10 μl DMSO were spotted onto the center of the disks. Plates were inverted and incubated at 30°C until halos were visible.
Overnight cultures (prestationary phase) were diluted in 50 ml fresh, prewarmed YPD to an OD of 0.05. Cells were then incubated at 30°C while shaking. 1-ml samples were taken every hour for 12 h, and doubling times were determined using Excel (Microsoft).
Table S1 shows S288c-derived strains used in the described studies. Fig. S1 shows analysis of the sensitivity of various S phase checkpoint mutants to Calcofluor white, images of the actin cytoskeleton in , , , and mutants upon treatment with HU, and the failure of Δ mutants to remove Sec4 from the bud tip during replication stress. Fig. S2 shows analysis of metabolic activity of HU-treated wild-type and Δ mutants during prolonged replication stress, images of -dependent HU-induced bud elongation in and cells. Fig. S3 shows an analysis of the budding index of wild-type, Δ, Δ , and mutants that shows that none of these mutants reenter the cell cycle during HU arrest, and a Western blot demonstrating hyperphosphorylation of Y19 of Cdk1 during HU arrest in and mutants. Fig. S4 shows that upstream regulators of septins are replication checkpoint proficient, extensive genetic interactions between and DNA replication, repair, checkpoint genes, examples of various septin phenotypes, and that HU-induced bud elongation in Δ mutants is not blocked by nocodazole. Online supplemental material is available at . |
The DNA damage response consists of a complex protein network that mediates the detection of damaged DNA and regulates multiple cellular processes (; ; ; ; ). In S, an evolutionarily conserved kinase cascade consisting of Mec1, Tel1, Rad53, and Dun1 is responsible for amplifying the DNA damage signal from DNA damage recognition enzymes and transducing such signal to downstream targets in the form of protein phosphorylation (; ). Although Mec1 and Tel1 are involved in sensing DNA damage (), Rad53 and Dun1 appear to function as effector kinases to regulate multiple cellular processes, such as the cell cycle, DNA replication, chromosome segregation, histone turnover, gene transcription, and possibly DNA repair (; de la Torre ; ; ; ; ). Rad53 is critical for cells to cope with various DNA damage stresses, as cells with Rad53 deletion or kinase-dead mutations are hypersensitive to genotoxic stresses (; ; ). In response to DNA damage or stalled replication forks, Rad53 is hyperphosphorylated and activated in a Mec1-dependent manner (; ; ). The activation of Rad53 is accompanied by its autophosphorylation, induced by its association with the hyperphosphorylated forms of the adaptor proteins Rad9 or Mrc1 (; ; ). Rad53 appears to be directly involved in the regulation of cell cycle, as its overexpression leads to cell cycle arrest even in the absence of exogenous DNA damage stresses (). Despite its importance, the targets of Rad53 in the DNA damage response are poorly known.
Rad53 consists of a central serine/threonine kinase domain, flanked by an N-terminal Forkhead associated 1 (FHA1) domain and a C-terminal FHA2 domain (). The FHA domain is found in a wide range of signaling proteins, with known roles in mediating protein–protein interactions through the binding of phosphorylated substrates (; ). DNA damage–induced interaction between Rad53 and hyperphosphorylated Rad9 is mediated by both the FHA1 and -2 domains of Rad53. Although differences in the binding specificity of both domains were found using an oriented phosphopeptide library approach (), the presence of either domain alone is sufficient for Rad53 activation (; ). Additionally, the FHA domains likely mediate the targeting of Rad53 to other proteins for their regulation by Rad53. For example, chromatin assembly protein Asf1 and phosphatase Ptc2 were shown to bind to the FHA1 domain of Rad53 in a phosphorylation-dependent manner (; ; ; ). Several other proteins have been shown to be regulated in a Rad53-dependent manner. Swi6 is an essential regulatory subunit of two different START-dependent transcription factors, Swi4 and Mbp1, which regulate the transcription of many genes involved in DNA replication (). Swi6 appears to undergo Rad53-dependent phosphorylation in response to DNA damage stresses and was suggested to be a substrate of Rad53 (). Dun1 was also shown to be phosphorylated and activated by Rad53 (). Furthermore, several other proteins were found to interact with Rad53, including Cdc7/Dbf4 (), Kap95 (), Yta7 (), Mdt1 (), and others (). Despite these studies, the mechanism by which Rad53 regulates cell growth in response to DNA damage and replication stress remains poorly understood.
The cell cycle of the budding yeast consists of highly coordinated events, including bud emergence, polarized cell growth, and protein trafficking, which are synchronized with the initiation and progression of DNA replication. After successful completion of DNA replication, mitosis and cytokinesis occur, and a new round of cell cycle begins (). At the initiation of S-phase, bud emergence is accompanied by the localization of bud site selection proteins and septins (; ). Septins are a family of conserved proteins that form filaments at the cortex of the mother-bud neck (). Localization of septins to the bud neck persists throughout the cell cycle except for disassembly and reassembly during G1 phase (). In budding yeast, genes encoding septins, i.e., Cdc3, Cdc10, Cdc11, and Cdc12, were identified through the isolation of temperature-sensitive mutations that prevented cytokinesis at restrictive temperature, resulting in the formation of chains of multinucleated and multibudded cells (). More recently, a fifth member of the septin family was identified, namely, Shs1 (seventh homologue of septin) (). Septins are known to perform important functions in spindle orientation, bud-site selection, the establishment and maintenance of polarized bud growth, the switch from polarized to isotropic bud growth, cell cycle, and morphogenesis checkpoints (; ; ; ; ). Increasingly, the regulatory role of septins in coordinating multiple steps in cell cycle progression is being revealed (; ; ).
In this study, we performed a proteomic screen to identify proteins that bind to the FHA domains of Rad53. This led to the finding that the FHA1 domain of Rad53 coordinates the interaction between Rad53 and a wide variety of proteins. In contrast, the FHA2 domain appears to have a rather specialized role in binding to Rad9 after DNA damage treatment. Among the FHA1 binding proteins, septins and their associated proteins represent a major functional group. We present evidence that Rad53 may play a role in the regulation of polarized cell growth in response to DNA replication stress.
A proteomic approach was developed to identify Rad53 FHA binding proteins present in yeast cell extracts. Various N-terminal (; ) fusion FHA domains bound to the IgG resin were used in the affinity purification. As shown in , the FHA binding proteins were purified using a tandem affinity purification (TAP) method (see Materials and methods). As a control, the same purification was performed using a mutant FHA protein carrying a point mutation previously shown to reduce its binding to phosphorylated substrates (R70A for FHA1 and R605A for FHA2; ). shows the purified FHA binding proteins. Although contaminant proteins, i.e., bands common in both wild type (WT) and mutant FHA purification, are still present, bands specific to the WT FHA1 can be seen. To identify the specific binding proteins of WT FHA domain, we used stable isotope labeling–based quantitative mass spectrometry, as illustrated in . Each purified sample was independently digested with trypsin and then labeled with a stable isotope containing N-isotag reagent (light N-isotag for WT sample and heavy N-isotag for mutant sample; see Materials and methods; ). The labeled samples were combined and analyzed by mass spectrometry for identification and quantification of proteins (). Proteins identified in the WT FHA purification, but not the mutant FHA purification, as determined by their isotope labeling, were then considered specific FHA binding proteins for further validation studies.
In a separate experiment, to identify DNA damage–induced changes in the FHA binding proteins, WT -FHA1 (or -FHA2) resin was used to purify proteins from untreated cells and methyl methanesulfonate (MMS)–treated cells (the eluted proteins had similar patterns like and are not depicted). The purified proteins from untreated cells were digested by trypsin and labeled with the light N-isotag, and those from MMS-treated cells were labeled with the heavy N-isotag. Quantitative mass spectrometry was again used to identify any MMS-induced changes in the specific binding proteins of the FHA domains.
The specific binding proteins of FHA domains of Rad53 are summarized in . Although most of them interact with the FHA1 domain of Rad53 independent of DNA damage, Rad9 and Mrc1 interact with the FHA1 domain only after MMS treatment. Rad9 was also found to bind to the FHA2 domain of Rad53 after MMS treatment, consistent with previous findings (; ). Additionally, Asf1 and Ptc2, both known FHA1 binding proteins (; ; ), were identified. The identification of these known FHA binding proteins validates the approach. Interestingly, Swi6 and its associated proteins, Swi4, Mbp1, and Whi5, were found to bind to the FHA1 domain of Rad53, further supporting a previously identified link between Rad53 and Swi6 (). Because the purification was performed under nondenaturing conditions, it is not surprising that protein complexes were purified and identified.
Although the FHA2 domain was found to bind only Rad9, the FHA1 domain of Rad53 binds to a wide variety of proteins, most of which are novel (). Interestingly, several cytosolic proteins were found, including the septins (Cdc10, Cdc11, Cdc12, Cdc3, and Shs1) and proteins involved in bud site selection (Bud3, Bud4, and Bud14), all of which localize to the bud neck. Based on the number of identified peptides for each protein identified (), a crude indicator of protein abundance, we deduce that the septins are among the more abundant FHA1 binding proteins. To confirm the specificity of FHA1 binding, protein extracts of strains containing TAP- or HA-tagged genes of interest were analyzed by affinity purification with either WT or mutant GST-FHA proteins immobilized on glutathione resins, and the bound protein was analyzed by Western blotting. confirms that in MMS-treated cells, Rad9 binds to both the FHA1 and -2 domains, whereas Mrc1 binds to only the FHA1 domain. The specificity of binding between septins and the FHA1 domain was similarly confirmed by Western blot analysis (). Furthermore, as shown in , deletion of genes encoding any of the septin-associated proteins Bud3, Bud4, or Bud14, as well as deletion of , did not affect the binding of the FHA1 domain to Cdc11. Deletion of , which results in cells with abnormal cell morphology (not depicted), was found to impair the ability of FHA1 domain to bind Cdc11 (). Collectively, the results show that the FHA1 domain most likely binds to septins directly in a Cdc10-dependent manner.
Binding specificities of most other FHA1 binding proteins have also been confirmed using the same approach, and the results are shown in Fig. S1 (available at ). In all cases, the R70A mutation, which is known to diminish the binding of the FHA1 domain to phosphopeptides (), greatly reduces the binding of the identified FHA1 binding proteins.
The finding that the septin complex binds to the FHA1 domain was unexpected because Rad53 is a known nuclear protein and septins are known to localize at the bud neck. We next asked whether a localization of Rad53 at the bud neck could be detected. As shown in , the majority of GFP-tagged WT Rad53 is nuclear, although some heterogeneous GFP signal was detected in the cytoplasm upon close inspection. It is difficult to observe the localization of WT Rad53 at the bud neck, especially in late S-phase, when the GFP signal from nuclear Rad53 is within close proximity to the bud neck and overwhelms any possible signal from the bud neck. To better visualize the localization of Rad53 outside of the nucleus, we generated a Rad53 mutant in which a putative nuclear localization signal (NLS) in the C-terminal region of Rad53 was removed (; see Materials and methods). Results in confirm the loss of predominant nuclear localization of the Rad53 NLS truncation mutant. In unsynchronized cells, the Rad53 NLS mutant was detected in the bud neck of a small percentage of the cells (∼8%; ). Interestingly, all of the cells showing positive Rad53 localization to the bud neck had large buds and a still undivided nucleus. We then examined the effect of DNA damage on the localization of the Rad53 NLS mutant. Exposure to MMS or hydroxyurea (HU) drastically increases the percentage of cells showing the localization of Rad53 to the bud neck, again in cells with large buds (). Closer inspection shows that the GFP signal of the Rad53 NLS mutant appears as a double ring–like pattern (, enlarged panel). Importantly, a Rad53 NLS mutant containing an additional FHA1 domain R70A mutation fails to localize to the septins, further indicating that the localization of Rad53 to bud neck is mediated through its FHA1 domain. We also analyzed cells at different times after the release from α-factor arrest and did not observe the same striking GFP signal at the septin ring as observed after HU or MMS treatment (unpublished data). Importantly, Rad53-GFP cells exhibited the same sensitivities to HU, MMS, or UV, compared with WT cells (unpublished data). shows the abundance levels of various GFP-tagged Rad53 mutant proteins, indicating that the lack of bud neck localization of Rad53 mutant containing both NLS truncation and R70A mutations was not due to its reduced protein abundance. Although Rad53-GFP still underwent MMS-induced hyperphosphorylation and slower gel shift, truncation of Rad53 NLS largely abolished its MMS-induced hyperphosphorylation (). Therefore, the majority of Rad53 needs to be nuclear for its proper activation.
Because the FHA1 domain of Rad53 interacts with many proteins, we asked what the role of the FHA1 domain of Rad53 might be. We examined the effect caused by overexpression of the FHA domains of Rad53 in yeast cells. Overexpression of the FHA1 domain was found to result in aberrant cell morphology, showing multiple elongated buds that fail to detach from the mother cell, even after zymolase treatment (). In contrast, overexpression of the R70A mutant FHA1 domain, or WT and the R605A mutant FHA2 domain, does not induce any such defect. shows the staining of a cell with overexpression of the FHA1 domain. It appears to be multinucleated and lacks any staining for septin Cdc11, suggesting that septin organization is disrupted by FHA1 overexpression.
We next asked whether any morphological defect could be observed in Δ cells. No major morphological defect was detected in the untreated Δ cells, although a closer analysis of Δ cells did reveal various defects in budding pattern and cell wall integrity (see Enserink et al. on p. ▪▪▪ of this issue). Upon HU treatment, a striking morphological defect characterized by an elongated bud was readily observed for Δ cells (). Such phenotype was observed in ∼30% of Δ and -KD cells after 14 h of chronic HU treatment. Together, the results suggest that Rad53 plays a role in the control of polarized cell growth in response to HU treatment, and its kinase activity appears to be essential for such regulation.
The morphology defect caused by overexpression of the FHA1 domain was found to be largely suppressed by deletion of Shs1 (), suggesting that may somehow function downstream of the FHA1-mediated binding of Rad53 to the septins. We then examined whether the Δ cells have any morphological defect in response to HU treatment. Interestingly, HU treatment of Δ cells induces multiple elongated buds that fail to detach (), even after zymolase treatment. Interestingly, Shs1 was efficiently phosphorylated by Rad53 in vitro (), raising the possibility that it may be a Rad53 substrate. Next, we asked whether Δ cells are hypersensitive to chronic treatment with HU. As shown in , although the Δ cells are not sensitive to chronic treatment with HU at 30°C, they are hypersensitive to HU treatment at 37°C. The Δ cells are almost as sensitive as cells, and such loss of viability at 37°C is specific to HU treatment, because Δ cells are not sensitive at 37°C in the absence of HU, or even in the presence of MMS (). Collectively, these results suggest that Shs1 may play an important role during the response to DNA replication stress, although the precise nature of such role is unknown.
#text
Strains used in this work are listed in Table S1 (available at ). All strains of TAP-tagged proteins were obtained from the Open Biosystems collection. For epitope tagging of Rad53, the RDKY5763 cells (ura3-52, leu2Δ1, trp1Δ63, his3Δ200, lys2ΔBgl, hom3-10, ade2Δ1, ade8, YEL069C∷URA3, and sml1∷TRP1) were used. For deletion or epitope tagging of Bud3, Bud4, Bud14, Cdc10, and Shs1, BY4741 and RDKY5763 cells were used. To generate endogenous C-terminal–tagged Cdc10-3xHA, Rad9-3xHA, and Rad53-GFP, we used standard homologous recombination technique and the pFA6a plasmids (a gift from M. Longtine, Oklahoma State University, Stillwater, OK; ). To make GFP-tagged Rad53 NLS truncation mutant, GFP was fused to the C-terminal end of endogenous Rad53 with concurrent removal of the amino acid 781 to the C terminus. To generate a kinase-dead mutant of Rad53, Rad53 was first cloned into pFA6a and mutated and then the mutant Rad53 was reintroduced back into the endogenous locus in cells. Correct integration and mutations were all verified by DNA sequencing.
Plasmids used in this work are listed in Table S2 (available at ). For overexpression studies, WT and mutant FHA1 (amino acid residues 2– 279) and FHA2 (amino acid residues 523–821) domains of Rad53 were cloned into pYES2/NT-C vector (Invitrogen) using BamHI and NotI restriction sites. For pull-down assays, the same FHA domains were subcloned into pGEX-4T1. To make the tag, a sequence containing the protein A and TEV cleavage site was first amplified from the plasmid pREP1 NT (a gift from K. Gould, Vanderbilt University School of Medicine, Nashville, TN; ) using a primer containing a 6xHis tag sequence and then inserted into the pET21a plasmid (Novagen) using NdeI and BamHI, resulting in the plasmid. Different FHA domains were then subcloned into the plasmid using the same restriction sites. Mutant FHA was generated using site-directed mutagenesis. In each case, the sequence is confirmed by DNA sequencing.
The -FHA1 (amino acid residues 2–279) and -FHA2 (amino acid residues 523–821) domains of Rad53 and their respective mutants (R70A and R605A) were expressed in BL21 cells and induced by 0.1 mM IPTG at 30°C for 3 h. Cells were harvested and lysed by sonication in TBS-N (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.2% NP-40) with protease inhibitors (5 mM EDTA, 1 mM PMSF, 0.2 mM benzamidine, 1 μM leupeptin, and 1.5 μM pepstatin). Cell debris was removed by centrifugation at 30,000 for 30 min. The cleared cell extract was incubated with IgG-Sepharose (GE Healthcare) for 2 h and washed extensively by TBS-N. Purity and abundance of bound -FHA proteins on the IgG resin were examined by either TEV protease cleavage or boiling in SDS sample buffer and subsequent SDS-PAGE analysis. Typically, the procedure resulted in a concentration of 5 μg FHA protein per μl of IgG resin. To make GST fusion proteins, GST-FHA proteins or GST-tagged septins were similarly expressed in BL21 cells. Glutathione resin was used for their purification according to manufacturer's instruction (GE Healthcare).
2 liters of yeast cells (BY4741) were grown in YPD medium to an OD of 1.5. Approximately 10 g of cells were broken in an ice-cooled bead beater with 40 ml lysis buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.2% NP-40, 0.5 mM DTT, 5 mM NaF, 10 mM β-glycerolphosphate, 1 mM sodium vanadate, 5 mM EDTA, 1 mM PMSF, 0.2 mM benzamidine, 1 μM leupeptin, and 1.5 μM pepstatin. Cell debris was removed by centrifugation at 30,000 for 30 min. Protein extract was divided into two equal fractions, each incubated with 0.1 ml of WT or the mutant -FHA containing IgG resin overnight at 4°C. The resins were then washed with 20 ml of lysis buffer and resuspended in 1.5 ml of lysis buffer without EDTA. The FHA domain was cleaved off by adding 100 units of TEV protease (Invitrogen) for 2 h at room temperature. Supernatant containing 6xHis-tagged FHA protein and its interacting proteins was collected and was further incubated with 0.1 ml of Ni-NTA resin (QIAGEN) for 1 h at room temperature. The Ni-NTA resin was washed with 10 ml of lysis buffer and then with 5 ml of TBS buffer. To elute the FHA binding proteins (but not the FHA domain), the Ni-NTA resin was incubated for 5 min at 80°C in 400 μl of an elution buffer containing 8 M urea, 100 mM Tris-HCl, pH 8.0, and 500 mM NaCl. Eluted proteins were reduced and alkylated before trypsin digestion, as described previously (). For MMS treatment, 0.05% of MMS was added to cells for 2 h before harvesting.
To confirm the binding specificity of FHA binding proteins, 50 ml of yeast cells containing a TAP- or HA-tagged gene of interest was grown in YPD to an OD of 1.0. Cells were harvested and broken by vortexing with glass beads. The cleared cell extract was then incubated with the same amount of WT or mutant GST-FHA proteins bound to glutathione resin. After binding, the resins were washed by 4 × 1 ml of TBS-N, boiled in SDS sample buffer, and subjected to Western blot analysis using anti-TAP antibody (Open Biosystems) or anti-HA antibody (Roche Applied Science).
Recombinant 6xHis-Rad53 was used to phosphorylate septins in vitro (). Approximately 5–10 μg of GST-fused Cdc3, Cdc10, Cdc11, Cdc12, or Shs1 were incubated with 100 ng of Rad53 in 20 μl of kinase buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.2% NP-40, 0.2 mM ATP, and 10 μC ATP) for 40 min at 30°C. After phosphorylation, the samples were boiled in SDS sample buffer with DTT for 5 min, and 10% of the sample was analyzed by SDS-PAGE and subjected to autoradiography. Plasmids encoding GST fusion of septins were a gift from J. Thorner (University of California, Berkeley, Berkeley, CA).
For labeling of the amino groups of peptides, we used an isotope-coded leucine-containing N-isotag reagents. The d0-tBoc-Leu-NHS reagent was purchased from Nova Biochem. The d10-tBoc-Leu-NHS was synthesized using the method described previously (). The light N-isotag, i.e., d0-tBoc-Leu-NHS, was used to label WT FHA binding proteins, whereas the d10-tBoc-leu-NHS was used to label mutant FHA binding proteins. The labeling procedure was performed as described previously ().
Samples were analyzed by μLC-ESI-MS/MS on a quadrupole ion trap mass spectrometer (Finnigan LCQ; Thermo Electron Corporation) as previously described (). For data analysis, SEQUEST was used for peptide identification, and the XPRESS and INTERACT software were used for quantitation as described previously (). The complete yeast database was used to analyze MS/MS spectra with no restriction on the enzyme used, and a variable modification of serine and threonine residues by phosphorylation was included in the database search. Only the top-matched, doubly tryptic peptides with high-quality MS/MS spectra were subsequently subjected to manual inspection. All proteins listed in were identified with at least two different peptides, both containing the light N-isotag labeling reagent, which was used to label WT FHA binding proteins.
For confocal imaging, yeast were grown to log phase (OD ∼0.1) and treated with various agents before imaging. Imaging was done on a spinning disk confocal (McBain Instruments) mounted live in minimal media on an inverted microscope (TE2000e; Nikon). Images were acquired using a 60× 1.4 NA Plan Apo objective lens with 1.5× auxiliary magnification (90× magnification total) using a charge-coupled device camera (Orca ER; Hamamatsu) with 2 × 2 binning. Acquisition parameters, shutters, and focus were controlled by MetaMorph software (Universal Imaging). All fluorescence images were acquired using the same settings, including laser intensity. Each fluorescence image presented is a maximum intensity projection of a z-series stack through the entire yeast cell (5–8 μm), whereas a single differential interference contrast image was acquired at the midpoint of the z stack. All imaging was conducted at room temperature (∼23°C).
Immunofluorescence samples were processed and mounted as previously described (), and images were acquired using an upright microscope (E800; Nikon) with conventional multimode optics. Differential interference contrast, DAPI, FITC, and Rhodamine images were acquired using a 60× 1.4 NA Plan Apo objective lens and MetaMorph software controlling the Orca ER charge-coupled device camera with 1 × 1 binning. Each fluorescence image presented is a maximum intensity projection of a z-series stack through the entire yeast cell (5–8 μm), whereas a single differential interference contrast image was acquired at the midpoint of the z stack. All imaging was conducted at room temperature (∼23°C). Images were processed in MetaMorph for brightness and contrast and minimal gamma adjustments.
Fig. S1 shows confirmation of the FHA1 binding proteins. Table S1 lists yeast strains used in this study. Table S2 lists plasmids used in this study. Online supplemental material is available at . |
Eukaryotic cells exhibit exquisite control over their architecture; this is critical for cell motility, division, and polarity. Networks of signaling proteins determine cell morphology by orchestrating cytoskeleton organization, membrane trafficking, and gene expression. Fundamentally important pathways are conserved between metazoans and the budding yeast (for review see ). In budding yeast, as in other eukaryotes, cell morphology is closely coupled with cell cycle progress. Bud emergence and growth in late G1 require establishment and maintenance of polarized growth through regulated organization of actin cytoskeleton assembly and membrane traffic. As cells pass into G2/M, buds depolarize to switch from apical to isotropic growth. The cytoskeleton is further reorganized during cytokinesis, when an actomyosin ring forms and contracts to separate mother and daughter cell cytoplasm. During cytokinesis, a septum is deposited between mother and daughter cells; this is destroyed a few minutes later, when cytokinesis is complete, resulting in mother/daughter separation. Temporal coordination of these events requires conserved signaling pathways but remains incompletely understood.
The control of cell morphology is also important for the determination of cell fate. Asymmetric segregation of molecules or structures that influence gene expression links a cell's differentiation status to its underlying structure. Budding yeast cells exhibit cell fate asymmetry that involves segregation of determinants to the daughter cell. The transcription factor Ace2 accumulates specifically in the daughter cell nucleus, where it induces expression of chitinases and glucanases required for septum destruction (; ; ; ). Ace2 asymmetry is also responsible for daughter-specific delay of G1 progression through an as-yet-unknown mechanism ().
The budding yeast regulation of Ace2 and morphogenesis (RAM) network is a recently discovered signaling pathway that controls cell fate asymmetry and polarized growth. Components of this pathway are conserved in a broad range of eukaryotes and are generally involved in the control of cell architecture (; ; ; ; ). The yeast network comprises six genes: , , , , , and (). Cells lacking any of these proteins exhibit two phenotypes: a failure to degrade the septum between mother and daughter, resulting in large groups of connected cells, and poor maintenance of polarized growth. The cell separation defect results from the mislocalization of Ace2 to both mother and daughter nuclei, resulting in the loss of Ace2-dependent transcription (; ). However, defective polarized growth is not attributable to loss of Ace2 function: cells lacking Ace2 can maintain polarized growth (). Therefore, the RAM network has separate roles in regulation of Ace2 and control of polarized growth.
The localization of RAM network proteins over the cell cycle reflects their dual roles. The proteins concentrate at sites of cell growth, such as the bud tip and the cortex of the expanding bud, and redistribute to the bud neck late in cell division, during telophase (; ; ). In addition to cell cortex localization, the Ndr/Lats family protein kinase Cbk1 and its conserved binding partner Mob2 localize to the daughter cell nucleus at the M–G1 transition. This localization requires Ace2, and nuclear localization of Ace2 is similarly dependent on Cbk1 and Mob2 (; ). Cbk1 and Mob2 lack canonical nuclear localization sequences, suggesting that the proteins associate with Ace2 and enter the nucleus as a complex.
Cbk1 kinase activity is essential for RAM network function. How is the enzyme controlled? Although Cbk1 protein levels remain constant, kinase activity fluctuates over the cell cycle with maximal specific activity during early bud growth and late mitosis (). The kinase's activity is low if other RAM network genes are deleted, suggesting that it functions downstream of the other components of the pathway ().
In addition, Cbk1 has two putative regulatory phosphorylation sites that are conserved among AGC group kinases (for review see ). Analogous sites are important for the in vivo function of Cbk1-related kinases in budding yeast (such as Dbf2), (such as and ), and mammals (such as Ndr and Lats; ; ; ; ; ; ). One site, found in the kinase domain activation loop, is likely autophosphorylated (). The other site, located in a hydrophobic motif immediately C-terminal to the kinase domain, is phosphorylated by an upstream kinase. Germinal center kinases, which are related to Ste20 and other p21-activated kinases, phosphorylate this site in several cases (; ; ). The role of phosphorylation at each site has been examined in vitro for the mammalian Ndr proteins. In , these phosphorylation sites are essential for function in vivo (; ). These findings are consistent with a prevailing model for the regulatory function of phosphorylation at these sites: they increase the kinase's specific activity. This view is supported primarily by enzymatic and structural studies of Akt/protein kinase B (PKB), an AGC family member (,; ).
We analyzed the in vivo function of these phosphorylation sites in Cbk1 and found that both were phosphorylated but that their function cannot be explained by simple regulation of the enzyme's catalytic activity. Mutation of one of these sites, which is in Cbk1's activation loop (“T-loop site”), severely compromised in vitro kinase activity but yielded only an intermediate phenotype for cell separation and polarized growth. In contrast, mutation of the C-terminal hydrophobic motif site (“CT-motif”) produced a protein with substantial kinase activity but an entirely null phenotype. Using phosphospecific antibodies, we found that both sites were regulated over the cell cycle; the CT-motif site's phosphorylation exhibited dramatic fluctuation, peaking before bud emergence and during cytokinesis. All RAM network components were essential for CT-motif modification; a subset was essential for T-loop phosphorylation, which we find is an intramolecular reaction in vivo. Cbk1 is extensively phosphorylated at additional sites, and dephosphorylation of these sites requires CT-motif phosphorylation. Interestingly, full phosphorylation of the C-terminal hydrophobic motif required Ace2, a downstream effector of the RAM network; thus, the regulatory target modulates activation of its upstream regulator. This study of Cbk1 phosphoregulation defines a novel role for CT-motif phosphorylation independent of kinase activation. It is this molecular event that bridges RAM network signaling with its phenotypic outputs of cell separation and polarized growth.
Ndr/Lats family kinases share two conserved phosphorylation sites that are important for kinase function (; ; ; ). One site is found in the T-loop of the kinase domain (serine 570), and the other lies in a short hydrophobic region C-terminal to the kinase domain, the CT-motif (threonine 743). To determine whether the putative phosphorylation sites are important for Cbk1's in vivo function, we mutated these sites to the nonphosphorylatable amino acid alanine and characterized the resulting phenotypes. We made these mutations, as well as substitution of an aspartic acid essential for catalytic activity, in HA-tagged Cbk1 expressed under the control of its endogenous promoter at the native chromosomal locus. We refer to the -HA-S570A construct as the T-loop allele, to the -HA-T743A construct as the CT-motif allele, and to the catalytically inactive -HA-D475A construct as the kinase-dead allele. These proteins are expressed at similar levels (see and ). Cells lacking functional Cbk1 have two distinct phenotypes, which have been reported previously and are shown in (; ; ; ; ). These cells fail to degrade the septum between mother and daughter cells after cytokinesis and, as a result, grow as clumps of >100 cells connected by chitin-rich junctions. In addition, cells carrying the kinase-dead allele or cells are spherical instead of ovoid because they cannot maintain polarized growth ().
The T-loop and CT-motif mutant alleles showed both phenotypes but to different degrees. Like cells expressing kinase-dead Cbk1, cells expressing the T-loop allele or CT-motif allele proteins grew as groups of cells connected by chitin (). We quantified the cell separation defect by counting the number of connected cells per cluster (). The separation defects of the CT-motif allele, kinase-dead allele, and were nearly indistinguishable (49.7 ± 12.1%, 63.3 ± 13.6%, and 64.0 ± 12.3% of cells in clumps containing >30 connected cells; , middle and right). In contrast, the T-loop mutant had an intermediate phenotype with most cells in clusters of <12, suggesting that phosphorylation of this site is only partially required for cell separation (, middle).
Cells expressing either the T-loop or CT-motif alleles were noticeably rounder than wild-type strains, suggesting that phosphorylation of these sites is required for maintenance of polarized growth (). To quantify this phenotype in mitotically growing cells, we determined the mean axial ratio by measuring the length and width of >200 cells for each strain in three independent trials (). This ratio measures the roundness of a cell, with a perfect circle equal to 1.0 and greater numbers representing deviation toward an ellipse. Wild-type and cells had a broad distribution of axial ratios, with means of 1.24 ± 0.13 and 1.21 ± 0.12, respectively. In contrast, the mean axial ratios of cells carrying the kinase-dead or -null alleles were 1.07 ± 0.06 and 1.07 ± 0.07, respectively. Cells with the CT-motif allele have a mean axial ratio of 1.09 ± 0.07, which is similar to both the kinase-dead and null mutants. In contrast, cells carrying the T-loop allele had an intermediate mean axial ratio of 1.13 ± 0.10, which is significantly different from both wild-type and null phenotypes, as well as the CT-motif mutant phenotype (using a nonparametric Mann-Whitney test, P < 0.0001).
Maintenance of polarized growth can also be assessed by observing mating projection formation in response to pheromone. Cells expressing Cbk1-HA formed mating projections normally, with characteristic chitin deposits at the base of an elongated projection and actin polarized into the tip (, top). In contrast, cells expressing the T-loop, CT-motif, or kinase-dead alleles formed abortive mating projections, similar to cells, which are unable to sustain polarized growth (; ). Abortive mating projections are characterized by slight bumps flanked by chitin deposits that lack the elongated projections and pronounced actin polarization of normal mating projections. Collectively, our observations suggest that phosphorylation of the T-loop and CT-motif sites play important yet distinct roles in cell separation and maintenance of polarized growth in vivo.
The phenotype of kinase-dead Cbk1 shows that the protein's enzymatic activity is critical for its function in both cell separation and polarized growth (), consistent with previous findings (). Therefore, the phenotypes of the nonphosphorylatable T-loop and CT-motif alleles could be due to reduction of normal Cbk1 kinase activity. To assess this, we characterized the relative specific activity of Cbk1 alleles in vitro, using quantitative detection with fluorescently labeled secondary antibodies to measure relative amounts of the immunoprecipitated enzyme. We evaluated autophosphorylation and histone H1 phosphorylation as previously described () and examined phosphorylation of a bacterially produced fragment of the transcription factor Ace2, a likely Cbk1 regulatory target. Surprisingly, wild-type and CT-motif mutant Cbk1 phosphorylated these substrates with comparable efficiency. In contrast, the T-loop mutant protein exhibits minimal kinase activity, only slightly higher than that of the kinase-dead allele (). Cbk1 function and kinase activity requires association with Mob2 (; ). We therefore tested all of the Cbk1 mutants and found that none were defective for Mob2 association (Fig. S1, available at ). In sum, these results show that the phenotypic differences between the T-loop and CT-motif alleles is not attributable to differences in kinase activity.
When expressed under the control of its endogenous promoter, Cbk1 localizes to sites of active cell growth at the cortex (; ; ). To determine whether the phenotypes of the T-loop, CT-motif, and kinase-dead mutant alleles were attributable to mislocalization, we visualized GFP-tagged versions of these proteins in live cells. All mutant Cbk1-GFP fusion proteins were present at the cell cortex of presumptive bud sites and the growing bud, like wild type (, left and middle). The mutant proteins also accumulated normally at the bud neck during cell division (, right). Thus, phosphorylation at the T-loop and CT-motif sites are not required for Cbk1's cortical localization.
Cbk1 also localizes to the daughter cell nucleus at the M–G1 transition (; ). Some nuclear localization of T-loop mutant was evident (, A [arrows] and B [nucleus delineated by localization of the ER marker Sec63-RFP and Cbk1-GFP nuclear localization indicated by arrows]). In contrast, we could not detect either the kinase-dead or the CT-motif mutant GFP fusion proteins in nuclei.
The nuclear accumulation of Cbk1 both coincides with and depends on Ace2 nuclear localization; in turn, Ace2 asymmetric localization requires Cbk1 kinase activity (; ). Thus, it is possible that cells carrying the CT-motif and kinase-dead alleles fail to recruit Ace2 to nuclei. Therefore, we examined Ace2-GFP localization in each of the mutant backgrounds (). We briefly incubated cells with rhodamine–concanavalin A, which binds stably to the cell wall, and allowed the cells to grow for ∼90 min in the absence of this vital stain. This specifically labeled mother cells and allowed us to distinguish between mother and daughter cells in the mutant strains, which grow as large clusters of connected cells. In wild-type cells, Ace2-GFP was distributed asymmetrically to the daughter cell nucleus, with relatively little cytoplasmic background in ∼90% of cells (). In the T-loop mutant, Ace2-GFP localized to daughter nuclei in ∼80% of cells. In contrast, in CT-motif mutant cells, Ace2-GFP localized faintly to both mother and daughter nuclei in ∼80% of cells (). This loss of Ace2-GFP asymmetry is similar to that seen in kinase-dead and cells. Ace2 expression levels do not vary appreciably in the mutant backgrounds (Fig. S2, available at ). Thus, both CT-motif phosphorylation and kinase activity are required for the proper and asymmetric distribution of Cbk1 and Ace2 to the daughter cell nucleus.
To determine whether Cbk1 is phosphorylated at the T-loop and CT-motif sites in vivo, we generated antibodies against the relevant phosphopeptides for each site. These antibodies strongly detected immunoprecipitated Cbk1-HA but only weakly recognized Cbk1-HA treated with λ phosphatase (). The phosphospecific antibodies are not general phosphoserine or phosphothreonine antibodies, as they did not recognize other highly phosphorylated proteins, such as GST-Ste20 (Fig. S3, available at ) or the mutant alleles for each site (). Thus, Cbk1 is phosphorylated in vivo at both the T-loop and CT-motif sites.
To determine whether phosphorylation of Cbk1's T-loop and CT-motif are independent regulatory inputs, we used the phosphospecific antibodies to detect modification of each site in the absence of the other's phosphorylation (). The T-loop mutant protein was detected by anti-pT743, the CT-motif phosphospecific antibody. The CT-motif mutant protein was detected by anti-pS570, the T-loop phosphospecific antibody, although the signal was decreased about fourfold relative to wild type. This suggests that each phosphorylation event can occur in the absence of the other, although CT-motif modification may promote maximal T-loop phosphorylation in vivo.
Kinase-dead Cbk1 was phosphorylated at the CT-motif site but not at the T-loop site, suggesting that the T-loop site is autophosphorylated in vivo (). This is consistent with the model proposed for other kinases related to Cbk1, such as hNDR (nuclear Dbf2-related kinase; ). Autophosphorylation may occur in cis as an intramolecular reaction or in trans as an intermolecular reaction between Cbk1 molecules. We constructed a diploid strain that expresses both wild-type Cbk1-GFP and the kinase-dead allele of Cbk1-HA to distinguish between these possibilities. As shown in , immunoprecipitated kinase-dead protein is not phosphorylated at the T-loop site by wild-type Cbk1 in vivo. Therefore, it is likely that Cbk1 modifies itself in vivo in cis; this is in contrast to the CT-motif site, which must be phosphorylated by a different kinase.
Cbk1 works with other RAM network components to control cell separation and polarized growth. Cbk1 kinase activity is low in cells lacking any other RAM network component, suggesting that activation of Cbk1 is among the final outputs of the pathway (; ). To determine whether the RAM network is required for phosphorylation of Cbk1's T-loop and CT-motif sites, we used phosphospecific antibodies in quantitative Western blotting to evaluate the modification of Cbk1-HA immunoprecipitated from cells lacking different RAM network genes (). We were able to detect a very low level of phosphorylation at the T-loop site using the anti-pS570 antibody in and strains (∼4 and ∼11% of wild-type phosphorylation). T-loop phosphorylation is detectable on Cbk1-HA immunoprecipitated from cells (∼12% of wild type) but is biased toward forms shifted to higher molecular weights. Phosphorylation of the T-loop site was absent in and backgrounds. The CT-motif–specific anti-pT743 antibody recognized only Cbk1-HA immunoprecipitated from wild-type cells and not from any of the strains lacking RAM network components. Therefore, all RAM network proteins are essential for CT-motif phosphorylation and only a subset is essential for phosphorylation of the T-loop site.
Cbk1 kinase activity is not reduced in cells, suggesting that Ace2 is an endpoint effector of the pathway; our data also support this (). In addition, Ace2 is not required for maintenance of polarized growth (; ; ). We used phosphospecific antibodies to detect Cbk1-HA immunoprecipitated from cells to determine whether Ace2 is important for phosphorylation at either the T-loop or CT-motif sites. T-loop phosphorylation is similar in wild-type and backgrounds (). Surprisingly, however, we found substantial reduction of Cbk1 CT-motif phosphorylation in cells (∼43% of wild type, measured by quantitative Western blotting with fluorescent secondary; see Materials and methods). RAM network genes were expressed at similar levels in wild-type and cells, as measured by RT-PCR (Fig. S4, available at ), indicating that loss of CT-motif modification is not an indirect effect of low RAM network gene expression. Therefore, full modification of Cbk1's critically important CT-motif site is reinforced or protected by the presence of this downstream effector.
Cbk1 functions in both polarized growth and cell separation, two temporally regulated processes. To assess the cell cycle–dependent control of phosphorylation at the T-loop and the CT-motif sites, we used the phosphospecific antibodies to monitor modification of these sites in synchronized cells (). T-loop phosphorylation fluctuated only slightly over the course of cell division: phosphorylation levels were generally high but diminished as cells entered S phase (). Before bud emergence, slower migrating forms of Cbk1 were highly phosphorylated at the T-loop site (). CT-motif phosphorylation is more dramatically regulated in a cell cycle–dependent manner (). Phosphorylation of this site peaked early, 45 min after release from G1 arrest, just before bud emergence. Phosphorylation then remained low until late in the cell cycle, as cytokinesis occurred. When the majority of cells had passed from M to G1, CT- motif phosphorylation decreased again. The peaks in CT-motif phosphorylation coincide with times of polarized growth and cell separation. In summary, Cbk1 is highly phosphorylated at the T-loop site over most of the cell cycle, whereas CT-motif phosphorylation coincides with periods of polarized growth and cell separation.
Gel mobility of Cbk1-HA immunoprecipitated at different times after release from pheromone arrest shifted markedly over the course of the experiment (, A and C, anti-HA signal). These changes in mobility suggest that the protein is dynamically phosphorylated at multiple sites as the cell cycle progresses. Both phosphospecific antibodies recognized multiple forms of Cbk1, indicating that these shifts are not solely attributable to phosphorylation of the T-loop and CT-motif sites. Consistent with this finding, phosphorylation site mutants and the kinase-dead allele were also shifted considerably relative to wild-type Cbk1 when separated by SDS-PAGE (). The CT-motif allele was biased most dramatically to slower migrating forms. These shifts were abolished by treatment with λ phosphatase, indicating that they are due to hyperphosphorylation. As indicated by the mobility shifting seen in both the mutants and over the cell cycle, there are likely additional sites of phosphorylation that are dynamically modified across the cell cycle; these as-yet-unknown sites may represent additional regulatory mechanisms for Cbk1 function.
We have shown that Cbk1 is controlled by phosphorylation of two highly conserved sites. Analogous sites have been characterized in other AGC kinases, most notably Akt/PKB, as well as more closely related enzymes, such as human Ndr and Lats and (; ; ; ; ). The prevailing view, based largely on studies of in vitro kinase activity, is that these modifications promote AGC kinase signaling primarily by increasing the target enzyme's catalytic activity. In contrast, we find that the relative functional importance of these sites in vivo is not directly related to their importance for Cbk1's enzymatic activity. Thus, our findings argue that this enzyme activation model does not fully explain the role of these regulatory inputs for Cbk1.
We find that the Cbk1 CT-motif allele has relatively normal kinase activity but exhibits severely defective polarized growth and failed cell separation. In contrast, the Cbk1 T-loop allele has extremely low kinase activity but a moderate phenotypic effect. How does this compare with the existing in vitro characterization of mammalian Ndr? Mammalian enzyme mutated at the T-loop site is similarly compromised for in vitro activity, suggesting that this modification functions similarly in yeast and metazoan enzymes (; ; ). An Ndr CT-motif mutant retains some activity, but this is 7–10-fold lower than the wild-type enzyme (; ; ). Furthermore, CT-motif phosphorylation clearly increases Ndr's in vitro activity (; ; ). It is possible that Cbk1 has inherently higher basal activity than Ndr in the absence of this modification, and our studies do not preclude the possibility that Cbk1's catalytic activity is increased when the enzyme is phosphorylated at the CT-motif. However, the relatively mild phenotype of the severely catalytically compromised T-loop allele indicates that Cbk1's in vivo function requires only minimal levels of the enzyme's catalytic activity. Therefore, our data strongly favor a revision of the prevailing hypothesis for regulation of Cbk1 homologues: that CT-motif phosphorylation has distinct regulatory functions in addition to promoting optimal catalytic conformation of the kinase domain.
What are these additional functions of Cbk1 CT-motif phosphorylation? Both T-loop and CT-motif mutant proteins localize normally to the cell cortex, indicating that neither modification is required to recruit the enzyme to its site of action; similarly, both mutant proteins interact with the conserved Cbk1 binding partner Mob2. Cbk1 CT-motif phosphorylation is, however, required for asymmetric localization of both the kinase and Ace2 to the daughter nucleus late in the cell cycle (). Asymmetric nuclear localization of Cbk1 and Ace2 is interdependent, and the proteins may enter the nucleus as a complex (; ; ). Thus, we propose that CT-motif phosphorylation is necessary for the formation of a productive Cbk1–Ace2 complex.
Intriguingly, the Cbk1 CT-motif mutant protein is dramatically hyperphosphorylated in vivo. This suggests that phosphorylation of the CT-motif site triggers dephosphorylation of Cbk1 at other sites, perhaps by recruiting phosphatases. Given the null phenotype of the CT-motif mutant, it is possible that Cbk1 hyperphosphorylation inhibits the kinase's interaction with downstream targets in vivo. Mammalian Ndr kinases are also phosphorylated at a third site immediately N-terminal to the kinase domain. This modification is important for Mob protein association and stimulation of kinase activity in vivo (), although mutation of the site does not dramatically affect basal in vitro kinase activity or Mob binding (; ). We have mutated Cbk1 at the corresponding site, serine 339, and find no phenotypic effects (unpublished data). Thus, modification of this site is not critical for Cbk1's in vivo function, although it remains possible that it is phosphorylated in vivo as part of a redundant regulatory mechanism. We are currently pursuing identification of the modified sites in hyperphosphorylated Cbk1 and investigating their functional significance.
The in vivo function of mammalian Ndr kinases is unknown; therefore, it is not presently possible to characterize the phenotypes of T-loop and CT-motif alleles of these enzymes. The physiological role of , the homologue of Cbk1, is better understood: it functions in morphogenesis of actin-rich projections as well as branching and tiling of sensory neuron dendrites (; ). Mutation of both T-loop and CT-motif sites renders , which is essential for embryonic development, nonfunctional (). Overexpression of the mutant alleles exerts a dominant-negative phenotype, with the single site mutants having a less extreme wing hair phenotype than the double mutant (). The dominant-negative phenotype of loss-of-function alleles is complex: intriguingly, kinase-inactive is dominant negative in neuron morphogenesis but not in wing hair development (; ). Although these overexpression studies are fundamentally different from our analysis of endogenously expressed alleles, it is clear that the regulatory phosphorylation sites are functionally important. We find that a Cbk1 double-mutant allele has a null phenotype equivalent to the single-mutant CT-motif allele; we do not observe dominant-negative effects in heterozygotes for any alleles at endogenous expression levels (unpublished data). If our analysis of the precise roles of Cbk1's regulatory sites applies more generally, we predict that T-loop alleles of the Ndr family kinases will have a less severe phenotype than CT-motif alleles when expressed under endogenous conditions. Furthermore, we speculate that, under these conditions, CT-motif modification of other Cbk1-related kinases may promote formation of functional complexes or dephosphorylation of other regulatory sites.
We have shown that all RAM network proteins are required for phosphorylation of Cbk1's CT-motif, suggesting that this regulatory event is a critical output of the pathway. Which kinase phosphorylates Cbk1's CT-motif site? Recent evidence suggests that germinal center kinases, which are related to p21-activated kinases, are responsible for phosphorylation of this site in other Ndr/Lats family kinases (; ; ). Kic1, a germinal center kinase–family kinase in the RAM network, is a good candidate for a CT-motif kinase, and we find that CT-motif phosphorylation is undetectable in cells lacking Kic1. However, as CT-motif modification depends on all RAM network proteins, Kic1's role may be indirect. The activity of endogenous Kic1 isolated from yeast is extremely low, and definitive in vitro assays have not yet been possible (unpublished data). The p21-activated kinase Ste20 does not phosphorylate Cbk1 in vitro (unpublished data), suggesting that Cbk1 is not a general target of this class of kinases. Furthermore, yeast cells lacking p21-activated kinase function do not exhibit defective polarized growth maintenance or septum degradation (). Therefore, phosphorylation of the CT-motif site by Kic1 may serve as the critical output of RAM network signaling that results in cell separation and polarized growth.
We find that Cbk1 most likely autophosphorylates itself in vivo at its T-loop site through an intramolecular reaction. In other AGC group kinases, such as Akt/PKB, the analogous modification is performed in trans by an upstream regulatory kinase. Although this is the first demonstration of such behavior for an Ndr/LATS kinase in vivo, it is consistent with previous in vitro observations. Ndr kinases clearly autophosphorylate at this site in vitro (; ), and biochemical data suggest that this modification can occur in cis (). We also find that T-loop autophosphorylation is only completely abolished in cells lacking Mob2 and Tao3. Because these are the only two RAM network proteins thus far found to coimmunoprecipitate with Cbk1, we suggest that Tao3 and Mob2 directly interact with Cbk1 to either help the kinase adopt a proper conformation for cis autophosphorylation or protect it from phosphatases that reverse the modification (; ). Consistent with this idea, previous reports have shown that Mob proteins are required for autophosphorylation of the T-loop site in Ndr (; ). Tao3 is also conserved in metazoans; we speculate that it plays a similar role in the regulation of Cbk1 homologues in these species.
The RAM network's control of the activity and asymmetric localization of Ace2 ensures that Ace2-driven genes involved in septum destruction are only highly expressed once in a cell's life: when it is a daughter. More specifically, Ace2 segregation and function depend on Cbk1's kinase activity (; ), and we have shown that Cbk1 phosphorylates the transcription factor in vitro. We are currently investigating the in vivo relevance of this modification. Our in vitro kinase assays show that Cbk1 CT-motif modification does not simply turn on the kinase's ability to phosphorylate Ace2. Rather, this suggests that phosphorylation at the CT-motif site promotes formation of a functional regulatory complex in vivo.
The logic of this system likely reflects the importance of temporally controlling Ace2 function. Septum destruction is among the final events of cell division, and Ace2-driven transcription of genes involved in the process is tightly regulated. Ace2 is likely negatively regulated through phosphorylation by mitotic Cdk (; ), and its nuclear entry requires activation of the mitotic exit network (). Cbk1's CT-motif phosphorylation is delayed until late M/G1. We speculate that Ace2's Cdk phosphorylation is reversed upon mitotic exit and that this coincides with modification of Cbk1. The co-occurrence of these events may create a short temporal window in which Cbk1 can act on Ace2 to promote its rapid accumulation and function in daughter cell nuclei.
Cbk1's phosphorylation is dynamic across the cell cycle. T-loop autophosphorylation is relatively constant, except a short period after bud emergence. In contrast, modification of the CT-motif fluctuates markedly, peaking just before bud emergence and again late in the cell cycle (). This coincides with periods of active polarized growth, as well as with cytokinesis and septum degradation after mitotic exit. Because the CT-motif phosphorylation requires all RAM network components, it is likely that this cell cycle–regulated modification is one of the key downstream outputs of this signaling pathway.
Based on our findings, we propose a speculative model for Cbk1's cycle of activation during M–G1 that consists of a priming phase followed by activation-dependent dephosphorylation. We propose that Cbk1 is initially phosphorylated at multiple priming sites, other than the T-loop and CT-motif, in a RAM network–dependent manner (, ). We suggest that these modifications both prevent Cbk1's action on downstream targets and promote phosphorylation of the kinase's CT-motif site, most likely by Kic1. This CT-motif modification may then trigger dephosphorylation of the initial priming sites, perhaps by recruiting phosphatases, producing Cbk1 that is fully active (, and ). This form of Cbk1 may then either productively interact with Ace2, which preserves CT-motif phosphorylation (, ), or become fully dephosphorylated and inactivated (, ). This putative kinase/target synergy may work to protect an activated pool of Cbk1 with CT-motif phosphorylation until Ace2's function is performed, ensuring maximal activation of both Cbk1 and Ace2 at the proper place and time.
Yeast strains were derived from S288C () and cultured in YPD media (1% yeast extract, 2% peptone, and 2% dextrose). Overnight cultures were diluted to OD 0.2 and grown 5 h at 30°C to OD 0.8 for use in all assays. Deletion mutants and C-terminal fluorescent tags (a gift from R. Tsien, University of California, San Diego, La Jolla, CA) and myc tags were generated using the primers listed in Table S1 (available at ; ; ). Cbk1 point mutants were generated using two fragment PCR based on genomic DNA from ELY140. The overlapping fragments were integrated at the endogenous locus into ELY132 and screened by PCR and sequencing.
To generate lysates for immunoprecipitation, ∼160 OD equivalents were resuspended in 2 ml ice-cold yeast lysis buffer (150 mM NaCl, 50 mM Tris, pH 7.4, 1% Triton X-100, 10% glycerol, 1 mM dithiothreitol, 60 mM β-glycerophosphate, 2 mM sodium orthovanadate, 10 mM sodium molybdate, 3 mM benzamidine, 1 mM AeBSF, 1 μg/ml pepstatin, 0.5 mM leupeptin, and 2 μg/ml chymostatin). The cell slurry was split into two 15-ml conical tubes, and 3-ml ice-cold glass beads were added to each. Cells were lysed by vortexing on ice seven times for 1 min, until >85% of cells were broken by microscopy. Lysates were cleared by centrifugation at 13.2 for 30 min, and the concentration of the resulting supernatant was obtained through protein assay (Bio-Rad Laboratories), using BSA for the standard curve. Protein concentrations were normalized to 35 μg/μl, and 2 ml of normalized lysates were incubated for 30 min on ice with 5.76 μg anti-HA (12CA5; a gift from R. Lamb, Northwestern University, Evanston, IL) or anti-myc (9E10; a gift from H. Folsch, Northwestern University, Evanston, IL) monoclonal antibodies. The lysates were then rotated 1.5 h at 4°C with 50 μl of 1:1 protein G bead slurry (Sigma-Aldrich), equilibrated in yeast lysis buffer. After incubation, beads were washed three times with yeast lysis buffer and three times with yeast wash buffer (omit Triton X-100, glycerol, and phosphatase inhibitors from lysis buffer). For phosphatase treatment, samples were washed once with λ phosphatase buffer and incubated at 37°C for 30 min in 50 μl λ phosphatase buffer with 1 μl λ phosphatase (New England Biolabs, Inc.) added. Samples were subsequently washed once in ice-cold yeast wash buffer and resuspended immediately in 2× SDS-PAGE sample buffer.
For all Western blots, proteins were resolved by SDS-PAGE and transferred as described previously (). Phosphospecific antibodies (Open Biosystems) were generated against the relevant phosphopeptides for the T-loop serine and the CT-motif threonine: CRLMAY(pS)TVGTPD and CPFIGY(pT)YSRFD, respectively. Western blotting was performed using BioTraceW polyvinylidene difluoride (Pall). Membranes were blocked for 30 min at room temperature with 10% BSA in Tris-buffered saline with 0.1% Tween (TBST). The phosphospecific antibodies were diluted 1:250 (anti-pS570) or 1:100 (anti-pT743) in 1% BSA TBST and incubated overnight at 4°C. The diluted anti-pT743 solution also contained the nonphosphorylated peptide at 1:250 (CPFIGYTYSRFD, from a stock solution of ∼0.6 mg/ml in 40% DMSO) to increase specificity for the phosphorylated protein. Blots were subsequently washed three times for 3 min with TBST. They were then incubated with Alexa 680– conjugated goat anti–rabbit secondary (Invitrogen) at 1:5,000 in TBST for 1 h at room temperature. After incubation, blots were washed six times for 5 min in TBST, and both were imaged and quantified on the Odyssey (v2.0; Li-Cor, Inc.). Images were processed using Photoshop (Adobe).
For kinase reactions, immunoprecipitates were stored overnight at 4°C in yeast lysis buffer with 35% glycerol. Beads were washed three additional times with kinase reaction buffer (20 mM Tris, pH 8.0, 150 mM NaCl, and 5 mM MnCl). Either 2.5 μg of histone H1 or the N terminus of Ace2 fused to maltose binding protein (a gift from D. Stillman, University of Utah, Salt Lake City, UT) was added in 30 μl kinase reaction buffer containing 20 μM cold ATP and 0.33 μCi/μL γ-P-ATP. Kinase reactions were allowed to proceed at room temperature for 60 min and stopped by addition of 7 μl of 5× SDS-PAGE sample buffer and 10-min incubation at 85°C. Proteins were separated by SDS-PAGE, and kinase activity was quantified using a Storm 860 and ImageQuant (Molecular Dynamics). Values obtained for kinase activity from the autoradiograph were normalized against anti-HA Western blot of the same immunoprecipitates so that kinase activity could be compared between samples.
Log phase cells from 4.8 liters of culture were arrested with 10 μM α factor (GenScript, Inc.). After incubation at 30°C for 2 h, >80% of cells exhibited mating projections. The cells were then harvested by centrifugation for 5 min at 25°C. Cells were released from arrest by washing twice with fresh media and then resuspended in fresh media. At each time point, cells were harvested by centrifugation and immediately frozen in liquid nitrogen. Samples of cells from each time point were also fixed by addition of formaldehyde to 5%, and these cells were imaged to determine budding index, providing an indication of cell cycle progression.
Staining of F-actin with rhodamine-phalloidin, cell wall chitin with calcofluor, and cell wall sugars with rhodamine-conjugated concanavalin A were done as described previously (; ). Fluorescence/differential interference contrast microscopy was performed using an Axiovert 200M (Carl Zeiss MicroImaging, Inc.) with either a Coolsnap HQ or Cascade II-512B camera (PhotoMetrics, Inc.). Images were obtained with an oil-immersion α plan-fluar 100×/1.45 NA (Carl Zeiss MicroImaging, Inc.), imaged in synthetic complete medium with dextrose. In scoring all phenotypes, cells were sonicated in a water bath four times for 15 s. The cell separation defect was scored as any cluster of more than two cells. Axial ratio was measured by hand using OpenLab (v4.0.4; Improvision) measurement recording program. Nuclear localization of Ace2-GFP was quantified as symmetric or asymmetric, using rhodamine conA labeling to distinguish mother and daughter cells. All subsequent statistical analysis was performed using Excel (Microsoft), Prism (GraphPad), or InStat (GraphPad).
Fig. S1 shows the coimmunoprecipitation of Mob2-13xmyc with Cbk1-HA, both wild type and mutants. Fig. S2 demonstrates that Ace2–protein A is expressed at similar levels across the Cbk1 mutants. Fig. S3 shows that the phosphospecific antibodies do not recognize other heavily phosphorylated proteins, such as GST-Ste20, purified from yeast. Fig. S4 demonstrates that deletion of Ace2 does not affect the transcript levels of RAM network proteins. Table S1 shows primers used for the construction of strains used in this study. Online supplemental material is available at . |
Insertion of transmembrane proteins into biological membranes generally requires the activity of protein-conducting channels that affect translocation of the exoplasmic polar domains and integration into the bilayer of the membrane-spanning segments of the nascent membrane protein (; ). Because the translocation channels themselves are membrane embedded, one might expect them to rely on copies of themselves for proper integration, and this indeed seems to be the case for the core translocation channels of the outer mitochondrial membrane (TOM40; ) and of the ER (Sec61α; ). Such a dependency of each newly synthesized component of a given translocation complex on functionally integrated copies of the same translocator may underlie the fidelity of the process of membrane expansion, by which each membrane serves as a template for its own growth. However, it also poses an evolutionary problem: how did biological membranes initially assemble the minimal membrane-integrated machinery required to permit the insertion of vital functional protein components? One possibility is that primitive membrane proteins can insert into lipid bilayers without assistance from protein-conducting channels. In support of this notion, a few proteins with short exoplasmic domains have been shown to be capable of transmembrane integration into protein-free lipid bilayers (; ). We refer to this type of integration as unassisted, meaning that the insertion does not require membrane proteins whether or not cytosolic chaperones are involved in maintaining the substrate protein in a conformation competent for integration.
Although most of the investigated proteins capable of unassisted integration are prokaryotic (), at least one eukaryotic protein, the ER form of cytochrome b (b5), inserts very efficiently with transmembrane topology into protein-free liposomes, provided that these have low cholesterol content (). b5 is a member of the group of C-tail–anchored (TA) proteins. Proteins of this diverse group play a variety of fundamental roles in membrane biogenesis and trafficking and are characterized by the presence of an N-terminal cytosolic region and of a single transmembrane domain (TMD) very close to the C terminus (). To qualify as a true TA protein, the C-terminal polar domain, which is downstream of the TMD, must not exceed 25–30 residues so that during synthesis, the hydrophobic TMD is not exposed to the cytosol until the polypeptide is terminated and released from the ribosome. Therefore, the nascent polypeptide does not have a chance to interact with signal recognition particle (SRP), which mediates cotranslational integration, and insertion into the bilayer must occur posttranslationally. In single-spanning proteins with longer C-terminal polar domains, the TMD becomes available to SRP while still in the nascent chain; these proteins can be cotranslationally integrated into the ER membrane, thus qualifying as bona fide type II proteins ().
In our previous work, unassisted transbilayer integration was demonstrated for a b5 construct tagged at its C terminus with a 19-residue sequence derived from the N terminus of bovine opsin, which provides an -glycosylation consensus site as well as an epitope (). The addition of this tag brought the length of the C-terminal polar domain to 28 residues, allowing us to monitor its translocation by a protease protection assay (). In the present study, we have investigated the limit of the length of the C-terminal polar domain appended to b5's TMD that can be translocated across protein-free lipid bilayers. Although it is clear that further lengthening of the C-terminal polar domain will result in a type II protein that can potentially use the SRP-dependent cotranslational Sec61-based pathway, our question is focused on the possible existence of an alternative overlapping, unassisted posttranslational pathway. Using in vitro and in vivo approaches in the mammalian and yeast systems, we show that the TMD of b5 can support the unassisted translocation of surprisingly long and differently charged C-terminal polar sequences with low or no energy requirements.
Although posttranslational integration is a hallmark of TA protein biogenesis, different TA proteins appear to have different requirements for membrane integration. Thus, although b5 undergoes unassisted transbilayer integration, the v-SNARE synaptobrevin 2 (Syb2) requires energy and a proteinaceous factor or factors of the ER membrane (; ; ; ). Because the unassisted translocation that we observed for our b5 constructs appeared to be relatively insensitive to the sequence of the lumenal domain, we investigated whether differences in the TMD underlie the different requirements for TA protein integration. We find that replacement of the b5 TMD with the more hydrophobic one of Syb2 or simply an increase in the hydrophobicity of b5's TMD obtained by a few amino acid substitutions results in constructs that show the same requirements for energy and for an ER proteinaceous component as native Syb. Conversely, a construct carrying a mutated Syb2 TMD with reduced hydrophobicity as well as the TA protein tyrosine phosphatase 1B (PTP1B), whose TMD, like the one of b5, is only moderately hydrophobic, inserts into lipid bilayers without assistance. Our results suggest a mechanism by which primitive membranes may have assembled in the absence of modern translocation machinery and identify TMD hydrophobicity as the feature that determines the different requirements for transmembrane integration of TA proteins.
In previous studies (; ; ; ), we demonstrated the posttranslational transmembrane integration into ER membranes of a b5 variant engineered to contain an -glycosylation site (derived from bovine opsin) near the C terminus (; and here renamed b5-ops-28). To investigate whether longer C-terminal polar domains can also be translocated in the posttranslational mode, we generated a battery of b5 constructs in which the C-terminal domain was progressively elongated to 125 amino acid residues. These constructs were obtained by fusing unrelated oligopeptides (derived from the N-terminal domain of bovine opsin, from the yeast protein Hsp150, or from the cytosolic domain of vesicular stomatitis virus glycoprotein [VSVG]) downstream of the original construct, b5-ops-28, or immediately after b5's TMD. The constructs were named according to the origin of the appended sequence and to the length of the entire C-terminal sequence downstream of the TMD (; and sequences in Table S1, available at ). For these extended constructs, beyond a certain length of the attached sequence, we might expect a shift to the cotranslational SRP-dependent mechanism of translocation because the TMD will emerge from the ribosome before chain termination and be recruited via SRP to the Sec61 translocon, as occurs with classic type II membrane proteins ().
To compare posttranslational and cotranslational translocation efficiencies, ER rough microsomes (RMs) were either added during translation of the various constructs or after the termination of protein synthesis and removal of ribosomes (). To follow translocation, we used our previously developed stringent assays (), which are based both on the M shift caused by -glycosylation of the translocated opsin sequence and on protection from protease K (PK) digestion of the translocated fragment. An aliquot of each sample was directly analyzed by SDS-PAGE to determine the amount of protein synthesized and the extent of glycosylation (; top, −PK), whereas the rest was digested with PK and subjected to immunoprecipitation to recover the protected fragment (PF; ; bottom, +PK; see for a full characterization of the assay).
As shown in , when the constructs were not exposed to RMs, only the nonglycosylated protein was detected (boxes indicate the nonglycosylated full-length protein or PF). Likewise, PFs were not recovered from samples incubated in the absence of membranes nor from samples incubated with membranes but digested with PK in the presence of detergent (Fig. S1 A, available at ). Instead, when RMs were present either during or after translation, a portion of each construct with lumenal domain up to 85 residues was converted to the glycosylated form (; asterisks indicate the glycosylated full-length protein or PF) except the construct b5-VSVG-33, which lacks an -glycosylation consensus sequence (). In agreement, these constructs, after co- or posttranslational incubation with RMs followed by PK treatment in the absence of detergent, generated a PF, most of which was glycosylated (again, with the exception of b5-VSVG-33). In contrast, the constructs bearing a sequence longer than 100 amino acids downstream of the TMD, although capable of cotranslational translocation, were no longer able to insert posttranslationally, as indicated by the lack of glycosylation and protection from proteolysis (, compare lane 14 with 15 and lane 17 with 18).
Quantitative comparison of co- and posttranslational translocation () revealed equal efficiencies for the constructs with C-terminal polar domains of up to 66 residues, whereas the construct with 85 lumenal residues showed slightly reduced efficiency, and those with >100 residues showed severely reduced efficiency in the post- versus the cotranslational mode. Thus, remarkably large domains of different sequence and net charge (; indicated above the lanes) are capable of posttranslational translocation. Interestingly, as previously shown for b5-ops-28 (), only the C terminus and not the N terminus of the recombinant proteins was translocated, as demonstrated by the lack of protection of b5's catalytic domain from PK digestion (unpublished data).
In a previous study, we showed that the transmembrane integration of b5-ops-28 occurs efficiently in yeast mutants that are defective in the function of the translocon or translocon accessory proteins (). In this study, we investigated whether the extended constructs could also translocate their C terminus across the ER of yeast mutants. As shown in , in the temperature-sensitive mutant, all of the analyzed recombinant proteins were fully or nearly fully glycosylated at the restrictive temperature, as indicated by the shift in electrophoretic mobility obtained after digestion with endoglycosidase H. In contrast, the glycosylation of carboxypeptidase Y (CPY), a protein that depends on the Sec machinery for its posttranslational translocation, was severely reduced at the restrictive temperature (, bottom), demonstrating inactivation of the mutant Sec61 protein by heat treatment. Similar results were obtained with another temperature-sensitive Sec61p mutant, (Table S2, available at ; and not depicted). Thus, Sec61-independent translocation in vivo appeared to be even more permissive than posttranslational translocation in vitro because even the b5-ops-125 construct was glycosylated in the yeast mutants.
In a previous study (), we demonstrated that b5-ops-28 can translocate its C terminus across protein-free phospholipid liposomes as efficiently as across RMs. Therefore, we investigated whether the extended constructs that retain the ability to posttranslationally translocate into RMs were also competent for transmembrane insertion into pure lipid vesicles prepared by the extrusion of a mixture of bovine liver phosphatidylcholine (PC)/phosphatidylethanolamine or PC alone (). Insertion into RMs was again assessed by glycosylation of the full-length protein (, top) and PF recovery (, bottom). In the case of liposomes, translocation of the C terminus cannot be assayed by glycosylation but is revealed by the protease protection assay (; nonglycosylated PFs are indicated by boxes in the bottom panel). Quite remarkably, all C-terminal domains capable of posttranslational translocation across RMs were translocated equally well across protein-free vesicles independently of length, sequence, and charge, and this was true even for the highly charged 85 residues of b5-ops-85 (, lanes 13–15). As expected, the 104- and 125-residue–long C-terminal tails were not observed to insert into protein-free liposomes (Fig. S1 B).
Although b5-ops-85 inserts equally well into RMs and liposomes, the extent of its translocation into both of these vesicles was less than that of the parent construct b5-ops-28. This lower efficiency might be the result of either heterogeneity in the translocation competence of the longer construct such that only a fraction of molecules can insert or a reduced translocation rate of equivalent b5-ops-85 molecules. As shown in the time course experiment in , the proportion of glycosylated molecules increased linearly with time for both b5-ops-85 and -28 during a period of 90 min, suggesting that in vitro–translated b5-ops-85 constitutes a homogeneous population of slowly translocating molecules.
Although b5 efficiently translocates its C terminus across the bilayer without assistance from ER proteins, other TA proteins, like Syb2, require one or more proteinaceous components of the ER for their integration (). Because the protein-independent mechanism seems to be quite permissive with regard to the C-terminal sequence to be translocated, we investigated the role of the cytosolic and TMD in determining the requirements for TA protein insertion. We excluded the involvement of b5's cytosolic N-terminal domain by replacing this region with GFP in the b5-opsin-85 construct and observing that the resulting protein was still able to efficiently translocate across pure lipid liposomes (Fig. S2, available at ). Therefore, we focused on the TMD, replacing the core of b5's TMD with the one of Syb2 (b5-Syb2-ops-28; ). As shown in , the new construct was as efficiently glycosylated and translocated across posttranslationally added RMs as the parent construct b5-ops-28. However, different from b5-ops-28, the construct with Syb2's TMD was unable to translocate its C terminus across protein-free liposomes (, bottom). b5-Syb2-ops-28, in contrast to the parent construct, generated two background bands after PK digestion in the absence of vesicles, which comigrated with the glycosylated and nonglycosylated PF (, lane 4); as can be seen in (lane 6), the intensity of the PF bands obtained after incubation with PC liposomes was similar to that of the background bands.
We next asked whether the different behaviors of the two recombinant proteins was caused by a specific amino acid sequence or, instead, was caused by differences in the physical/chemical characteristics of the two TMDs. As reported in , the degree of the hydrophobicity of b5's TMD is considerably lower than that of Syb2's. To ascertain whether this difference in the hydrophobicity of the TMDs could account for the different translocation requirements of the two model proteins, we generated a construct with a TMD based on the one of b5 but modified by the substitution of four residues with more hydrophobic ones (b5-HH-ops-28; and ). As shown in , when tested in its ability to posttranslationally translocate across RMs or liposomes, this construct showed the same behavior as the b5-Syb2-ops-28 chimera, inserting efficiently into RMs but not into protein-free vesicles.
To confirm that TMD hydrophobicity and not a specific sequence is the principal parameter determining competence for unassisted insertion, we produced two further constructs: b5-scrambled-ops-28, in which the order of residues in b5's TMD was changed, and b5-Syb2mut-ops-28, in which the TMD of Syb2 was mutated to become less hydrophobic (). As shown in , both of these constructs were able to insert into protein-free liposomes (lanes 7 and 12), albeit with somewhat reduced efficiency compared with the parent construct b5-ops-28 (lane 3).
Previous studies have reported that Syb2 depends on a microsomal trypsin-sensitive component for its insertion (; ; ). We asked whether Syb2's TMD is sufficient to confer this trypsin sensitivity and, if so, whether the increased hydrophobicity is the basis for this effect. RMs were treated with increasing concentrations of trypsin at 4°C to digest exposed proteins; efficacy of the treatment was verified by immunoblotting with antibodies against two ER integral membrane proteins, Sec61β and ribophorin I. As shown in , Sec61β was severely depleted already by treatment with 1 μg/ml trypsin, whereas ribophorin I was more resistant to the treatment and showed partial conversion to a shorter form lacking the cytosolic domain at higher trypsin concentration. The trypsin-treated or mock-treated RMs were tested for transmembrane integration of the parent b5-ops-28 and of the two constructs with modified TMD (). As expected, trypsin treatment had no effect on b5-ops-28's ability to translocate its C terminus. In contrast, for the two constructs with more hydrophobic TMDs, even a mild trypsin digestion inactivated a microsomal component that is required for insertion of the constructs (, lanes 8–10 and 13–15).
To investigate whether the ability to integrate into lipid bilayers without assistance from membrane proteins is a peculiarity of b5 or whether it is shared with other TA proteins, we analyzed the TMD sequence of 27 ER resident TA proteins (Table S3; available at ), searching for unassisted translocation candidates on the basis of low TMD hydrophobicity. For analysis, we selected PTP1B, whose TMD is similar in length and hydrophobicity to the one of b5 (). PTP1B is a prototypic tyrosine phosphatase anchored to the ER membrane () and is involved in the regulation of numerous signaling events (). To adapt PTB1B to our assays, we appended the opsin epitope at its C terminus to obtain the PTP1B-ops-35 construct ( and Table S1). As shown in , this PTP1B variant behaved exactly like b5-ops-28 in that it posttranslationally inserted into RMs with the same efficiency as in cotranslational conditions, as judged by the appearance of a glycosylated form and by the recovery of a glycosylated PF, and it inserted to the same extent into pure lipid liposomes, as shown by the presence of the nonglycosylated PF. Based on the ratio of glycosylated to nonglycosylated full-length protein, the efficiency of translocation of PTP1B-ops-35's C terminus was roughly the same as that of b5-ops-28 (, left).
To further compare the mechanisms of the transmembrane integration of b5 and PTP1B, we analyzed the effect of the cholesterol content of liposomes on insertion efficiency (). Our previous work demonstrated that b5-ops-28 insertion is sharply inhibited by cholesterol levels only slightly higher than those normally found in the ER (). When in vitro–synthesized PTP1B-opsin-35 was incubated with liposomes containing increasing proportions of cholesterol, translocation was supported only by cholesterol-poor ones, as is the case for b5-ops-28 (). Thus, the unassisted insertion pathway described for b5 can also be exploited by other TA proteins whose TMDs have a similar degree of hydrophobicity.
Although the unassisted translocation process itself must be nucleotide independent, energy consumption by chaperones could be necessary to maintain the translocation substrates in a competent conformation. Because this possibility appeared especially likely in the case of the constructs with extended C-terminal domains, we compared the energy requirements for the integration of the longest construct capable of unassisted translocation (b5-ops-85) with those of the parent construct b5-ops-28. Nucleotides were depleted from translated samples by gel filtration, and samples were further diluted into a buffer compatible with b5 integration to reduce the final ATP concentration to 3 nM. As shown in , severe nucleotide depletion had no effect on the efficiency of the translocation of b5-ops-28 (compare lane 1 with 6) and, remarkably, had only a minor effect on that of the extended variant b5-ops-85 (compare lane 7 with 12) into either RMs (top panels) or liposomes (bottom panels). Furthermore, the addition of ATP, GTP, or both to restore nucleotide levels to those of the reticulocyte lysate was without effect on the translocation of b5-ops-28 (, lanes 2–4) and was mildly stimulatory on b5-ops-85 (, lanes 8–10). These results indicate that unassisted translocation not only of b5's C terminus but also of an extended polar domain appended to b5's TMD has extremely low energy requirements.
At variance with b5, Syb2's insertion is reported to be energy dependent (; ). We asked whether Syb2's TMD alone is responsible for this effect by testing the transmembrane integration of the chimera b5-Syb2-ops-28 after ATP depletion. As shown in , after nucleotide depletion, insertion of the construct was severely reduced (lane 1). In agreement with and but at variance with , the readdition of ATP stimulated integration much more effectively than GTP. Thus, Syb2's TMD confers to the chimera all of the insertion requirements of Syb2 itself.
Membrane proteins generally integrate into the lipid bilayer cotranslationally: during synthesis, the signal peptide or the first hydrophobic membrane-anchoring sequence to emerge from the ribosome is recognized by SRP, which delivers the nascent chain–ribosome complex to the translocation machinery of the ER membrane (). The insertion pathway followed by TA proteins represents an exception to this rule because members of this class of proteins lack an N-terminal signal sequence, and their membrane-anchoring sequence is too close to the C terminus to become cotranslationally available to SRP (). Given the important functions of TA proteins in fundamental cellular processes, the molecular mechanisms underlying this posttranslational mode of transmembrane protein topogenesis are currently being studied in several laboratories.
The insertion pathway of a few TA proteins has been investigated in some detail, and important differences have been reported, as exemplified by Syb2 and b5. There is a general consensus that Syb2 requires a proteinaceous component of the ER membrane and energy to associate with RMs (; , ; , ), although there is disagreement regarding the nature of the involved protein and of the energy donor (ATP or GTP). In contrast, b5 integration occurs without assistance from any microsomal protein () and appears to have very low energy requirements (; ).
In our previous work, we used a b5 variant carrying at its C terminus a 19-residue sequence derived from bovine opsin, which permitted us to monitor translocation by protection from proteolysis and/or glycosylation of the short translocated sequence (). The total length of the domain translocated without assistance, including the seven C-terminal residues of b5 itself, two linker residues, and the opsin sequence, was 28 residues. Two important questions were raised from this study: what the maximum size is of the C-terminal polar domain that can be translocated without assistance and what the basis is for the reported differences in the insertion requirements for different TA proteins. In the present study, we address both of these issues.
As for the first question, we find that unexpectedly long sequences (up to 85 residues) appended to b5's C terminus can be translocated across protein-free bilayers. Consistent with these in vitro observations, constructs with extended lumenal domains could efficiently transfer their C terminus into the ER lumen of yeast mutants defective in translocon function. The process of unassisted integration appeared to be relatively insensitive to the nature of the sequence to be translocated. The three classes of sequences that were attached to b5's C terminus and that translocated without assistance by membrane proteins in this study had different charge and different intracellular localizations when in their normal context: whereas the Hsp150 and opsin sequence are exoplasmic, the VSVG sequence was derived from the cytosolic tail of the viral glycoprotein. The VSVG tag differed from the other two also in its net positive charge, arguing against any role of an electrochemical gradient in the translocation process. Furthermore, the b5-VSVG-33 construct lacked the 19-residue opsin tag and the six C-terminal residues of b5, both of which are present in the other b5-based constructs, excluding a role for these residues in unassisted translocation.
Although unassisted transmembrane integration was relatively insensitive to the nature of sequences <100 residues appended to b5's C terminus and was unchanged by substitution of the cytosolic domain, it was dramatically affected by the properties of the TMD, pinpointing the basis of the differing requirements for transmembrane topogenesis of different TA proteins (the second question raised above). Substitution of b5's TMD with the more hydrophobic one of Syb2 conferred on the chimera all of the properties reported for native Syb2: incapacity to insert into liposomes, sharp inhibition of posttranslational integration by trypsin treatment of microsomes, and a requirement for energy (; , ; ). These altered requirements for insertion were the result of the increased hydrophobicity of Syb2's TMD and not of a specific amino acid sequence because an increase in the hydrophobicity of b5's TMD conferred by four amino acid substitutions was sufficient to completely block the unassisted translocation of even a short (28 residues) lumenal domain, mutations that decreased the hydrophobicity of Syb2's TMD conferred to the resulting construct the capability of unassisted insertion, and scrambling the order of residues in b5's TMD generated a construct that was still capable of insertion into liposomes. In addition, we ruled out the possibility that the capacity for unassisted transmembrane integration is a peculiar feature of b5 by demonstrating the same capability of a nonrelated TA protein, PTP1B, which shares with b5 only the moderate hydrophobicity of its TMD.
Two other TA proteins whose membrane association has been investigated with binding assays, Nyv-1p () and Sec61β (), have requirements similar to those of Syb2, whereas another one, Bcl2, appears to be more like b5 (). The behavior of these three TA proteins is predictable on the basis of the hydrophobicity of their TMDs (Table S3). Thus, TMD hydrophobicity appears to be a reliable predictor of whether or not assistance for transmembrane integration of a given TA protein is required.
Why should a more hydrophobic TMD preclude unassisted translocation? A possible explanation is that newly synthesized TA proteins with very hydrophobic TMDs require an ER protein and energy for delivery to the bilayer in a translocation-competent form rather than for the translocation step itself. It is known that the poor water solubility of hydrophobic peptides constitutes a major problem for their assembly into preformed lipid bilayers (; ). TA proteins with TMDs of limited hydrophobicity may be endowed with a certain degree of water solubility, which would allow them to directly access the bilayer. In contrast, TA proteins with very hydrophobic TMDs are probably rapidly sequestered into insoluble aggregates unless assisted by a chaperone, which might then require interaction with a cytosolically exposed ER protein to deliver the substrate to the bilayer. According to one study, this chaperone function is fulfilled by SRP and SRP receptor, acting in concert via a novel posttranslational mechanism (). However, the observation supporting this conclusion is restricted to a small fraction of truncated Syb2 and Sec61β nascent chains immediately after their release with puromycin. With naturally terminated Syb2 polypeptide incubated for longer times with RMs, a requirement for SRP receptor was not detected, and, in agreement with our data, ATP but not GTP was necessary for association of the polypeptide to RM membranes (). Thus, there appears to be more than one delivery pathway of TA proteins to the ER.
The finding reported here that a polar domain of nearly 100 residues placed downstream of an appropriate TMD can be translocated across protein-free bilayers was quite unexpected. How can such large polar domains make it across the hydrophobic core of the lipid bilayer? The free energy released upon insertion of the TMD (; ) could possibly provide the driving force to overcome the kinetic energy barrier to translocation of the polar domain. Thus, the TMD of b5 and of other TA proteins endowed with a similar capacity for unassisted insertion would function as nanosyringes, catalyzing the translocation of their own C-terminal polar domain, provided that the lipid bilayer to be crossed is sufficiently disordered and that the length of the polar domain is not excessive. Indeed, increasing bilayer order by the addition of cholesterol () results in sharp inhibition of the process (; and this study); analogously, the constructs with long lumenal domains, for which the kinetic barrier to translocation is presumably higher, integrate more slowly than those with shorter ones. Above a certain length of the C-terminal domain, the free energy of insertion of the TMD is presumably no longer sufficient to overcome the activation energy barrier opposing translocation.
An important question concerns the folding status of the C-terminal domain that is translocated without assistance. Structural experiments on the N-terminal domain of opsin have revealed that it easily folds to yield a compact structure and that this folding also occurs in the absence of -glycosylation (). In contrast, the 19-residue repeat domain of Hsp150 appended in one or more copies to the Hsp constructs is unstructured (). Thus, we expected that these two sequences might have different requirements for translocation because of a different folding status. Instead, the very low energy requirements for the b5-ops-85 construct argue against an active unfolding process occurring before translocation. It is possible that the opsin domain was not folded under our experimental conditions or, alternatively, that it has sufficient conformational flexibility to spontaneously oscillate between folded and unfolded states. The problem of the conformation of the lumenal domain in the unassisted translocation pathway described here is the subject of ongoing investigation in our laboratory.
In addition to suggesting a mode of evolution of biomembranes, we believe that the results presented here have implications for contemporary membrane biogenesis and physiology. At present, we cannot rule out that TA proteins capable of transmembrane integration into protein-free liposomes in vitro are assisted in vivo by proteinaceous factors that accelerate their insertion, as found for M13 and Pf3 phage coat proteins (). However, the equal efficiency of insertion into RMs and protein-free liposomes of PTP1B and of all constructs based on b5's TMDs as well as our previous demonstration of the absence in microsomal extracts of any protein stimulating b5-ops-28 translocation () strongly suggest that the unassisted pathway characterized in vitro occurs also in vivo. Unassisted insertion may be relevant not only for TA proteins and for some bacterial proteins () but also for thylakoid membrane biogenesis (). Furthermore, given its potential overlap with the cotranslational pathway, it could represent a salvage mechanism for the insertion of some type II membrane proteins. It is well known that hydrophobicity is the principal parameter determining the affinity of signal sequences/anchors for SRP (); thus, exactly those type II proteins that are more likely to escape from the cotranslational pathway would be better substrates for an unassisted salvage pathway. Finally, phenomena similar to the one reported here could underlie voltage- () and lipid ()-dependent posttranslational rearrangements of membrane proteins that involve the translocation of large domains across the bilayer. Unconventional mechanisms of membrane biogenesis like the one reported here are thus likely to be of general significance and are likely to attract increasing attention in the coming years.
The constructs used in this study are illustrated in . Sequences of the TMDs and of the C-terminal lumenal domains are reported in and S1, respectively. All cDNAs for transcription/translation were cloned in the pGEM4 vector under the SP6 promoter and checked by sequencing. The recombinant plasmids were obtained by inserting cassettes of paired oligonucleotides or of PCR-amplified fragments into the parent plasmid b5-ops-28 (called b5-Nglyc in a previous study []) or into its derived extended constructs. The cDNA coding for PTP1B, which was a gift from J. Chernoff (Fox Chase Cancer Center, Philadelphia, PA; ), was subcloned into pGEM4 and modified to contain the C-terminal opsin tag.
Transcription, translation in the reticulocyte lysate system (Promega), translocation reactions with RMs (a gift of R.S. Hegde, National Institute of Child Health and Human Development, Bethesda, MD) or liposomes prepared by extrusion, digestions with PK, immunoprecipitation with antiopsin monoclonal antibody (a gift of P. Hargrave, University of Florida, Gainesville, FL) used at 8 μg/ml or anti-VSVG polyclonal antibody used at 7 μg IgG/ml (Sigma-Aldrich), and SDS-PAGE analysis were performed as described previously (). In each experiment, samples contained equal amounts of phospholipids, as specified in the figure legends; phospholipid phosphorus in RMs was assayed according to , whereas phospholipid concentration in liposome suspensions was assessed by measuring the recovery of H-PC (GE Healthcare) included in trace amounts in the lipid mixtures before essication and extrusion. Incubations with vesicles were performed for 1 h unless specified otherwise. Dried gels were either exposed to film or to a phosphorimager screen (Storm; GE Healthcare). All quantifications were performed with the Storm phosphorimager using ImageQuant software (GE Healthcare). For comparisons of translocation efficiency based on the amount of PF generated, intensity values of the PF bands were normalized to the amount of the total full-length product generated in the corresponding translation reaction. In all figures, the numbers on the side of the panels indicate the position and molecular mass (in kilodaltons) of size markers.
90 μl of each translation reaction was gel filtered on Sephadex-25 fine columns (0.8 × 4 cm; Bio-Rad Laboratories) equilibrated in a buffer suitable for translocation (translocation buffer [TB]; 50 mM Hepes, pH 7.2, 250 mM sorbitol, 70 mM KOAc, 5 mM KEGTA, 2.5 mM Mg(OAc), and 2 mM DTT). 15 fractions of 90 μl were collected, and an aliquot of each was analyzed by SDS-PAGE to identify those with the maximum recovery of the radioactive protein. These were then assayed for ATP by the luciferin-luciferase procedure as previously described () and used in the translocation assays after appropriate dilution in TB.
Four equivalents of RMs (see for a definition) were incubated for 1 h at 4°C with increasing concentrations of trypsin in a final volume of 50 μl trypsin buffer (50 mM triethanolamine–acetic acid, pH 7.5, 250 mM sucrose, and 1 mM DTT). Digestion was terminated by the addition of PMSF to 1 mM and aprotinin to 47 mTIU/μl for 15 min at 4°C. Samples were then adjusted to 500 mM KOAc in a final volume of 200 μl, and RMs were sedimented at 63,000 rpm for 30 min in a rotor (TLA 100.3; Beckman Coulter), resuspended in 200 μl trypsin buffer, sedimented again, and finally resuspended at 0.5 equivalents/μl in storage buffer (50 mM Hepes, pH 7.2, 250 mM sucrose, and 2 mM DTT). The effect of the trypsin treatment was assessed by Western blotting/ECL (SuperSignal West Pico; Pierce Chemical Co.) with antiribophorin I (antibody RIL3; a gift from G. Kreibich, New York University School of Medicine, New York, NY; ) and anti-Sec61β (provided by R.S. Hegde; ) antibodies.
The b5-ops constructs described in and in and S1 were subcloned between the promoter and the terminator in the yeast shuttle vector pFL26. The resulting plasmids coding for b5-ops-28, -47, -85, and -125 were designated pKTH5013, pKTH5126, pKTH5181, and pKTH5237, respectively. The control yeast strain H1689 was created by integrating pKTH5013 into the locus of strain H245 (see Table S2 for yeast strains). The strains were created by integrating the plasmids pKTH5103, pKTH5126, pKTH5181, and pKTH5237 into the locus of the parental strain H257, generating strains H2270, H2237, H2240, and H2113, respectively. The strains were created by transforming the same plasmids into H1693, creating strains H1698, H2148, H2151, and H2116, respectively (Table S2).
The aforementioned described yeast cells were grown in synthetic complete medium in full (2%) glucose at the permissive temperature (24°C). Expression of the b5 constructs and parallel inactivation of Sec61p were obtained by up-regulating the promoter in 2% raffinose (low glucose) medium at the restrictive temperature (38°C) for 1 h. Metabolic labeling of proteins with [S]methionine/cysteine (GE Healthcare), lysis, immunoprecipitation with antiopsin and anti-CPY antibodies, and endoglycosidase H digestion were performed as previously described ().
Fig. S1 shows the detergent controls for all of the b5 constructs and the lack of transmembrane integration of b5-hsp-104 and b5-ops-125 into liposomes. Fig. S2 shows the lack of effect of the substitution of b5's catalytic domain on unassisted translocation. Table S1 provides sequence information on the lumenal domains of all of the constructs analyzed in this study. Table S2 is a list of the yeast strains used for the in vivo studies, and Table S3 is a list of 27 TA proteins analyzed for TMD hydrophobicity. Online supplemental material is available at . |
Reentry into the cell cycle requires integration of signals from several redox-dependent processes (). For example, production of hydrogen peroxide (HO) is required for mitogenic signaling in response to EGF, bFGF, PDGF, and thrombospondin2 (; ). One mechanism by which HO acts in mitogenic signaling is through the transient oxidation of cysteine residues present in signaling targets such as the phosphatases protein tyrosine phosphatase 1B and PTEN (phosphatase and tensin homologue on chromosome 10), which regulate signaling through the extracellular signal–related kinase (ERK) 1/2 and PI3-kinase–Akt pathways, respectively (for review see ).
Given the prominent role of oxidants in cell cycle reentry, the G0–G1 transition can be considered an oxidative phase of the cell cycle, as suggested by a recent study on metabolic cycles in yeast (). However, although production of HO in response to growth factors is required for cell cycle reentry (), high levels of HO during the G0–G1 transition cause cell cycle arrest. In serum-stimulated mouse lung epithelial cells, as in many other cell types (for review see ), signals from the ERK1/2 and PI3-kinase–Akt pathways are integrated temporally at the level of expression of cyclin D1 (, ; ). Recently, we showed that pathways regulating expression of cyclin D1 are targeted by reactive oxygen species (ROS) and reactive nitrogen species, resulting in cell cycle arrest (, ; ). Arrest can be bypassed by loading cells with catalase (), supporting the notion that intracellular levels of HO represent one mechanism for redox-dependent control of cell cycle progression.
Peroxiredoxins (Prxs) are a highly abundant family of widely expressed antioxidant enzymes (for reviews see ; ; ). Because PrxI interacts with c-Abl () and c-Myc (; ) and PrxII modulates signaling through the PDGF receptor (), Prxs have emerged as important factors that link ROS metabolism to redox-dependent signaling events. All Prxs use a redox-active peroxidatic cysteine to attack peroxide substrates, resulting in the formation of a cysteine sulfenic acid (Cys-SOH). As is typical for 2-Cys Prxs, PrxI and -II are obligate homodimers, and in these enzymes the Cys-SOH of the peroxidatic cysteine in one subunit is attacked by a resolving cysteine in the neighboring subunit, resulting in an intersubunit disulfide bond. In mammalian cells, the intersubunit disulfide is reduced by thioredoxin (Trx), which is then regenerated by Trx reductase (TrxR) using reducing equivalents from NAD(P)H (). Calcium concentration, pH, and oxidation state influence the assembly of 2-Cys Prx dimers into decamers, and decamers into high molecular mass oligomers (for reviews see ; ; ). Recent work also provides evidence for a link between structural transitions in the oligomeric state of Prxs and their peroxidase and protein chaperone activities (; ; ).
In contrast to prokaryotic homologues, eukaryotic 2-Cys Prxs have a particularly interesting biochemical characteristic in that they are readily inactivated by their own substrate, HO. Because of a C-terminal domain that induces a kinetic pause in the catalytic cycle, the peroxidatic cysteine of PrxI and -II is susceptible to hyperoxidation, leading to the formation of sulfinic acid (Cys-SOH), which cannot participate in disulfide bond formation with the resolving cysteine (). Inactivation through hyperoxidation has been proposed to allow HO to accumulate to substantial levels, thereby facilitating redox-dependent signaling, a concept known as the “floodgate” hypothesis (). The fact that the sulfinic acid form of 2-Cys Prxs is not a terminal end product but can be reduced in an ATP-dependent manner by sulfinyl reductases, such as sulfiredoxins (; ) and p53-inducible sestrins (), suggests that Prx-SOH may participate in regulatory signaling loops.
We tested the relevance of the floodgate hypothesis during mitogenesis by investigating the connection between the oxidative state of Prx isoforms and cell cycle entry and arrest. Our studies indicate that widespread inactivation of PrxI and -II by hyperoxidation is not a facet of normal mitogenic signaling. Rather, examination of dose-dependent responses to fluxes of HO demonstrate that cell cycle arrest in response to oxidative stress correlates with recruitment of PrxII-SOH into cytoplasmic oligomers and that recovery of cell proliferation occurs after Prx-SOH is reduced. Unexpectedly, transient overexpression of PrxI and -II led to increased levels of hyperoxidized Prxs in response to oxidative stress and failed to protect cells from arrest. We propose that Prx-SOH functions in stress response pathways that warn cells of perturbations in oxidant metabolism and thereby contribute to oxidant-induced cell cycle arrest.
To examine the oxidation state of Prxs during mitogenic signaling, mouse C10 lung epithelial cells were collected in G0 by serum deprivation, and the formation of Prx-SOH in response to serum stimulation was assessed using an antibody specific for Prx-SOH. Prx-SOH was not detected above background levels in cells stimulated for 15 min with medium containing serum concentrations from 2 to 20%, a range that induces dose-dependent induction of tyrosine phosphorylation (, lanes 2–5), activation of the ERK1/2 and PI-3 kinase–Akt mitogenic signaling pathways, and expression of cyclin D1 (). These results indicate that normal mitogenic signaling does not require inactivation of Prxs by hyperoxidation, in agreement with a recent report on the role of PrxII in PDGF signaling ().
To further explore Prx oxidation in cell cycle control, we adopted an experimental paradigm that utilizes a dose-dependent HO generating system to evoke transient cell cycle arrest (). C10 cells were synchronized in G0 by serum deprivation and induced to reenter the cell cycle by adding medium containing 10% FBS with or without glucose oxidase (GOx). In complete medium with glucose and 10% FBS, GOx caused the dose-dependent production of HO in a linear fashion for at least 8 h (). For example, in complete medium, 5.0 mU/ml GOx generated ∼10 μM HO/h.
During the first 6 h of serum stimulation, 1.0 or 2.5 mU/ml GOx had little effect on the expression of cyclin D1, whereas doses of 5.0 mU/ml or greater blocked expression of cyclin D1 (, lanes 7–9). In response to continuous exposure to 1.0 mU/ml GOx, the levels of activated ERK1/2 were similar to the serum control, cyclin D1 was expressed, and hyperoxidized 2-Cys Prxs were not observed (, lane 5), suggesting that C10 cells are able to metabolize considerable amounts of exogenous HO during the G0–G1 transition without accumulating hyperoxidized 2-Cys Prxs. At 2.5 mU/ml, levels of phospho-ERK1/2 were unaffected, Prx-SOH was barely detectable after 6 h of exposure, and cyclin D1 was expressed at nearly normal levels. In contrast, at 5.0 mU/ml, hyperoxidized Prx-SOH accumulated to substantial levels and cyclin D1 was not expressed (, lane 7). Concentrations of GOx ≥10.0 mU/ml induced accumulation of hyperoxidized Prx-SOH, caused hyperactivation of ERK1/2, and blocked expression of cyclin D1 (, lanes 8 and 9).
We previously showed that termination of ERK1/2 signaling after 3 h of exposure to the highest dose of GOx (15 mU/ml) restores expression of cyclin D1 but not cell proliferation (). Hence, prolonged activation of ERK1/2 is a useful marker of oxidant-induced arrest at the G0–G1 transition of the cell cycle. Although GOx influenced the levels of phospho-ERK1/2 in a dose-dependent manner as before, it did not induce phosphorylation of JNK in synchronized cells at any dose (, lanes 5–9). In asynchronous cells, activation of JNK in C10 cells by HO is associated with cell death ().
To determine if retroreduction of Prx-SOH prevented the accumulation of Prx-SOH, serum-stimulated cells were treated with 1-chloro-2,4-dinitrobenzene (DNCB), with or with out GOx. DNCB depletes cells of reduced glutathione (GSH) and blocks reduction of Trx by inhibiting TrxR (), thereby impairing the ability of Trx and GSH to participate in the retroreduction of Prx-SOH to catalytically active forms. Within 10 min, 5 μM DNCB caused a 90% reduction in GSH levels that persisted for at least 3 h (unpublished data).
In the absence of GOx, DNCB blocked the ability of serum to induce expression of cyclin D1 but did not prevent phosphorylation of ERK1/2 (, lane 4) or cause the accumulation of hyperoxidized Prxs. In contrast, DNCB markedly sensitized 2-Cys Prxs to hyperoxidation by GOx (, compare lanes 6–9 with lanes 10–13), suggesting that Prx retroreduction pathways are active during cell cycle reentry. Although phospho-ERK1/2 levels were increased in cells treated with GOx and enhanced in cells treated with DNCB and GOx, only with DNCB were high concentrations of GOx able to induce phosphorylation of JNK (, lanes 10–13).
Cell proliferation was then examined in serum-stimulated cells treated with DNCB and/or GOx (). GOx and/or DNCB were added to serum-stimulated cells, and proliferation was examined over a 3-d period without changing the culture media. C10 cells exposed to 1.0 or 2.5 mU/ml GOx proliferated as well as untreated controls, whereas those exposed to doses of GOx ≥5.0 mU/ml failed to proliferate by 3 d (). Greater than 70% of cells arrested in response to all but the highest dose of GOx (15.0 mU/ml) remained viable for at least 3 d ( and not depicted). Caspase 3 was not activated in serum-stimulated cells at any dose of GOx, although it was readily activated after exposure to GOx by staurosporin (unpublished data), indicating that proapoptotic pathways were functional in arrested C10 cells. Cells treated with DNCB alone recovered slowly (), whereas cells treated with DNCB and any dose of GOx did not proliferate (not depicted).
Although DNCB sensitized Prxs to hyperoxidation by GOx, it did not sensitize Prxs to hyperoxidation in response to serum at any time point. Together, these studies indicate that formation of Prx-SOH may not be required for mitogenic signaling during the G0–G1 transition of the cell cycle. In contrast, dose-response experiments with GOx revealed a sharp transition from unimpeded cell proliferation to cell cycle arrest that occurred between concentrations of 2.5 and 5.0 mU/ml, and that arrest was reflected in failure to express cyclin D1.
Transitions between dimers, decamers, and high molecular mass oligomers of Prxs are governed by oxidation state (; ), phosphorylation during G2/M (; ), and other parameters (for review see ). To study the oxidation state of 2-Cys Prxs under various conditions, an immunoblotting method was devised to detect the relative amounts of reduced or oxidized Prx (Prx-SH, Prx-SOH, or Prx-S-S-Prx) versus hyperoxidized Prx (Prx-SOH). With this method, it was possible to estimate the fraction of catalytically active PrxI and -II despite the limitation that the Prx-SOH antibody recognizes hyperoxidized PrxI and -II with equivalent efficiency.
When extracts were resolved by standard SDS-PAGE, total PrxI and -II levels detected by immunoblotting and quantified by densitometry varied less than ±8% during the first 6 h after serum stimulation, with or without GOx (). When probed first for Prx-SOH and then for either PrxI or -II after stripping the membrane, immunoblotting produced reciprocal signals that reflected the fraction of PrxI or -II that was not catalytically inactivated versus the fraction that was inactivated by hyperoxidation. Using densitometry, the levels of reduced/oxidized PrxI (), reduced/oxidized PrxII (), and Prx-SOH () were estimated as a function of GOx concentration after 3 h of exposure and after 3 h of recovery in fresh medium (). At 2.5 mU GOx/ml, >85% of PrxI was hyperoxidized after a 3-h exposure (, lane 6). After recovery, <50% of PrxI was hyperoxidized, and the reduction in Prx-SOH levels () was accompanied by recovery of the signal for reduced PrxI (, lane 15; and ), confirming the activity of retroreduction pathways in C10 cells. PrxII appeared to be less sensitive to hyperoxidation than PrxI; at 2.5 mU/ml GOx (, lane 6), only ∼25% of PrxII had been inactivated by 3 h (). At 10 or 15 mU/ml, both PrxI and -II were quantitatively hyperoxidized (, compare lanes 8 and 9 with lanes 17 and 18), and little signal for reduced PrxI and -II was regained after a 3-h recovery period (). In cells treated with GOx, expression of cyclin D1 was inversely correlated with the levels of Prx-SOH ().
To assess the relationship between Prx hyperoxidation and cellular redox status, GSH levels were measured as a function of GOx concentration after exposure and recovery (). A considerable drop in GSH levels was not observed at 3 h until concentrations of GOx exceeded 5.0 mU/ml, and at all concentrations of GOx, GSH levels increased after recovery in fresh medium (). These results agree well with a report that shows PrxII is hyperoxidized in response to levels of HO that do not inhibit the TrxR–Trx system or deplete cells of GSH (). Hence, cells treated with 5.0 mU/ml GOx for 3 h that retained near normal levels of GSH underwent transient cell cycle arrest, whereas those treated with either 10 or 15 mU/ml GOx that accumulated hyperoxidized PrxI and -II that could not be reduced after 3 h of recovery (), perhaps because of low GSH levels (), were not able to proliferate.
When assessed under standard conditions, the total levels of PrxI and -II did not change during the first 6 h of serum stimulation (). When samples were denatured in the presence of SDS, but without reducing agents to preserve disulfide bonds, gel electrophoresis showed that both PrxI (, lane 1) and PrxII (lane 7) from serum-starved cells were partitioned between 23-kD Prx-SH/Prx-SOH monomers and 38-kD Prx-S-S-Prx homodimers. Upon addition of serum, the levels of PrxI (, lanes 2–6) and Prx II (lanes 8–12) monomers decreased, and PrxI and -II homodimers with intersubunit disulfide bonds increased (, lanes 2–6 and 8–12, respectively). After exposure to 15 mU/ml GOx, all dimers with intersubunit disulfide bonds were lost by 30 min, and only hyperoxidized PrxI and -II monomers were detected for the duration of the experiment (, lanes 13–17; and not depicted). Because homodimers with intersubunit disulfide bonds are produced only during peroxide catalysis (), these results indicate that PrxI and -II metabolize HO produced in response to serum stimulation. Upon hyperoxidation, a condition in which intersubunit disulfide bonds cannot form, only Prx-SOH monomers were observed, as expected.
At 2.5 mU/ml GOx, 85% of PrxI was hyperoxidized, and yet cells expressed cyclin D1 and proliferated normally. In contrast, cells treated with 5.0 mU/ml GOx did not express cyclin D1 or proliferate. To better understand this difference, native gel electrophoresis was used to examine the effect of GOx on the oligomerization state of PrxI and -II. When cell extracts were resolved by electrophoresis in the absence of reducing agents and SDS, immunoblotting indicated that PrxI was organized exclusively in complexes >660 kD (unpublished data). In contrast, PrxII was detected in two sets of bands that we refer to as A–A′ and B–B′ (). Compared with the mobility of native molecular mass markers, A–A′ migrated with an apparent molecular mass of ∼66 kD and B–B′ with a mass of ∼140 kD. Although similar PrxII complexes have been observed in other cell types (), the precise constituents of these complexes are not known.
In extracts of serum-starved cells, bands A and B were the predominant form of PrxII (, lane 1). Addition of DNCB or FBS alone for 3 h did not change the mobility of PrxII on native gels (, lanes 2 and 3), but together DNCB and FBS increased the signal of band B′ (lane 4). Because FBS and DNCB do not induce Prx hyperoxidation (), changes in band B may reflect structural transitions during formation of PrxII-S-S-PrxII dimers during peroxide metabolism (), in agreement with studies that show PrxII metabolizes HO produced in response to growth factors () and terminates HO-activated signaling by phospholipase D1 ().
In response to exposure to 1.0 or 2.5 mU/ml GOx, band B′ increased in abundance relative to band B, perhaps reflecting increased engagement of the PrxII 140-kD complex in peroxide metabolism (, lanes 5 and 6). At concentrations of GOx of 5.0 mU/ml or higher, bands B and B′ disappeared, band A decreased, and band A′ appeared (, lanes 7–9). As observed in , DNCB shifted the dose response for the A–A′ and B–B′ complexes to lower concentrations of GOx (, lanes 10–14).
When reprobed for Prx-SOH, little hyperoxidized PrxII was observed for cells treated with 1.0 mU/ml GOx (, lane 5), whereas hyperoxidized Prx-SOH was observed to comigrate with band B′ in extracts from cells treated with 2.5 mU/ml GOx (, lane 6). At concentrations of GOx ≥5.0 mU/ml, Prx-SOH was incorporated into several discrete high molecular mass complexes (HMCs) with apparent molecular masses >500 kD and considerable levels of A′ accumulated (, lanes 7–9). Recruitment of Prx-SOH into HMCs correlated with loss of signal from the PrxII B–B'complex (, lanes 7–9).
In time course experiments, the A–A′ and B–B′ complexes responded to serum stimulation and cell proliferation and, during recovery from exposure, to 5.0 mU/ml GOx. The levels of the 140-kD B–B′ complex fluctuated during the first 12 h of serum stimulation (, lanes 1–6) and increased markedly in abundance as cells reached confluence 48–96 h later (lanes 8–10). As cells reached confluence, increases in the A–A′ also were observed (, lanes 8–10). Serum stimulation and cell proliferation for >3 d caused no change in the signal for total Prx-SOH detected under reducing and denaturing conditions or Prx-SOH in HMCs detected by native gel electrophoresis (, lanes 2–10). The PrxII complexes were largely unaffected by exposing cells to 2.5 mU/ml GOx for the first 3 h of serum stimulation (, lanes 1–9), even though substantial levels of Prx-SOH were observed under these conditions (, lanes 1–4) and the cultures took slightly longer to reach confluence. Note that 2.5 mU/ml GOx did not increase HMCs containing Prx-SOH.
At 5.0 mU/ml GOx, the B–B′ complex was not observed during the 3-h exposure, HMCs containing Prx-SOH increased in abundance, and cyclin D1 was not expressed (, lanes 1 and 2). After GOx was removed at 3 h, total Prx-SOH levels were reduced over time, and Prx-SOH in HMCs returned to background levels (, lanes 3–9). As signal for Prx-SOH diminished in HMCs, A′ was lost, the B–B′ complex reappeared, and cyclin D1 was expressed (, lanes 5–9). By 96 h, the HMCs and PrxII A–A′ and B–B′ complexes observed by native gel electrophoresis were identical in extracts from cells exposed to all three conditions, even though proliferation to confluence was delayed in cells treated with 5.0 mU/ml GOx (e.g., total cellular protein at 72 h was ∼50% of the 10% FBS control).
Immunofluorescence confocal microscopy was used to localize Prx-SOH within C10 cells treated with various doses of GOx. In all cells, the Prx-SOH antibody reacted with the cell nucleus, but this signal did not correlate with the level of Prx hyperoxidation detected by immunoblotting. In cells treated with 1.0 mU/ml GOx for 3 h, immunostaining was occasionally observed in small patches at the cell periphery (), and this pattern was more obvious in cells treated with 2.5 mU/ml GOx (). At 5.0 mU/ml, GOx staining was observed in a filamentous pattern in the cell cytoplasm (). Prx-SOH in cytoplasmic filaments was particularly evident in cells treated with 10.0 mU/ml GOx, and at 15 mU/ml GOx, staining was prominent around the cell periphery (). At higher doses of GOx, the peripheral Prx-SOH staining pattern correlated with changes in morphology that included a considerable increase in cell diameter. A filamentous cytoplasmic staining pattern for Prx-SOH was not observed in asynchronous cells at any dose of GOx ( and not depicted).
Up-regulation of PrxI is thought to counteract the effects of enhanced oxidant production in tumor cells and thereby promote cell survival and proliferation (; ). To test the effects of Prx expression on responses to GOx, we generated expression vectors for HA-tagged PrxI, PrxII, and PrxII-ΔC, a robust mutant of PrxII that is 100-fold less sensitive to inactivation by HO (; ). HA-PrxI interacts with endogenous PrxI in coimmunoprecipitation experiments, and HA-PrxI and -PrxII are hyperoxidized in response to GOx and reduced during recovery (unpublished data), indicating that HA-tagged Prxs function in peroxide metabolism in a manner similar to their endogenous counterparts. C10 cells were first transfected with expression constructs, and 24 h later the cultures were trypsinized and cells were plated at identical cell densities and synchronized by serum deprivation for 72 h as before. The transfected and serum-starved cell cultures were then treated with 5.0 mU/ml GOx as before.
In synchronized cells, immunoblotting showed HA-PrxI (, lanes 10–12) and HA-PrxII (lanes 13–15) were expressed at levels about fourfold that of their endogenous counterparts. Because of addition of the HA epitope tag and deletion of the PrxII C-terminal domain, HA-PrxII-ΔC comigrated with endogenous PrxII. As compared with untransfected cells (, lane 3) or vector controls (lane 6), expression of catalase (lane 9) and the robust PrxII-ΔC mutant (lane 18) reduced but did not eliminate Prx-SOH levels generated in response to GOx during a 3-h exposure, with 3 h of recovery period as before. HA-PrxI (, lane 12) and HA-PrxII (, lane 15) were hyperoxidized under these conditions and thereby increased the total cellular levels of Prx-SOH as measured by densitometry (). After recovery, expression of HA-PrxI or -PrxII did not reduce the levels of phospho-ERK1/2 or promote expression of cyclin D1 (). Although cells expressing catalase (, lanes 7–9) or PrxII-ΔC (, lanes 16–18) showed lower levels of total Prx-SOH and pERK1/2 after recovery, cells had not expressed cyclin D1 or resumed proliferation by this time. Expression of HA-PrxI or -PrxII did not affect expression of cyclin D1 in response to serum alone (, lanes 11 and 14). When cells treated with 5.0 mU/ml GOx were examined after 72 h of recovery, cells expressing HA-PrxI and -PrxII proliferated in a manner similar to vector controls, whereas cells expressing PrxII-ΔC resumed proliferation earlier during recovery (). Thus, as in serum-stimulated cells, the accumulation of Prx-SOH in cells overexpressing PrxI or -II was correlated with delays in cell cycle progression during recovery.
The susceptibility of 2-Cys Prxs to inactivation by hyperoxidation is highly conserved in eukaryotes, inspiring the hypothesis that the Prx inactivation loop evolved to support peroxide-dependent signaling (). Here, we have examined the relationship between the oxidation state of PrxI and -II and transition from G0 into G1, a portion of the cell cycle known to respond to peroxide-dependent signaling (). Based on the presence of homodimers containing intersubunit disulfide bonds that are generated only during the Prx catalytic cycle, both PrxI and -II appear to metabolize HO produced in response to serum stimulation (), although the source of HO, rate of catalysis, and sites of metabolism are unknown. In cells treated with DNCB (), which depletes cells of GSH and disrupts both Trx and GSH-dependent steps in the retroreduction cycle (), Prxs were sensitized to hyperoxidation by GOx. Nonetheless, in the presence of DNCB, hyperoxidation of Prxs was not observed in serum-stimulated cells at any time point, indicating that hyperoxidation of PrxI or -II may not be required during mitogenic signaling, a result that is in agreement with studies on the role of PrxII in PDGF signaling (). Rather, our studies suggest that oligomers of hyperoxidized PrxII play a role in cell cycle arrest.
Hyperoxidized PrxI accumulated more rapidly in response to exogenous fluxes of HO than did hyperoxidized PrxII (), but levels of PrxI-SOH did not correlate with arrest. In contrast to PrxI, cell cycle progression, arrest, and recovery were correlated with changes in the oligomeric state of PrxII. As assessed by native gel electrophoresis, PrxII existed in two complexes of ∼66 kD (A–A′) and ∼140 kD (B–B′). As cells proliferated to confluence, both A–A′ and B–B′ increased in abundance and B–B′ increased in complexity (). In confluent cells, the B–B′ complex encompassed three distinct bands, suggesting recruitment of additional factors as cells exited the cell cycle, a matter presently under investigation.
In GOx dose-response experiments, C10 cells were able to accumulate substantial levels of hyperoxidized PrxI or -II during mitogenic signaling without marked effects on cell cycle progression. For example, exposure to 2.5 mU/ml GOx resulted in nearly complete hyperoxidation of PrxI and considerable levels of hyperoxidized PrxII, yet C10 cells were able to express cyclin D1 and proliferate. At these levels of exposure, GSH levels were unaffected, and hyperoxidized PrxI and -II were readily reduced once GOx was removed (). At levels of GOx that induced transient cell cycle arrest upstream of cyclin D1, but did not alter GSH levels, the B–B′ PrxII complexes disappeared and hyperoxidized PrxII appeared to be incorporated into HMCs. When oxidative stress was terminated, Prx-SOH in HMCs was readily reduced, the B–B′ complex reappeared, and cells resumed expression of cyclin D1 and cell proliferation. In contrast to the rate of hyperoxidation of PrxII seen in response to GOx (), the dose-dependent structural transitions in PrxII were abrupt (), suggesting a threshold effect for delimiting choices between cell cycle progression and arrest. Prx-dependent thresholds that regulate responses to increasing doses of HO have been observed in yeast (; ).
Electron microscopy shows that in vitro PrxII decamers are able to stack up on one another in an oblique fashion, forming short filaments (). Immunostaining showed that Prx-SOH becomes organized in filamentous structures in the cytoplasm of serum-stimulated C10 cells (). Because this was not observed in asynchronous cells, recruitment of Prx-SOH oligomers into cytoplasmic filaments may be linked to a process active in serum-stimulated cells, such as actin stress fiber formation, or reflect acquisition of chaperone function by hyperoxidized PrxII (). Linking the organization of PrxII to actin stress fiber formation is an attractive possibility, for actin stress fiber formation is a redox-dependent process that regulates signaling through ERK1/2 and expression of cyclin D1 ().
Elevated expression of PrxI, PrxII, and robust mutants of these enzymes has been shown to protect cells against oxidative stress (; ), but these studies were not conducted in synchronized cells. In our previous studies, we have observed very different responses to oxidative stress that depend on cell cycle position and cell density (; ; ; ). Differential sensitivity may be related to wiring of MAPK pathways, for JNK is not activated by HO in synchronized C10 cells (), whereas it is readily activated by HO in asynchronous C10 cells at levels that result in hyperoxidation of <20% of PrxI (; unpublished data).
In synchronized cells, a fourfold increase in expression of HA-PrxI and -PrxII relative to endogenous PrxI and -II did not reduce the level of hyperoxidized endogenous PrxI or -II in response to GOx but, rather, resulted in increased levels of total cellular Prx-SOH. Expression of HA-PrxI and -PrxII also did not promote cell proliferation during recovery (). Together with the GOx dose-response studies, these results indicate that oligomers of hyperoxidized Prx-SOH may be sensed as an anti-mitogenic signal.
Although the propensity of eukaryotic 2-Cys Prxs to be inactivated by HO may provide a “floodgate” for permitting HO to accumulate for redox-dependent signaling, our data provide evidence for an additional hypothesis for the conservation of the inactivation shunt in mammals. Rather than simply buffering intracellular peroxide, Prx enzymes may continuously interpret and report peroxide levels, using their redox and oligomeric states as posttranslational modifications to interface with and modulate redox-sensitive cellular events (). Thus, Prxs may serve as highly sensitive peroxide dosimeters that link oxidant metabolism to a variety of redox-dependent processes required for cell cycle reentry. Upon serum stimulation, these enzymes become engaged in metabolizing HO produced in response to activation of growth factor receptors, actin stress fiber formation, cell migration, and other processes. If oxidant metabolism goes awry or the cell is exposed to threshold levels of exogenous ROS, structural transitions regulated by hyperoxidation would terminate the Prx catalytic cycle, thereby interrupting interactions with regulatory factors or disrupting redox cycling of other factors. Alternatively, PrxII-SOH oligomers may be sensed directly as an anti-mitogenic signal. Linking Prx hyperoxidation to cell cycle progression would allow cells to respond to perturbations in peroxide homeostasis well before depletion of GSH or disruption of the TrxR–Trx system.
It is intriguing that p53 is activated by oxidative stress and that downstream targets of p53 include factors that influence cellular redox state, including sestrins that regenerate Prx activity (). Retroreduction of Prxs by sulfinyl reductases is a reasonable facet of stress responses only if restoration of Prx activity contributes to recovery of activity in cell signaling pathways, for degradation by the proteasome would be equally effective in ridding cells of hyperoxidized Prxs. Hence, 2-Cys Prxs may provide an exquisitely sensitive, widely distributed, and dose-dependent “smoke alarm” for alerting cells to oxidative stress.
C10 mouse lung epithelial cells () were cultured, synchronized, and stimulated with serum as described previously (). For oxidant exposures, recombinant GOx (Roche) in 10 mM phosphate buffer, pH 7.4, was diluted in medium immediately before use. Levels of HO generated by GOx in complete medium were determined as described previously (). DNCB (Sigma-Aldrich) was dissolved in DMSO and used at a final concentration of 5 μM. For growth curves, cells were plated in duplicate in 6-well dishes and treated as described (see Results) before trypsinization and counting with a hemocytometer.
Cell extracts were prepared with NP-40 lysis buffer (150 mM NaCl, 1.0% NP-40, 50 mM Tris, pH 8.0, 1 μg/ml leupeptin, 1 μg/ml aprotinin, 1 mM NaF, 1 mM NaVO, and 1 mM PMSF) or passive lysis buffer (Promega) as noted. At harvest, cells in 60-mM dishes were washed once with cold PBS, pH 7.4, 100 μl of lysis buffer were added, and lysates were collected by scraping with a rubber policeman. The insoluble fraction was pelleted by centrifugation in a microfuge for 5 min, and the protein concentration of the soluble fraction was determined using a protein assay (Bio-Rad Laboratories).
Total levels of PrxI and -II were determined using SDS-PAGE and immunoblotting conditions as described previously (). For assessing relative levels of PrxI and -II that were not hyperoxidized, blots were first probed for Prx-SOH using anti-PrxSO antibody (Lab Frontier) and then stripped at 50°C for 15 min in 62.5 mM Tris, pH 6.8, 2% SDS, and 100 mM β-mercaptoethanol. Stripped blots were washed several times with TBS/T, blocked in 5% nonfat milk in TBS/T for 30 min, and reprobed with anti-PrxI or anti-PrxII antibody. In contrast to probing for Prx-SOH first, probing for PrxI or -II before stripping and reprobing with Prx-SOH antibody did not influence detection of Prx-SOH isoforms.
Antibodies for PrxI (LF-PA0001), PrxII (LF-PA0007), and Prx-SOH/SO (LF-PA0004) were obtained from Lab Frontier. Antibodies to ERK1/2, phospho-ERK1/2, and phospho-JNK were obtained from Cell Signaling Technologies. Anti-phosphotyrosine mouse monoclonal 4G10 was purchased from Upstate Cell Signaling Solutions, anti–cyclin D1 (sc-450) was purchased from Santa Cruz Biotechnology, Inc., and anti-actin from Sigma-Aldrich. Mouse monoclonal anti-HA 12CA5 was a gift from E. Harlow (Harvard University, Cambridge, MA).
Full-length coding sequences for human PrxI and -II were recovered with BamHI ends from pET-17 (Novagen) vectors () using PCR and the following primer sets: PrxI forward, 5′-cgcggatccatgtcttcaggaaatg-3′; PrxI reverse, 5′-cgcggatcctcacttctgcttgg-3′; PrxII forward, 5′-cgcggatccatggcctccggtaacg-3′; PrxII reverse, 5′-gcgggatccctaattgtgtttggag-3′. PrxII-ΔC was generated from a previously described pET-19 (Novagen) PrxII construct () by introducing a stop codon at D188 using the QuikChange site-mutagenesis kit (Stratagene) with the following primers: forward, 5′-GACACGATTAAGCCCAACGTGTAGGACAGCAAGGAATATTTC-3′; reverse, 5′-GAAATATTCCTTGCTGTCCTACACGTTGGGCTTAATCGTGTC-3′. PrxII-ΔC was subcloned from the PrxII pET-19 vector using the PrxII primers listed. PCR products were cloned first using topo-TA cloning vector pCR2.1 (Invitrogen). Positive clones were digested with BamHI, and fragments were subcloned into pCMV-HA to introduce an N-terminal HA epitope tag. The pZeoSV-catalase expression vector () was a gift from D. Lambeth (Emory University, Atlanta, GA). Expression constructs were propagated in DH5α cells and prepared for transfection by alkaline lysis and sedimentation to equilibrium in CsCl. Asynchronous C10 cells at 70% confluence in 60-mm plates were cotransfected with Prx expression plasmids and an EGFP expression vector (pEGFP-N2; CLONTECH Laboratories, Inc.) using Lipofectamine 2000 (Invitrogen) according to manufacturer's protocols. Based on EGFP expression, transfection efficiency was routinely >70%.
C10 cells were lysed in 1% Triton, 50 mM Hepes, 250 mM NaCl, 10% glycerol, 1.5 mM MgCl, 1 mM PMSF, 1 mM EGTA, 2 mM NaVO, 10 μg/ml aprotinin, and 10 μg/ml leupeptin, pH 7.4. GSH was measured as previously described with some modifications (). In brief, samples were mixed 1:1 with 2 mM monobromobimane (Thiolyte; Calbiochem) in 50 mM -ethylmorpholine, pH 8.0, and incubated at RT for 5 min in the dark. Trichloroacetic acid was added to the reaction mixture to a final concentration of 5%. Samples were centrifuged at 3,000 for 5 min, and supernatants were injected onto a Waters Symmetree C-18 column (150 × 4.5 mm). The GSH-monobromobimane adduct was eluted with 10% CHCN/0.25% glacial acetic acid and detected by fluorescence emission of 480 nm after excitation at 395 nm.
C10 cells were plated on glass coverslips in 100-mm tissue culture dishes, synchronized or allowed to grow asynchronously to 70% confluence, and treated as indicated. Coverslips were rinsed with PBS, fixed with 3% paraformaldehyde for 15 min at RT, and washed several times with PBS, and cells were permeabilized with 0.1% Triton X-100 in PBS for 15 min at RT. After gentle washing, coverslips were blocked for 1 h at RT with 10% normal goat serum in PBS and incubated with 1 μg/ml Prx-SOH antibody in PBS with 1% BSA overnight at 4°C. Alexa Fluor 594 (Invitrogen) conjugated goat anti–rabbit secondary antibody at 1 μg/ml in PBS was added for 25 min at RT in the dark. Coverslips were mounted on slides, and images were generated at RT using a confocal scanning laser microscope (MRC 1024 ES; Bio-Rad Laboratories) on a stand (BX50; Olympus), using a 40× Plan-Apo lens (Olympus) with a 0.95 NA and a correction collar. Digital images were collected with Laser Sharp Capture Software (Bio-Rad Laboratories) and processed as black-and-white images. Contrast was adjusted using Photoshop (Adobe). |
Wnts play key roles in cell polarization and migration during vertebrate gastrulation, by signaling through a noncanonical pathway similar to the Frizzled (Fz) signaling pathway that determines epithelial planar cell polarity (PCP) in (). An essential step during Fz/PCP-driven cell polarization in is the localization of PCP components, including the receptor Fz, to specific sites of the cell cortex (). Such subcellular localization of vertebrate PCP components during Wnt-dependent cell polarization and migration has not yet been reported, and the cellular mechanisms by which Wnt/PCP signaling acts remain poorly understood.
Increasing evidence suggests that noncanonical Wnts control cell migration by regulating cell adhesion. Ectopic Wnt5 expression decreases cell adhesion in cultures of dissociated dorsal mesoderm from gastrulas (). Additionally, “knocking down” the presumed Wnt11 receptor Frizzled 7 (Fz7) in embryos causes defects in germ layer separation at the onset of gastrulation (). Fz7 appears to function in this process by interacting with paraxial protocadherin C to control differential adhesiveness between the germ layers (; ). Wnt11 itself has recently been shown to modulate the de-adhesion forces needed to separate zebrafish mesendodermal progenitors from substrates decorated with fibronectin and E-cadherin (; ).
Our previous work shows that during zebrafish gastrulation Wnt11 is required for the polarization and coherent migration of prechordal plate progenitors (, ). The prechordal plate derives from mesodermal and endodermal cells (mesendoderm) that internalize at the dorsalmost germ ring margin, where the embryonic organizer (shield) forms, and then migrate as a coherent group of mesenchymal cells along the overlying ectodermal layer toward the animal pole (; ). We recently provided evidence that Wnt11 controls cell cohesion of prechordal plate progenitors by modulating the subcellular localization of E-cadherin in these cells (). Although such a mechanism could serve to globally regulate cell cohesion, it remains to be established whether Wnt11 possesses a more direct function in the local control of cell contact behavior.
In this study, we show that Wnt11 controls cell contact persistence of gastrulating zebrafish cells at a local level by determining the subcellular distribution of PCP components at the plasma membrane. We find that at cell contact points, Wnt11 triggers the accumulation of its receptor, Fz7, on apposing plasma membranes, along with the intracellular mediator Dsh and Wnt11 itself. These Wnt11-induced Fz7 accumulations increase cell contact persistence in a manner that is dependent on the activity of Flamingo (Fmi), which is an atypical cadherin, but independent of further downstream signaling by RhoA and Rok2. This work suggests that Wnt11 directly controls cell adhesion through local interactions with Fz7, Dsh, and Fmi at cell contacts.
To address the effects of Wnt11 signaling on cell behavior within the zebrafish embryo, we first fused fluorescent proteins (CFP and YFP) to the C terminus of Wnt11 pathway components, including Wnt11, its receptor Fz7, and its intracellular signaling mediator Dsh. and fusion constructs were designed as previously described (; ). Wnt11-YFP activity was confirmed by its rescue of the mutant phenotype (Table S1). Fusions of the zebrafish and homologues (; ; ; ) behave identically in terms of subcellular localization; therefore, all data described in this study were obtained with tagged and untagged versions of (referred to as ). To visualize Wnt11 pathway components in a simple cellular context, we injected our fusion constructs into one-cell–stage embryos and analyzed the resulting expression in animal pole blastoderm cells of pregastrula-stage embryos (30% epiboly; 5 h after fertilization [hpf]). Using this “animal pole assay,” we avoided endogenous expression of and , although maternal is present (Fig. S1, available at ).
To determine if Wnt11 regulates the subcellular distribution of Fz7, we analyzed Fz7-YFP localization in the presence of various amounts of Wnt11 under the animal pole assay conditions. Independent of Wnt11, Fz7-YFP localized uniformly at the plasma membrane and to cytoplasmic “puncta” (). In the presence of low amounts of (5–20 pg mRNA/embryo), however, Fz7-YFP accumulated into patches on the plasma membrane (). This effect appeared to be Fz7 specific, as the uniform distribution of membrane-bound YFP (Lyn-YFP), or another unrelated transmembrane receptor for FGF8 (FGF8R), did not change upon Wnt11 expression (Figs. S2 and S3, available at ). Moreover, Wnt11 induced Fz7-YFP accumulations in a concentration-dependent manner, with increased amounts of Wnt11 enhancing both the number and size of Fz7-YFP accumulations (). In contrast, high amounts of the mutant allele of (40 pg mRNA; , which encodes an inactive C-terminal truncated version of Wnt11 (), did not induce Fz7 accumulations (unpublished data). Additionally, high levels of Wnt3a, which is a canonical Wnt signal (50–100 pg mRNA/embryo), did not result in recognizable changes of Fz7 distribution at the plasma membrane (unpublished data). These data indicate that Fz7 accumulation is both dependent on and specific to Wnt11 activity. Wnt11 is therefore able to influence Fz7 subcellular localization, a potential mechanism to locally regulate cell behavior.
Wnt11 has previously been shown to bind Fz7 (). To test the expectation that Wnt11 colocalizes with Fz7, we analyzed the subcellular localization of Wnt11-YFP along with Fz7-CFP in the animal pole assay (). To distinguish between cell-autonomous and nonautonomous Wnt11 activity, we also transplanted Wnt11-YFP–expressing cells into the animal pole of host embryos ubiquitously expressing either the plasma membrane marker Lyn-CFP () or Fz7-CFP (). In the absence of Fz7 overexpression, Wnt11-YFP was predominantly found in cytoplasmic structures of Wnt11-producing cells, but could also be detected in small puncta at the plasma membrane of both Wnt11-producing () and -receiving cells (). In contrast, in the presence of Fz7-CFP, Wnt11-YFP was found in both producing () and receiving cells () to be enriched at regions that colocalized with Fz7-CFP accumulations. These Wnt11-YFP accumulations were larger than the plasma membrane puncta found in the absence of exogenous Fz7 ( and G–I). These observations suggest that Wnt11, by binding to Fz7, induces both cell-autonomously and nonautonomously local accumulations of Fz7 at the plasma membrane.
Wnt11-induced Fz7 accumulations predominantly localized to cell contacts, suggesting that Wnt11 locally functions at those sites. To investigate whether Fz7 accumulations form on both contacting plasma membranes or whether they are restricted to one cell, we transplanted Fz7-YFP–expressing cells into host embryos expressing both Wnt11 and Fz7-CFP (). Consistent with accumulation on both cells, Wnt11-induced Fz7 accumulations located between Fz7-CFP–positive host cells and Fz7-YFP–positive donor cells contained both Fz7-YFP and -CFP (). Importantly, Wnt11 induced Fz7 accumulations at host–host and donor–donor cell contact sites (; and not depicted), confirming our previous observation () that Wnt11 acts both cell-autonomously and nonautonomously.
A crucial step in noncanonical Wnt signaling is the relocalization of Dsh from the cytoplasm to the plasma membrane, an event that also occurs in response to Fz7 overexpression (; ). To test if Wnt11-induced Fz7 accumulations are sites of local Wnt11 activity, we looked for changes in Dsh localization. When coexpressed with Fz7, Dsh-YFP localized uniformly to the plasma membrane, as expected (). However, when Dsh-CFP was coexpressed with Fz7-YFP and Wnt11 (20 pg mRNA), it localized to the resulting Fz7-YFP accumulations (). This suggests that Wnt11-induced Fz7 accumulations are sites of Wnt11 activity.
Wnt11 and Fz7 have previously been implicated in cell polarization and adhesion during and zebrafish gastrulation (; , ; ; ; ). To test whether Wnt11-induced Fz7 accumulations influence cell morphology or adhesion, we recorded 3D time-lapse videos of cells in the animal pole assay. We found that, on average, Fz7-YFP accumulations at the first time point of the video occupied only 70% (± 3.6% SEM) of the total cell contact length, whereas at the last time point before separation, 96% (± 2.2% SEM) was occupied (). Consistent with this, when two cells containing accumulated Fz7-YFP at their contact site moved apart, cells separated last at the site of Fz7 accumulation (; Video 1, available at ). Intriguingly, these separating cells also showed a deformation of their plasma membranes toward the remaining cell contact (), suggesting that the Fz7 accumulation site is resistant to separation. In contrast, cells uniformly expressing either Fz7-YFP or a membrane-tethered version of YFP (Lyn-YFP) separated more evenly along their cell contacts (; Videos 3 and 4). To quantify these observations, we measured the angle between contacting plasma membranes at the last time point before cell separation (illustrated in ). Consistent with a pronounced deformation, we found a significantly bigger angle in the presence of Wnt11 (90 ± 4.4° SEM; P < 0.05), as compared with cells separating in the absence of Wnt11 (Fz7-YFP, 70 ± 3.73° SEM; Lyn-YFP, 65 ± 3.2° SEM; ). We also observed similar separation behavior by Fz7- and Wnt11-overexpressing epiblast cells in the germ ring of shield-stage embryos (6 hpf; Video 2), where endogenous Wnt11 is expressed and active (). Overall, our observations suggest that Wnt11-induced Fz7 accumulations increase the persistence of cell contacts in the immediate area.
To analyze the influence of Wnt11-induced Fz7 accumulation on cell contact persistence, we determined the rate of cell contact shrinkage during separation and the time of cell separation in the presence and absence of Fz7 accumulation (). Although control cells had rapid shrinkage of their cell contact length during the last phase of separation (Fz7-YFP, 1.5 ± 0.19 μm/min SEM; Lyn-YFP, 1.4 ± 0.15 μm/min SEM; and ), cells containing Fz7 accumulations showed a significantly slower shrinkage rate (0.65 ± 0.06 μm/min SEM; P < 0.05; and ). Additionally, during the timeframe of our videos (75 min), we saw consistently longer contact times and a higher percentage of cell contacts that did not separate in the presence of Fz7 accumulations (, G and G′). These combined differences in contact time and cell separation are given in as the ratio of cells displaying contact times longer than 30 min (including nonseparating cells) versus cells separating within the first 30 min of the video. In sum, Wnt11-induced Fz7 accumulations are associated with a local increase in cell contact persistence, most likely reflecting stronger cell adhesion at these contact sites.
Wnt11-induced Fz7 accumulations could be the cause or consequence of increased cell contact persistence. To distinguish between these two possibilities, we compared the rate of shrinkage of cell contact length to that of local Fz7 accumulations. If Fz7 accumulations are only secondary to increased cell adhesion, the length of cell contacts and Fz7 accumulations should shrink at a similar rate. In contrast, if Fz7 accumulations cause increased cell contact persistence, the rate by which Fz7 accumulations shrink should be considerably lower than the rate by which the cell contacts shrink. Consistent with Wnt11-induced Fz7 accumulations causing increased contact persistence, we found that Fz7 accumulations shrink slower than the cell contacts ().
In , Fz and Dsh are proposed to form a signaling complex with the atypical cadherin Flamingo (Fmi) that locally directs cytoskeletal reorganization and/or adhesion (for review see ). Furthermore, in zebrafish, Wnt11 and Fmi appear to cooperatively control gastrulation movements (). To determine if Fmi plays a role in the local increase of cell contact persistence resulting from Fz7 accumulation, we first analyzed zebrafish Fmi2 (also named Celsr2) localization in the animal pole assay. Fmi2 is maternally provided and zygotically expressed throughout gastrulation (unpublished data). In embryos expressing YFP-tagged Fmi2 (Fmi2-YFP), Fmi2-YFP localized at the plasma membrane and accumulated at cell contacts both with and without exogenous Wnt11 (). Interestingly, coexpression of Fmi2-YFP with Fz7-CFP in the absence of Wnt11 resulted in the preferential accumulation of Fz7-CFP in places of increased Fmi2-YFP signal (). Thus, Fmi2 is able to induce accumulation of itself and Fz7 independently of Wnt11. In addition, Wnt11-CFP localized to Fmi2-YFP accumulation when coexpressed with untagged Fz7, indicating that all three components colocalize (). Fmi2 is therefore a likely component of Fz7/Wnt11 accumulations, possibly acting to regulate cell adhesion at cell contacts.
Using gain- and loss-of-function experiments, we investigated the role of Fmi in cell contact behavior in the animal pole assay. Cell contact persistence, as measured by the time needed for cells to separate and the percentage of nonseparating cells, was enhanced in cells expressing only Fmi2-YFP compared with cells expressing Wnt11 and Fz7 (, G and H–H″; Video 5, available at ). To reduce endogenous Fmi activity, we injected morpholinos (MOs) targeted against the three genes (, , and ; 4 ng/embryo each; unpublished data). In Fmi-MO–injected embryos, Wnt11/Fz7-mediated cell contact persistence was reduced (, H–H″; Video 6), and the shrinkage rate of cell contacts during the last phase of separation was significantly increased (0.88 μm/min ± 0.06 SEM; P < 0.05; , I and I′). Loss of Fmi function at sites of Fz7 accumulation therefore causes cells to separate faster, suggesting that Fmi contributes to Wnt11-induced Fz7 accumulation control of cell contact persistence.
Wnt11-induced Fz7 accumulations could directly affect cell contact persistence or act via downstream mediators of Wnt11 signaling, such as RhoA and Rok2 (for review see ). To distinguish between these possibilities, we monitored the effects of Wnt11-induced Fz7 accumulations on cell contact behavior within the animal pole assay when RhoA and Rok activity is decreased. Using MOs, we “knocked down” zebrafish and , as they are expressed during gastrulation and function downstream of Wnt11 (; ). To further decrease RhoA/Rok function, we incubated -MO–injected embryos in the specific Rok inhibitor Y-27632 before image acquisition. Loss of RhoA/Rok function did not significantly interfere with the formation or effect of Fz7 accumulations on cell contact persistence (contact shrinkage rate in last phase of separation, 0.75 ± 0.14 μm/min SEM; P = 0.2; Fig. S4, available at ). This suggests that Wnt11 downstream signaling through RhoA and Rok is not directly involved in mediating enhancement of contact persistence by Wnt11-induced Fz7 accumulation.
We next wanted to determine if Wnt11 locally controls cell adhesion in gastrulating cells and whether Fz7 accumulation plays a role in such a process. We first sought to monitor endogenous Wnt11 and Fz7 activity in gastrulating cells of shield stage embryos (6 hpf) by expressing moderate amounts of Dsh-YFP in these cells. We used Dsh-YFP because (a) Fz7 recruits and colocalizes with Dsh at the plasma membrane, as seen in the animal pole assay (), (b) a fluorescently tagged live marker gene is needed to dynamically monitor endogenous Fz7 plasma membrane accumulation, and (c) moderate expression (60 pg mRNA) had no obvious effect on tissue morphogenesis during gastrulation (not depicted). We analyzed Dsh-YFP localized in both epiblast cells, which endogenously express and (; ; ; ; ; ), and hypoblast cells, which endogenously express and require Wnt11 for proper cell polarization and coherent migration (, ).
Dsh-YFP was observed in epiblast and hypoblast cells as puncta, both in the cytoplasm and at the plasma membrane (; and not depicted), most likely reflecting dynamic Dsh assemblies (). Localization of Dsh-YFP puncta to the plasma membrane was seen to be regulated by endogenous Wnt11 signaling, as the number of plasma membrane puncta per cell within the shield-stage germ ring (6 hpf) was reduced in mutants (quantified in 3D; P < 0.05; ). Additionally, the decrease in Dsh-YFP puncta at the plasma membrane in mutant embryos was rescued by injecting 20 pg mRNA (P < 0.05; ). As expected, if Wnt11 is locally recruiting Dsh, we found that in embryos with ubiquitous Dsh-CFP and mosaic Wnt11-YFP expression, secreted Wnt11-YFP colocalizes with Dsh-CFP at the plasma membrane of both producing and receiving cells (, E–G′). This recruitment and colocalization of Wnt11 and Dsh in puncta is presumably through the function of endogenous Fz. Consistent with this, we found Dsh puncta at shield stage to localize to both contacting plasma membranes, as was shown for Fz7 in the animal pole assay (6–7 hpf; ). Overall, Dsh-YFP puncta at the plasma membrane appear to highlight subcellular sites of endogenous Wnt11 and Fz7 activity.
It is possible that these subcellular sites of endogenous Wnt11/Fz7 activity modulate contact persistence of gastrulating cells, similar to the Wnt11-induced Fz7 accumulations in the animal pole assay. To address this possibility, we observed both puncta and separation behavior of ectodermal and mesendodermal cells within the germ ring of late shield-stage embryos (7 hpf). As Dsh puncta at the plasma membrane are only partially reduced in mutants (), it is unlikely that all Dsh puncta correspond to endogenous Wnt11-induced Fz7 accumulations. Thus, to specifically address the role of Wnt11 activity under endogenous Fz7 conditions, Wnt11-YFP puncta at the plasma membrane were followed in gastrulating cells that also express the membrane marker GPI-anchored RFP (mem-RFP). We found that Wnt11 puncta localized to contact sites of separating cells and remained at the last contact point before separation ( and Video 7, available at ). To compare the effect of Wnt11 puncta on contact persistence with that of Wnt11-induced Fz7 accumulations, we also expressed Wnt11-YFP and mem-RFP in the animal pole assay. By determining the contact time of separating cells and the percentage of nonseparating cells, we found that cells containing Wnt11 puncta at their contact sites exhibit increased contact persistence compared with cells expressing Lyn-YFP or Fz7-YFP (; and Video 8). This Wnt11 enhancement of contact persistence was considerably smaller than that of Wnt11-induced Fz7 accumulation, an expected result given the small size of Wnt11 puncta (). In addition, Wnt11 puncta initially occupied only 14 ± 2% SEM of the total cell contact length, but covered 57 ± 7% SEM at the last time point before separation. This indicates that Wnt11 puncta persist at the last point of cell contact, similar to Fz7 accumulations ().
We previously found that Wnt11 controls directed and coherent movements of prechordal plate progenitors during gastrulation (). To determine whether Wnt11 coordinates directed prechordal plate migration by regulating cell contact persistence, we compared the separation behavior of these cells in wild-type versus mutant embryos. As expected, we saw shorter contact times in mutants (; and Videos 9 and 10, available at ), and less compaction of the tissue (, F′ and G′). These results indicate that endogenous Wnt11 function is required for cell contact persistence within the forming prechordal plate, a potential mechanism to facilitate directed and coherent movement during gastrulation.
Central to this study is our finding that Wnt11 induces the accumulation of its receptor Fz7 and its intracellular mediator Dsh at distinct sites of cell contacts. Direct binding of Wnt11 to Fz7 may be required to trigger this accumulation, as we found that the C-terminally truncated inactive version of Wnt11 potentially lacking the binding site to Fz (Slb/Wnt11; ) is not sufficient. This is consistent with previous data indicating that XWnt11 and XFz7 biochemically interact in (). In addition, relocation of Dsh from the cytoplasm to sites of Fz7 accumulation presumably occurs through direct binding of Dsh to Fz7 ().
Significantly, we demonstrate that Wnt11-induced Fz7 accumulations locally increase the persistence of cell contacts. This is based on the observations (a) that separating cells containing Fz7 accumulations at their contact sites break their cell contact last at sites of Fz7 accumulation, (b) that these cells remain in contact longer and exhibit deformation of their membranes toward the point of cell contact, and (c) that during the course of cell separation the size of Fz7 accumulations shrink slower than the total cell contact length. Therefore, we propose that Fz7 accumulations are adhesive subdomains within cell contacts.
It is unlikely that Fz7 or Wnt11 molecules directly modulate cell adhesion, nor do our data support a critical role of downstream signaling through Wnt11 effectors such as RhoA and Rok2 in this process. One possibility is that Wnt11-induced Fz7-accumulations interact with an adhesion molecule mediating these adhesive functions. Several lines of evidence point to the involvement of the atypical cadherin Fmi, a key component of PCP signaling in (). Previous work showed a functional interaction between Fmi and Wnt11 in controlling convergent extension movements during zebrafish gastrulation (). In addition, Fmi exhibits adhesive function in cultures of dissociated zebrafish embryonic cells, as well as in S2 cell cultures (; ; unpublished results). We not only found that Fmi colocalizes with Fz7 accumulations but also discovered that endogenous Fmi function is required for the cell contact persistence mediated by Fz7 accumulation. Collectively, our data strongly suggests that Fmi plays a key role in controlling cell adhesion at sites of Wnt11-induced Fz7 accumulation.
In , localization of Fmi to proximal and distal cortices of pupal wing epithelial cells is essential for proper Fz and Dsh localization to the distal cortex (; ; ), where a potential protein complex of Fmi, Fz, and Dsh mediates cell polarization (). Our similar finding of Fz7, Dsh, and Fmi2 colocalizing at cell contacts in zebrafish points to a shared role for subcellular distribution and function of these PCP components in and zebrafish. Although this is an attractive hypothesis, there are also differences between PCP signaling in and zebrafish. First, no Wnt ligand in PCP signaling has yet been identified in . Second, although Fz7, Dsh, and Fmi2 colocalize on both contacting plasma membranes in zebrafish, only Fmi localizes to both membranes in . Despite these differences, Fmi might serve as a common effector for PCP signaling by recruiting other PCP components to specific sites on the plasma membrane and directly regulating cell adhesion. This notion is supported by our observation in zebrafish that Fmi can promote Fz7 accumulation at cell contact sites independently of Wnt11, suggesting that Fmi, through its effect on Fz7, can determine sites of local Wnt11 activity.
How does the function of Wnt11/Fz7 in controlling cell contact persistence relate to previous findings about the morphogenetic activity of Wnt/PCP signaling in zebrafish and ? It has been proposed that Wnt/PCP signaling controls cadherin-mediated rearrangement of cells as they polarize. In , PCP signaling mediates cadherin recycling required for the packing of wing epithelial cells into hexagonal arrays, a process accompanied by asymmetric localization of PCP components (). Similarly in zebrafish, Wnt11-mediated E-cadherin endocytosis and recycling controls cohesion of prechordal plate progenitors required for directed and coherent cell migration (). Thus, there is a basis for cadherin endocytosis/recycling serving as a common mechanism by which Wnt/PCP signaling globally regulates cell cohesion and junctional remodeling. However, Wnt/PCP components may also directly control cell contact dynamics in vertebrates. Based on this study, we propose that in zebrafish Wnt11 locally controls cell adhesion by interacting with Fz7, Dsh, and Fmi at the plasma membrane. Whether the effects of Wnt11 on Fz7 accumulation and E-cadherin endocytosis/recycling are independent activities of Wnt11 in controlling cell adhesion remains to be established.
But what is the relevance of our findings from the animal pole assay for the endogenous function of Wnt11 during zebrafish gastrulation? Our observation that Dsh recruits to sites of secreted Wnt11 puncta at the plasma membrane under endogenous levels of Fz suggests that these Dsh puncta mark sites of endogenous Wnt11/Fz activity. This is also supported by the fact that the number of Dsh puncta at the plasma membrane of germ ring cells is controlled by endogenous Wnt11. Furthermore, we found that Wnt11/Dsh puncta at cell contact sites display properties similar to Wnt11-induced Fz7 accumulations in the animal pole assay. In particular, Wnt11/Dsh puncta (a) localize to both contacting plasma membranes, (b) occupy the last point of cell contact, and (c) enhance the persistence of cell contacts. Relating these findings to gastrulation movements, we determined that, as expected, the persistence of cell contacts between migrating prechordal plate progenitors is decreased in mutant embryos. This is consistent with previous data showing that mutant prechordal plate progenitors migrate less coherently and directed (, ). In total, these findings support a critical endogenous role for Wnt11/Fz7/Dsh accumulations in promoting the local persistence of cell contacts, a mechanism by which Wnt11 controls coherent cell migration.
Interestingly, recent studies provide evidence for a connection between noncanonical Wnt signaling and focal adhesion points. In it has been shown that the transmembrane heparan sulfate proteoglycan Syndecan4 interacts with Fz7 and Dsh. Furthermore, binding of Syndecan4 to fibronectin facilitates Fz7 recruitment of Dsh to the plasma membrane and, subsequently, the activation of noncanonical Wnt signaling (; ). Considering the requirement for Syndecan4 during convergent extension movements (), as well as the colocalization of Dsh and Fz with focal adhesion components in cultured cells (), it is conceivable that Wnt11/Dsh puncta at the plasma membrane share similarities in composition and function with focal adhesion points. Future experiments addressing the localization and function of Syndecan4 and other focal adhesion components, in respect to Wnt11/Dsh puncta, will provide insight into the role of Wnt/PCP signaling in regulating local cell adhesion in different developmental processes.
Fish maintenance and embryo collection was performed as previously described (). For mutant analysis, we used homozygous embryos ().
Fz7a/Fz7b-YFP/CFP fusions were constructed as C-terminal fusions of zebrafish (NM_131139; ) and (NM_170763; ; ) cDNA fused to either or (CLONTECH Laboratories, Inc.) as previously described for Fz-GFP fusion (). Full-length cDNA of zebrafish (NM_170763; (; ) was amplified by PCR using a cDNA library of shield- and bud-stage embryos, followed by subcloning into pCS2+ expression vector. Wnt11-YFP/CFP fusions were constructed as C-terminal fusions of zebrafish cDNA () with either (an EYFP-derivative that was provided by A. Miyawaki, RIKEN, Saitama, Japan) or (an improved eCFP-derivative; ). A linker sequence of nine amino acids (EFSSGSIDG) was introduced between Wnt11 and YFP/CFP. For Dsh-YFP/CFP fusion constructs, a C-terminal fusion of cDNA to either or was generated by replacing the with of a previously described dsh-egfp fusion construct (). For Lyn-YFP/CFP fusions (Lyn-YFP/CFP), the myristoylation membrane localization signal of Lyn tyrosine kinase was fused to the N terminus of either eCFP or venusYFP. The membrane-RFP (memRFP) construct was provided by A. Siekmann (Max-Planck-Institute for Cell Biology and Genetics, Dresden, Germany) and encodes a monomeric form of RFP, which is targeted to the plasma membrane by a GPI anchor. Zebrafish , the orthologue of mouse , was cloned from a zebrafish gastrula library, and a full-length version of Fmi2 was fused to the N terminus of venusYFP and cloned into the + to generate . Details of cloning zebrafish will be published elsewhere (FCB and MT, manuscript in preparation). Zebrafish FGF8 receptor fused to RFP was provided by M. Kolanczyk and M. Brand (Max-Planck-Institute for Cell Biology and Genetics, Dresden, Germany).
For mRNA synthesis, pCS2+ expression vectors containing the cDNAs for different constructs were linearized by Not1 restriction digest and transcribed by Sp6 mRNA polymerase, as previously described (). Description of mRNA amounts per experiment is stated in the figure legends. For Fmi-YFP expression, plasmid DNA was injected in two-cell–stage embryos. To knock down RhoA function, embryos were injected with a mix of 8 ng of MO oligonucleotides directed against the zebrafish (5′-TCTTGCG AATTGCTGCCATATTTGC-3′) and (5′-AGCTTCTTACGGATAG CTGCCAT-3′) genes. and morphant embryos exhibited similar convergent extension defects, as previously reported (). For knocking down Fmi activity, we used MOs targeting the start codons , , and , as follows: MO (5′-CATGGTGTAAAACTCCGCAAACAGG-3′), MO (5′-CATCCATATCACTGGTAATTCCATG-3′), and MO (5′-CAAAGAGCAACAAATCCCCCTTCAT-3′)
Dechorionated embryos (4 hpf) were incubated with E2-media solution containing 50 μM of the Rok-specific inhibitor Y-27632 (Tocris Bioscience). For time-lapse imaging, inhibitor-treated embryos were mounted in 1% Agarose containing the inhibitor and covered with inhibitor containing E2-media.
In situ stainings were performed as previously described (; ). In situ probes were synthesized from cDNA for , , and (; ; ), using a DIG RNA-labeling kit (Roche).
For localization studies, embryos were fixed in 4% PFA in PBS overnight at 4°C, followed by washing in PBST (PBS + 0.1% Tween20), dechorionation, and mounting in Agarose-coated dishes in PBST. Confocal images were acquired at room temperature using a TCS-SP2 confocal microscope (HCX APO L 63×/0.9 W UVI dipping objective and confocal software; both Leica). For Fig. S2, a LSM-META confocal microscope with Achroplan 40×/0.8 W dipping objective was used (Carl Zeiss MicroImaging, Inc.). For colocalization studies, CFP/YFP or GFP/RFP channels were acquired by sequential scanning. Confocal images were analyzed and quantified using Volocity 3.0 (Improvision), LSM 5 Image Browser (Carl Zeiss MicroImaging, Inc.), and ImageJ v. 1.29–1.32 (). To analyze significance, P values were determined in Microsoft Excel (unpaired test, two-tailed distribution). For time-lapse imaging, embryos were dechorionated and mounted in 1% low melting point Agarose covered with embryo media, and imaging was performed at room temperature using water-dipping lenses. Single-color, time-lapse imaging was performed by two-photon microscopy using the Radiance system and a Plan Apo 60×/1.2 WI objective (both Bio-Rad Laboratories), as previously described (). For two-color, time-lapse imaging of cells expressing YFP and RFP signals, a LSM 405-/594-nm confocal microscope with C-Apochromat 63×/1.2 W correlation objective was used (Carl Zeiss MicroImaging, Inc.). YFP and RFP channels were acquired by simultaneous scanning using 488-/594-nm laser lines for excitation, and 490-/590-nm NFT filters and BP505-580-/LP610-nm filters for separation and detection of the signals. To acquire 3D time-lapse videos, defined z stacks with a step size of 1.6–1.8 μm were recorded over a time interval of ∼145 s (depending on the exact size of the z stack). 3D time-lapse videos were analyzed using Volocity 3.0 or LSM 5 Image Browser (a) by randomly choosing cell contacts from the first time frame of the videos and tracking their separation behavior in 2D (by selecting one z section/time point) until the end of the video or until cells were completely separated (“forward tracking”; ; 6, H–H″; and 8, B–D and H–J) and (b) by selecting cell contacts that separated within the timeframe of 75 min and tracking them backward from the time point of complete separation until the first timeframe of the videos (“backward tracking”; I, 6 I′, and S4). Differential interference contrast time-lapse videos showing prechordal plate morphogenesis at the shield stage (6 hpf) of embryos were taken using an Axioplan2 microscope (Carl Zeiss MicroImaging, Inc.) and a 40× 0.8 WPh2 Achroplan objective, as previously described (). To acquire 3D time-lapse videos, defined z stacks with a step size of 3–4 μm were recorded in time intervals of 30 s using Openlab imaging software (Improvison). To analyze the contact time of cells, a 4D version of Image was used to track randomly chosen cell contacts from the first time point of the videos in 3D until cells completely separated.
Table S1 lists the biological activities of Wnt11-YFP compared with Wnt11. Fig. S1 shows the expression patterns of , , and . Fig. S2 shows the quantification of intensity increase at Fz7 accumulations. Fig. S3 shows distribution of FGF8 receptor–RFP at Fz7 accumulations. Fig. S4 shows quantification of separation dynamics in the absence of RhoA/Rok2 function. Video1 shows separating cells expressing Fz7-YFP and Wnt11. Video2 shows separation of gastrula-stage cells expressing Fz7-YFP and Wnt11. Video 3 shows separating cells expressing Fz7-YFP. Video 4 shows separating cells expressing Lyn-YFP. Video 5 shows separating pregastrula-stage cells expressing Fmi2-YFP. Video 6 shows separating pregastrula-stage morphant cells expressing Fz7-YFP and Wnt11. Video 7 shows separating gastrula stage cells expressing Wnt11-YFP and memRFP. Video 8 shows separating pregastrula stage cells expressing Wnt11-YFP and memRFP. Video 9 shows morphogenesis of wild-type prechordal plate progenitors. Video 10 shows morphogenesis of mutant prechordal plate progenitors. Online supplemental material is available at . |
Microvilli are slender F-actin–containing structures on the apical surface of many epithelial cells. Perhaps the best-studied examples are the densely packed microvilli of brush borders on intestinal and kidney proximal tubule epithelial cells, where microvilli are believed to increase the surface area for absorption. Less well ordered microvilli, having a somewhat different complement of actin binding proteins, are found on the apical aspect of other epithelial cells, such as the placental syncytiotrophoblast (; ). In neither case is the regulation of these structures understood.
One protein common to both types of microvilli is ezrin, a member of the ezrin/radixin/moesin (ERM) family of membrane-cytoskeletal linking proteins. Members of this family have an ∼300-residue N-terminal 4.1 ERM (FERM) domain, followed by a central ∼150-residue region predicted to be largely α-helical, and terminate in an ∼100-residue domain known as the C-ERMAD (C-terminal ERM association domain), as it has the ability to bind the FERM domain of all family members (); an F-actin binding site lies in the last ∼30 residues of the C-ERMAD (; ). ERM proteins are conformationally regulated, as the F-actin binding site in the C-ERMAD, and some sites for association with membrane proteins in the FERM domain, are masked in dormant molecules when these two domains are associated (). Activation to release the association and expose these binding sites can occur through PIP2 binding and subsequent phosphorylation of a C-terminal threonine (567 in ezrin), found on the interface between the FERM domain and the C-ERMAD (; ; ; ; ).
The cytoskeletal-membrane linking properties of ERM proteins are attributed to their ability to bind F-actin through their C-terminal domain and membrane proteins, such as CD44, CD43, and ICAM-1-3, through their FERM domain (; ; ; , ; ; ). In addition to this direct linkage with transmembrane proteins, the FERM domain binds the related scaffolding proteins EBP50 (ERM binding phosphoprotein of 50 kD)/NHERF1 and E3KARP/NHERF2, proteins enriched in epithelial microvilli (; ). These proteins consist of two N-terminal PDZ (postsynaptic density/95-discs large/zona occludens-1) domains and a C-terminal region that binds tightly to isolated FERM domains (). The EBP50 binding site on the FERM domain lies on the same surface occupied by the last helix of the C-ERMAD in the dormant protein, thereby providing a physical explanation for its masking in dormant ezrin (; ). EBP50 binds the C-terminal tails of many transmembrane proteins, including the CFTR, the β2-adrenergic receptor, and the PDGF receptor to regulate aspects of their function (; ; ; ; ; ; ). Previously, we identified EPI64 (EBP50-PDZ interactor of 64 kD) from extracts of placental microvilli, the richest known source of both ezrin and EBP50, as a protein that binds the PDZ domains of EBP50 (). EPI64 is a cytosolic protein of 508 residues containing an N-terminal Tre-2/Bub2/Cdc16 (TBC) domain and ending in the sequence DTYL, which binds preferentially to the first PDZ domain of EBP50 ().
The current model based on biochemical interactions () suggests that microvilli should exhibit a uniform distribution of phosphorylated ezrin, EBP50, and EPI64 along their length. We used high-resolution fluorescence microscopy to assess this prediction and found that distinct subdomains exist in microvilli whose distribution can be regulated by EPI64. By exploring the effects of expressing different domains of EPI64, we found that when EPI64 is unable to be linked to ezrin, cells have a reduced number of microvilli. We have traced this phenomenon to the mislocalization of EPI64's TBC domain, which we find binds directly to Arf6-GTP.
To localize specific proteins in microvilli by immunofluorescence microscopy, 10–15 confocal sections 0.2 μm apart covering just the apical aspect of stained human JEG-3 syncytiotrophoblast cells were collected and merged. When the plasma membrane was stained with fluorescently tagged WGA and ezrin localized with specific antibodies, a high degree of colocalization was found. Comparing these with phalloidin staining for F-actin, the microvillar rootlet extending into the cell body was clearly visible, and this allowed us to determine the orientation of microvilli in triple-label studies (). To quantitate the degree of colocalization of three markers, the fluorescent intensity along the length of a minimum of 25 microvilli was determined, and the mean was assessed. It should be noted that because microvilli vary in length (from ∼0.75 to 1.25 μm), this mean gives a measure of colocalization but does not accurately reflect sharp transitions seen in individual microvilli (Fig. S1, available at ). As expected, the distribution of ezrin closely follows that of the plasma membrane ().
When the distribution of T567 phosphorylated ezrin (pERM) was compared with the plasma membrane, it was found that they were largely overlapping but that the pERM localization was slightly more enriched in the distal portion of microvilli and reduced toward their base (, arrows; and Fig. S1). Although this is a subtle effect in the means (), it is also seen in other cell types, such as intestinal Caco-2 epithelial cells and porcine kidney proximal tubule LLC-PK1 cells (not depicted). In similar analyses, EBP50 was found to precisely colocalize with the plasma membrane–like ezrin ().
In contrast to the localizations of ezrin and EBP50 in JEG-3 cell microvilli, the localization of EPI64 was more variable. Traces of individual microvilli never showed a clear colocalization with other markers but, rather, were enriched in different regions of individual microvilli ( and Fig. S1). As such, the mean distribution curve shows no specific localization, though the peak of the curve is near the microvilli bases. Remarkably, EPI64 was sometimes enriched in the region of the microvillus rootlet with no ezrin or plasma membrane colocalization.
To begin examining how microvillar subdomains arise, we investigated the effects of overexpressing EPI64 and EBP50. Overexpressing EPI64 resulted in the majority of microvilli exhibiting enrichment of EPI64 in a region immediately below ezrin and associated with the microvillar rootlet (, arrows). Moreover, overexpressing EPI64 caused EBP50 to now be largely coincident with EPI64 and the distribution of both EBP50 and EPI64 to be distinct from that of ezrin (). Thus, overexpressing EPI64 results in redistribution of EBP50 and enhances the distinction between microvillar subdomains. In contrast to EPI64, overexpressing EBP50 had no discernable effect on the distribution of ezrin, pERM, EBP50, or EPI64 in JEG-3 cells (unpublished data). We therefore focused our studies on the role of EPI64.
To further investigate the role of EPI64, we generated several N-terminally tagged mutants (). Because the C-terminal DTYL sequence of EPI64 binds the PDZ domains of EBP50, we made a mutant (EPI64-LA) with an additional C-terminal alanine known to abolish this interaction (). Some TBC domains are known to accelerate the catalytic activity of Rab GTPases through a catalytic arginine finger motif (; ), so we mutated this residue, arginine 160 in EPI64, to generate EPI64-R160A. In addition, we made an N-terminal construct (EPI64-NT) containing the TBC domain and a C-terminal construct (EPI64-CT) consisting of the C-terminal 189 residues. The expression of these constructs modestly enhanced the level of EPI64 expression as revealed by Western blotting ().
Overexpressing EPI64-LA resulted in the reduction or loss of microvilli from JEG-3 cells as seen by ezrin or F-actin staining, without noticeably affecting the levels of ezrin, pERM, or EBP50 in these cells (). Instead of being enriched in the apical aspect of cells, like overexpressed wild-type EPI64, EPI64-LA was found diffuse in the cytoplasm (). To quantify the microvillar loss phenotype, at least 200 transfected cells were assessed for the presence of microvilli, with 46% of the transfected cells lacking or exhibiting greatly reduced numbers of microvilli, compared with 25% of untransfected or wild-type EPI64 overexpressing cells that had reduced numbers of microvilli. In addition, a small percentage of EPI64 and EPI64-LA overexpressing cells contained vacuoles (), which are discussed in the next section. In contrast to EPI64-LA, EPI64-R160A overexpression was similar to overexpressing wild-type EPI64.
To investigate the mechanism of microvillar loss by EPI64-LA, we overexpressed domains of the protein. Overexpressing EPI64-CT resulted in a reduction of microvilli and an increase in the number of ruffling cells (), with the ruffles enriched in the transfected protein (not depicted). Similar to overexpressing EPI64-LA, overexpressing EPI64-NT, containing the TBC domain, resulted in microvillar loss. A similar effect was seen in cells expressing a mutant TBC domain, EPI64-NT-R160A (). To explore whether the loss of microvilli resulted from a defect in cell polarity, the localization of the tight junction marker ZO-1 and the adherens junction marker E-cadherin was examined in cells overexpressing EPI64 or EPI64-LA. In neither case did cell junctions appear significantly different between transfected and nontransfected cells (Fig. S3, available at ).
The data imply that mislocalizing the TBC domain from the apical region results in loss of microvilli. Because perturbing the binding of EBP50 to ezrin should reproduce this effect, we explored the consequences of overexpressing EBP50-PDZ1-PDZ2, which lacks the C-terminal ezrin binding site. Consistent with this notion, whereas overexpressing wild-type EBP50 had no effect on microvilli, overexpressing EBP50-PDZ1-PDZ2 resulted in the loss of microvilli (). Moreover, knock down of EBP50 by siRNA treatment for 48 h also resulted in the loss of surface microvilli, whereas treatment with control siRNA did not (). Interestingly, the microvilli located above the adherens junctions were the most resistant to EBP50 knockdown, perhaps suggesting that the lateral border provides a signal for microvillus formation that can normally be propagated across the apical surface.
Overexpressing the full-length EPI64 constructs caused a small percentage of cells to contain intracellular vacuoles often decorated by EPI64 and F-actin both in JEG-3 cells but more dramatically in HeLa cells (). Because similar F-actin– decorated structures are seen in cells overexpressing dominant-active Arf6 (), we cotransfected tagged EPI64 and Arf6 and found that the vacuoles were also enriched in Arf6 (), suggesting a relationship between Arf6 and EPI64. We thus examined cells expressing EPI64 and wild type or dominant-active (Q67L) or dominant-negative (T27N) mutants of Arf6. EPI64 colocalized on vacuoles with both Arf6 and Arf6 Q67L, but not with Arf6 T27N (). An enrichment of endogenous EPI64 is seen in the region of vacuoles produced by Arf6 Q67L overexpression and on areas of the plasma membrane to which it localizes, suggesting that Arf6 may have the ability to recruit EPI64 to these regions ().
It has been found that the TBC domain of binds Arf6-GDP (). Thus, we explored the possibility that Arf6 binds to the TBC domain of EPI64. Cells were transfected with Xpress-tagged EPI64 and HA-tagged Arf6, Arf6 T27N, or Arf6 Q67L. Immunoprecipitates of EPI64 contained a small amount of wild-type Arf6, were enriched in Arf6 Q67L, and lacked Arf6 T27N (). Immunoprecipitates from cells expressing EPI64-LA or -NT recovered both Arf6 and Arf6 Q67L efficiently, whereas Arf6 T27N was absent (). Notably, there was a consistent increase in the relative amount of wild-type Arf6 recovered by these constructs when compared with EPI64, indicating that the mutated proteins have increased access, or bind more efficiently, to Arf6-GTP. In addition, both constructs also decorated Arf6 Q67L–induced vacuoles, as is seen for EPI64 (unpublished data). In coimmunoprecipitation experiments, EPI64-R160A behaved indistinguishably from EPI64 (), suggesting that the R160A mutation has no effect on Arf6 binding.
An in vitro binding assay using purified recombinant proteins was used to determine whether the interaction between EPI64 and Arf6-GTP is direct (). Immobilized EPI64 bound Arf6-GTP with a greater affinity than Arf6-GDP, whereas no binding was observed to immobilized BSA. Collectively, these results indicate that, unlike the TBC domain of , which preferentially binds Arf6-GDP (), the TBC domain of EPI64 specifically binds Arf6-GTP.
To evaluate whether EPI64 might be a GTPase activating protein (GAP) for Arf6, we assessed the level of Arf6-GTP in cells by making use of the ability of GST-GGA3 to bind selectively to Arf6-GTP (). Cells were transfected with HA-tagged Arf6, together with full-length EPI64 or mutants, and lysed. Lysates were incubated with GST-GGA3 beads, and the amount of bound Arf6-GTP was determined. Expression of all constructs containing the TBC domain did not reduce the level of Arf6-GTP and, in fact, enhanced it, when compared with the vector alone control (), making it unlikely that the TBC domain of EPI64 is a GAP for Arf6.
The connection between EPI64 and Arf6 prompted us to explore the effect of expressing wild-type, dominant-active, or dominant-negative Arf6 on microvilli of JEG-3 cells. Overexpressing wild-type Arf6 and Arf6 Q67L both led to microvillar loss, whereas the dominant-negative Arf6-T27N had no effect (). Dominant-active Arf6 also greatly enhanced the percentage of cells with vacuoles, as reported previously (). In cells transfected to express HA-Arf6 yet still containing microvilli, the Arf6 colocalized with ezrin and EPI64 in the apical microvilli (), confirming observations of Arf6 at the apical surface of epithelial cells ().
We have uncovered two unexpected findings relating to the organization of the apical domain of epithelial cells. First, using high-resolution light microscopy, we found that the microvilli of epithelial cells do not have a uniform composition but, rather, have distinguishable subdomains that are influenced by mild overexpression of EPI64. Second, we have found that expressing an EPI64 construct unable to bind the PDZ domains of EBP50 results in a dramatic loss of microvilli. Moreover, knock down of EBP50, or expression of an EBP50 construct that binds EPI64 but not ezrin, also results in the loss of microvilli. Thus, our studies suggest that mislocalization of EPI64 results in loss of microvilli. Because a major recognizable feature in EPI64 is the presence of a TBC domain that we show binds directly to Arf6-GTP, it is likely that Arf6 is also involved, directly or indirectly, in regulating microvilli on cells.
As far as we are aware, all previous studies have suggested that microvilli are of uniform cytoskeletal composition along their length. Here, we demonstrate that phospho-T567 ezrin is enriched toward the distal ends of microvilli, whereas total ezrin seems to be uniformly distributed, colocalizing with the plasma membrane marker WGA. Additionally, EPI64 is not evenly distributed in microvilli but appears to be enriched toward the middle of some and at the base of others. What could account for this uneven distribution? It has been shown that the microvilli of cultured cells are not static structures but, rather, undergo growth and retraction over a period of minutes (). The actin filaments in microvilli, in which the barbed ends are associated with the tip, are presumably treadmilling, as has been documented in other systems (; ). For example, ezrin might become activated at the microvillus tip, associate with the newly assembled filaments there, and ride down the microvillus with treadmilling actin and become dephosphorylated as it nears the base. In this way, its membrane-cytoskeletal linking role might be inactivated. The finding that EBP50 colocalizes with ezrin along the membrane-associated part of the microvillus, extending slightly beyond the region of pERM, was also initially surprising because EBP50 is proposed to only bind activated ezrin molecules (). Perhaps as ezrin becomes dephosphorylated it takes time to return to its closed form (depending on the rates of dissociation from actin and the membrane) and so retains some ability to bind EBP50 after dephosphorylation.
Analysis of cells overexpressing EPI64 reinforces the idea that microvilli have different domains: ezrin is found colocalizing with the plasma membrane of microvilli, whereas EBP50 and EPI64 are now enriched on the rootlet ∼0.25 μm below the membrane; how this enrichment might occur remains unknown. To investigate how EPI64 alters the distribution of proteins in microvilli, we began to study its functional domains. To evaluate the importance of the C-terminal linkage of EPI64 to the PDZ domains of EBP50, we examined cells expressing EPI64 with an additional single alanine residue to abolish binding to EBP50. Surprisingly, the transfected cells lost microvilli, indicating that the proper localization of EPI64 is necessary for microvillar maintenance. By expressing domains of EPI64, we found that expressing any construct mislocalizing the TBC domain resulted in microvillar loss.
Sequence comparisons suggest that the TBC domain of EPI64 might have RabGAP activity (). However, biochemical studies did not identify any GAP activity on several purified Rab proteins using full-length recombinant EPI64 (unpublished data; see the following paragraph). A clue to the possible role of the TBC domain came from two sources. First, we noticed that in 5–10% of cells overexpressing EPI64 vacuoles formed, often associated with F-actin, highly reminiscent of the vacuoles seen in dominant-active Arf6-expressing cells (). Second, it was recently reported that the TBC domain of binds to Arf6-GDP (). Immunoprecipitation experiments and in vitro studies using recombinant proteins revealed that the TBC domain of EPI64 also binds Arf6, but with a preference for Arf6-GTP over Arf6-GDP. Initially, we wondered if EPI64 was a GAP for Arf6. RabGAPs have a conserved arginine residue necessary for enhancing the hydrolysis of the γ-phosphate of GTP (). Cells expressing EPI64-R160A, a construct in which this catalytic arginine was mutated to alanine, showed no reduction in microvilli, and the level of Arf6 associated with EPI64 was not diminished. Indeed, overexpression of either EPI64 or EPI64-R160A resulted in an enhanced level of Arf6- GTP. Therefore, EPI64 is not a GAP for Arf6 and, because EPI64 has a preference for Arf6-GTP, it is not a guanine nucleotide exchange factor for Arf6 either. At present, we believe that EPI64 is an effector of Arf6-GTP and that, by binding Arf6-GTP, it protects it from inactivation by a GAP.
While this manuscript was in the final stages of review, reported the important finding that EPI64 is a GAP for Rab27A and that this activity is abolished by a R160K mutation. Combining their results with ours implies that the TBC domain of EPI64 has two distinct functions: as a RabGAP for Rab27A and as a binding domain for Arf6-GTP that is unaffected by the R160A mutation.
Why does expression of EPI64 defective in binding EBP50 result in microvillar loss? Loss of microvilli is also seen in cells transfected to express just the PDZ domains of EBP50 or in cells with EBP50 levels reduced by siRNA treatment, thereby breaking the linkage to ezrin. Furthermore, overexpressing wild-type Arf6 or its mutant that binds GTP constitutively also results in microvillar loss. Because overexpression of the dominant-active Arf6 results in the formation of actin-containing vacuoles in the majority of cells, it is possible that the loss of microvilli is simply due to sequestration of F-actin internally. However, we do not think this is very likely, as overexpression of wild-type Arf6, which induces a very much lower level of vacuoles, also results in microvillar loss, including cells where no vacuoles are evident. Although these data suggest a correlation between elevated levels of mislocalized Arf6-GTP and the loss of microvilli, we also attempted to explore the consequences of Arf6 depletion by siRNA. Although we were able to knock down expression of transfected HA-Arf6 about fivefold, no effect on microvilli was seen in siRNA-treated HA-Arf6 expressing or otherwise wild-type cells (unpublished data). Whether these data suggest that microvilli can exist in the absence of Arf6 function, which is consistent with the lack of interference by expression of the dominant-negative Arf6-T27N mutant, or that Arf6 was not sufficiently reduced, remains an open question. If Arf6 is directly relevant to the regulation of microvilli, an attractive scenario is that Arf6-GDP is returned to the plasma membrane through a membrane-recycling pathway, where it is acted upon at the plasma membrane by Arf6 guanine nucleotide exchange factors such as ARNO or EFA6 (, ) and then captured by EPI64. Microvillar formation or maintenance might first require capture of Arf6-GTP by EPI64, followed by Arf6 inactivation, when it is brought in close proximity to a microvillar ARF-GAP. Consistent with this model, we found that the association of Arf6-GTP with EPI64 was elevated in mutants defective in microvillar localization. Moreover, in a preliminary report, a novel Arf6 GAP (ARF-GAP) was shown to bind the activated form of ezrin (Wu, F., H. Deng, R. Zhou, Z. Guo, Z. Fan, K. Yuan, X. Cao, J. Forte, and X. Yao. 2005. American Society for Cell Biology Annual Meeting. Abstr. 1804).
Among the known effectors of Arf6 is phosphatidylinositol-4 phosphate 5 kinase-α (PIP5Kα), which generates PIP2 at the plasma membrane (). Excess PIP5Kα, by either overexpression of the protein or overexpression of dominant-active Arf6, induces the formation of vacuoles rich in PIP2 and F-actin (). This appears to be a consequence of greatly enhanced endocytosis and decreased membrane recycling through an actin-dependent mechanism, thereby depleting PIP2 from the plasma membrane, ultimately resulting in multiple phenotypic consequences (; ). The similar phenotype conferred by EPI64 overexpression suggests that it drives this system by enhancing the level of Arf6-GTP, to hyperactivate PIP5Kα to generate excess PIP2. Our results suggest that localized capturing or cycling of EPI64-Arf6-GTP is necessary for the maintenance of microvilli. If this is the case, one possibility is that PIP5K has to be locally activated to maintain microvilli. As it has been reported that PIP2 binding to ezrin is the first step in its activation (), one attractive model is that transient locally elevated levels of PIP2 might be responsible for initiating or maintaining active ezrin necessary for microvilli.
In summary, we have shown that microvilli have subdomains and uncovered an involvement of EPI64 in regulating the presence of microvilli on the cell surface. We have also shown that the EPI64's TBC domain binds Arf6-GTP, which is the second example of a TBC domain interacting with Arf6, indicating that the function of TBC domains may extend beyond their RabGAP activities to encompass other small GTPases involved in membrane trafficking. A recent report suggested that actin organization and clathrin-mediated endocytosis at the apical surface of polarized epithelial cells are regulated by Arf6 (). The collected data suggest that EPI64 integrates actin cytoskeletal organization and membrane trafficking events mediated by Arf6 and Rab27A, thereby placing it at a pivotal point in balancing these processes. It will be fascinating to further unravel the connection between microvillar structure, endocytosis, and these proteins; such studies are under way.
JEG-3 and HeLa cells (American Type Culture Collection) were maintained in a 5% CO humidified atmosphere at 37°C in MEM (Invitrogen) with 10% FBS (Invitrogen).
Antisera and affinity-purified antibodies against human ezrin and EBP50 were described previously (; ). Hexahistidine-tagged EPI64 was expressed in insect cells, purified, and used to elicit antibody production in rabbits. The affinity-purified antibody recognized a single band at 64 kD in total extracts of several cultured cells (Fig. S2, available at ). Specific antibodies for phospho-T567 ezrin/T564 radixin/T558 moesin were obtained from Cell Signaling Technology or generated against the peptide CRDKYK(Tp)LRQIR (). Antibodies against the Xpress epitope were purchased from Invitrogen (anti-Xpress) or Santa Cruz Biotechnology, Inc. (OmniProbe M-21). Anti-HA (HA.11) was obtained from Covance Research Products. Mouse anti-Arf6 was provided by C. D'Souza-Schorey (University of Notre Dame, Notre Dame, Indiana). Mouse anti–ZO-1 and anti–E-cadherin were obtained from BD Biosciences. The antibody against tubulin (N356) was purchased from GE Healthcare. Donkey anti–rabbit Alexa Fluor 488, goat anti–mouse Alexa Fluor 568, goat anti–rabbit Rhodamine Red-X, fluorescently conjugated phalloidin, and WGA were obtained from Invitrogen. Donkey anti–rabbit and anti–mouse Cy5 were obtained from Jackson ImmunoResearch Laboratories. Goat anti–rabbit HRP was obtained from MP Biomedicals. GTPγS and GDPβS were obtained from Sigma-Aldrich. The siRNAs targeting human EBP50 (5′-CGGCGAAAACGTGGAGAAG-3′) and Luciferase GL2 (5′-CGUACGCGGAAUACUUCGA-3′) were obtained from Dharmacon.
Protein samples were separated by SDS-PAGE and transferred to Immobilon-P (Millipore). Western blotting was done as described previously (). For immunofluorescence, cells grown on glass coverslips were fixed in 3.7% formaldehyde/PBS for 10 min at room temperature. Cells were permeabilized in 0.2% Triton X-100/PBS for 5 min at room temperature, rinsed in PBS, and incubated with primary antibodies in PBS/2% FBS. After washing in PBS, secondary antibodies and additional markers (phalloidin and/or WGA) were added in PBS/2% FBS (unless WGA was included, in which case, FBS was omitted). Cells were mounted on glass slides in Vectashield (Vector Laboratories) and observed on a microscope (Eclipse TE-2000U; Nikon) using a 100× 1.4 NA lens (Nikon) on a confocal imaging system (UltraView LCI; PerkinElmer). Z-series images of single focal planes were obtained at 0.2-μm steps through the cells on a 12-bit digital output charge-coupled device camera (C4742-95-12ERG; Hamamatsu). Maximum projections of apical sections and identical contrast enhancements to all images were performed using ImageJ (NIH).
Fluorescence intensity grayscale levels along the length of at least five random microvilli on each of at least five different cells were measured by selecting a linear region of interest along the length of each microvillus from its tip to actin rootlet (or the opposite end of staining if not stained for actin) using the UltraView software and recorded in Excel (Microsoft). R () was used to analyze protein distribution trends in the microvilli of each dataset. In brief, the mean intensity of the dataset was subtracted from each recorded point to center the numbers on a common point on the y axis (y = 0). These new numbers were used to generate a scatter plot of all data points followed by the application of a smooth curve fitted by LOESS linear regression.
Xpress-tagged EPI64 and EPI64-LA constructs in pcDNA 3.1 His/Xpress were described previously (). Xpress-tagged EPI64 constructs expressing the N-terminal (residues 1–324, including the TBC domain) or C-terminal halves (residues 319–508) of EPI64 were generated by PCR. Xpress-tagged EPI64-R160A and EPI64-NT-R160A were made by PCR mutagenesis. Xpress-tagged EBP50 constructs were derived by subcloning from previously described GST constructs () into pcDNA3.1 His/Xpress. HA-tagged Arf6 constructs were a gift from J. Donaldson (National Institutes of Health, Bethesda, MD; ). The construct expressing GST-GGA3 was a gift from J. Bonifacino (National Institutes of Health, Bethesda, MD; ).
Polyethylenimine (PEI) has been shown to be a nontoxic transfection reagent (). JEG-3 cells were seeded in 60-mm dishes to be 60–70% confluent at the time of transfection. 1 μg of DNA was mixed with 50 μl of serum and antibiotic-free MEM and 5 μl of 1 mg/ml PEI Linear (mol wt 25,000; Polysciences, Inc.), previously dissolved in water at room temperature for 10 min. Cells were washed with serum and antibiotic-free MEM and overlaid with the DNA/MEM/PEI mixture in 2 ml of serum and antibiotic-free MEM and incubated under normal growing conditions. After 4 h, 1.6 ml of serum and antibiotic-free medium and 400 μl of FBS was added to the cells. Cells were allowed to recover for 24–48 h. For 6-h transfections, cells 80–90% confluent were washed and incubated for 2 h with the DNA/MEM/PEI mixture before the supplementation of additional MEM and FBS. Under the conditions used, ∼50% of the cells were transfected.
JEG-3 cells were seeded in 60-mm dishes to be 30% confluent at the time of transfection. Cells were washed twice with PBS before adding 4 ml of OptiMEM medium (Invitrogen). 2.5 μl of 20 μM Block-iT Alexa Fluor Red Fluorescent Oligo (Invitrogen) and 2.5 μl of 20 μM duplex siRNA targeting either EBP50 or Luciferase (GL2) for control were diluted into 500 μl OptiMEM. 5 μl of Lipofectamine RNAiMAX (Invitrogen) was diluted into 500 μl OptiMEM per reaction, mixed with diluted siRNA, and incubated at room temperature for 15 min. The resulting 1-ml volume of Lipofectamine complexed with siRNA was added to cells and incubated at 37°C. After 4 h, the medium was replaced with 4 ml of MEM with 10% FBS and -glutamine. 48 h after transfection, cells were either fixed and stained for immunofluorescence or lysed for Western blot analysis.
Transfected JEG-3 cells were stained for ezrin, actin, and the expressed protein (for identifying transfected cells) to indicate the overall apical membrane phenotype. At least 200 cells were counted for each experiment and scored as normal, lacking/few microvilli, or ruffling. In instances where cells were ruffling but still had normal-looking microvilli, cells were scored as ruffling. Where surface structures did not clearly fall into the above categories, cells were scored as having abnormal microvilli. The siRNA-treated JEG-3 cells containing Alexa Fluor red fluorescent oligo (to identify transfected cells) and siRNA targeting either EBP50 or GL2 were stained for ezrin. 150 cells were counted for each experiment and scored as described.
After a 6-h transfection in 100-mm dishes, JEG-3 cells were washed in prewarmed PBS at 37°C and treated for 2 min at 37°C with 5 ml of freshly made 1.25 mM DSP in PBS (Pierce Chemical Co.) to cross-link proteins. Cells were then washed in prewarmed TBS (50 mM Tris, pH 7.4, and 150 mM NaCl) at 37°C and incubated in 10 ml of fresh TBS for an additional 15 min at 37°C to quench the cross-linking reaction. Each dish was lysed in 1 ml of ice-cold RIPA buffer (0.1% SDS, 1% Triton X-100, 1% deoxycholate, 150 mM NaCl, 1 mM EDTA, and 25 mM Tris, pH 7.4) with protease inhibitors and clarified by centrifugation at 100,000 in a TLA 100.3 rotor for 10 min. After removal of 100 μl of clarified supernatant, as a total lysate fraction, the remainder was added to a 25 μl volume of protein A Sepharose (Sigma-Aldrich) and 5 μg of anti-Xpress OmniProbe M-21 and incubated at 4°C for 2 h with gentle inversion. Bead-bound proteins were washed with ice-cold RIPA buffer and eluted by boiling in 50 μl of 2× SDS sample buffer.
Hexahistidine-tagged EPI64 was purified from insect cells using standard methods and coupled to CNBr-activated Sepharose beads at ∼1 mg/ml as described previously (). Purified myristoylated Arf6 was provided by P. Randazzo (National Cancer Institute, Bethesda, MD; ). Purified Arf6 was subjected to nucleotide exchange in the presence of GTPγS or GDPβS as described previously (). A molar excess of Arf6-GTP or -GDP was incubated with 25 μl of EPI64 or BSA beads for 2 h at 4°C with gentle inversion in binding buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 2 mM MgCl, and 1 mM DTT). The bound protein was then washed extensively with ice-cold binding buffer and eluted by boiling in 50 μl of 2× SDS sample buffer. The eluted protein was separated by 12% SDS-PAGE and analyzed by Western blotting using anti-Arf6 antibodies.
A GST-GGA3 binding assay was used to determine the relative levels of Arf6-GTP (). In brief, the different EPI64 constructs were cotransfected with HA-tagged Arf6 into JEG-3 cells grown on 100-mm dishes. After 18–24 h, cells were lysed in 1 ml of ice-cold GGA3 buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 2 mM MgCl, 1% wt/vol Triton X-100, and protease inhibitors). Lysates were cleared by centrifugation at 16,000 for 5 min at 4°C, and the resulting supernatant was applied to 50 μg of GST-GGA3 previously immobilized on glutathione Sepharose for 1 h at 4°C with continuous gentle inversion to bind Arf6-GTP to GGA3. The beads were washed in ice-cold GGA3 buffer, and proteins were eluted by boiling in sample buffer, followed by SDS-PAGE and Western blotting.
Fig. S1 shows sample plots of individual microvillar protein distributions for ezrin, pERM, and EPI64 with respect to the actin cytoskeleton and plasma membrane. Fig. S2 shows a Western blot of cell lysates demonstrating the specificity of the EPI64 antiserum compared with those detecting ezrin and EBP50. Fig. S3 shows immunofluorescent images of JEG-3 cells overexpressing EPI64 or EPI64-LA counterstained for the cell junction markers ZO-1 and E-cadherin. Online supplemental material is available at . |
The multivesicular body (MVB) sorting pathway provides a mechanism for the lysosomal degradation of transmembrane proteins and plays a critical role in a diverse range of processes, including growth factor receptor down-regulation (), antigen presentation (), developmental signaling (; ; ), and the budding of enveloped viruses (). The proteins that constitute the MVB sorting machinery were identified by a genetic screen in yeast for mutants that missort an MVB cargo (). Most of the mutants isolated were class E mutants, which accumulate enlarged endosomes and exhibit defects in the formation of MVB vesicles. Further characterization of the class E Vps proteins led to the identification of three high molecular weight cytoplasmic complexes that function in MVB sorting, the endosomal sorting complex required for transport (ESCRT) complexes I, II, and III (; ,).
The ESCRT-I complex (Vps23, -28, and -37) is recruited to endosomes by Vps27, which interacts with ubiquitinated cargo and initiates the MVB sorting reaction (). ESCRT-I also interacts with ubiquitinated cargo via the UEV domain of Vps23 (). Genetic studies indicate that ESCRT-II (Vps36, -22, and -25) functions downstream of ESCRT-I (). ESCRT-II interacts with ubiquitinated cargo via the NZF domain of Vps36, and with phosphatidylinositol 3-phosphate (PtdIns3P) via the GRAM-like ubiquitin binding in EAP45 domain (; ). ESCRT-II then recruits the ESCRT-III subunits (Snf7, Vps20, -2, and -24) to the endosome, where they oligomerize to form the ESCRT-III complex (; ). ESCRT-III, in turn, recruits accessory factors such as Bro1 (), which, in turn, recruits Doa4 (), the deubiquitinating enzyme that removes ubiquitin from MVB cargo before their sorting into MVB vesicles. ESCRT-III also recruits the AAA-type ATPase Vps4, which catalyzes the disassembly of the ESCRT machinery and recycles the ESCRT complexes into the cytosol to allow further rounds of cargo sorting (; ).
Recent studies on the architecture of the ESCRT machinery have enhanced our understanding of how the ESCRT complexes assemble and interact with ubiquitinated cargo and phosphoinositides (). The structures of the core complexes of yeast ESCRT-I and -II have been determined (; ; ; ). The interaction between ESCRT-I and -II has been mapped to the C-terminal domain of Vps28 and the NZF-1 domain of Vps36 (). Even though it has been shown in vitro that ESCRT-I and -II can form a stable complex in solution, no such complex has yet been detected in cytosolic extracts from yeast cells, suggesting that this interaction is tightly regulated. We report the identification of a new component of the ESCRT-I complex, multivesicular body sorting factor of 12 kD (Mvb12). We show that Mvb12 plays a role in assembling ESCRT-I into an oligomeric complex in the cytosol. In doing so, Mvb12 prevents premature assembly of ESCRT-I and -II to ensure their ordered and sequential recruitment onto the endosomal membrane during MVB sorting.
To identify novel regulators of the ESCRT machinery, we searched the Genome Database for ORFs that show endosomal localization. We found eight uncharacterized ORFs and then tested to see if any of these ORFs are required for MVB sorting by examining the localization of a biosynthetic MVB cargo, the vacuolar hydrolase carboxypeptidase S (CPS), in deletion mutants lacking each of these ORFs. In wild-type cells, GFP-CPS is sorted into the MVB pathway and accumulates in the lumen of the vacuole (). In contrast, GFP-CPS accumulates on enlarged endosomes, as well as on the vacuolar-limiting membrane in ESCRT deletion mutants such as (). Interestingly, deletion of one of the uncharacterized ORFs, YGR206W, results in a defect in MVB sorting, and GFP-CPS is missorted to the limiting membrane of the vacuole (). We also examined the localization of the endocytic cargo Ste2, the yeast pheromone receptor, in the mutant. In wild-type cells, Ste2-GFP is endocytosed from the plasma membrane and sorted into the lumen of the vacuole. Sorting of Ste2-GFP is partially impaired in the mutant, as Ste2-GFP localized to both perivacuolar puncta and the vacuolar lumen (). The partial defect in MVB sorting, as well as the lack of observable enlarged endosomes, suggests that the Ygr206w protein is likely not a core component of the ESCRT machinery, but instead functions as a regulator of the ESCRT complexes. Sequence analysis indicates that YGR206W encodes a small protein with no well-characterized domains or motifs. We confirmed the predicted molecular weight and named the protein Mvb12. Mvb12 shares partial sequence homology with a set of genes in fungi and mammalian genomes, although the homology with mammalian genes is weak (Fig. S1, available at ).
To confirm the endosomal localization of Mvb12, we constructed a C-terminal GFP fusion of Mvb12 under the control of its endogenous promoter. Mvb12-GFP localized to the cytoplasm, as well as to perivacuolar puncta, many of which colocalized with dsRed-FYVE (EEA1), which is an endosomal marker (). The endosomal association of Mvb12 is likely transient and peripheral because the majority of Mvb12 was recovered from the cytosolic fraction in subcellular fractionation experiments, similar to the fractionation profile of ESCRT-I (unpublished data). We then tested the localization of Mvb12-GFP in different ESCRT deletion mutants. Interestingly, Mvb12-GFP failed to localize to endosomes and, instead, completely redistributed to the cytoplasm in the Vps23 (ESCRT-I) deletion mutant (). However, the endosomal localization of Mvb12-GFP is not dependent on downstream components of the MVB sorting machinery, including Vps36 (ESCRT-II), Snf7 (ESCRT-III), or Vps4 (). Collectively, these data indicate that the localization of Mvb12 to endosomes is dependent on ESCRT-I and suggest that Mvb12 functions either in a complex with or downstream from ESCRT-I in the MVB sorting pathway.
We next tested if Mvb12 is in a complex with ESCRT-I. By gel filtration chromatography, Mvb12 cofractionated with Vps23, suggesting that Mvb12 is likely a new component of ESCRT-I (). To confirm this, we performed a coimmunoprecipitation experiment in which we immunoprecipitated Flag-tagged Mvb12, and then probed for Vps23, which is a component of ESCRT-I (). Mvb12 coimmunoprecipitated with Vps23, indicating that Mvb12 is a new component of ESCRT-I. These data are consistent with the recent genome-wide analyses of protein–protein interactions in the yeast proteome that identified the Ygr206w protein as a new component of the ESCRT-I complex (; ).
To determine the effect of deleting on the size and stability of the ESCRT-I complex, we analyzed the elution profile of Vps23 by gel filtration. Deletion of did not appear to affect the stability of Vps23, but did result in a change in the elution profile of Vps23, reducing its apparent molecular weight from ∼350 kD in wild-type cells to ∼250 kD in cells (). Unexpectedly, the elution profile of Vps36, which is a subunit of ESCRT-II, shifted from 150 kD in wild-type cells to ∼250 kD in cells (). The shift in the apparent molecular weight of both ESCRT-I and -II to ∼250 kD in cells was quantitative, with most (if not all) of the cytoplasmic ESCRT-I and -II now eluting at this size (). Because ESCRT-I and -II are known to interact with each other in vitro (), we suspected that the ∼250-kD complex might be a heterodimeric complex of ESCRT-I (∼100 kD) and -II (∼150 kD). This hypothesis is supported by the coimmunoprecipitation of Vps23 (ESCRT-I) with Vps36 (ESCRT-II) from cell extracts, confirming the identity of the ∼250-kD species as a heterodimeric complex of ESCRT-I and -II (). To further characterize the ESCRT-I–II heterodimer, we generated a double mutant to disrupt ESCRT-I–II interactions. The size of ESCRT-I, based on the elution profiles of Vps23 and Vps37, was determined to be ∼100 kD in the double mutant (). The ∼100-kD species of ESCRT-I observed in the mutant most likely represents the monomeric form of ESCRT-I, as this size is consistent with the molecular weight of –expressed monomeric ESCRT-I determined by equilibrium analytical ultracentrifugation (; ).
The presence of a ∼100-kD ESCRT-I complex in mutant cells suggests that the cytosolic ∼350-kD species observed in wild-type cells may represent an oligomer of the ESCRT-I core complex (Vps23, -28, and -37). To test this hypothesis, we coexpressed Vps23-Flag and -Myc in the same cells and probed their ability to associate by performing coimmunoprecipitations. Cytosolic fractions were prepared from cell lysates and subjected to immunoprecipitation under native conditions using antibodies specific for the Flag tag. Vps23-Myc coimmunoprecipitated with Vps23-Flag in cytosolic fractions from wild-type cells, indicating that there are at least two copies of Vps23 in the ∼350-kD ESCRT-I complex (). In contrast, Vps23-Myc did not coimmunoprecipitate with Vps23-Flag in cytosolic fractions from cells (). These results suggest that the ∼350-kD species is likely an oligomer of the ESCRT-I core complex. Deletion of results in the dissociation of the oligomer into monomers of the ESCRT-I core complex, indicating that Mvb12 is required in assembling the ESCRT-I oligomer.
In some cases, molecular weight estimates obtained by gel filtration chromatography is dependent on the shape of the protein–protein complex being analyzed; therefore, we used a second technique to analyze the molecular weight of yeast native ESCRT-I. Blue native gels (BN-PAGE) provide a reliable and reproducible means of estimating the molecular weight of native protein complexes. As shown in , BN-PAGE analysis of cytosolic ESCRT-I complex from yeast revealed the presence of a protein complex whose molecular weight is ∼300 kD. In contrast, the stable monomeric ESCRT-I purified from migrated at a significantly lower molecular weight of ∼80 kD, which is consistent with the molecular weight of ∼85 kD determined independently using equilibrium analytical ultracentrifugation (; ). The clear difference in the migration pattern of yeast ESCRT-I versus monomeric ESCRT-I from on BN-PAGE strongly suggests that the yeast complex is an oligomer. To get a rough estimate of whether the yeast ESCRT-I oligomer is a dimer, trimer, or a higher-order species, we induced oligomerization of the ESCRT-I monomer using the chemical cross-linker bis(sulfosuccinimidyl)suberate (BS). As shown in , brief exposure to the chemical cross-linker generated clearly distinguishable dimeric and trimeric forms of the ESCRT-I, suggesting that the ESCRT-I core complex has an inherent propensity to oligomerize. Interestingly, the molecular weight of the yeast complex coincides best with that of the trimeric ESCRT-I. Collectively, the results of the coimmunoprecipitations and BN-PAGE strongly indicate that the ∼350-kD species observed in yeast represents an oligomeric complex consisting of at least two (but likely three) ESCRT-I core complexes (Vps23, -28, and -37).
To further test the possibility that ESCRT-I is a trimer, we performed a new coimmunoprecipitation experiment using a strain expressing two different tagged versions of Vps23, Myc-tagged chromosomal Vps23 and Flag-tagged Vps23 expressed from a low-copy plasmid. Using an antibody specific for Vps23, we were able to detect both tagged versions of Vps23, and we found that the plasmid-derived Vps23-Flag was expressed at lower levels (∼50–60%) than the chromosomal Vps23-Myc (unpublished data). As such, the differential expression of the two different (Myc- and Flag-tagged) pools of Vps23 provides an indirect way to examine the stoichiometry of the ESCRT-I complex by comparing the ratio of their incorporation into the oligomer. Consistent with our trimer prediction, immunoprecipitation using a Flag antibody coimmunoprecipitated Vps23-Myc. The ratio of Vps23-Myc to -Flag in the immunoprecipitate, as detected by an antibody specific for Vps23, was determined to be ∼2:1 by densitometric analysis (Fig. S3, available at ), suggesting that there are three copies of Vps23 in the ESCRT-I complex. Collectively, our data are consistent with the cytosolic ESCRT-I complex being a trimer.
Vps23 is predicted to contain a coiled-coil domain between residues 212–304 (). Coiled-coil domains are known to mediate protein oligomerization (). Bioinformatic analyses suggest that the coiled- coil domain of Vps23 has a high probability of forming a trimer (Fig. S2, available at ). Several signature heptad repeats were identified in the coiled-coil region of Vps23 and hydrophobic–hydrophilic mutations were introduced at positions a and d of the heptad repeat (abcdefg) to disrupt coiled-coil interactions. In addition to the point mutants, two coiled-coil domain deletion mutants were also generated to test the functional role of this region in Vps23/ESCRT-I oligomer formation (). Gel filtration analysis of the coiled-coil domain mutants Vps23(M283D/L286D; ), Vps23(Δ206-256), and Vps23(Δ257-299; not depicted) displayed the formation of a ∼250-kD complex, which is consistent with the ESCRT-I–II heterodimer observed in cells, demonstrating the importance of the coiled-coil region in maintaining the integrity of the ESCRT-I oligomer. Similar to the strain, all of the Vps23 coiled-coil mutants displayed defects in MVB sorting with GFP-CPS missorted and colocalized with FM4-64 at the limiting membrane of the vacuole (). We also tested if the coiled-coil domain of Vps23 corresponds to the Mvb12 binding site by coimmunoprecipitation experiments. Both deletions and point mutations in the coiled-coil domain of Vps23 abolished interactions with Mvb12 (), suggesting that Mvb12 binds to the coiled-coil region. We further tested if this interaction between Mvb12 and Vps23 is direct by in vitro pull down assays. Mvb12 binds to His-tagged Vps23, but not to His-tagged Vps23(Δ206-299) missing the coiled-coil domain, indicating that Mvb12 binds to the coiled-coil domain of Vps23 (). These data suggest that the ESCRT-I oligomer is formed by the coiled-coil domains of Vps23 molecules and is stabilized by Mvb12.
The ESCRT complexes appear to function in a sequential manner at the surface of the endosome during the formation of MVB vesicles and the sorting of cargo into these vesicles. Assembly of the ESCRT machinery on the endosomal membrane is regulated by both ubiquitin and lipid (PtdIns3P) signals and requires the coordinated recruitment of all ESCRT complexes and associated proteins (; ; ). Interactions among the yeast ESCRT complexes and their mammalian homologues have been extensively investigated by yeast two-hybrid analysis (; ; ). ESCRT-I has been shown to interact with the Vps27 via interactions between the UEV domain of Vps23 and the PTVP motif in Vps27 (; ). Two-hybrid studies have indicated that ESCRT-I interacts with both ESCRT-II and -III components (). The interaction between ESCRT-I and -II has been characterized biochemically and mapped to Vps28 (ESCRT-I) and Vps36 (ESCRT-II; ; ). In this study, we identify Mvb12 as a new component of ESCRT-I that interacts with the coiled-coil domain of Vps23 and stabilizes ESCRT-I as an inactive, soluble oligomer. In doing so, Mvb12 negatively regulates the interaction between ESCRT-I and -II. This ensures the ordered recruitment of ESCRT-I and -II and spatially restricts ESCRT-I–II assembly to the surface of the endosome. Premature assembly of the ESCRT-I–II complex in the cytosol of cells likely interferes with the normal sequence of ESCRT-mediated cargo sorting events on the endosome, resulting in a defect in MVB sorting.
Both the yeast and mammalian ESCRT-I complexes have an apparent molecular weight of ∼350 kD by gel filtration (; ). The structure of the ESCRT-I core complex has been determined, consisting of Vps23, Vps28, and Vps37 in a 1:1:1 stoichiometry (; ). In these studies, a truncated form of -expressed ESCRT-I was used and determined to be ∼85 kD by analytical ultracentrifugation, which is consistent with the monomeric form of ESCRT-I (; ). The difference between the size of –expressed ESCRT-I and yeast native ESCRT-I suggests that the native ESCRT-I complex is an oligomer of the ESCRT-I core complex. Indeed, the ability of ESCRT-I to assemble into clearly distinguishable dimeric and trimeric forms when treated with a chemical cross-linker () strongly suggests that ESCRT-I monomers have a propensity to associate with each other and assemble into an oligomer. What, then, is the oligomeric status of the ESCRT-I complex in yeast? Is it a dimer, trimer, or a higher order oligomer? The results of a variety of approaches, including gel filtration, BN-PAGE, and coimmunoprecipitation experiments, presented in this study strongly argue that the ∼350-kD species observed in yeast is an oligomer that likely contains three copies of the ESCRT-I core complex.
The oligomer model for the yeast native ESCRT-I complex predicted by gel filtration chromatography and BN-PAGE is also strongly supported by the analysis of the coiled-coil domain of Vps23 using the MultiCoil prediction program (). Coiled-coil domains are known to mediate protein oligomerization in several cases including transcription factors and viral proteins. To predict the oligomeric state of Vps23/ ESCRT-I, we used the MultiCoil prediction program to analyze the Vps23 sequence. The predicted dimeric or trimeric state of this domain was calculated and the probability of each residue in a dimeric (blue) or trimeric (red) coiled-coil was plotted against residue number (Fig. S2 A). The program predicted a coiled-coil region from residue 212–304. This region has a higher probability of forming a trimeric coiled-coil than a dimeric coiled-coil. As a control, we used this program to analyze gp41p of HIV, the trimeric structure of which has been determined by NMR and x-ray crystallography. The MultiCoil program also predicts a higher probability of trimer formation for the coiled-coil region of gp41p (Fig. S2 B). Typically, trimers are not packed as tightly as dimers, rendering them less stable and better suited for regulation. Thus, an ESCRT-I trimer would be more suitable for assembly/disassembly regulation. Furthermore, mutations of key residues within the coiled-coil domain of Vps23, as predicted by the MultiCoil program, recapitulated the phenotype of the strain (). The ESCRT-I oligomer is destabilized by either Mvb12 deletion or Vps23 coiled-coil domain mutations. Collectively, our analyses of the coiled-coil domain of Vps23 indicate that the coiled-coil domain of Vps23 is required for the oligomerization of the ESCRT-I complex.
Although the oligomeric status of the yeast ESCRT-I complex has been established, it is not clear what role Mvb12 plays in the oligomerization process. Although it is difficult to address this question, there are some interesting clues in the data presented here. Our data indicate that Mvb12 () and the coiled-coil domain of Vps23 () both appear to be essential in mediating the formation of the ∼350-kD oligomer, and that Mvb12 interacts with the Vps23 coiled-coil domain (). However, the fact that we cannot detect the oligomeric form of ESCRT-I in the coiled-coil mutants of Vps23 suggests that the coiled-coil domains are critical in bringing the monomers together (possibly via hydrophobic interactions), at which point Mvb12 probably functions as a molecular clamp or adaptor that tethers three ESCRT-I core complexes together in a stable oligomer. Therefore, Mvb12 appears to play a structural role in stabilizing the ESCRT-I oligomer, as well as a regulatory role in preventing ESCRT-I from interacting with ESCRT-II in the cytosol. Consistent with the role of Mvb12 as a regulator of ESCRT-I function, loss of Mvb12 does not result in an absolute block in MVB sorting, as is seen in mutants lacking the core components of ESCRT-I.
The results of previous genetic and biochemical studies have suggested a model for the sequential assembly of the ESCRT complexes on the endosomal membrane. In this model, membrane-bound ESCRT-I recruits -II, which recruits and activates downstream ESCRT machinery, resulting in cargo sorting into MVB vesicles. If Mvb12 is a negative regulator of the interaction between ESCRT-I and -II, one would expect that the membrane association of ESCRT-I would be accompanied by the dissociation of Mvb12, allowing ESCRT-I to recruit -II. Consistent with this, subcellular fractionation experiments show that Vps23 and Mvb12 exhibit different levels of membrane association (Fig. S4, available at ). Although a small pool of Mvb12 can be detected on the endosomal membrane, which is consistent with the partial Mvb12-GFP localization to endosomes, Vps23 has a larger membrane-associated pool (∼25% of total protein) than Mvb12 (∼8%), and the difference in membrane association is enhanced in cells (∼73% for Vps23 vs. ∼10% for Mvb12). The significant difference in the membrane association of Vps23 versus Mvb12 argues that the majority of ESCRT-I on the endosome is not bound to Mvb12. Thus, the function of Mvb12 as a clamp for the ESCRT-I oligomer appears to be restricted to the cytosol. It is generally difficult to predict the stoichiometry of individual constituents in a native, higher-ordered oligomer. However, in vitro analyses of the interaction between –expressed ESCRT-I and Mvb12 suggest that Mvb12 exists in a 1:1 stoichiometry with other ESCRT-I subunits (unpublished data; Williams, R., personal communication).
Based on our data, we propose the following working model for the function of Mvb12 (). In the cytosol, Mvb12 associates with and stabilizes ESCRT-I in an oligomeric and inactive state. ESCRT-I is recruited to the endosomal membrane by Vps27 via the interaction of the UEV domain of Vps23 with the PTVP motif of Vps27 and with ubiquitinated cargo. A conformational switch in ESCRT-I results in the dissociation of Mvb12, as well as the disassembly of the ESCRT-I oligomer into monomers. Consequently, the ESCRT-II binding site within ESCRT-I (Vps28 C-terminal domain) is exposed, allowing ESCRT-I to recruit ESCRT-II to the endosome. This leads to the activation of ESCRT-II and the recruitment of downstream ESCRT machinery (ESCRT-III and Vps4), culminating in the formation of MVB vesicles and the sorting of cargo into these vesicles. Ultimately, it will require detailed structural and biochemical analyses of the native ESCRT-I complex to add mechanistic details and further refine this model.
DNA encoding a HA, Myc, and Flag epitope were introduced just before the stop codon in the chromosomal copy of , , and , respectively. , , , , and sequences were then amplified from genomic DNA. The SpeI–SacI- digested PCR products of and were ligated with SpeI–SacI-digested pRS414 and pRS415 to generate pTC74 and pTC75. The SmaI–XhoI-digested PCR products of were ligated with SmaI–XhoI-digested pGEX4T-1 to generate pTC76 expression vector. The NheI–XhoI-digested PCR products of were ligated with NheI–XhoI-digested pET23b to generate pSJ033 expression vector. The SpeI–SacI-digested PCR products of , , and were ligated with SpeI–SacI-digested pRS416 to generate pSJ102, pSJ106, and pSJ111, respectively. Vps23 coiled-coil point mutations were introduced into pSJ102 by QuikChange mutagenesis (Stratagene) to generate pSJ140(Vps23). Vps23 coiled-coil deletion mutants were constructed by inverse PCR and blunt-end ligation using pSJ102 as the template to generate pSJ135(Vps23) and pSJ137(Vps23), using pSJ033 as the template to generate pSJ143(Vps23) expression plasmid. pGO45 GFP–CPS, dsRed-FYVE, and Ste2-GFP have been previously described (; ). The following yeast strains were constructed for this study: TCY246 (SEY6210.1; ); TCY257 (SEY6210.1; ); TCY274 (SEY6210.1; ); SJY027 (SEY6210.1; ); and SJY030 (SEY6210.1; , ). The following yeast strains were previously described (,) SEY6210 (
); EEY6-2 (SEY6210; ); MBY30 (SEY6210; ); MBY3 (SEY6210; ); and MBY24 (SEY6210.1; ).
of 0.4–0.6; some were labeled with FM4-64 for vacuolar membrane staining and resuspended in medium for visualization. Visualization of cells was performed on a fluorescence microscope (Axiovert S1002TV; Carl Zeiss MicroImaging, Inc.) equipped with FITC and rhodamine filters, captured with a digital camera (CH350 CCD; Photometrix), and deconvolved using Delta Vision software (Applied Precision, Inc.). Results presented were based on observations of >120 cells.
For gel filtration analysis, yeast cells were spheroplasted and lysed in PBS containing 0.1 mM AEBSF, 1 μg/ml pepstatin A, 1 μg/ml leupeptin, 1 mM benzamide, and protease inhibitor cocktail (Complete; Roche). The lysate was precleared for 5 min at 300 , followed by a 100,000 centrifugation. The resulting lysate was loaded onto a Sephacryl S300 column (16/60; GE Healthcare) and separated with PBS. Fractions were analyzed by Western blotting using anti-HA, anti-Flag monoclonal antibodies, and anti-Vps23 polyclonal antibody.
Soluble extracts were prepared from a yeast strain expressing Myc-tagged Vps23. Approximately 90 μl of the yeast extract or purified ESCRT-I was mixed with 10 μl of 10× sample buffer (5% Coomassie brilliant blue G-250, 100 mM Bis-Tris, pH 7.0, 500 mM [ɛ-aminohexanoic acid], and 10% glycerol). In the samples where cross-linking was performed, purified ESCRT-I was incubated with 1 mM BS at room temperature for 30 min. The samples were then analyzed using a 4–15% gradient gel using the same previously described technique ().
Immunoprecipitations under native conditions were performed essentially as previously described ().
Equal amount of –expressed Vps23-His and Vps23(Δ206-299)-His proteins were incubated with Ni-NTA beads for 1 h in Ni-NTA Bind Buffer (Novagen), and then incubated with an equal amount of -expressed Mvb12-Flag proteins for an additional 4 h at room temperature. The beads were washed three times with 1 ml Ni-NTA Bind Buffer to remove unbound material. Bound proteins were eluted by incubating with 1 ml 0.5 M acetic acid for 30 min. Proteins conjugated to the beads were eluted with 1 ml elution buffer for 30 min. 10% TCA was added to all samples to precipitate proteins. Acetone-washed protein samples were boiled in sample buffer and analyzed by 10% PAGE and Western blotting using anti-Flag monoclonal antibodies and anti-Vps23 polyclonal antibody.
Fig. S1 shows a multiple sequence alignment of Mvb12 with its putative homologues. Fig. S2 displays the result of the MultiCoil analysis of the coiled-coil domain of Vps23. Fig. S3 displays data that supports a model in which ESCRT-I forms a trimer in the cytosol. Fig. S4 shows that Mvb12 likely dissociates from ESCRT-I upon membrane association. Online supplemental material is available at . |
In eukaryotic cells, most short-lived proteins are degraded by the ubiquitin system (; ; ). Modification of cellular proteins with Lys48-linked polyubiquitin chains leads to their degradation by the 26S proteasome. Ubiquitylation also participates in the down-regulation of plasma membrane proteins, including many receptors and transporters. In this case, the targeted proteins are modified by either a single ubiquitin or short Lys63-linked ubiquitin oligomers (; ). These modifications do not target substrate molecules to the proteasome. Instead, they promote endocytosis of the tagged membrane proteins and their trafficking to the lysosome (equivalent to the yeast vacuole). Ubiquitin-modified proteins first sort to the limiting membrane of the late endosome, which invaginates at multiple sites, leading to the accumulation of internal vesicles. The membrane proteins accumulate in these vesicles, but ubiquitin does not. The mature late endosome structure is called a multivesicular body (MVB), which then fuses with the lysosome, resulting in breakdown of the internal vesicles by lysosomal lipases and proteases (; ).
Ubiquitin is a relatively stable protein in yeast despite its covalent linkage to many proteins destined for proteasomal or vacuolar degradation (). This is possible because ubiquitin-protein modification is transient. Deubiquitylating enzymes (DUBs) release ubiquitin from polyubiquitin conjugates by cleaving the isopeptide bond between the ubiquitins in a chain or at the ubiquitin C terminus linked to substrate. The yeast DUB family consists of at least 20 members, including 16 in the ubiquitin-specific processing protease (UBP) subfamily (; ; ).
Mechanisms regulating DUB activity in the cell are only beginning to be analyzed. Subcellular localization, posttranslational modification, and interaction with regulatory factors are all likely to play important roles in DUB regulation (). For instance, several DUBs associate with the 26S proteasome, and this association is required for full activity. Rpn11/POH1, a DUB of the MPN/JAMM class, is an integral subunit of the proteasome 19S regulatory complex (; ). Its deubiquitylating activity requires interaction with other subunits of the 19S complex. Ubp6 binds to the 19S subunit Rpn1, and this interaction strongly stimulates Ubp6 enzymatic activity (). Thus, the activities of both Rpn11 and Ubp6 are delimited to their desired site of action, the proteasome. Other DUBs may not need to be regulated in this manner because they have intrinsically high substrate specificity. For instance, isopeptidase T (IsoT) and its yeast orthologue Ubp14 regenerate free ubiquitin from unanchored polyubiquitin chains, but no activity is seen toward polyubiquitin-protein conjugates (; ). This specificity was traced to an IsoT/Ubp14 element called the ZnF-UBP or DAUP domain (), which is necessary for ubiquitin binding and forms a pocket around the free C-terminal tail of the ubiquitin chain ().
Doa4, a 926-residue yeast DUB of the UBP class, contributes to the release of ubiquitin from ubiquitin-protein conjugates destined for degradation (; ; ; ). The enzyme acts primarily at the late endosome membrane, although genetic and biochemical data suggest that it may have additional roles linked to the proteasome (; ). The inactivation of Doa4 leads to severe phenotypic abnormalities, including the depletion of free ubiquitin, accumulation of apparent proteolytic remnants attached to short ubiquitin chains, defects in the proteolysis of both proteasomal and vacuolar substrates, hypersensitivity to amino acid analogues such as canavanine, and a strong sporulation defect.
Multiple observations indicate that Doa4 is responsible for deubiquitylating membrane proteins at the MVB. First, mutations interact genetically with mutations in class E vacuolar protein-sorting (VPS) factors, which are essential for maturation of the late endosome into MVBs (). Inactivation of these factors compromises MVB vesiculation, leading to the accumulation of stacked membrane cisternae known as the class E compartment (), and causes strong suppression of the phenotypic abnormalities associated with mutations. Second, Doa4 colocalizes with components of the large MVB-localized endosomal sorting complex required for transport III (ESCRT-III; ). Specifically, in yeast cells lacking the AAA ATPase Vps4, Doa4 accumulates in the class E compartment along with the ESCRT-III factors Vps24/Did3 and Snf7/Vps32/Did1 (). Third, transmembrane proteins in transit to the vacuole from either the cell surface or the Golgi accumulate in ubiquitinated forms in cells (; ). Finally, Doa4 is able to bind to the class E VPS protein Bro1, which associates directly with the Snf7 subunit of ESCRT-III; Bro1 may help recruit Doa4 to the late endosome ().
Both endocytic and biosynthetic cargoes need to be ubiquitylated in order to be transported efficiently to the vacuolar interior. However, Doa4 must remove the ubiquitin tag before or during cargo protein movement into the internal vesicles of the MVB to prevent ubiquitin from getting degraded along with the conjugated cargo protein. Therefore, Doa4 enzyme activity needs to be tightly controlled. In this study, we demonstrate that binding to the late endosome plays an important role in regulating Doa4 function. The N-terminal region of Doa4 is both necessary and sufficient for this binding. We identified four short conserved sequence blocks within this domain. From a deletion analysis of two of these motifs (boxes A and B), we determined that although these elements have no role in Doa4 expression or catalytic activity, they are critical for its physiological function. Deletion of either box A or B causes a phenotype equivalent to a complete loss of Doa4, and this correlates with the inability of the mutant proteins to interact with the late endosome. Notably, an N-terminal fragment of Ubp5, another yeast DUB, directs Ubp5 to the yeast bud neck, but when this segment is replaced with an N-terminal fragment of Doa4, the chimera relocalizes to the late endosome, where it can partially substitute for Doa4. These data show that the N-terminal extensions of Doa4 and Ubp5 determine their respective cellular distributions and that this contributes to the functional specialization of these paralogues.
Several normally soluble ESCRT-III factors, including Snf7 (Vps32/Did1) and Vps24 (Did3), concentrate in the class E compartment in yeast mutants with a defective Vps4 ATPase (). A Doa4-GFP fusion protein, which is primarily cytosolic in wild-type cells, also concentrates in several spots (usually from one to three per cell) adjacent to the vacuole (). Colocalization of Doa4-GFP and Vps24-HA by indirect immunofluorescence in cells suggested that these spots were class E compartments, but low cellular levels of Doa4-GFP made the detection of foci difficult. To enhance detection sensitivity and to verify the colocalization of endogenous Doa4 with ESCRT-III, we constructed wild-type and strains expressing Doa4 tagged with nine c-myc epitopes. Cells were costained with antibodies against endogenous Snf7 and c-myc. Despite the low cellular concentration of Doa4, bright foci were detected with the anti-myc antibody in cells, and these foci did indeed coincide with those of the ESCRT-III subunit Snf7 adjacent to vacuoles ().
The UBP subfamily of DUBs, to which Doa4 belongs, comprises a diverse set of cysteine proteases with two well-conserved motifs, the Cys and His boxes, which include all of the active site residues (; ). Many UBPs have long N-terminal segments (and occasionally C-terminal ones) that extend from the core catalytic domain but are of generally unknown function. These extensions may have regulatory roles. In a well-studied example, the N-terminal domain of the human UBP called herpesvirus-associated ubiquitin-specific protease ([HAUSP] USP7) binds the p53 tumor suppressor, allowing HAUSP to cleave polyubiquitin-p53 conjugates and, thereby, limit p53 degradation (). Previously, we had shown that the N-terminal noncatalytic region of Doa4 (∼560 residues) conferred Doa4 function on the catalytic domain of Ubp5, as did a shorter, 310-residue N-terminal Doa4 fragment ().
We hypothesized that the N-terminal noncatalytic domain of Doa4 might target the enzyme to the late endosome membrane. To test this model, we expressed C-terminally truncated versions of Doa4 fused to GFP in wild-type and mutant cells and examined their cellular localization (). Anti-GFP immunoblot analysis showed proteins of the expected sizes that were expressed at levels similar to or slightly above that of the full-length Doa4-GFP fusion protein (). The Doa4-, Doa4-, and Doa4-GFP fusion proteins were all concentrated at the class E compartment in a strain ( and Fig. S1, available at ). In cells transformed with the Doa4-GFP derivative, Doa4 foci were still observed, but they were smaller than those of the aforementioned fusions. In contrast, localization of Doa4-GFP to the late endosome was not observed. The modest differences in mutant protein expression levels did not correlate with class E localization. Collectively, the deletion analysis demonstrated that the first 208 residues of Doa4 are sufficient for its endosomal localization, whereas residues between 129 and 208 are necessary for the interaction. We named the N-terminal 208-residue segment the late endosome localization (LEL) domain or signal.
We previously isolated the orthologue by virtue of its ability to suppress the phenotype of a mutant (). Overall, the enzyme is 43% identical to its counterpart. The catalytic domains are the most closely related (64% identity). Several other potential Doa4 orthologues have been identified recently. These include potential ORFs from (86% identity), (88% identity), and (46% identity; ). Doa4 is also 41% identical to the Ubp5 paralogue, but, surprisingly, we could detect no functional overlap between these two enzymes even when the latter was expressed from a high copy plasmid ().
The LEL domain of Doa4 includes four short motifs that are well conserved among the Doa4 orthologues. We named these boxes A, B, C, and D (). In contrast, Ubp5 has diverged substantially in these regions. For instance, in box A, there is a negatively charged Glu at the fourth position, where there is a positively charged Lys in all of the Doa4 orthologues (and an uncharged Gln in the potential orthologue from ); a bulky aliphatic at the sixth position but an Ala in Ubp5; an Asp at the seventh position versus Glu in Ubp5; and a Leu at the ninth residue, where there is a Trp in Ubp5. The potential significance of box D for endosomal localization is suggested by the impaired capacity of Doa4-GFP, which lacks box D, to concentrate at the class E compartment (). To address the functional relevance of these motifs more fully, we undertook a detailed analysis of alleles bearing deletions of the two most conserved ones, boxes A and B.
Mutant alleles with precise deletions of box A (residues 51–63) or B (residues 94–100) were generated in the pDOA4-GFP plasmid backbone, yielding pDOA4ΔA- and pDOA4ΔB-GFP. The plasmids were transformed into a strain, and expression of the mutant proteins at or above wild-type levels was confirmed by anti-GFP immunoblot analysis (). The and mutants were tested by multiple assays. Although wild-type MHY501 cells can survive exposure to the arginine analogue canavanine at concentrations up to 1.5 μg/ml, a mutant (MHY623) cannot form colonies even at canavanine levels as low as 0.4 μg/ml. As expected, the wild-type Doa4-GFP fusion protein completely suppressed the canavanine sensitivity of the mutant. Neither the nor alleles allowed survival on plates with 0.4 μg/ml canavanine ().
Another prominent feature of the mutant is the accumulation of small ubiquitin-containing species, which appear to be proteolytic remnants attached to short ubiquitin chains, and strongly reduced levels of free ubiquitin (; ). The and alleles also could not suppress these aberrations and, in fact, caused a further increase in levels of the short-chain species (). This dominant-negative effect might reflect an interaction of the Doa4 mutants with a protein that binds both Doa4 and another DUB that normally is able to compensate weakly for Doa4 loss.
To examine the role of boxes A and B in Doa4-dependent proteolysis, we used pulse-chase analysis to measure degradation rates of three well-characterized ubiquitin system substrates in the and strains. We used two model test substrates, Ub-P–β-galactosidase and L–β-galactosidase, and the naturally short-lived transcription factor Matα2. Degradation of all three substrates was severely impaired in both mutants ().
Monoubiquitin addition to transmembrane biosynthetic cargo destined for the vacuole lumen serves as a signal for cargo sorting to the invaginating vesicles of MVBs; Doa4 is responsible for removing ubiquitin before cargo internalization into MVB vesicles (). A model substrate for this is the vacuolar hydrolase carboxypeptidase S (CPS), which is synthesized as an integral membrane protein. In the vacuole, the CPS precursor (pCPS) is cleaved from its transmembrane anchor by resident hydrolases to yield the mature lumenal form (). Doa4 removes the ubiquitin tag from pCPS at the late endosome (). We investigated the contribution of Doa4 boxes A and B to this process. Anti-CPS immunoblot analysis was performed in transformants of a mutant (). Inactivation of the Pep4 and Prb1 vacuolar proteases blocks virtually all vacuolar proteolysis, including the cleavage of pCPS, and allows the detection of ubiquitylated proteins that enter the vacuole. Monoubiquitylated pCPS (pCPS-Ub) accumulated in cells lacking Doa4 as expected (, first lane), and this was suppressed by pDOA4-GFP (, second lane). In contrast, neither pDOA4ΔA- nor pDOA4ΔB-GFP was able to suppress the accumulation of pCPS-Ub (, third and fourth lanes). We conclude that boxes A and B are required for the Doa4-mediated deubiquitylation of protein cargo in the MVB.
The failure of the box A/B mutants to deubiquitylate pCPS was not caused by a defect in Doa4 catalytic activity. Mutant versions of Doa4 were expressed in along with the ubiquitin fusion substrate Ub-M–β-galactosidase (). Cleavage of ubiquitin from M–β-galactosidase by the mutants was indistinguishable from wild-type Doa4 in this assay. Therefore, these short N-terminal deletions are unlikely to have disrupted overall Doa4 folding or catalytic activity and must instead impair some other feature of Doa4 function in the cell.
Boxes A and B are the most conserved elements in the LEL domain, which targets Doa4 to the late endosome (), so we tested whether these elements are necessary for this localization function. Indeed, both Doa4ΔA- and Doa4ΔB-GFP failed to concentrate in the class E compartment in cells (). Therefore, boxes A and B are required for Doa4 association with the late endosome.
As noted above, has an enzyme, Ubp5, that is much more closely related to Doa4 than any other DUB in this species. The similarity between Doa4 and Ubp5 extends over most of their lengths, including the more divergent N-terminal domains (). However, no functional overlap between the two proteins has been detected. Potentially, this separation of function could derive, at least in part, from the segregation of Doa4 and Ubp5 to different cellular compartments. The cellular distribution of Ubp5 has not been reported. Therefore, we tagged Ubp5 with GFP and examined its localization by intrinsic GFP fluorescence. The protein was seen throughout the cytoplasm and nucleus, but a fraction concentrated at the bud neck and incipient bud sites ().
Given that Ubp5 and Doa4 are most diverged in their N termini and that the LEL signal in Doa4 is in its N-terminal extension, we asked whether the distinct localization of Ubp5 was also caused by determinants in its N-terminal region. Residues 1–35 of Ubp5 have little obvious similarity to Doa4, whereas residues 36–162, which align with the LEL region of Doa4, have diverged in the regions corresponding to boxes A–D (). Like full-length Ubp5-GFP, a fraction of Ubp5-GFP localized to the bud neck in dividing yeast cells, but the Ubp5-GFP derivative no longer concentrated there (). These results suggest that sequences within the Ubp5 segment from residues 36 to 162 are necessary for the bud neck localization of Ubp5.
We determined whether the N-terminal domain of Doa4 could redirect Ubp5 to the late endosome and, if so, whether this correlated with any ability to provide Doa4 function. Two chimeras were constructed: one with the Doa4 LEL domain (residues 1–208) replacing the corresponding region of the Ubp5 backbone (Doa4-Ubp5-GFP) and the other with the full Doa4 extension placed upstream of the Ubp5 catalytic domain (Doa4-Ubp5-GFP). Both fusions showed predominantly class E compartment localization in cells (). Interestingly, the LEL domain of Doa4 fused to Ubp5 provided partial Doa4 activity; Doa4Ubp5-GFP was able to suppress, albeit incompletely, the hypersensitivity of the strain to canavanine (, sectors 2 and 5). Ubp5-GFP, which lacks the Doa4 LEL signal, does not localize to the late endosome (unpublished data). Thus, the N-terminal domains determine the cellular localization of Doa4 and Ubp5, and the LEL signal of Doa4 can confer Doa4 function on a distinct DUB.
Endosomal localization of Doa4, as determined by trapping of the protein in the class E compartment in mutants, requires the ESCRT-III components Snf7 and Vps24 (). In another strain background, Snf7 was also found to be necessary for such Doa4 localization but Vps24 was not (). At the same time, the latter study implicated another class E VPS factor, Bro1, in recruiting Doa4 to the endosome membrane, and the overexpression of Doa4 rescued defects associated with a mutation. Mutations in Bro1 cause accumulation of the class E compartment and mislocalization of protein cargo there (). Bro1 itself accumulates in the class E compartment in cells defective for the Vps4 ATPase.
Given that previous studies had already yielded some apparent differences between VPS factor requirements for Doa4 localization to the endosomes (; ), we examined the function of Bro1 in more detail. First, we analyzed free ubiquitin and ubiquitin-conjugate profiles in cells deleted for the gene. Strikingly, a -like depletion of free ubiquitin and accumulation of small ubiquitin-containing species was observed in the mutant (). Inactivation of Vps4 efficiently suppresses the phenotypic abnormalities of cells (). Similarly, strong suppression of the aberrant ubiquitin profile of was seen in a double mutant (). These data are consistent with a close link between Bro1 and Doa4 function.
If Bro1 were necessary for recruiting Doa4 to the late endosome, cells should no longer concentrate Doa4-GFP in the class E compartment, as reported previously (). Unexpectedly, we found that the presence or absence of Bro1 did not appear to alter the ability of Doa4-GFP to localize to the class E compartment in cells ( and Fig. S2, available at ). These data suggested that yeast might have more than one Doa4 receptor for the endosome, although Doa4 function in the MVB pathway appears to require Bro1 (because the loss of Bro1 mimicked the loss of Doa4 phenotypically). Conceivably, an alternative Doa4 receptor or cofactor with lower affinity for Doa4 might partially substitute for Bro1 when Doa4 levels are elevated, as when Doa4-GFP is expressed from a multicopy plasmid (). The overexpression of Doa4 did appear to suppress, albeit only very weakly, the defect observed by antiubiquitin blotting ().
The strain background used here differed from that used in the earlier Bro1 study (), which might also contribute to the differences in genetic requirements for Doa4 localization at the late endosome. We note several other differences. In the wild-type strain, CPS localizes exclusively within the vacuolar lumen, but in our background, a considerable fraction of CPS is found at the vacuolar membrane in wild-type cells (unpublished data). In addition, Bro1 did not cofractionate with either soluble or membrane-bound forms of Doa4 (see Discussion), and we failed to detect a two-hybrid interaction between two proteins (unpublished data).
Ubiquitin-modified proteins need to be deubiquitylated in a regulated manner to ensure that the signaling function of ubiquitin is not abrogated prematurely. On the other hand, deubiquitylation of target proteins may be required to switch them to distinct physiological states, and it is also needed to maintain sufficient levels of active, free ubiquitin in the cell. For instance, ubiquitin attachment to a cell surface receptor is a signal for its endocytosis, but if Doa4 were to remove the ubiquitin while the protein was still at the plasma membrane, the endocytic signal would be short circuited. Doa4 appears to only be recruited to ubiquitylated membrane proteins after they have reached the late endosome (or it may only be activated once there), where it cleaves off the ubiquitin moieties before membrane vesiculation into the MVB. This allows the recovery of ubiquitin and might also serve as a signal for other trafficking steps.
The present structure-function study of the Doa4 DUB has demonstrated that an N-terminal segment of its noncatalytic domain is both necessary and sufficient for directing Doa4 to the late endosome. The LEL signal is also able to redirect a functionally distinct yeast DUB, Ubp5, to the late endosomal membrane, allowing partial recovery of Doa4 function. The Ubp5 N-terminal domain normally localizes Ubp5 to the bud neck and incipient bud sites. Thus, the noncatalytic domains of these enzymes restrict them to distinct cellular sites, helping to define their functional specificity.
Yeast strains used in this study are listed in . The strains used for recombinant DNA work were JM101 and TOP10. Yeast and bacterial media were prepared as described previously, and standard yeast and bacterial molecular genetic methods were used (). Monoclonal mouse antibodies against ubiquitin, GFP, and the c-myc epitope were purchased from Covance. Polyclonal rabbit antibody against β-galactosidase was purchased from MP Biomedicals. Rabbit antibodies against CPS and Snf7 were gifts from D. Katzmann (Mayo Clinic, Rochester, MN). Antibody against Matα2 was described previously ().
boxes A and B were individually deleted by a two-step PCR-based approach (). A YCplac33 plasmid carrying a 6,149-bp KpnI–PstI yeast genomic fragment including the gene (pDOA4-8) was used as a template (). Amplified DNA fragments were cloned into pGEM-T (Promega). The resultant plasmids were digested with AgeI and BglII (restriction sites had been introduced into the respective flanking PCR primers), and the mutant DNA fragments were cloned into AgeI–BglII-digested pDOA4-GFP, a 2-μm plasmid encoding a functional fusion of Doa4 with enhanced GFP. All constructs were verified by DNA sequencing.
The chromosomal gene was inactivated by PCR-based gene deletion using the pFA6a-TRP1 plasmid () as a PCR template to create a DNA fragment that directed replacement of the chromosomal ORF with the gene by homologous recombination in MHY606 diploid yeast cells. Trp transformants were checked for the deletion allele by PCR. The heterozygous diploids were sporulated, and tetrads were dissected.
An analogous approach was used for constructing plasmids encoding GFP fusions with different portions of Doa4. pFA6a-GFP(S65T)-HIS3MX6 was used as a template (). GFP fusions were made after Doa4 residues 128, 190, 208, 232, and 560 by recombining in MHY501 yeast cells PCR-amplified - DNA fragments that had sequence identity upstream of the indicated codons and downstream of the stop codon with a cotransformed YEplac195-DOA4 () plasmid. His Ura colonies were screened for recombinant plasmids by PCR, and the plasmids were recovered in and reintroduced into various yeast strains for localization studies.
To construct plasmids that expressed full-length Ubp5 or the Doa4-Ubp5 chimera as fusions with GFP, the respective DUB ORFs were PCR amplified from either yeast genomic DNA or a plasmid encoding an HA-tagged Doa4-Ubp5 chimera, which was described previously (). PCR products were digested with BamHI and HindIII and were ligated into BamHI–HindIII-restricted pUG35 (/; Euroscarf). The Doa4-Ubp5 chimera was made by a PCR gap repair method that was developed previously (). The pUG35-UBP5 plasmid was gapped by digestion with AflII and SphI, the desired segment was amplified by PCR, and the two DNA fragments were cotransformed into MHY501 yeast cells. Recombinant plasmids were recovered in bacteria and sequenced to verify that they had the expected exchange of and sequences.
Plasmids encoding Ubp5- and Ubp5-GFP were constructed by a variation of the PCR gap repair method. pUG35-UBP5 was gapped with BamHI and AflII, and the large DNA fragment was isolated. Three oligonucleotides were designed as follows: an upstream 46-nucleotide primer was derived from the pUG35 sequence and included the start codon of the gene. The 5′ sequences of two downstream 65-mer oligonucleotides matched the noncoding strand of downstream of the codons for Pro35 and Ala162, respectively. Upstream and downstream pairs of primers had a 22-base overlap at their 3′ ends that straddled the deletion endpoints. The primers were annealed and extended with Taq DNA polymerase, resulting in an 89-bp double-stranded fragment that was introduced into yeast cells along with gapped pUG35-UBP5. Recombinant plasmids from Ura colonies were recovered in , and the deletions were verified by DNA sequencing.
Pulse-chase analysis of protein degradation was conducted as described previously (). Cells were labeled for 5–10 min with [S]TransLabel (MP Biochemicals). SDS-PAGE gels were dried and analyzed using a Storm Phosphorimager and ImageQuant software (GE Healthcare).
Western immunoblot analyses were performed as described previously (). For antiubiquitin immunoblotting, cells were grown at 30°C to midlogarithmic phase and lysed in SDS gel loading buffer by heating to 100°C for 10 min; cleared supernatants were loaded onto 16% tricine polyacrylamide gels, and proteins were transferred to Immobilon-P membranes (Millipore). Membranes were boiled for 30 min in water before incubation with antiubiquitin antibody. Antibody binding was detected using ECL reagents (GE Healthcare). For anti-GFP and -CPS immunoblot analysis, cell extracts were prepared by lysis with 0.2 M NaOH and 0.2% mercaptoethanol for 10 min on ice (). Trichloroacetic acid was added to a final concentration of 5%, and the samples were incubated for an additional 10 min on ice. Precipitated proteins were collected by centrifugation, neutralized, dissolved in loading buffer, loaded onto 10% SDS-polyacrylamide gels, and analyzed by immunoblotting.
Cellular distributions of Doa4 and Snf7 were examined in fixed yeast cells by indirect immunofluorescence as described previously (). After overnight incubation with primary antibodies, cells were washed and incubated for 1 h with secondary antibodies (Oregon green anti–mouse and Texas red anti–rabbit IgG conjugates; Invitrogen). GFP and FM 4-64 fluorescence in live cells was imaged as described previously (). Samples were viewed on a fluorescence microscope (Axioscope; Carl Zeiss MicroImaging, Inc.) with a plan-Apochromat 100× NA 1.4 objective lens equipped with a CCD camera (Axiocam; Carl Zeiss MicroImaging, Inc.). Images were processed using Openlab software (version 3.1.5; Improvision). All experiments were conducted at room temperature.
Fig. S1 shows the colocalization of Doa4- and Doa4-GFP with the vital dye FM 4-64 to the class E compartment in cells. Fig. S2 demonstrates continued class E compartment localization of Doa4-GFP in Δ cells costained with FM 4-64. Online supplemental material is available at . |
In vertebrate striated muscle cells, the most peripherally located myofibrils are attached to the sarcolemma through costameres, structures compositionally and functionally similar to focal adhesions (; ). Costameres are thought to laterally transmit the force of muscle contraction across the cell membrane to the ECM and serve to keep sarcomeres in register. The protein assemblies that compose the costameres are located beneath the Z-disks of peripheral myofibrils. Some components of focal adhesions (), including α integrin (), have also been found located at peripheral M-lines. For both focal adhesions and Z-disk costameres, integrins are coupled to cytoskeletal actin filaments and myofibrillar thin filaments, respectively. However, the means of attaching myosin thick filaments to the muscle cell membrane is unknown.
In muscle, the actin thin filaments are attached to dense bodies (Z-disk analogues) and the myosin thick filaments are organized around M-lines (for review see ). All the dense bodies and M-lines appear to be anchored to the cell membrane and, thus, also serve the same function as vertebrate costameres. In , clustered on the cytoplasmic side of the sarcolemma at the base of dense bodies and M-lines, is a complex of proteins associated with the cytoplasmic tail of PAT-3 (β-integrin). These proteins include UNC-112 (Mig-2), PAT-4 (integrin-linked kinase), PAT-6 (actopaxin), and UNC-97 (PINCH; ). At the dense bodies, vinculin, α-actinin, and talin likely link integrins to actin thin filaments. However, at the M-lines, the identity of the molecule or molecules that directly link the membrane-proximal integrin complex to the myosin thick filaments is unknown. Among UNC-97–interacting molecules is UNC-98, a 310-residue protein containing four C2H2 Zn fingers that localizes by antibodies to the M-lines (; ; ). The interaction between UNC-97 and UNC-98 requires all four Zn fingers of UNC-98 ().
To identify additional functional partners of UNC-98 at the M-line, we screened a yeast two-hybrid library, using as bait the N-terminal, non-Zn finger–containing 112 residues of UNC-98 (). 33 positive clones were identified encoding 18 unique proteins that interact with the N terminus of UNC-98 (Table S1, available at ). Three of the confirmed clones encoded myosin heavy chain (MHC) A, a body wall muscle–specific myosin.
contains four different muscle MHC genes, each encoding a different myosin isoform, A–D (; ; ). All four heavy chains have a similar structure, including a myosin head domain, IQ domains, and a coiled-coil domain (). In addition, the body wall muscle–specific isoforms, MHC A and B, have an ∼30-residue-long C-terminal nonhelical region. The positive clones identified in the screen encoded this nonhelical tail piece and a portion of the coiled-coil domain. To determine whether the N terminus of UNC-98 interacts specifically with MHC A, prey plasmids were generated encoding the analogous region of MHC B, C, and D (). The N terminus of UNC-98 interacts with the C terminus of MHC A but not with the equivalent regions of MHC B, C, and D in the yeast two-hybrid system (). This result is consistent with the lack of expression of UNC-98 in the pharynx (), where MHC C and D are specifically expressed. Moreover, the interaction of UNC-98 with MHC A and not MHC B is consistent with the different localizations of the two myosins in thick filaments of body wall muscle: MHC B to the polar regions and MHC A to the central region (), near the M-line localization of UNC-98.
To narrow the critical region of MHC A required for interaction with UNC-98, additional prey plasmids encoding a series of deletion derivatives of the C terminus of MHC A were tested (). As shown in , the N terminus of UNC-98 interacts with the C-terminal 200 residues of MHC A, including the nonhelical region and a portion of the coiled-coil rod (MHC A). Although the 32-residue nonhelical tail contributes to this binding (absence of binding when this region is removed; see MHC A), it is not sufficient for this binding (absence of binding when just this region is tested; see MHC A). The nonhelical region of MHC A, which may be phosphorylated (), is not required for MHC A to initiate thick filament assembly (). It is possible that the nonhelical region protrudes from the surface of the thick filament shaft and interacts with other proteins, such as UNC-98.
To provide additional evidence that UNC-98 interacts with MHC A, in vitro protein interaction was shown using an ELISA assay. Wild-type myosin II (including MHC A) showed saturable binding to both full-length and the N-terminal portion of UNC-98 expressed in (). To obtain evidence that UNC-98 is associated with MHC A in vivo, we sought to determine whether UNC-98 copurifies with native thick filaments using established procedures (; ; Fig. S1, B and C, available at ). Fractions were taken at each step of the preparation and analyzed by Western blot using antibodies specific to the N terminus of UNC-98 (Fig. S1, A and C). UNC-98 is prominent in the fraction in which thick filaments pellet, indicating that UNC-98 copurifies with thick filaments. In contrast to UNC-98, UNC-97 does not copurify with thick filaments (Fig. S1 C). To further isolate intact thick filaments, a fraction containing thick filaments, thin filaments, and ribosomes was fractionated on a sucrose density gradient. Fractions were collected from the bottom of the gradient (starting with S1) and were immunoblotted. Sucrose gradient fractions S3–S5 that contain both myosin and paramyosin () also contain UNC-98 (), indicating that UNC-98 copurifies with thick filaments.
Because UNC-98 interacts with both UNC-97 and MHC A, we asked what effect loss of function of , (encodes MHC A), or would have on the in vivo localization of these proteins. To facilitate these studies, antibodies to UNC-97 were generated (Fig. S2, available at ). is a strong loss-of-function splice site mutation () that results in a greatly reduced level of a truncated UNC-98 mutant protein (Fig. S1 A). As shown in , compared with wild type, in , neither the membrane-associated attachment complexes (visualized by anti–UNC-97) nor the A-bands (visualized with anti–MHC A) are organized in their normal sharply defined patterns. In fact, MHC A seems to be no longer restricted to straight A-bands and sometimes appears to cross over rows of dense bodies (, enlarged view in merged panel of ).
As noted above, the N-terminal 112 residues of UNC-98 interact with MHC A, a region of UNC-98 that is not required for interaction with UNC-97 (). Transgenic overexpression of the N terminus of UNC-98 as a GFP fusion protein in a wild-type background results in abnormal aggregates that contain MHC A and the N terminus of UNC-98 (). In contrast, UNC-97 is properly localized (). This suggests that an excess of the N terminus of UNC-98 competes with endogenous UNC-98 for binding with MHC A, interfering to some degree with the interaction of intact UNC-98 and MHC A. In contrast, when the C-terminal portion of UNC-98 containing all four Zn fingers, which is not necessary for binding with MHC A, is overexpressed in a wild-type background, MHC A and UNC-97 are properly localized (). Because overexpression of the myosin binding portion of UNC-98 can disrupt the normal localization of MHC A, this suggests that MHC A localization depends on UNC-98.
We wished to determine the effect of loss-of-function mutations in , the gene encoding MHC A, on myofibril organization in adults; however, available loss-of-function mutations are embryonic lethal (). Therefore, a strain was used in which a mutant was rescued by a transgenic array containing copies of the wild-type gene translationally fused to GFP (). Extrachromasomal arrays are occasionally lost upon cell division during development in . This resulting “mosaic expression” allowed visualization of body wall muscle cells lacking expression in a viable adult animal.
As shown in , in adult body wall muscle cells that lack MHC A, UNC-98 aggregates especially at the ends of the spindle-shaped cells and is not associated with focal adhesions. In contrast, in these cells, UNC-97 is not found in aggregates and is still localized to membrane-proximal regions, but in an abnormal pattern. Given that UNC-98 aggregates in cells lacking MHC A, interaction between UNC-98 and MHC A must be critical for anchorage of UNC-98 to thick filaments. The different degree of disruption of UNC-97 and UNC-98 in cells lacking MHC A is consistent with the idea that UNC-97 and UNC-98 can primarily exist in different protein complexes. UNC-97 is part of a four-protein complex associated with the cytoplasmic tail of β-integrin (), whereas thick filaments contain UNC-98 () but not UNC-97 (Fig. S1 C). The somewhat disrupted organization of UNC-97 in cells lacking MHC A can be explained by considering that the organization of integrins (and integrin-associated proteins) is directed by transmembrane signals arising from both inside and outside the cell. When MHC A (and thick filaments) are lost, at least some signals originating from the inside of the cell are lost, and thus the organization of UNC-97 is affected.
What is the effect of loss of function of on UNC-98 and MHC A? The previously reported loss-of-function mutation produces a slightly truncated protein of approximately normal abundance that retains the UNC-98 binding region, and thus it was not suitable for our studies (Fig. S2 A). Therefore, RNAi was used to examine the loss of function of . Bacteria expressing double-stranded RNA for were fed to worms beginning at the L1 larval stage to avoid embryonic lethality (). The resulting adult animals were then stained with anti– UNC-97. As shown in , some muscle cells have normally localized UNC-97, whereas other muscle cells show reduced levels of UNC-97 that is poorly organized. Significantly, in the cells showing reduced UNC-97, UNC-98 is aggregated and MHC A is mislocalized (). This suggests that the interaction of UNC-98 with UNC-97 allows its attachment to anchored focal adhesion structures. UNC-98, when properly localized at the base of the M-lines, via its interaction with UNC-97, recruits MHC A to the center of the A-band (the M-line). This interpretation is supported by the following data. When the N terminus of UNC-98, the portion of UNC-98 that has been shown not to bind UNC-97 (), is overexpressed in a wild-type background, it is diffuse within the myofibril and unable to correctly localize to focal adhesion structures. However, UNC-97 is normally localized to the dense bodies and M-lines ().
The results are consistent with a model in which UNC-98 acts as a molecular bridge between UNC-97 under the muscle cell membrane and MHC A at the M-line (). Previous studies suggest that myofibril assembly is directed by signals arising from outside the muscle cell. This was first demonstrated by showing that weak alleles of (later shown to encode an ECM protein) show retardation of myofibril assembly (). The assembly process begins with the localization of UNC-52 (perlecan) in the ECM and PAT-2 and -3 (integrins) in the muscle cell membrane, clustering at the bases of future M-lines and dense bodies (; ). This is believed to be followed by an association of the cytoplasmic tail of PAT-3 (β-integrin) with a complex of proteins that includes UNC-97 (PINCH). Previously, it was shown that the first two LIM domains of UNC-97 interact with the four C2H2 Zn fingers of UNC-98 and that UNC-98 is localized to M-lines (). In this study, our data indicate that the N-terminal portion of UNC-98 interacts specifically with the C-terminal tails of MHC A, but not MHC B (). This result is consistent with the fact that in body wall muscle M-line proteins are likely to be specifically associated with MHC A, but not MHC B, as MHC A is localized to the middle portion of thick filaments (). Supporting evidence for an interaction between UNC-98 and MHC A was provided by showing that UNC-98 interacts with purified myosin in vitro and copurifies with thick filaments (). By using antibodies to probe loss-of-function mutants and RNAi animals, it was shown that the localization of UNC-98 and MHC A are dependent on each other and on UNC-97 ( and ).
Another model for the data is that UNC-98 has a signaling function, shuttling between the integrin-associated complex near the cell membrane and the thick filaments in the A-band. Using standard immunoprecipitation buffers, UNC-98 is poorly solubilized from whole worms (unpublished data). This suggests that if a shuttling or non–thick filament–attached fraction were present, it is at low quantities.
There are two possibilities as to why the mutation does not result in a more severe disorganization of MHC A. First, the allele used, although the most severe allele of the three alleles, is not a molecular null () and some, albeit truncated UNC-98 protein, can be seen by immunoblot (Fig. S1 A). Even by RNAi for , the phenotype is not more severe than any of the mutant alleles, and a substantial amount of UNC-98 protein can be found by Western blot (unpublished data). Second, the pathway we have revealed in which UNC-98 links integrin complexes to thick filaments may be only one of several pathways that link the plasma membrane to thick filaments. For example, UNC-97 (PINCH) may interact with proteins other than UNC-98 that directly interact with myosin. Indeed, UNC-96, whose mutant phenotype is very similar to that of UNC-98 and is localized to M-lines and copurifies with thick filaments (; ), is linked to UNC-97 through two novel LIM domain proteins (unpublished data). Additionally, other members of the integrin-associated complex (UNC-112, PAT-4, and PAT-6) may also interact with proteins that link to thick filaments. Finally, the thick filaments of peripheral myofibrils may be linked to the muscle cell membrane through other proteins, such as dystrophin, spectrin, and vinculin. In mammalian skeletal muscle, these three proteins have been localized to M-lines of peripheral myofibrils ().
Linkage of thick filaments to integrin adhesion complexes at the M-line likely plays a role in transmission of contractile forces across the cell membrane to the ECM. Although an obvious vertebrate homologue of UNC-98 cannot be discerned, given its membership in a very large Zn finger protein family, it is expected that functional homologues of UNC-98 do exist. It is proposed that in vertebrate muscle, a similar mechanism of linkage between integrins and myosin thick filaments occurs at the M-lines of peripheral myofibrils.
The following strains were used in this study: wild-type N2; GB246, ; N2; ; N2; (); RW1596, (); NL2099, (); HE130, ; and HE110, (). RW1596 was provided by P. Hoppe (Western Michigan University, Kalamazoo, MI) and R. Waterston (University of Washington, Seattle, WA). NL2099, HE130, and HE110 were obtained from the Genetics Center.
Strain PJ69-4A containing pGDBU-C1 () with a cDNA insert (cDNA library provided by R. Barstead, Oklahoma Medical Research Foundation, Oklahoma City, OK) for expression of aa 1–112 of UNC-98 was used for screening (named pGDBU98-4c). Four million yeast colonies were screened, and interactors were identified as previously described (). Of 759 colonies activating the HIS3 reporter, 94 activated the ADE2 reporter. These positive clones were retransformed into pGDBU98-4c, confirming 33 positives, which were sequenced. Preys were designed using pGAD-C1 () to express MHC A, aa 1636–1937;, aa 1636–1870;, aa 1938–1969;, aa 1871–1969; and, aa 1771–1969 (); MHC B (aa 1632–1963); MHC C (aa 1639–1947); and MHC D (aa 1630–1938).
UNC-98 aa 1–112 and aa 1–310 were expressed using pET-24a, and UNC-97 (aa 146–201; the least conserved LIM domain) was expressed using both pET-24a and pGEX-6p-1 (GE Healthcare). The plasmids were transformed into BL21 (DE3)-RIL (Stratagene) and induced, and the proteins were purified as described previously (, ). Rabbits were immunized with 97-LIM3 (Spring Valley Laboratories, Inc.) to obtain Benian-16 antiserum. 97LIM-3∷GST and aa 1–112 of UNC-98 were induced and used to affinity purify Benian-16 and EU131 (), generating APBenian-16 and NAPEU131.
Total myosin II from wild-type was prepared as described by and except that the final step used a HiPrep 16/60 Sephacryl S-300 column. Fractions containing myosin were combined, and the concentration was determined. The ELISA was performed using the procedures described in with the following alterations: plates were coated with 100 μl of myosin at 50 μM, incubated in 100 μl UNC-98 aa 1–112 or aa 1–310 (in 50 mM Tris, pH 7.5) at 0–1 μM, and reacted with 75 μl of anti–UNC-98 (APEU131) at 1:1,000.
75 μg of wild-type, , and extracts and 50 μg of wild-type and extracts were separated and transblotted. The UNC-98 blot was exposed to antibodies affinity purified with full-length UNC-98 (APEU131) at 1:300 and aa 1–112 of UNC-98 (NAPEU131) at 1:1,000. The UNC-97 blot was exposed to anti–UNC-97 (APBenian-16) at 1:200. The proteins were visualized with HRP-conjugated secondary antibodies (1:10,000) and ECL (GE Healthcare).
Thick filaments from wild-type animals were purified as previously described (; ). Proteins from each step of the procedure were separated on a 4–15% SDS-PAGE gel and transblotted. The blot was exposed to anti–UNC-98 (NAPEU131) at 1:200 or anti–UNC-97 (APBenian-16) at 1:100. The supernatant from the 5,000- spin was fractionated by a 19–38% sucrose gradient. Fractions collected from the bottom of the gradient were loaded onto duplicate SDS-PAGE gels and transblotted. One blot was exposed to anti-actin (C4) at 1:2,500 (MP Biomedicals), anti-paramyosin (5–23) at 1:1,200, and anti-MHC B (5–8) 1:5,000 (); the other was exposed to anti–UNC-98 (NAPEU131) at 1:100. Proteins were visualized as described above.
Embryos from animals were suspended in S medium overnight to synchronize L1 larvae (). L1 worms were fed bacteria () expressing double-stranded RNA targeting (Ahringer clone F14D12.2; Geneservice Ltd) until they reached young adulthood and were fixed.
Wild-type animals were costained with anti–UNC-97 (APBenian-16; 1:100) and either anti–α-actinin (MH35) at 1:200 () or anti–UNC-89 (MH42) at 1:200 () using the procedures of . Alexa 488– and Alexa 647–conjugated secondary antibodies (Invitrogen) were used at 1:200. Images were captured with a confocal microscopy system (Radiance 2100; Bio-Rad Laboratories, Inc.) and displayed using LaserSharp2000 software. Using procedures described in , N2, , , N2; , animals, and N2; animals were stained using anti–MHC A (5–6) at 1:400 (), anti–UNC-98 (APEU131) at 1:200 (), and anti–UNC-97 (APBenian-16) at 1:100. FITC and Cy-3–conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) were used at 1:400. Images were captured with a deconvolution microscopy system described in and processed using Photoshop (Adobe).
Table S1 is a summary of prey clones recovered from a yeast two- hybrid screen using the N terminus of UNC-98 as bait. Fig. S1 shows verification of the specificity of UNC-98 antibodies and demonstration that UNC-98, but not UNC-97, copurifies with thick filaments. Fig. S2 shows that anti–UNC-97 antibodies recognize a protein of ∼40 kD that localizes to M-lines and dense bodies of wild-type muscle. Online supplemental material is available at . |
The extracellular signal-regulated kinase (ERK) cascade consists of the kinases Ras-activated factor (RAF), MEK (MAPK and ERK kinase), and ERK. They are coupled to a great variety of upstream activators and downstream effectors that regulate proliferation, differentiation, and survival in multicellular organisms. Mammalian cells contain three members of the RAF family (Raf-1, B-Raf, and A-Raf), two different MEK proteins (MEK1 and MEK2), and two ERK proteins (ERK1 and ERK2). These kinase isoforms appear very similar with regard to their structural and biochemical properties and, thus, do not reveal how the ERK cascade acquires signaling specificity to execute context-specific physiological functions ().
Signaling specificity could be mediated by scaffold and adaptor proteins that trigger the formation of specific signaling complexes at different subcellular locations (; ). Two scaffold proteins are known to facilitate ERK activation in mammalian cells: the kinase suppressor of Ras 1 (KSR1) and MEK1 partner (MP1). KSR1 was identified as a positive modulator of Ras/MAPK signaling (; ). Upon EGF stimulation, KSR1 is recruited to the plasma membrane, where it enhances MEK and ERK activation (). MP1 was identified in a yeast two-hybrid screen as a specific binding partner of MEK1 (). MP1 is recruited to late endosomes by the adaptor protein p14 (). MP1 and p14 are structurally almost identical and form a very stable heterodimeric complex () that is required for ERK activation on endosomes (; ). However, the biological significance of p14–MP1-facilitated ERK signaling was not known, and it was unclear whether KSR and the p14–MP1 complex would function in a redundant manner.
In this study, we show that the p14–MP1-MEK1 complex is specifically required to regulate endosomal traffic and cellular proliferation. Conditional gene targeting of in mice reveals its essential function during early embryogenesis and skin development. These findings demonstrate a crucial function of the p14–MP1-MEK1 signaling complex in the regulation of tissue homeostasis.
We investigated the biological function of the endosomal adaptor protein p14 by generating mice that carry a floxed allele. The single mouse gene (geneID 83409) is located on chromosome 3 and is ubiquitously expressed (). Exons 1–4 of were flanked by P sites to create a conditional allele (Fig. S1, A–C; available at ).
mice were intercrossed.
mice were identified in a total of >200 offspring ().
mice were born at a 2:1 ratio over their wild-type (wt) littermates, which was a clear indication of embryonic lethality.
intercrosses were analyzed at different embryonic stages. No phenotypic abnormalities could be detected at embryonic day (E) 6.5. At E8.5, ∼25% of the embryos ( = 50) were grossly growth retarded with severe developmental defects and were homozygous mutants (). By E10.5, no
embryos were detected ( = 70; ).
embryos was caused by placental defects, we used Mox2Cre (MORE) mice to delete specifically in the epiblast (). Epiblast-restricted p14 deletion also caused embryonic lethality before E10.5 ( = 86; ).
mice was not caused by placental defects but was the result of defects in the developing embryo.
mouse embryonic fibroblasts (MEFs) to determine the cellular function of p14.
MEFs were infected with an adenovirus- expressing Cre. These MEFs (
MEFs) were devoid of p14 protein () and mRNA (not depicted). Interestingly, the protein levels of the p14 interaction partner MP1 were also considerably reduced in the absence of p14, whereas protein levels of ERK1/2 were not affected ().
MEFs, whereas myc-MP1 mislocalized to the cytoplasm in
MEFs (). Reexpression of p14 restored MP1 protein levels (, third lane) and its endosomal localization (). Thus, p14 is essential to recruit MP1 to late endosomes, which, in turn, is required for efficient and sustained EGF-induced MEK and ERK signaling ().
The p14–MP1 heterodimer interacts with MEK1 (; ; ). Because MEK1 has been implicated in regulating endosomal dynamics and Golgi disassembly during mitosis (; ), we next asked whether the p14–MP1-MEK1 complex regulates endosomal transport.
and
MEFs (). However, late endosomes, multivesicular bodies (MVBs; lysobisphosphatidic acid [LBPA]), and lysosomes (LAMP1) were displaced to the cell periphery (, arrows). Reexpression of p14 restored the perinuclear localization of late endosomes (). A mutant p14caax, which resides at the plasma membrane (), did not restore proper endosomal localization (). This finding demonstrates a crucial role of p14 in the regulation of late endosomes. To investigate whether the positioning of late endosomes requires p14–MP1-MEK1 signaling, we determined the localization of MVBs and lysosomes in MEFs in which was deleted (
MEFs; ).
MEFs (, arrows), which is reminiscent of their mislocalization in
MEFs.
MEFs ().
An endosome distribution analysis was performed to determine the position of late endosomes and lysosomes (LAMP1) relative to the nucleus ( and Fig. S2 B, available at ).
MEFs, ∼80% of late endosomes and lysosomes were located in a perinuclear region (within 20 μm of the nucleus), and 20% were >20 μm away.
and
MEFs were >20 μm away from the nucleus. However, the total number of late endosomes and lysosomes was not changed ( and Fig. S2 B). These findings indicate that the p14–MP1-MEK1 complex but not KSR1 is required to regulate the distribution of late endosomes.
To determine whether the p14–MP1-MEK1 signaling complex is required for efficient transport from early endosomes to late endosomes and lysosomes, we used different endocytic cargos. The p14–MP1-MEK1 complex does not regulate the uptake or endosomal traffic of transferrin or dextran (Fig. S2 A). EGF-induced endocytosis of the EGF receptor (EGFR) into early endosomes was not affected (, 10 min).
MEFs, EGF colocalized with LAMP1-positive late endosomes (, 30 min; arrows).
and
MEFs (, 30 min). To further assess the defect in endocytic EGFR traffic, we used quantitative immunoblot analysis of EGFR degradation ( and Fig. S2 C). 60 min after EGF stimulation, >60% of total EGFR was degraded in the
MEF, whereas only 30% of total EGFR was degraded in the
MEF ( and Fig. S2 B). Together, these findings show that late endosomal sorting of activated cell surface receptors is a specific function of the p14–MP1-MEK1 signaling complex.
We next addressed how altered late endocytic traffic and reduced ERK signaling would affect tissue homeostasis. Because EGFR and ERK signaling are critical regulators of epidermal proliferation and differentiation, we performed the conditional deletion of in the epidermis. were crossed with K5-Cre2 transgenic mice () and bred further to generate ;K5-Cre2 () animals. PCR analysis from epidermal DNA and Western blot analysis from epidermal lysates demonstrated that was specifically deleted in the epidermis but not in the dermis ( and Fig. S1, D and E). mice were born alive but died shortly after birth. E18.5 embryos were alive but displayed dramatic skin defects. skin appeared erythemic and moist as compared with control littermates (). The epidermis consisted of only a few (four or less) cell layers, and nucleated cells were frequently found in the uppermost cell layer (, arrow). The stratum corneum and granular layers were not defined. This indicated compromised terminal differentiation, which resulted in a fatal skin barrier defect () and rapid dehydration, finally causing the perinatal death of the mice. These findings demonstrated an essential function of p14 in the development of the epidermis.
Immunofluorescence () and Western blot analysis from epidermal lysates () demonstrated that p14 is specifically required for MEK and ERK activation in the epidermis but does not affect the p38 or JNK pathway. Because p14–MP1-MEK1 signaling is required to regulate transport of the EGFR to late endosomes, we next asked whether the fatal failure of epidermal development is caused by aberrant EGFR traffic. Consistent with previously published results (), the EGFR was expressed in the basal cell layer of epidermis (, inset). However, in epidermis, EGFR expression was not restricted to the basal cell layer and extended frequently into suprabasal cell layers (, inset), indicating an impaired degradation of EGFR. The failure to down-regulate the EGFR in the suprabasal cell layers resulted in unscheduled and strong suprabasal keratin 6 expression (). Keratin 6 expression is known to be induced by suprabasal EGFR expression (). The expression of keratins 14, 10, and 1 was only mildly affected (Fig. S1 G). Thus, impairment of late endosomal transport and subsequent suprabasal accumulation of the EGFR might result in unscheduled keratin 6 expression, which caused the disastrous failure of the epidermal development of animals.
Induction of keratin 6 indicates a pathological status of the epidermis and is frequently associated with hyperproliferation. However, the epidermis was much thinner compared with the epidermis (). Cell death was not increased as monitored by TUNEL analysis and activated caspase-3 immunofluorescence staining (Fig. S1 H). Therefore, we investigated whether cell cycle progression was affected in the epidermis.
To detect keratinocytes in S phase, pregnant animals at 18.5 d of gestation were injected with BrdU. 1 h later, embryonic skin was collected and analyzed by immunofluorescence analysis. BrdU-positive cells resided in the basal cell layer of and epidermis. The number of BrdU-positive cells in the epidermis was reduced to 54% (). The mitotic index of the epidermis was determined by immunofluorescence microscopy with antiphosphohistone H3 antibody. Mitotic cells localized to the basal layer of and epidermis. The number of mitotic cells in the epidermis was reduced to 50% (). These findings suggested that endosomal p14–MP1-MEK1 signaling regulates proliferation in the epidermis.
Next, we asked whether the regulation of proliferation was a general and cell-autonomous function. Isolated keratinocytes exhibited no or extremely poor growth. Therefore, we used MEFs to determine whether p14 regulates proliferation.
and
MEFs were plated and grown for 3 d.
MEFs grew twice as fast compared with
MEFs ().
MEFs ().
To address whether cellular proliferation requires the endosomal localization of p14, we used the p14caax mutant, which efficiently retargets the p14–MP1 complex from endosomes to the plasma membrane ().
MEFs (). This indicates that only an endosomal p14–MP1 complex provides certain spatial information that is required for cell cycle progression. These findings substantiated a link between endosomal signaling and the regulation of proliferation.
To determine the mitotic index, MEFs were growth arrested by contact inhibition and 48-h serum starvation. MEFs were released from the growth arrest by subconfluent replating in either FCS- or EGF-containing medium (EGF data is not depicted; results were similar to those from FCS-treated cells).
MEFs had increased from 5 to 24%, whereas the number of mitotic
MEFs changed from 4 to 6% ().
MEFs (not depicted).
MEFs fully restored mitogenic proliferation (), showing that endosomal p14–MP1-MEK1 signaling regulates proliferation in a cell-autonomous manner. The defects of mitogenic entry were also reflected by reduced BrdU incorporation after mitogenic stimulation (Fig. S1 F), indicating a delay in S-phase entry. Next, we determined how the loss of p14 would affect entry into mitosis upon mitogenic stimulation using propidium iodide FACS analysis. 6 h after release from the growth arrest into FCS-containing medium, 44% of the
MEFs had entered S phase.
MEFs had entered S phase, and 70% were still in G1 phase ().
The HSV-Tk cassette for negative selection and an FRT-NEO-FRT P cassette were gifts from M. Busslinger (Institute of Molecular Pathology, Vienna, Austria). The IRES-hrGFP-SV40polyA cassette was from Vitality hrGFP (Stratagene). The genomic locus of p14 was amplified from HM-1 (E14.1 derivative) DNA with Herculase (Stratagene). All cassettes and the genomic locus for the targeting vector were cloned into pSP64 vector with a modified polylinker. The 5′ P site was inserted before exon 1. The IRES-driven hrGFP was inserted to follow p14 promoter activity upon deletion. The 5-kb-long arm was generated using three PCRs: primers (all sequences are listed in the supplemental material, available at ) 152–153, 154–181, and 180–193. The intermediate fragment, the four exons, and poly-A was amplified by PCR using primers 188–194 and was flanked 5′ by a P site. The short arm of the targeting vector was generated by PCR using the primers 184–185 and cloned 3′ of the FRT-NEO-FRT P and IRES-hrGFP cassette. Gene targeting was performed in HM-1 (E14.1 derivative) embryonic stem cells by electroporating the linearized targeting construct. For selection, 300 μg/ml G418 was used, and clones were screened by PCR and Southern blot analysis. Chimeric mice were created by the injection of two independent targeted embryonic stem cell clones into C57BL/6 blastocysts.
For genotyping, DNA was extracted from tails according to standard protocols. For genotyping of the epidermis tail, the epidermis was separated from dermis by dispase II (Roche) digest, and DNA was extracted according to standard protocols. For Southern blots, 10 μg KpnI-digested DNA were probed by using a 450-bp external probe (primers 219–220). Primer sequences are listed in the supplemental material. Western blots were performed as previously described ().
Isolated skin pieces were embedded in optimal cutting temperature–Tissue-Tek on dry ice. Immunofluorescence was performed on 6-μm frozen sections as previously described () and were analyzed using a confocal microscope (LSM510 Meta; Carl Zeiss MicroImaging, Inc.; for details, see supplemental material). For semithin section microscopy, skin was fixed with glutaraldehyde (2.5% vol/vol in 0.1 M sodium cacodylate buffer, pH 7.4) followed by unbuffered aqueous osmium tetroxide (1% wt/vol) and unbuffered aqueous uranyl acetate (0.5% wt/vol). Specimens were embedded in Epon epoxy resin. 500-nm semithin sections were stained with Toluidine blue. Immunofluorescence on MEFs was performed as previously described () and was analyzed using a microscope (Axioplan2 or confocal LSM510 Meta; Carl Zeiss MicroImaging, Inc.; for details, see supplemental material).
Primary antibodies were used according to the manufacturer's instructions. Anti-ERK1/2, antiphospho-ERK1/2, antiphospho-AKT, antiphospho-MEK1/2, antiphospho-p38, and antiphospho-JNK1/2 were purchased from Cell Signaling Technology. Antikeratins 1, 6, 10, and 14 were obtained from Covance. Anti–β-integrin 4 and anti–mouse Lamp1 antibody were purchased from BD Biosciences. Anti-Ki67 was obtained from Novacastra, anti-BrdU was purchased from Roche, and antiphosphohistone 3 was purchased from Upstate Biotechnology. Anti-EGFR and -EEA1 antibodies were obtained from Fitzgerald and Santa Cruz Biotechnology, Inc., and anti-LAMP1 was purchased from BD Biosciences. The LBPA antibody was a gift from J. Gruenberg (University of Geneva, Geneva, Switzerland). Anti-p14 and -MP1 antibodies were described previously (). Fluorescently labeled secondary antibodies are listed in the supplemental materials. MEFs were stimulated with 50 ng/ml AlexaFluor594-transferrin or 100 ng/ml of fluorescently labeled EGF for 10 and 30 min or were incubated with 3 mg/ml AlexaFluor488-dextran for 15 and 45 min.
Pregnant animals at day 18.5 of gestation were injected intraperitoneally with 1 ml/100 g bodyweight of a 10-mM BrdU (Roche) solution. Embryonic skin was removed and embedded in optimal cutting temperature– Tissue-Tek. Tissue culture cells were incubated for 30 min with 10 μM BrdU. Samples were processed according to the manufacturer's instructions.
To separate the epidermis from dermis, skin was incubated (dermis facing down) in Dispase II (Roche) for 30 min at 32°C. To extract proteins, isolated epidermis was incubated in protein lysis buffer (50 mM Tris-HCl, 200 mM NaCl, 1% Triton X-100, 10% glycerol, 5 mM NaPO, 2 mM NaOV, 50 mM NaF, 20 mM β-glycerophosphate, and protease inhibitors) and mechanically disrupted using a mixer mill tissue homogenizer (MM 301; Retsch).
MEFs were generated from day 13.5
embryos.
MEFs were immortalized with E1A retrovirus and were subsequently infected with either a control adenovirus or with an adenovirus-expressing Cre (gift from M. Cotten, GPC Biotech, Munich, Germany). Single-cell clones were selected and scored for p14 protein by Western blotting. For experiments involving starvation and mitogenic stimulation, MEFs were grown on fibronectin-coated dishes (10 μg/ml fibronectin). MEFs were growth arrested by contact inhibition and 48-h serum starvation. MEFs were released from the growth arrest by subconfluent replating in FCS-containing medium. KSR1 and KSR1 MEFs were provided by A.S. Shaw (Washington University School of Medicine, St. Louis, MO; ). MEK1 and MEK1 MEFs were gifts from M. Baccarini (University of Vienna and Medical University of Vienna, Vienna, Austria; ; ).
Embryos were incubated for 8 h at 37°C in staining solution (1.3 mM MgCl, 100 mM NaPO, 3 mM KFe(CN), 3 mM KFe(CN), and 1 mg/ml X-Gal, pH 4.5, with HCl) according to .
Mouse p14 and the p14caax cDNAs were subcloned into the retroviral transfer vector MMP (). For retroviral gene transfer, embryonic fibroblasts were transduced at an MOI of 5 in the presence of 8 μg/ml polybrene (Sigma-Aldrich).
The epidermis of at least four different mice for the respective phenotype was analyzed by counting positive cells per random field of view (at least 25). The mean was calculated with a SD in a confidence interval of P < 0.001. Growth curves of MEFs were evaluated in three independent experiments.
Images of LAMP1 immunofluorescence were acquired using a CCD camera (AxioCam HRC; Carl Zeiss MicroImaging, Inc.) on an epifluorescence microscope (Axiovert; Carl Zeiss MicroImaging, Inc.) at 16-bit data depth. Image analysis was performed with MATLAB software (The MathWorks) using custom-designed scripts. Nuclei were identified using the DAPI nuclear counterstain. The position of the geometric center of the nucleus was calculated for each cell and used as a parameter representative of the cell center. Individual endosomes were identified on the basis of their fluorescence intensities using a local thresholding approach. Endosome distances were binned in units of 2 μm in the range of 0 to 40 μm (from the nuclear center). The number of endosomes in each bin was used for further analysis. 24
(1,804 late endosomes), 20
(3,351 late endosomes), and 19
cell (1,795 late endosomes) and 21
MEFs were analyzed. Results are given in relative frequencies (percentages) for each distance. Extreme cells (Fig. S2 B) were excluded from evaluation. A chi-square independence test was performed on the original data (sum of all endosomes per bin). Endosome distance distributions were found to be significantly different (P < 0.001).
An imaging microscope (Axioplan2; Carl Zeiss MicroImaging, Inc.) with a 100× NA 1.3 oil objective (AxioPlan NeoFluar; Carl Zeiss MicroImaging, Inc.) was used for all images shown in and Fig. S2 B. For the acquisition of images, a camera (AxioCam HRc; Carl Zeiss MicroImaging, Inc.) and AxioVs40V4.5.0.0 software (Carl Zeiss MicroImaging, Inc.) were used. All fluorochromes (AlexaFluor568 donkey anti–goat IgG [H+L], AlexaFluor568 goat anti–mouse IgG [H+L], and AlexaFluor594 goat anti–rat IgG [H+L]) used in these two figures were purchased from Invitrogen.
A confocal laser-scanning microscope (LSM510 Meta; Carl Zeiss MicroImaging, Inc.) with a 63× plan-Apochromat NA 1.4 oil objective (phosphohistone 3; Carl Zeiss MicroImaging, Inc.) was used for all images shown in , , , , and in Figs. S1 (G and H) and S2 A. For the acquisition of images, LSM Image Examiner software (version 3.1.0.117; Carl Zeiss MicroImaging, Inc.) was used. All fluorochromes (AlexaFluor488-dextran [mol wt of 10,000 kD; anionic fixable], AlexaFluor594-transferrin, AlexaFluor488-streptavidin complexed to EGF and biotinylated streptavidin, AlexaFluor488 goat anti–rabbit IgG [H+L], AlexaFluor568 goat anti–mouse IgG [H+L], AlexaFluor594 goat anti–rat IgG [H+L], AlexaFluor568 donkey anti–goat IgG [H+L], AlexaFluor568 donkey anti–goat IgG [H+L], and AlexaFluor568 donkey anti–goat IgG [H+L]) used in these figures were purchased from Invitrogen.
The following primers were used: 152 (TGCGGTTTATTAGTAGTTGTGGTC), 153 (GTGCTACTGCATCGATCCTCTGTA), 154 (TACAGAGGATCGATGGCAGTAGCAC), 181 (GAAACGGTTGTGTAGTTCAGT), 180 (GTGCAATTCTGGAGCAGCTTC), 193 (AGCCTCTTGCTCTCCCTCAGT), 188 (GGGCACTGGGCAGCCCTGCCA), 194 (GCGACTGTAGGGGTGTGTTGG), 184 (GGCTTCTCCAGTGCTGTGTCT), 185 (AGACCTGCACCTGGCTCCTCT), 215 (GGTGACTACAACTCCCAGGCG), 202 (CAAGGGCATGCATAGATG), 218 (GAGTGGTTTCCTCGGGAGGAT), 208 (AATGGCCTCAACTCTCAGCTT), 170 (AGCTGGTTGCCGAACAGGATG), 219 (TGCACATGCATCCTGCCTGCT), and 220 (CTCATGGCAGGCAGGTGACTA).
The supplemental material contains a description and characterization of the conditional p14 allele, marker analysis of the p14-deficient epidermis, and a characterization of fluid phase endocytosis and recycling of transferrin in p14- and MEK1-deficient MEFs. Fig. S1 shows the targeting strategy, PCR genotyping, and Southern and Western blotting as well as differentiation and apoptosis markers in the epidermis. Fig. S2 shows that p14 is not required for transferrin receptor recycling and fluid phase endocytosis but is required for late endosomal positioning and EGFR degradation. Online supplemental material is available at . |
Normal human cells gradually lose replicative capacity after a finite number of cell divisions as a result of telomere shortening and irreversibly enter the growth-arrested state called replicative senescence (). This process is accompanied by a series of changes, including cell enlargement, increased acidic β-galactosida se activity (senescence-associated β-galactosidase [SA–β-gal]), and alterations of chromatin structures. Replicative senescence has been suggested to represent certain aspects of organismal aging ().
It is known that cells also enter a state that is indistinguishable from replicative senescence after exposure to sublethal doses of various types of stress, including reactive oxygen species, oncogenic activation, and inappropriate culture conditions. Such a stress-induced state is called premature senescence or stress-induced senescence (; ). Accordingly, cellular senescence refers to both replicative and premature senescence. Although the stimuli inducing the two types of senescence appear to be considerably diverse (telomere dependent and independent), the stress-induced MAPK p38 plays a key role in signaling such diverse stimuli to induce both types of senescence ().
Cellular senescence is not a condition that simply causes organismal aging to endanger the survival of individuals. Instead, it has been hypothesized that cellular senescence plays an adaptive role in preventing cells from immortalization and neoplastic transformation in vitro and plays a role in tumor suppression in vivo (; ). Indeed, it was recently reported that senescent cells exist in several types of premalignant tumors in vivo and that cellular senescence is required to prevent tumorigenesis in vivo (). Therefore, cellular senescence acts as a barrier to tumorigenesis as apoptosis does by stably suppressing the growth of stressed cells.
The contribution of cellular senescence to tumor suppression depends on the stable maintenance of growth arrest. Alterations of chromatin structure are believed to account for the irreversible nature of the senescent state (). In the nucleus, DNA is wrapped around histone octamers to form nucleosomes, an array of which, in turn, is packed into higher ordered chromatin structures by the association of linker histone H1 and other chromatin proteins (). Senescent cells form characteristic heterochromatin structures called senescence-associated heterochromatic foci (SAHFs; ; ). SAHFs contain methylated histone H3 at Lys9 (H3(K9)me) and heterochromatin protein 1, which are typically detected in heterochromatin. It was also reported that H3(K9)me is enriched at proliferation-promoting gene promoters specifically in senescent cells that are concomitant with the appearance of SAHFs but not in quiescent cells (). Furthermore, Suv39h1, a histone methyltransferase for H3K9, is required for oncogene-induced senescence in mouse lymphocytes and for the suppression of lymphoma (). Therefore, it is suggested that senescent cells maintain the growth-arrested state, at least in part, by stably forming heterochromatin at proliferation-promoting gene loci.
The molecular mechanism of SAHF formation remains poorly understood. Although histone chaperones Asf1a and histone regulation A have been reported to play a role in SAHF formation (), it is not known how these chaperones induce SAHFs. In this study, we have found an unexpected role of histone H1 in cellular senescence.
SAHFs are characterized by numerous, highly condensed granular DNAs spreading throughout the nucleus, as shown in WI-38 cells that were induced to senesce by the retroviral expression of oncogenic Ras (RasG12V; , bottom). SAHFs were observed in >80% of RasG12V-transfected cells as early as day 1 after drug selection. However, the degree of chromatin condensation appeared to increase during the following 7 d, during which the senescent state was stably maintained.
Because SAHFs are similar in size to inactive X chromosomes in WI-38 female normal fibroblasts (, top), we hypothesized that each granular structure originates from individual chromosomes. It is known that individual chromosomes occupy nonoverlapping and discrete spaces called chromosome territories (). Previously, it was reported that centromeres are localized at the periphery of chromosome territories in quiescent human lymphocytes and human fibroblasts (). We examined the number and position of centromeres in RasG12V-transfected senescent WI-38 cells using antibodies recognizing centromere protein B (CENP-B). Approximately 85% of CENP-B foci were colocalized with SAHF signals (). Conversely, ∼70% of SAHFs contained one CENP-B focus at their periphery (). These results suggest that individual SAHFs originate from individual chromosomes.
We also performed FISH analysis using a probe specific for chromosome 12. Optical sections of the nucleus of RasG12V-transfected senescent WI-38 cells at 0.4-μm intervals showed that two chromosome 12 signals completely overlapped with two SAHFs in all of the five senescent cells exained (). Together, we conclude that each SAHF represents a heterochromatinized portion of an individual chromosome.
It has been proposed that in interphase nuclei, individual chromosomes are surrounded by the interchromosome domain (ICD), where transcription and RNA splicing take place (). To investigate which regions in the nucleus are responsible for transcription in SAHF-positive senescent cells, RNA polymerase II was subjected to immunofluorescence (IF) analysis using specific antibodies (). In mock-transfected WI-38 cells, anti-RNA polymerase II signals were diffusely distributed throughout the nucleus. In RasG12V-transfected senescent WI-38 cells, RNA polymerase II was specifically localized in the DAPI-negative regions in a mutually exclusive manner with SAHFs (, bottom). Quantitative analyses revealed that ∼90% of SAHFs excluded anti-RNA polymerase II signals (). This result is not caused by the inability of the antibody to penetrate SAHFs because anti-H3(K9)me antibody reacted with SAHFs in IF experiments (see ).
Next, we performed in situ labeling of nascent RNAs to visualize the transcription sites (). RasG12V-transfected senescent WI-38 cells were permeabilized with 0.05% Triton X-100 and incubated with transcription buffer containing bromo-labeled UTP (BrU). The incorporation of BrU into nascent RNAs was visualized by IF analysis using anti-BrdU antibody. Although the shape of SAHFs was not well preserved after the transcription reaction, small dots of anti-BrU signals were exclusively localized in the DAPI-negative regions (). Interestingly, relatively large dots of anti-BrU signals were localized at the surface of SAHFs (, bottom enlarged images). Together, we conclude that in SAHF-positive senescent cells, active transcription occurs at the DAPI-negative inter-SAHF regions, particularly at the surface of SAHFs. This result is consistent with a previous report that SAHFs do not contain active sites for transcription () and with our hypothesis that the DAPI-negative regions correspond to ICDs (see Discussion).
Three types of condensed chromatin are characterized by specific posttranslational modifications of core histones: phosphorylations of histone H3 at Ser10 and Ser28 (H3(S10)ph and H3(S28)ph) in mitotic chromosomes, phosphorylation of histone H2B at Ser14 (H2B(S14)ph) in apoptotic chromatin, and methylation of histone H3 at Lys9 (H3(K9)me) in transcriptionally inactive heterochromatin (). It has been argued that SAHFs represent transcriptionally inactive heterochromatin because SAHFs are colocalized with H3(K9)me and heterochromatin protein 1 (; ).
To investigate to which of these distinct chromatin condensations SAHFs are related, we examined the posttranslational modifications of histone H2B and histone H3 in RasG12V-transfected senescent WI-38 cells (). In IF experiments, neither anti-H3(S10)ph, anti-H3(S28)ph, nor anti-H2B(S14)ph antibodies stained SAHFs in senescent WI-38 cells, whereas they stained mitotic WI-38 chromosomes and apoptotic HeLa nuclei (). Similarly, in immunoblotting experiments, neither anti-H3(S10)ph nor anti-H3(S28)ph antibodies detected substantial protein bands in acid-extracted (AE) histone fractions derived from mock- or RasG12V-transfected WI-38 cells, whereas strong signals were observed in histone fractions from colcemid-treated HeLa cells (). Although anti-H2B(S14)ph signals were observed in histone fractions derived from mock- and RasG12V-transfected WI-38 cells, they were much weaker than those observed in colcemid- or UV-treated HeLa-derived histone fractions (). These results indicate that SAHFs are a type of chromatin condensation that is distinct from those found in mitotic chromosomes or apoptotic cells. Anti-H3(K9)me antibody reacted with SAHFs in senescent cells as well as the whole nuclear regions in mock-transfected cells (), as previously reported (). Importantly, however, immunoblotting analysis of AE fractions using anti-H3(K9)me antibody revealed that the total amount of H3(K9)me per unit cell (calibrated by anti-H2B signals) was not substantially different between mock- and RasG12V-transfected WI-38 cells (), suggesting that the apparently strong anti-H3(K9)me signals at SAHFs reflect the condensation of DNAs.
To analyze the molecular mechanisms involved in SAHF formation, we examined chromatin proteins in mock- and RasG12V-transfected WI-38 cells. The fractionation scheme used in this study is shown in . SDS-PAGE and Coomassie staining of the fractionated proteins revealed several protein bands showing distinct patterns in mock-transfected and senescent cells (). Notably, a protein band having an apparent molecular mass of 37 kD (hereafter referred to as a 37-kD protein) was found in chromatin-rich fraction N2 derived from mock-transfected cells but not from senescent cells (, arrow). Mass spectrometric analysis identified this protein as histone H1 ().
To confirm that the 37-kD protein is histone H1, we performed immunoblotting analyses using two different antihistone H1 antibodies prepared in rabbit or sheep (). Although the anti-H1 rabbit antibody detected a single band, the anti-H1 sheep antibody detected two bands showing apparent molecular masses of 37 kD (, lanes 4, 5, 7, and 8; arrowheads). This is probably the result of different specificities of these antibodies for histone H1 variants. In both cases, no anti-H1 signals were detected in fraction N2 derived from RasG12V-transfected senescent WI-38 cells (, lanes 6 and 9).
To further confirm that histone H1 is lost from senescent cell chromatin, histones were prepared from fraction N2 by acid extraction (). Histone H1 prepared from colcemid-treated HeLa cells (, lanes 2 and 6) showed lower mobility than that prepared from nontreated HeLa cells (, lanes 1 and 5) and mock-transfected WI-38 cells (, lanes 3 and 7), which is consistent with the hyperphosphorylation of histone H1 at metaphase. Histone H1 was not detected by Coomassie staining (, lane 4) or by anti-H1 sheep antibody (, lane 8). Collectively, we conclude that histone H1 is lost from the chromatin of RasG12V-transfected senescent WI-38 cells.
To determine whether histone H1 existed in any other fractions of RasG12V-transfected senescent WI-38 cells, we performed immunoblotting analyses of whole cell extracts (WCEs) using anti-H1 sheep antibody (). WCEs were prepared from mock-transfected, RasG12V-transfected, and serum-starved quiescent (0.1% FBS) WI-38 cells by directly suspending cell pellets in SDS sample buffer. No 37-kD histone H1 band was detected in the WCE from senescent cells (, lane 3). Time-course experiments showed that the majority of histone H1 was lost between days 2 and 3 after the completion of drug selection in the RasG12V-transfected cells (). These results indicate that histone H1 is lost from the entire cell in the Ras-induced senescence.
We extended the experiments to include other types of normal human fibroblasts (MRC-5, IMR-90, and BJ; ). The amount of histone H1 was moderately decreased in RasG12V-transfected MRC-5 and IMR-90 cells (, lanes 4 and 6) but not in BJ cells (, lane 8). Importantly, as shown in , the extent of histone H1 reduction is well correlated with the frequency of SAHF-positive cells after RasG12V expression. We also examined histone H1 in cellular senescence induced by other types of stimuli. Histone H1 was lost from or considerably decreased in senescent cells induced by extensive culturing (replicative senescence) or by the ectopic expression of a constitutive active form of MKK6 (MKK6EE) that activates the stress-induced MAPK p38 (; ; ). Moreover, the extent of histone H1 reduction was correlated with the frequency of SAHF-positive cells. Together, the loss of histone H1 is observed not only in RasG12V-transfected WI-38 cells but also in other types of RasG12V-transfected normal human fibroblasts and in senescent cells induced by other stimuli, extending its correlation with SAHF formation.
The expressions of most histone genes are coupled with DNA replication (). As senescent cells do not undergo DNA replication (; ), it is expected that histone H1 is transcriptionally repressed. Indeed, RT-PCR analysis showed that the mRNA expression of , one of the major histone H1 genes encoding histone H1.3, was repressed in RasG12V-transfected senescent WI-38 cells compared with young proliferating cells (10% FBS; , lanes 1 and 3). This raised the possibility that the loss of histone H1 from senescent cell chromatin is simply caused by the repression of de novo histone H1 synthesis. We found that the mRNA expression of was not detected in quiescent WI-38 cells that had been serum starved for 1 wk (note that 1 wk is long enough for RasG12V expression to cause histone H1 loss; and , lane 2), as expected from the observation that the fraction of quiescent cells in S phase was very small (, right). Nevertheless, the serum-starved quiescent WI-38 cells contained the same amount of chromatin-bound histone H1 protein as proliferating WI-38 cells cultured with 10% FBS (), arguing that repression of the de novo histone H1 synthesis itself is not sufficient for the loss of histone H1 protein from chromatin. Therefore, the loss of histone H1 from senescent cells is most likely posttranslationally regulated.
We next examined whether or not the ectopic expression of histone H1 prevented RasG12V-induced senescence. Young proliferating WI-38 cells were first infected with retrovirus (driven by the weak long terminal repeat promoter or the strong cytomegalovirus promoter) or lentivirus (driven by the strong EF1α promoter) expressing untagged histone H1 and were infected with RasG12V-expressing retrovirus (). The ectopic expression of histone H1 only slightly increased the amount of chromatin-bound histone H1 in RasG12V-transfected cells (, compare lane 2 with lanes 4, 6, and 8). However, it did not substantially prevent the RasG12V-induced loss of histone H1, growth arrest, or SAHF formation () even when histone H1 was expressed by the strong cytomegalovirus and EF1α promoters. Together, these experiments do not conclude whether or not histone H1 loss has any causative role in cellular senescence because we could not manipulate the histone H1 level sufficiently to induce phenotypic changes.
We next examined whether the ectopic expression of various types of recombinant histone H1 alone had any effects on WI-38 cells. Young proliferating WI-38 cells were transfected with one of four retroviruses expressing histone H1 fused to EGFP or HA (, top). Fluorescence microscopy revealed that all histone H1 fusion proteins were localized in nuclei and colocalized with mitotic chromosomes ( and Fig. S3 A, available at ), indicating their proficient chromatin binding in nuclei. Unexpectedly, cells expressing N-terminal EGFP-H1 fusion proteins (N fusions 1 and 2) showed decreased levels of chromatin-bound endogenous histone H1 compared with cells expressing C fusion or HA-H1 (). N fusions 1 and 2 have different linker amino acid sequences between EGFP and the histone H1 open reading frame (see Materials and methods) but showed the same phenotypes in WI-38 cells.
Interestingly, concomitant with the decreases in histone H1 protein levels, the expression of N fusions caused severe growth defects in young WI-38 cells as revealed by the growth curves () and the BrdU incorporation assay (, top). N fusions also induced a large, flat morphology and increased the percentage of SA–β-gal–positive cells (Fig. S3 B and , bottom), suggesting that the ectopic expression of N fusions induced premature senescence in WI-38 cells. Consistent with this idea, cells expressing N fusions showed decreased levels of phosphorylated Rb and cyclin A and increased levels of p21 and phosphorylated p53 (), which are typical molecular markers for senescent cells. In contrast, the ectopic expression of C fusion or HA-H1 did not induce senescence markers or decrease the levels of endogenous histone H1. The ectopic expression of untagged histone H1 slightly increased the levels of histone H1 protein but did not induce senescence phenotypes in WI-38 cells either (, lane 3). Collectively, these results suggest that the induction of senescence phenotypes is correlated with the decreased levels of chromatin-bound endogenous histone H1. Notably, N fusions did not induce the accumulation of p16 or SAHF formation (, DNA).
We found that a protein band showing an apparent molecular mass of 22 kD was accumulated in the AE fraction prepared from RasG12V-transfected senescent WI-38 cells (, arrowhead). Mass spectrometric analysis and immunoblotting analysis identified this protein as high mobility group A2 (HMGA2; ). Similar observations have been reported recently (). Time-course experiments indicated that the accumulation of chromatin-bound HMGA2 was inversely correlated with the loss of histone H1 from chromatin (, lanes 5–8). Interestingly, the amount of chromatin-bound HMGA2 was also increased in senescent WI-38 cells expressing both histone H1 and RasG12V (). We also showed that macroH2A was incorporated into chromatin in RasG12V-transfected cells as previously reported (, lanes 5–8; ).
To investigate the role of macroH2A and HMGA2 in senescence phenotype induction and SAHF formation, young proliferating WI-38 cells were transfected with HA-tagged HMGA2 (HA-HMGA2) or FLAG-tagged macroH2A (macroH2A–FLAG; ). The amounts of chromatin-bound HA-HMGA2 and macroH2A–FLAG were comparable with those of endogenous proteins observed in RasG12V-transfected cells (, lanes 2–4; and Fig. S4 A, available at ). Furthermore, IF analysis revealed that both HA-HMGA2 and macroH2A–FLAG were localized in nuclei (Fig. S4 B and not depicted). The expression of HA-HMGA2 or macroH2A–FLAG did not induce senescence phenotypes, including growth defects, a large, flat morphology, or SAHF formation (, bottom; Fig. S4, C and D; and not depicted).
To investigate whether the overexpression of HMGA2 or macroH2A collaborates with the reduced endogenous histone H1 level found in N fusion–expressing cells, young WI-38 cells expressing HA-HMGA2 or macroH2A–FLAG were further transfected with the retrovirus expressing N fusion 2 and were scored for SAHF-positive cells (). The coexpression of HA-HMGA2 and N fusion 2 significantly increased the percentage of SAHF-positive cells (P < 0.001), whereas the coexpression of macroH2A–FLAG and N fusion 2 did not. The coexpression of HA-HMGA2 and N fusion 1 led to similar results, but the coexpression of HA-HMGA2 and HA-H1 did not (Fig. S5), suggesting that the removal of endogenous histone H1 from chromatin by N-fusion histone H1 rather than the production of recombinant histone H1 itself is important for SAHF formation. Together, these results indicate that the simultaneous overexpression of HA-HMGA2 and N-fusion histone H1 leads to SAHF formation. However, cells coexpressing HA-HMGA2 and N fusion 2 did not show the accumulation of p16, at least up to day 3 after drug selection (unpublished data).
#text
The following primary antibodies were used: anti-RNA polymerase II (8WG16), anti-HA (16B12), and anti-HMGA2 (Covance); anticyclin A, anti-p16, anti-p21, anti-p53 (Bp53-12), anti–H-Ras, and anti-MKK6 (Santa Cruz Biotechnology, Inc.); antiphospho-S10 histone H3, antiphospho-S14 histone H2B, antidimethyl-K9 histone H3, antihistone H2B, and antihistone macroH2A1 (Upstate Biotechnology); anti-BrdU (BMC9318; Roche); antiactin (Chemicon); antihistone H1 sheep (Abcam); anti-Rb (G3-245; BD Biosciences); antiphospho-S15 p53 (Cell Signaling); anti–CENP-B (provided by H. Masumoto, Nagoya University, Nagoya, Japan); antiphospho-S28 histone H3 (provided by T. Urano, Nagoya University); and antihistone H1 rabbit (provided by K. Ohsumi, Tokyo Institute of Technology, Kanagawa, Japan).
In situ labeling of nascent RNAs was performed as previously described () with slight modifications. In brief, cells were rinsed with glycerol buffer (20 mM Tris-HCl, pH 7.5, 5 mM MgCl 0.5 mM EGTA, 25% glycerol, and Complete protease inhibitors [Roche]) containing 0.05% Triton X-100 at 4°C for 1 min. Then, the cells were incubated in transcription buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl, 0.5 mM EGTA, 100 mM KCl, 25% glycerol, 25 μM S-adenosylmethionine [New England Biolabs, Inc.], 0.5 mM each of BrU [Sigma-Aldrich], ATP, CTP, and GTP, 8 U/ml RNase inhibitor [Takara], and Complete protease inhibitors) at 35°C for 10 min. Incorporated BrU was visualized by IF analysis using anti-BrdU antibody. For BrdU incorporation assay, cells were cultured in DME supplemented with 100 μM BrdU for 5 h, fixed with 2% formaldehyde at 4°C for 15 min, and permeabilized with TB at room temperature for 5 min. After treatment with 1.5 N HCl at room temperature for 10 min, incorporated BrdU was visualized by IF analysis using anti-BrdU antibody.
FISH was performed using a chromosome 12–specific probe conjugated with Cy-3 (StarFISH; Cambio). FISH on metaphase spreads was performed according to the manufacturer's protocol. FISH on senescent cells was performed as described previously ().
Cells were harvested and incubated in buffer A (40 mM Tris-HCl, pH 7.4, 1 mM EDTA, 0.5 mM DTT, 1 mM sodium orthovanadate, 1 mM sodium fluoride, 100 nM okadaic acid, and Complete protease inhibitors) containing 10 mM NaCl at 4°C for 10 min. Then, the cells were passed through a 26- (mock-transfected cells) or 23-gauge (RasG12V-transfected cells) needle ∼20 times to disrupt the cell membrane and were centrifuged at 1,000 for 10 min at 4°C. Supernatants were removed (fraction C), and pellets were incubated in buffer A containing 150 mM NaCl and 0.2% Triton X-100 at 4°C for 10 min. After centrifugation at 1,000 for 10 min at 4°C, supernatants were removed (fraction N1). Pellets (fraction N2) were resuspended in the SDS sample buffer (125 mM Tris-HCl, pH 6.8, 4% SDS, and 20% glycerol, colored with bromophenol blue) and sonicated. For acid extraction of histones, fraction N2 was incubated with 200 mM sulfuric acid at 4°C for 30 min and centrifuged at 20,000 for 10 min at 4°C. Supernatants were supplemented with 1/4 vol of 100% trichloroacetic acid and incubated at 4°C for 10 min. After centrifugation at 20,000 for 5 min at 4°C, pellets were rinsed with cold acetone, dried, and resuspended in the SDS sample buffer. For the preparation of cell lysates (
), cells were suspended in lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% NP-40, 1 mM EDTA, 1 mM DTT, 1 mM sodium orthovanadate, 1.25 mM sodium fluoride, 100 nM okadaic acid, 10 mM β-glycerophosphate, and Complete protease inhibitors) at 4°C for 30 min. After centrifugation at 20,000 at 4°C for 10 min, the supernatants were used for immunoblotting analyses using anti-Rb, cyclin A, p16, p21, p53, P-p53, Ras, and actin antibodies. Pellets were suspended in the SDS sample buffer and used for immunoblotting analyses using anti-H1, HMGA2, and H2B antibodies.
RT-PCR was performed with an RNA PCR kit (Takara) according to the manufacturer's protocol. PCR primers used in this study are (5′-CAAGCCTAGGAAGCCTGCTG-3′ and 5′-AAAGGGGAACGTCCCGCCAG-3′) and (5′-ATTTGGTCGTATTGGGCGCCTGGTC-3′ and 5′-TTGTCATACTTCTCATGGTTCACAC-3′).
For SA–β-gal staining, cells were fixed with 0.5% glutaraldehyde in PBS at room temperature for 10 min and incubated in staining solution (0.65 mg/ml X-Gal [5-bromo-4-chloro-3-indolyl-β--galactoside], 5 mM KFe(CN), 5 mM KFe(CN), and 1 mM MgCl in PBS, pH 6.0) at 37°C overnight.
Fig. S1 shows the specificities of FISH signals and anti-BrdU IF signals. Fig. S2 shows the specificity of antihistone H1 sheep antibody. Fig. S3 shows the colocalization of N-fusion H1 and mitotic chromosomes and the enlargement of WI-38 cells expressing N-fusion H1. Fig. S4 shows the effect of HA-HMGA2 expression on cell growth and SAHF formation. Fig. S5 shows the effect of N-fusion 1 expression on SAHF formation. Online supplemental material is available at . |
Control of spindle length is important for different reasons during mitotic and meiotic cell divisions. Mitotic spindles of most animals and fungi lengthen in a process called anaphase B. This spindle lengthening provides a mechanism for moving sister chromatids apart that is redundant with anaphase A. The meiotic spindles of animal oocytes, however, are different because they mediate highly asymmetric cell divisions in which chromosomes are discarded in nonviable division products called polar bodies. The female meiotic spindles of mouse do not lengthen during anaphase, and a mutation causing anaphase B–like spindle elongation results in a large polar body and a correspondingly small egg (). This result indicates that restriction of meiotic spindle length may be important for preserving egg volume and egg contents required by the embryo. Indeed, the meiotic spindles of zygotes actually shorten in a coordinated, anaphase-promoting complex-dependent manner during meiosis I and II (, ).
Results published during the last 10 yr indicate that a discrete set of spindle-lengthening and spindle-shortening mechanisms can control overall spindle length. In many organisms, the plus ends of kinetochore fiber microtubules polymerize at the same rate that the minus ends depolymerize (). When plus-end polymerization is blocked with taxol () or by depletion of the kinetochore protein, CLASP (CLIP170-associated protein; ), spindles shorten because of continued minus-end depolymerization. Conversely, when minus-end depolymerization is blocked experimentally, spindles can elongate continuously (). In addition, spindles elongate excessively in the absence of the kinetochore-associated microtubule depolymerase, KLP67A (; ). Another major lengthening mechanism is outward sliding of overlapping antiparallel microtubules mediated by kinesin-5 family members (; ). In unperturbed embryonic spindles, a cell cycle–regulated cessation of minus-end depolymerization coincides with the initiation of anaphase B spindle elongation. In this case, the rate of kinesin-5–driven outward sliding is matched by the rate of minus-end depolymerization until this depolymerization stops (). Evidence has also been presented for two other spindle-shortening mechanisms, inward sliding of overlapping antiparallel microtubules by kinesin-14 family members (; ) and an undefined “tensile element” that squeezes inward and buckles microtubule bundles during nocodazole-induced spindle shortening ().
None of these mechanisms appear to be universally used for spindle-length control. For example, cessation of minus-end depolymerization does not alter spindle length in human U2OS cells () and minus-end depolymerization does not occur at all in some fungal cells (; ). Separation of spindle poles occurs in mitotic embryos even after laser cutting of spindle microtubules (). This outward movement of spindle poles is thought to be driven by a cortical motor protein pulling on astral microtubules that extend from the spindle pole toward the cortex. Thus, mitotic spindles can “elongate” in the absence of an outward sliding mechanism. This may explain why the sole kinesin-5 in this species is not essential (). In contrast, budding yeast mitotic spindles can elongate in the complete absence of cortical pulling forces (), and this species' kinesin-5 family members are essential ().
The goal of this study was to determine whether the microtubule-severing protein, katanin, has a conserved role in spindle-length control. Katanin is a heterodimeric protein consisting of an AAA ATPase subunit called p60 and an accessory subunit called p80. The ATPase subunits from sea urchin () and human () can sever microtubules on their own, but this activity is enhanced by the p80 subunit. Katanin's microtubule-severing activity and its conserved localization at the spindle poles of sea urchins () and vertebrates () make it a strong candidate for playing a role in spindle-length control.
We previously reported that inhibition of katanin in mammalian cells causes a change in the distribution of γ-tubulin and a reduction in the rate of nocodazole-induced spindle microtubule depolymerization as assayed by decreasing fluorescence intensity of YFP-tubulin (). Here, we report that katanin inhibition slows the rate of nocodazole-induced spindle shortening in mitotic mammalian cells and blocks the second of two meiotic spindle shortening phases in .
Previous studies have demonstrated that inhibition of microtubule polymerization with taxol () or nocodazole () can induce mitotic spindle shortening and thus reveal spindle-shortening forces that were balanced by polymerization-dependent spindle-lengthening forces before drug treatment. To test whether katanin contributes to inward forces in the mitotic spindle, we measured the velocity of nocodazole-induced spindle shortening in cells expressing YFP-tubulin and CFP fusions to each of two katanin inhibitors. We previously found that these inhibitors slowed the rate of nocodazole-induced microtubule depolymerization as assayed by the rate of decrease in YFP-tubulin fluorescence intensity but did not quantify spindle-shortening velocities (). Nocodazole-induced spindle shortening was biphasic with an early, fast phase and a later, slow phase. As shown in , both con80, a p80 katanin fragment that displaces endogenous p60 katanin from spindle poles (), and Ploop-p60, a point mutant of p60 katanin that inhibits the in vitro microtubule-severing activity of wild-type katanin (), slowed both the early and late phases of nocodazole-induced spindle shortening relative to control cells expressing CFP alone. In contrast, a CFP fusion to a p60 katanin subunit that does not inhibit severing by wild-type katanin, ΔN-Ploop-p60 (), did not slow the rate of spindle shortening relative to control cells expressing CFP alone. These results indicate that katanin-mediated microtubule severing promotes spindle shortening when polymerization is blocked. We propose that microtubule severing in untreated spindles generates arrays of short, overlapping microtubules near spindle poles. Nocodazole-mediated depolymerization would lead to a rapid loss of microtubule overlap that is required for the kinesin 5–mediated outward sliding that opposes spindle-shortening mechanisms. In the absence of microtubule-severing activity at spindle poles, microtubules would be longer, overlap zones would be longer, and more time would be required for nocodazole to cause a loss of overlap (see Discussion).
To test the generality of katanin's role in spindle-length control and to examine spindle shortening in the absence of drugs, we continued our analysis with female meiotic spindles. These spindles were chosen because they are the only spindles that have well-documented and dramatic shortening phases () and they are the only spindles that have a demonstrated requirement for a katanin homologue, MEI-1–MEI-2 (; ).
In wild-type , MI and MII spindles shorten from steady-state metaphase lengths of 7.7 and 6.2 μm, respectively, to minimum lengths of 2.8 and 2.4 μm (; and ). After reaching minimal length, the spindles narrow and lengthen to form a midbody that extends into the polar body. As shown in , spindle shortening could occur by either of two general mechanisms, inward sliding of antiparallel microtubules or depolymerization of microtubule minus ends at spindle poles. In a sliding-only mechanism, microtubule density increases as the spindle shortens. If spindle shortening proceeded only by minus-end depolymerization or by a combination of inward sliding and plus-end depolymerization, microtubule density would not increase.
To distinguish between these possibilities, changes in microtubule density were monitored by obtaining the mean fluorescence intensity of GFP∷tubulin (mean pixel value) in the spindle throughout meiosis (see Materials and methods). In the example shown in , GFP∷tubulin fluorescence intensity remained constant when spindle length was constant (, 0–1.5 min) but began to increase dramatically as the spindle shortened (, 1.5–5 min). This increase in microtubule density abruptly switched to decreasing microtubule density when spindle length reached 56% of its starting metaphase length (, 5–7.5 min). The switch between increasing and decreasing microtubule density began within 1 min of spindle rotation from parallel to perpendicular relative to the cortex. We previously demonstrated that homologue separation initiates 0.7 min after spindle rotation (). After the transition from increasing to decreasing density, the spindle continued to shorten to 33% of its starting length. Similar results, obtained in 12 out of 12 experiments () using wild-type during either MI or MII, indicated that there are two sequential and mechanistically distinct phases of meiotic spindle shortening. The first phase most likely occurs by inward sliding of antiparallel microtubules that stops just before rotation, perhaps because at that point the microtubules are completely overlapped. Microtubule density increases uniformly in all parts of the spindle during the first phase (unpublished data), possibly because the microtubule overlap zones are heterogeneous in length and position within the spindle. The second shortening phase occurs after rotation, during chromosome segregation, and is accompanied by net microtubule depolymerization.
To test the role of microtubule severing in meiotic spindle shortening, we chose a partial-loss-of-function mutation in the p80 katanin homologue, , because null mutants of or the p60 katanin homologue, , do not assemble bipolar spindles (), and failure of these spindles to shrink () might be an indirect result of a lack of antiparallel microtubules. The missense mutant has a reduced amount of MEI-2 protein (). We tested the in vitro microtubule-severing activity of purified MEI-1 for the first time () and found that MEI-2 is absolutely required for the microtubule-severing activity of MEI-1. Thus, MEI-1 differs from katanin catalytic subunits from sea urchin (), human (), or (), each of which sever microtubules on their own. The MEI-2 dependence of MEI-1 suggests that meiotic spindles should have a reduced amount of microtubule-severing activity. worms were previously reported to lay viable embryos with large polar bodies (), suggesting that these worms retain sufficient MEI-1–MEI-2 activity to assemble bipolar spindles but that these spindles might be unusually long at polar body induction.
Time-lapse imaging of a strain expressing GFP∷tubulin revealed that both the maximum and the minimum meiotic spindle lengths were considerably greater than in wild type (). Wild-type spindles maintain a constant metaphase length before initiating shortening (, ). If this steady-state length were the result of a balance between katanin-dependent shortening forces and katanin-independent lengthening forces, one might expect to see a period of spindle lengthening during metaphase in embryos. Instead, spindles maintained a constant length before initiating shortening in nine out of nine embryos that remained parallel to the plane of focus for 3 or more minutes during metaphase. This suggests that the increased maximum length of spindles is a result of a defect during spindle assembly rather than a change in a balance of forces during metaphase.
The abnormally long metaphase spindles of embryos shortened at a wild-type velocity of −0.95 μm/min to 6.2 μm for MI or 5.8 μm for MII during a period of increasing microtubule density but maintained a constant length as microtubule density decreased ( and ). The percentage of spindle shortening in mutants (MI, 64%; MII, 63%) is similar to the percentage of shortening that is observed for wild-type spindles at the switch to decreasing microtubule density (MI, 58%; MII, 63%; see ), a transition that occurs nearly concurrently with spindle rotation. At this point, the velocity in drops to −0.01 μm/min compared with −0.74 μm/min for wild type, and consequently the mutant spindle length remains at 65% of maximum, whereas that of wild type decreases to 38%. These results suggests that spindles undergo the initial, sliding-dependent phase of shortening but do not undergo the second phase of shortening, which occurs after rotation in wild-type embryos. Consistent with this interpretation, chromosome separation in spindles initiated within 1 min after spindle shortening ceased or microtubule density began decreasing ( = 3), just as separation initiated within 1 min after rotation in wild type ( = 14; ).
In contrast to wild-type spindles, which rotated at the time that microtubule density switches from increasing to decreasing, spindles did not rotate (). The spindle remained parallel to the cortex as the cortex invaginated and, in the example shown in , only later turned toward the cell interior by the invagination of the cortex. The greatest angle at which spindles interacted with the cortex was highly variable in contrast to wild-type spindles, which rotated to a near perfect 90° angle (). Although it is possible that the MEI-2 gene product is directly required for spindle rotation, a second hypothesis is that rotation can only occur when the spindle has shortened to a critical length. Polar bodies in embryos were larger and more variable in size than in wild-type embryos (), suggesting that the wild-type function of spindle shortening and rotation is to minimize polar body size and thus preserve embryo volume.
To more carefully compare changes in microtubule density during the late stage of spindle shortening, we used a strain expressing both GFP∷tubulin and mCherry∷histone to monitor spindle changes initiating at anaphase onset. Changes in the fluorescence intensity profiles down the pole–pole axis of a wild-type and a spindle are shown in . In wild-type spindles, microtubule density in the spindle poles decreased, whereas microtubule density increased between the separating chromosomes. In contrast, spindles exhibited a nearly uniform microtubule density down their pole–pole length and the relative density at the poles versus the midzone did not change over time. Similar results were obtained in 11 out of 11 wild-type embryos and 13 out of 13 spindles. Thus, the microtubule-severing activity of MEI-1–MEI-2 katanin is required for the redistribution of microtubules from the spindle poles to the spindle midzone. Both wild-type () and spindles () decrease in overall microtubule density after anaphase onset, indicating the additional action of katanin-independent microtubule depolymerizers in the reduction of overall microtubule density.
It could be that MEI-1–MEI-2 located at spindle poles is suddenly activated at anaphase onset, and its microtubule-severing activity promotes disassembly of spindle pole microtubules directly. It is also plausible that microtubule-severing activity during spindle assembly generates a wild-type spindle composed of overlapping short microtubules that are transported from the poles to the midzone after anaphase onset. In this case, spindles would fail to undergo normal postanaphase morphogenesis simply because they were composed of excessively long microtubules. To test when MEI-1–MEI-2 katanin might be acting, we examined the localization of GFP∷MEI-1 by time-lapse imaging ( and Fig. S2, available at ). GFP∷MEI-1 was localized uniformly on spindle microtubules during early stages of MI spindle assembly (not depicted) and subsequently concentrated on both spindle poles and on chromosomes (). The metaphase localization pattern was similar to that previously shown by immunofluorescence for both MEI-1 () and MEI-2 (). GFP∷MEI-1 fluorescence intensity was the most intense and focused just after spindle rotation and when chromosomes had just reached the poles (). After this stage, GFP∷MEI-1 fluorescence decreased substantially as the midbody lengthened ( and Fig. S2) and was then associated predominantly with chromosomes. The presence of GFP∷MEI-1 at spindle poles just after rotation is consistent with a direct role of microtubule severing in the second phase of spindle shortening, and the decreasing intensity of GFP∷MEI-1 at the end of this phase is consistent with the decreasing density of microtubules at spindle poles ().
We previously reported that spindles do not undergo timely translocation to the cortex (). This translocation defect might indicate a direct role for microtubule severing in kinesin-dependent spindle translocation () or it might be an indirect consequence of the lack of antiparallel microtubule organization in spindles. To test whether the spindle translocation defect in mutants is simply a consequence of disorganized spindles, spindle translocation was examined in worms expressing either GFP∷tubulin or GFP∷histone. KLP-18 was chosen for this analysis because depletion of this motor was previously shown to have meiosis-specific spindle organization defects (). As shown in , spindles, like wild-type spindles, translocated to the cortex either before the embryo exited the spermatheca or within 2 min of spermatheca exit. As previously reported (), spindles reached the cortex 9.9 ± 2.4 min after spermatheca exit. Surprisingly, this strong translocation defect in mutants was completely suppressed by RNAi-mediated depletion of KLP-18 (). These results indicate that bipolar spindle organization is not required for timely spindle translocation, the translocation defect in mutants is a consequence of a KLP-18–dependent process, and katanin is not directly required for spindle translocation. We suggest that in the absence of katanin, rigid, KLP-18–dependent microtubule bundles extend from the disorganized meiotic spindles and sterically inhibit movement toward the cortex.
Although the spindle translocation defect in mutants was suppressed by , bipolar spindles were still not assembled in double mutants, indicating a direct role for katanin in spindle assembly. One possible role for katanin in spindle assembly was suggested by work on the plant cortical microtubule cytoskeleton, where katanin is required for formation of organized parallel/antiparallel microtubule bundles (), just as it is in meiosis. Plant cortical microtubules are nucleated from microtubule-associated γ-tubulin complexes at a discrete 42° angle from preexisting microtubules (). It is intriguing to speculate that katanin-mediated severing might be required to release microtubules from these branched networks to allow parallel bundling. If this were the case with acentrosomal meiotic spindles, then γ-tubulin (RNAi) might partially suppress the need for katanin in spindle assembly. An alternative hypothesis is based on the fact that microtubule severing and microtubule nucleation both have the same net effect of increasing microtubule number. In this hypothesis, katanin and γ-tubulin might have partially overlapping roles in increasing microtubule number early in spindle assembly. In this case, γ-tubulin (RNAi) would be expected to worsen the phenotype of a mutant.
To test these hypotheses, GFP∷tubulin-expressing worms were fed bacteria expressing (γ-tubulin) or (γ-tubulin ring complex [γ-TuRC] subunit) double-stranded RNA for 40–44 h. This amount of time is required to deplete maternal protein from the syncytial gonad. Longer feeding regimens resulted in oogenesis defects and failures in ovulation. Meiotic spindle morphology appeared wild type in or worms (), consistent with previous reports of normal polar body size and number (). Microtubule density, as determined from the ratio of mean GFP∷tubulin fluorescence in the spindle to fluorescence in the cytoplasm, was similar between wild-type, , , and spindles but was dramatically reduced in ; or ; double-mutant spindles (). This synthetic effect was not observed in ; double-mutant worms (Fig. S3, available at ), indicating that the enhancement was specific. Neither nor suppressed the spindle morphology defect in worms. Because and enhanced rather than suppressed the spindle morphology defect in worms, these results suggest that MEI-1 and γ-tubulin play redundant roles rather than antagonistic roles in promoting increasing microtubule density during spindle assembly. In addition, because the mutant completely lacks function, we can conclude that MEI-1 and γ-tubulin are acting in parallel rather than sequentially.
If this hypothesis were correct, then MEI-1 and γ-tubulin should both be present at the time and place that meiotic spindle assembly initiates. As shown in , meiotic spindle assembly initiates at nuclear envelope breakdown, when a diffuse cloud of microtubules fills the volume of the perforated nucleus. These microtubules then organize into a much smaller assembly of dense microtubule bundles by the time the oocyte ovulates into the spermatheca (; ). By the time the embryo exits the spermatheca and enters the uterus, a bipolar, metaphase-length spindle is present within a vesicle-free zone visible by differential interference contrast (). Before NEBD, GFP∷TBG-1 became progressively more concentrated on the nuclear envelope as oocyte maturation progressed (). At nuclear envelope breakdown, GFP∷TBG-1 fluorescence entered the nuclear region, forming a diffuse cloud that might indicate association with either microtubules or vesicles derived from the nuclear envelope. This large, diffuse area of γ-tubulin enrichment did not contract in diameter nor increase in local intensity (), as did the α-tubulin (). By the time the embryo exited the spermatheca into the uterus, no discrete localization of GFP∷TBG-1 was observed in the vesicle-free zone that contains the metaphase I spindle (). GFP∷MEI-1 was also distributed throughout the volume of the perforated nucleus immediately after nuclear envelope breakdown (). GFP∷MEI-1 differed from GFP∷TBG-1 in that it became highly concentrated on chromosomes and spindle poles during spindle assembly and remained concentrated on these structures throughout meiosis (, , and Fig. S2). The transient presence of microtubules, MEI-1, and γ-tubulin within the perforated nucleus immediately after nuclear envelope breakdown is consistent with the hypothesis that microtubule severing and γ-TuRC–templated nucleation provide redundant mechanisms for increasing microtubule number during the initial stages of spindle assembly.
Our results demonstrate that inhibition of katanin inhibits spindle-shortening processes in fibroblast mitotic spindles and meiotic spindles. Mitotic spindles of cultured mammalian cells do not normally shorten, but intrinsic shortening processes can be revealed when microtubule polymerization is blocked with nocodazole or taxol (; ; ). These shortening processes are opposed by kinesin-5–dependent outward sliding forces (), which depend on bundles of overlapping microtubules. We propose that untreated CV1 spindles are composed of discontinuous arrays of overlapping microtubules generated by a combination of microtubule severing near the poles and dynamic instability (; ). Short microtubules result in short overlap regions, which are lost rapidly during nocodazole treatment. The rapid loss of outward sliding forces allows rapid spindle shortening, which is possibly due to the inward squeezing “tensile element” (). In this scenario, spindles are composed of more contiguous long microtubules after katanin inhibition, and these longer microtubules have long overlap zones that oppose spindle shortening forces for a longer time period during nocodazole treatment ().
female meiotic spindles naturally undergo dramatic shortening cycles during each anaphase. Our results indicate that this shortening is initially driven by katanin-independent spindle shortening that is accompanied by increasing microtubule density and may be driven by inward sliding of antiparallel microtubules (). This phase is followed by a katanin-dependent phase in which microtubule density decreases at poles and increases in the midzone. We favor a model for late shortening in which katanin, localized at spindle poles, severs pole-proximal microtubules into short fragments that are then transported to the midzone (). The second phase of spindle shortening is also accompanied by a katanin-independent decrease in the overall density of spindle microtubules that may be caused by other microtubule-destabilizing proteins.
The increased steady-state metaphase spindle length in the partial-loss-of-function katanin mutant is consistent with a model in which wild-type meiotic spindles are assembled as an array of discontinuous, overlapping, short microtubules (), just as we propose for CV1 mitotic spindles (). A reduced rate of microtubule severing results in a longer spindle because mean microtubule length is longer. This increased microtubule length would limit the extent of spindle shortening by the early, inward sliding mechanism (; and ). Our model is qualitatively consistent with the EM tomography results of , who found many short microtubules in a wild-type spindle but only long microtubules in a spindle. It is more difficult to resolve our model quantitatively with the EM tomography data, however, because more than half the microtubules in this tomogram extended beyond the tomogram and thus could not be measured.
The finding of numerous short microtubules in a wild-type meiotic spindle and the lack of short microtubules in a spindle () is also qualitatively consistent with our finding that katanin and γ-TuRC have partially redundant roles in promoting microtubule density increase during spindle assembly. Our results with GFP-tubulin fluorescence intensity measurements revealed microtubule densities that were similar between wild-type metaphase spindles and spindles (). We propose that this difference is small because γ-TuRC–mediated nucleation can increase microtubule number in the absence of severing activity. This result is consistent with the small (25%) difference in polymer mass determined from fixed immunofluorescence analysis by .
Perhaps the most perplexing question remaining is why mutants do not assemble extremely long bipolar spindles similar to what is seen with the partial-loss-of-function mutant. We offer two hypotheses explaining the absolute requirement for katanin in meiotic spindle assembly. In the first hypothesis, MEI-1–MEI-2 complexes have an additional, nonsevering function, such as cross-bridging antiparallel microtubules or recruiting other essential proteins to the spindle. In this scenario, spindles actually have no microtubule-severing activity but can assemble bipolar spindles with a low concentration of MEI-1–MEI-2 cross-bridging activity. This cross-bridging activity may be more critical in noncentrosomal microtubule arrays. The nonlinear dependence of microtubule-severing activity on katanin concentration () supports the notion that a reduced amount of MEI-2 could result in a complete loss of severing activity. In the second hypothesis, mean microtubule length increases as katanin concentration decreases until a critical stage is reached where microtubules are too long to allow spindle assembly. In the meiotic embryo, the relative activities of other critical regulators of microtubule dynamics and motility would be optimized for assembling spindles from short microtubules. These other microtubule-binding proteins would then assemble aberrant structures from long microtubules.
The conserved localization of katanin at spindle poles of many species and the similarity of spindle-length phenotypes in CV1 cells and meiotic spindles suggests that katanin regulates mitotic and meiotic spindle length in many species. The generality of katanin's relative role in the assembly of mitotic versus meiotic spindles remains to be elucidated. Our dominant-negative inhibitors did not prevent mitotic spindle assembly in CV1 cells (); however, genome sequences reveal that human and mouse each have three distinct katanin catalytic subunits (unpublished data) as well as the related microtubule-severing ATPase, spastin. These different microtubule-severing proteins could easily have unique roles at different times in development.
CV1 cells with an integrated YFP-tubulin construct were transfected on 25-mm coverslips with Lipofectamine Plus (Invitrogen) using CFP-X plasmid constructs, which were described previously (). 12–20 h after transfection, coverslips were assembled into perfusion chambers maintained at 37°C. Cells were imaged with a microscope (Microphot SA; Nikon) equipped with a 60× Plan Apo 1.4 objective, a charge-coupled device (CCD; Quantix KAF1400; Photometrics), and Ludl shutter controlled by IP Lab Spectrum software. Mitotic CFP-expressing cells were identified quickly using reduced illumination. A single CFP fluorescence image was captured using fixed neutral density filters, exposure, and gain so that the expression level could be estimated. 50 μl of culture medium containing 80 μM nocodazole was then added gently to the edge of the perfusion chamber and shuttered, and time-lapse acquisition of YFP fluorescence images was initiated. Diffusion of nocodazole to the imaged cell did not appear to be rate limiting, as the fastest changes in spindle length always occurred in the first 5 s of the image sequence.
A cDNA encoding the long isoform of MEI-1 () was cloned as a 6-his–tagged fusion along with an untagged MEI-2 cDNA into pFastBac Dual (Invitrogen), expressed in Sf9 cells using the Bac to Bac System (Invitrogen) and partially purified by Ni-chelate chromatography.
A 1:10 mixture of tetra-methy-rhodamine–labeled and unlabeled porcine brain tubulin was polymerized and stabilized with 20 μM taxol. Microtubules were immobilized on the surfaces of perfusion chambers at a final concentration of 0.1 μM tubulin dimer using an –expressed G234A mutant of the first 560 amino acids of ubiquitous human kinesin heavy chain. Preparations of purified MEI-1 or MEI-1–MEI-2 complexes were perfused at final concentrations ranging from 0.5 to 5.7 μM in 20 mM K-Hepes, pH 7.5, 0.1 mM EGTA, and 2 mM MgSO with ATP or ADP at 1.8 mM. Reactions were stopped at time intervals by perfusion with glutaraldehyde, and images of fixed microtubules were captured with a CCD.
In this study, wild type indicates one of several integrated transgenic strains. The integrated GFP∷tubulin strain WH204 () was used for the studies shown in and and and . The integrated GFP∷tubulin strain AZ244 () was consistently brighter than WH204 and was used for the studies shown in , , and . The integrated GFP∷histone H2b strain AZ212 () was used in . The integrated GFP∷tubulin, mCherry∷histone strain, OD57, was the wild type used for . OD57 was generated by particle bombardment of a promoter vector () engineered to express a histone H2b fusion to a version of mCherry () that included four introns and preferred codons. This modified mCherry DNA was synthesized by Molecular Cloning Laboratories. The mCherry∷histone strain was crossed with the GFP∷tubulin- expressing strain, OD4, to obtain OD57. worms were obtained from P. Mains (University of Calgary, Calgary, Canada) and were crossed with WH204 to obtain worms expressing GFP∷tubulin, with AZ212 to obtain worms expressing GFP∷histone H2b and with OD57 to obtain worms expressing both GFP∷tubulin and mCherry∷histone. The allele used in this study was the nonsense mutant (). Homozygous worms expressing GFP∷tubulin and used for the studies shown in were described previously (). For data shown in , homozygous worms expressing a higher level of GFP∷tubulin were derived from crosses between HR1069 (/hT2 I; hT2/+ III] and AZ244. The GFP∷MEI-1-expressing strain was EU1065 (). Western blots with anti–MEI-1 antibody indicated that this transgene is expressed at a much lower level than endogenous MEI-1 (not depicted), indicating that the localization in is not due to overexpression. OD44, the GFP∷γ tubulin-expressing strain used in , was provided by A. Desai (University of California, San Diego, La Jolla, CA). strains were maintained at 16°C and shifted to 20°C for 24 h before filming. EU1065 was maintained at 25°C to prevent germline silencing. All other strains were maintained at 20°C.
Adult hermaphrodites were anesthetized with Tricaine/tetramisole as described previously (; ) and gently mounted between a coverslip and a thin agarose pad on a slide. Mineral oil or petroleum jelly was used to reduce evaporation at the edge of the coverslip. Images for , , and ; histograms in ; most of ; and data for all tables were acquired on a microscope (Microphot SA; Nikon) as described for CV1 cell imaging. Images in and , parts of , and Fig. S2 were acquired with a spinning-disk confocal microscope (Perkin-Elmer) equipped with an 100× Plan Apo 1.35 objective (Olympus), CCD (Orca ER; Hamamatsu), and Slidebook acquisition software. All quantitative analysis was performed with IP Lab Spectrum software (Scanalytics).
Mean microtubule density in the entire spindle as shown in , , and was determined in IP Lab Spectrum by highlighting the entire spindle as a “segment” and determining the mean pixel value within the entire segment. This measurement is proportional to microtubule density as opposed to microtubule mass, which would be proportional to the sum of the values of all of the pixels within a spindle. Three separate segments that encompassed most of one spindle pole, most of the middle quarter of the spindle, and a band between the spindle pole and the middle quarter of the spindle were used to repeat this analysis. During the period of overall density increase shown in and , density (mean pixel value) increased uniformly in these three segments (not depicted). After spindle rotation, mean pixel values in these three segments behaved differently (not depicted), so line scan analysis () was used to illustrate these more complex intensity changes.
Line scans of fluorescence intensity shown in were generated in the following manner. GFP∷tubulin labeled discrete bundles of microtubules running from pole to pole. Exactly four of these bundles are visible in the wild-type spindle shown in . The ROI (region of interest) line tool in IP Lab Spectrum was used to draw a line down the length of one of these bundles. For wild type, a straight line was used. For , an irregular line was drawn to follow one of the irregular microtubule bundles. The ROI boundary function of IP Lab was then used to determine the pixel value at each point along this line, and these values were plotted as a function of spindle length in DeltaGraph. mCherry-histone–labeled chromosomes were between the microtubule bundles. Thus, a second line was drawn from pole to pole that passed over one chromosome pair to plot the anaphase movement of one chromosome pair.
Depletion of KLP-18 was accomplished by soaking worms in double-stranded RNA produced from cDNA clone yk745f12 provided by Y. Kohara (National Institute of Genetics, Mishima, Japan) using previously described methods (). γ-TuRC subunits were depleted by RNAi by feeding using clones III-5K19 () and III-1F05 () (MRC Gene Services; ).
Fig. S1 shows purification of recombinant MEI-1–MEI-2. Fig. S2 shows that katanin localizes at spindle poles at the time of katanin-dependent microtubule redistribution. Fig. S3 shows that ; double mutants do not show the synthetic effect on microtubule density observed in ; double mutants. Online supplemental material is available at . |
Most proteins in chloroplasts are encoded by the nuclear genome and synthesized in the cytosol as precursors with N-terminal targeting signals called transit peptides. Precursor import into chloroplasts is mediated by a protein translocon, which is composed of the Toc (translocon at the outer envelope membrane of chloroplasts) and the Tic (translocon at the inner envelope membrane of chloroplasts) proteins and stromal chaperones (; ). During import, transit peptides of precursors first interact with the Toc and then with the Tic proteins (). When sufficient ATP is present in the stroma, precursors are translocated across the inner envelope membrane into the stroma () and the transit peptide is removed by the stromal processing peptidase during the translocation (; ; ; ). Although many Tic and Toc proteins have been identified, the interactions among the translocon components and the mechanistic steps of the import process are largely unknown.
Tic110 is the major Tic protein identified (; ). Tic110 has an N-terminal membrane anchor and a large stroma-located hydrophilic domain (; ). The N-terminal half of the Tic110 stromal domain binds transit peptides directly. Tic110 is therefore thought to be the stroma-side receptor for transit peptides and the first protein that binds precursors as they emerge from the inner membrane channel (, ). Hsp93 (ClpC) belongs to the Clp/Hsp100 subfamily of the AAA+ (ATPase associated with various cellular activities) family of proteins (; ). It is present both in a soluble form in the stroma and in an inner membrane–tethered form (; ). It is proposed to function as the motor for chloroplast protein translocation, as translocation requires ATP hydrolysis in the stroma () and Hsp93 is the only known ATPase stably associated with the entire translocon complex ().
Tic40 has a similar topology to Tic110. The stroma-located hydrophilic domain of Tic40 is composed of a tetratricopeptide repeat (TPR) domain followed by a C-terminal domain homologous to the C terminus of cochaperones Sti1p/Hop (Hsp70/Hsp90-organizing protein) and Hip (Hsp70-interacting protein; ; ). TPR domains, which consist of highly degenerated 34-amino-acid repeats, are present in proteins of diverse functions and are known to mediate protein–protein interactions (; ; ). In contrast, no molecular function had been assigned to the Hip and Hop C-terminal domain, even though this domain is highly conserved from yeast (Sti1p) to human (Hip and Hop) to plants (Tic40).
Tic40 has been shown to function at a similar stage of import to Tic110 and Hsp93 (; ). -null mutants are extremely pale green and retarded in development but still viable, suggesting that Tic40 plays a nonessential stimulatory/regulatory role during import. Chloroplasts isolated from mutants are normal in binding of precursors to the outer envelope membrane but are specifically defective in precursor translocation across the inner envelope membrane. Precursors tend to be released from mutant chloroplasts before the completion of translocation. Based on these results, we have proposed that Tic40 may function as a cochaperone to coordinate the action of Tic110 and Hsp93 (). This hypothesis predicts that the Tic40 will physically interact with Tic110 and/or Hsp93 and the interaction will have regulatory or mechanical consequence to the import of precursor proteins into the stroma. In this work, we have analyzed the molecular interactions among Tic40, Tic110, and Hsp93. We provide evidence to show that Tic40 regulates transit peptide–Tic110 interaction and Hsp93 ATP hydrolysis.
We first investigated the interaction between Tic40 and Tic110. We tested whether they could directly interact with each other by yeast two-hybrid assays. Pea cDNAs encoding the entire stroma-located hydrophilic domain (Tic40S; ), TPR subdomain (Tic40TPR), and C-terminal Sti1p/Hop/Hip-homologous subdomain (Tic40Hip/Hop, hereafter referred to as the Hip/Hop domain) were fused with the GAL4 DNA-binding domain. Pea cDNAs encoding the entire stroma-located hydrophilic domain (Tic110S) and the N- and C-terminal parts of this stromal domain (Tic110N and Tic110C; ) were fused with the GAL4 activation domain. As shown in , Tic40S interacted with Tic110S. This interaction is most likely due to the interaction between Tic40TPR and Tic110C (). We further confirmed this interaction by in vitro pull-down assays. GST fused with Tic40 stroma l hydrophilic domain, the TPR subdomain, or the Hip/Hop subdomain was produced and purified from (GST-atTic40S, -atTic40TPR, and -atTic40Hip/Hop, respectively; ). They were incubated with Tic110S fused to a C-terminal His tag (atTic110S-His; identical to atTic110 in ). As shown in , GST-atTic40S and -atTic40TPR, but not GST- atTic40Hip/Hop or GST, could pull down atTic110-His.
We next investigated whether Tic40 had a direct interaction with precursors. GST-atTic40S was incubated with in vitro–translated S-labeled precursor to the small subunit of ribulose-1,5-bisphosphate carboxylase oxygenase (prRBCS), a representative chloroplast stromal protein imported from the cytosol. Recombinant atTic110S-His was used as a positive control, and the 33-kD subunit of the oxygen-evolving complex of the thylakoid membrane fused to a His tag (OE33-His) was used as a negative control. As shown in , atTic110S-His indeed preferentially bound to prRBCS as previously reported (). GST-atTic40S did not bind to prRBCS or mature RBCS, suggesting that there is no direct contact between Tic40 and the importing substrates.
Because both transit peptides () and Tic40 () bind to Tic110, we investigated whether these two bindings affect each other. Preincubation of atTic110S-His with prRBCS, but not mature RBCS, increased the amount of atTic110S-His associated with GST-atTic40S (, lanes 2 and 3). This stimulatory effect could also be observed by incubating Tic110 with a synthetic peptide corresponding to the transit peptide of the precursor to ferrodoxin (prFD; , lane 2; ). A control peptide, SynB2, had no effect (, lane 3). SynB2 has an artificially designed sequence that possesses a similar number of charges but a different conformation from a typical mitochondrial presequence, and it does not function as a mitochondrial presequence (). It is also discriminated by Toc75 from genuine chloroplast transit peptides (). These results suggest that binding of Tic110 to transit peptides increases the affinity of Tic110 to Tic40.
We then investigated what happened to transit peptides bound by Tic110 after Tic40 binding. We tried to load Tic110 with as many transit peptides as possible by incubating atTic110S-His with 10-fold excess of H-labeled prFD transit peptides and then recovered atTic110S-His with bound transit peptides by metal-affinity resin. GST-atTic40S or GST control proteins were added to the resin suspension in increasing amounts. Increasing amounts of H were detected in the supernatant as increasing amounts of GST-atTic40S were added (), suggesting that GST-atTic40S caused the release of transit peptides from atTic110S-His. These data suggest that binding of transit peptides by Tic110 recruits Tic40 to Tic110. Binding of Tic40 to Tic110 then causes transit peptides to be released from Tic110.
Processing of transit peptides occurs when the importing precursors first emerge from the inner membrane while precursors are still bound to the inner membrane translocon (; ; ). If binding of Tic40 causes the release of transit peptides from Tic110 in vivo, then in a mutant, the release of transit peptides from Tic110, and therefore the processing of transit peptides, should be delayed and the amount of mature RBCS associated with Tic110 should be reduced. We thus compared the import of [S]prRBCS into isolated wild-type chloroplasts and chloroplasts isolated from the mutant, which is a null mutant of Tic40 caused by transfer DNA insertion into the gene (). The import reaction was terminated at various time points by diluting with cold import buffer. Chloroplasts were reisolated and treated with 0.5 mM dithiobis(succinimidyl)propionate cross-linker to stabilize protein–complex interaction. Chloroplasts were lysed, and membrane fractions were isolated and analyzed (). The amount of precursor proteins associated with the membranes was the same in the mutant and wild-type chloroplasts (, top). However, the appearance of processed mature proteins in the membranes was greatly delayed in the mutant chloroplasts. We have previously shown that bound precursors tend to be released from mutant chloroplasts before translocation is completed (). Therefore, the fact that no accumulation of precursors was observed when production of processed mature proteins was delayed may be due to the release of the untranslocated precursors.
When the total membranes after import were further solubilized for immunoprecipitation, anti-Tic110 antibodies immunoprecipitated a high amount of mature RBCS in the wild-type chloroplasts (, bottom). Before immunoprecipitation, the ratio of RBCS to prRBCS was 12:1 at the 30-min import time point, and when precipitated by anti-Tic110 antibodies, the ratio increased to 29:1 (). This result indicated that there was a population of RBCS that was specifically associated with Tic110 and Tic110-associated translocon complexes. This population was likely absent in because in the mutant chloroplasts the ratio of RBCS to prRBCS remained similar in total membranes and in the population precipitated by anti-Tic110 antibodies (). It is possible that in the mutant chloroplasts, by the time transit peptides are spontaneously released by Tic110, the mature region of the precursor proteins has been translocated further into the stroma.
Because Hsp93 is the only chaperone molecule stably associated with the entire translocon complex, its ATPase activity should be important for precursor translocation. We therefore investigated which component that may have contact with Hsp93 could affect its ATPase activity. Recombinant pea Hsp93 with an N-terminal His tag, His-Hsp93 (), was assayed for its ability to hydrolyze ATP. Addition of prFD transit peptides, Tic110, or Tic40 all had an approximately twofold stimulation on Hsp93 ATP hydrolysis compared with Hsp93 alone (). It is not clear whether such a small stimulation was specific or merely due to the presence of unfolded proteins. We further analyzed the effect of Tic40 subdomains. GST-atTic40TPR or -atTic40Hip/Hop was incubated with Hsp93. Interestingly, it was the Tic40 Hip/Hop domain, not the TPR domain, that had a clear stimulatory effect on Hsp93 ATP hydrolysis (). Various combinations of atTic110S-His plus domains of Tic40 or prFD transit peptides were also tested, and no additional stimulatory effect was observed (unpublished data). In support of the importance of the Tic40 Hip/Hop domain, mutant plants could be complemented by a full-length cDNA encoding the Tic40 precursor, but not by the same construct with the Hip/Hop domain deleted (unpublished data).
To further confirm the stimulatory function of the Tic40 Hip/Hop domain, three mutants of GST-atTic40Hip/Hop were generated. In the Hip/Hop domain of human Hip, the sequence DPEV occurs twice. Mutation of these two sequences to APAV inhibits progesterone receptor complex assembly with Hsp90 (). At the corresponding positions, atTic40 has NPDV and NPRV. (The subscript number stands for the residue number. It is equivalent to residue 396 and 405 if the first residue of atTic40 precursor, instead of mature protein, is counted as residue 1.) Another similar sequence of NPMN is located further downstream. We mutated each of the asparagines to alanine. Limited proteolysis () and secondary structure predictions (not depicted) indicated that the mutations did not affect the conformation of the three GST-atTic40Hip/Hop mutants. When assayed for stimulation of Hsp93 ATP hydrolysis, the N320A and N329A mutations abolished the stimulatory activity of GST-atTic40Hip/Hop, whereas the N342A mutation had no effect ().
The mutant is extremely pale and small but still viable (), indicating that Tic40 plays a nonessential stimulatory/regulatory role during import. If Tic40 functions in stimulating Hsp93 ATP hydrolysis in vivo, either by promoting ATP/ADP exchange or by activating Hsp93 ATPase activity, it is likely that the protein import defect of mutants can be partially compensated by increasing the available amount of ATP in the stroma. We therefore preloaded isolated wild-type and mutant chloroplasts with 3 mM ATP and performed import experiments. Without ATP preloading, the amount of mature RBCS imported by the mutant chloroplasts was 26% of the wild-type chloroplasts. With ATP preloading, the mutant was increased to 38% of wild type (). We further confirmed the physiological significance of the results by testing whether Hsp93 ATPase activity was different in the two ATP concentrations before and after ATP preloading. Dark-adapted chloroplasts at the beginning of import have a stromal ATP concentration of ∼0.07 mM (; ; see Materials and methods). Preincubating chloroplasts in 3 mM ATP for 5 min at room temperature should increase the stromal ATP concentration to ∼2.1 mM (). The ATPase activity of Hsp93 was indeed increased significantly from an ATP concentration of 0.07 to 2.1 mM (). Stimulation of Hsp93 ATP hydrolysis by Tic40 provides another possible explanation for the release of untranslocated precursors from the mutant chloroplasts: a rapid ATP hydrolysis by Hsp93, stimulated by Tic40, may be important for a unidirectional translocation of precursors into the stroma.
We next investigated whether the nucleotide state of Hsp93 affected its association with Tic40. Hsp93 was preincubated with ATP, ADP, or the nonhydrolyzable ATP analogue adenylylimidodiphosphate (AMP-PNP). To mimic the in vivo situation, prFD transit peptides, atTic110S-His, and GST-atTic40S were added to the reaction. Tic40-containing protein complexes were recovered using glutathione resin. The amount of Hsp93 recovered with Tic40 was higher in the presence of ATP or AMP-PNP and lower in the presence of ADP (), suggesting that Hsp93 may associate with Tic40 in its ATP state and dissociate from Tic40 after ATP is hydrolyzed to ADP. This result also suggests that Tic40 most likely functions in stimulating Hsp93 ATPase activity instead of facilitating ATP/ADP exchange.
Stimulation of Hsp93 ATPase activity was only observed with the Tic40 Hip/Hop domain, not with the entire Tic40 stromal domain containing both the Hip/Hop and the TPR domains (GST-Tic40S; ). It is possible that the Tic40 stromal domain is normally in a closed conformation in which the TPR domain shields the Hip/Hop domain. Binding of the Tic40 TPR domain to Tic110 causes a conformational change in Tic40, exposes the Hip/Hop domain, and allows the Hip/Hop domain to stimulate Hsp93. This mechanism would be similar to that found for protein phosphatase 5, in which its TPR domain normally engages with its phosphatase domain. Binding of the TPR domain to Hsp90 dissociates the TPR domain from the phosphatase domain and permits substrate access to the phosphatase domain ().
A working model for the sequential steps of protein translocation into the chloroplast stroma is proposed here (): as the transit peptide of a precursor protein emerges from the inner envelope membrane channel, it is bound by the N-terminal part of the Tic110 stromal domain (). This binding causes a conformational change in Tic110 and recruits Tic40TPR binding to Tic110 (). Binding of Tic40TPR to Tic110 causes release of the transit peptide from Tic110, freeing the transit peptide for cleavage by the stromal processing peptidase. Binding of Tic40TPR to Tic110 also unshields the Tic40 Hip/Hop domain, which then stimulates ATP hydrolysis by Hsp93 (). The energy of ATP hydrolysis by Hsp93 is most likely used to translocate the processed mature protein into the stroma (). Analyses of similar AAA+ proteins suggest that these proteins actively push substrates through its axial channel using conformational changes caused by ATP hydrolysis (; ; ). Hsp93-ADP may then dissociate from Tic40 (). Tic110 may also dissociate from Tic40 when there is no transit peptide bound. Under normal growth conditions in the light in which the stromal ATP concentration is high, Hsp93 may soon be reloaded with ATP and be ready for the next round of precursor translocation after Tic110 binds another incoming transit peptide. In our model, ATP is hydrolyzed only after the Tic40 TPR domain binds to Tic110, which only occurs after Tic110 binds to transit peptides. Therefore, Tic40 may function like a timing device to coordinate the sequential steps of translocation.
The cochaperones Hip, Hop, and Tic40 all contain the TPR and Hip/Hop domains. The function of the TPR domains has been addressed extensively through structural and functional studies (; ). In contrast, no molecular function had been assigned to the Hip/Hop domain. Our data suggest that the Hip/Hop domain of Tic40 functions as an ATPase stimulation protein for Hsp93. Because the Hip/Hop domain is highly conserved, its identification as an ATPase stimulation protein may have implications in the functional mechanisms of Hip and Hop. It would be interesting to investigate whether the Hip/Hop domain in Hip and Hop can also stimulate the ATPase activity of Hsp70 and/or Hsp90. In addition, in vivo, whether the Tic40Hip/Hop domain has a stable association with Hsp93 like GroEL with GroES, or only interacts transiently with Hsp93 like DnaK with DnaJ, and which one of the two Hsp93 ATPase domains is activated by Tic40Hip/Hop, all require further investigations.
A possible similarity between protein translocation into the chloroplast stroma and retrotranslocation of misfolded proteins from the ER lumen is noted here. ER retrotranslocation requires the cytosolic ATPase p97/Cdc48, which is an AAA+ family protein, like Hsp93. VIMP (VCP interacting membrane protein), an ER-membrane protein with a sizeable cytosolic domain, acts as a receptor for p97, similar to Tic110 (). VIMP links p97 to Derlin-1, which is a proposed component of the ER retrotranslocation channel (; ). Derlin-1 is ∼25 kD with four predicted α-helical transmembrane domains, a size and structure very similar to Tic20, the proposed protein translocation channel across the chloroplast inner envelope membrane (). Therefore, both systems contain an AAA+ type motor, a membrane protein with a large soluble domain as the receptor for the motor and a channel of a similar size and structure. It is possible that the substrates of both systems, the chloroplast precursors after being translocated across the outer membrane and the misfolded ER proteins after being recruited to Derlin-1, are in a similar folding state. Tic40 may have been evolved to meet some specific requirements of chloroplasts to increase the efficiency of precursor processing and translocation.
All constructs used for the yeast two-hybrid analyses were made by PCR amplifying the respective DNA fragment using a forward primer that added an NcoI site and a reverse primer that added an EcoRI site to the amplified fragment. The fragment was digested and cloned into the NcoI–EcoRI site of plasmid pACT2 (for the activation-domain constructs) or pAS2-1 (for the binding-domain constructs). Other constructs were made similarly with the following specifications: GST fusion constructs of atTic40 were made in the BamHI–EcoRI site of pGEX-5X-1 (GE Healthcare), His-Hsp93 was made in the NdeI–SalI site of pET28a (Invitrogen), and OE33-His (At5g66570, amino acids 86–332 with the first amino acid of the precursor as 1) was made in the NdeI–EcoRI site of pET22b (Invitrogen).
Two-hybrid assays were performed using the MatchMaker-II system (CLONTECH Laboratories, Inc.) according to the manufacturer's instructions. Except where specified in the figure legends, pull-down assays using His-tagged proteins were performed using TALON (CLONTECH Laboratories, Inc.) resin by incubating the resin with the protein mixture at 4°C for 2 h. Pelleted resin was washed thrice with PBS (10 mM NaHPO, 1.8 mM KHPO, 140 mM NaCl, and 2.7 mM KCl) containing 1 mM imidazole, and bound proteins were eluted with 50 mM imidazole in 50 mM Tris-HCl, pH 8.0. Pull-down assays using GST-tagged proteins were performed using Glutatione-4B resin (GE Healthcare) by incubating the resin with the protein mixture at 4°C for 30 min to 1 h. Pelleted resin was washed thrice with PBS, and bound proteins were eluted with 10 mM glutathione in 50 mM Tris-HCl, pH 8.0.
Recombinant prRBCS and RBCS were overexpressed and purified from (). Antibodies against Tic110 were generated as described previously (). Antibodies against Tic40 were produced by subcloning the coding region of atTic40 residues 227–447 into pET22b (Invitrogen), overexpressing the recombinant protein with a C-terminal His tag in , and injecting the purified recombinant proteins into mice. Anti-Hsp93 and anti-OE33 antibodies were generated by immunizing rabbits with the His-Hsp93 and OE33-His recombinant proteins described in the previous paragraph.
The first 34 amino acids of the prFD transit peptide without the initiation methionine (sequence ASTLSTLSVSASLLPKQQPMVASSLPTNMGQALF; ) and the SynB2 peptide (sequence MLSRQQSQRQSRQQSQRQSRYLL; ) were synthesized and purified by HPLC to >80% purity by SynPep. 1.5 mg of the prFD peptide was dissolved in 50 μl of water and mixed with 20 μl of boric acid (adjusted to pH 8.5 with NaOH), 20 μl of 0.5 M formaldehyde, and 15 μl of H-labeled 0.12 M NaBH. The reaction was incubated on ice for 15 min and quenched by adding 50 μl of 1 M Tris-HCl, pH 7.5. H-labeled peptide and H-labeled NaBH were separated on a Sephadex G-10 column with 0.1 M NHHCO. The peptide fractions were collected and dried.
[S]prRBCS and [S]RBCS were synthesized by in vitro transcription and translation (). Growth of plants, isolation of chloroplasts from the WS2 wild type and the mutant, import assays and isolation, solubilization, and immunoprecipitation of membrane fraction from chloroplasts after import were performed as described previously (). Stromal ATP concentrations used in is calculated as follows: internal ATP concentration of dark-adapted chloroplasts is ∼1.82 nmol/mg chlorophyll (; ). With a stroma volume of 25 μl/mg chlorophyll (), 1.82 nmol/mg chlorophyll is ∼0.07 mM. At room temperature for 5 min, ATP translocation across the envelope can achieve ∼70% (). Preincubating chloroplasts in 3 mM ATP for 5 min at room temperature should increase the stromal ATP concentration to ∼2.1 mM. Quantifications of gel images were performed on LAS-1000plus pictrography 3000 (Fuji).
A 16-μl solution containing 0.5 μM of His-Hsp93 in 75 mM Hepes-KOH, pH 7.0, and 4 mM MgCl was mixed with 4 μl of 1 mM ATP containing 0.078 μl of γ-[P]ATP (5,000 Ci/nmol) and incubated at 37°C for 30 min. The reaction was stopped by adding 2 μl of 0.5 M EDTA. 0.35 μl of the reaction was spotted on a TLC plate. The plate was air-dried, and ATP and Pi were separated by a solution containing 0.5 M LiCl and 1 M formic acid. The TLC plate was dried. The amounts of ATP and Pi were quantified by a PhosphorImager (FLA-5000; Fuji). When other proteins were added to the reaction, they were present at the same concentration as His-Hsp93, except for the prFD transit peptide, which was present at a 10× concentration of other proteins. |
Mitochondria and ER of eukaryotic cells form two intertwined endomembrane networks, and their dynamic interaction controls metabolic flow, protein transport, intracellular Ca signaling, and cell death (; ; ; ; ; ). Mitochondrial Ca uptake, via a yet to be identified Ca channel of the inner mitochondrial membrane (the mitochondrial Ca uniporter), regulates processes as diverse as aerobic metabolism (), release of caspase cofactors (), and feedback control of neighboring ER or plasma membrane Ca channels (; ). A corollary of the efficient mitochondrial Ca uptake during IP-induced Ca release is the close apposition of ER and outer mitochondrial membranes (OMM; ; ; ; ). The molecular determinants of this crosstalk, however, are still largely unknown (). Recently, PACS2, which is an ER-associated vesicular-sorting protein, was proposed to link the ER to mitochondria (). The knockdown of PACS2 led to stress-mediated uncoupling of the organelles, which was also reflected by the inhibition of Ca signal transmission.
On the other side, the abundant OMM channel voltage-dependent anion channel 1 (VDAC1) was also suggested to participate in the interaction. It was shown to be present at ER–mitochondrial contacts and to mediate Ca channeling to the intermembrane space from the high [Ca] microdomain generated by the opening of the inositol 1,4,5-trisphosphate receptor (IPR; ; ). In addition, VDAC1 mediates metabolic flow through the OMM, forming an ATP microdomain close to the ER and sarcoplasmic reticulum Ca ATPases (SERCAs; ; ), and both VDAC1 and VDAC2 take part in metabolic and apoptotic protein complexes (; ; ).
The transfer and assembly of components of cellular protein complexes were shown to be assisted by molecular chaperones, adding a novel function to their role in nascent protein folding (; ). Accordingly, Ca binding–, heat shock–, and glucose-regulated chaperone family members are abundantly present along the Ca transfer axis, linking the ER and mitochondrial networks. Well known examples are the Ca-binding chaperones of the ER lumen (), immunophilins interacting with ER Ca-release channels and the mitochondrial permeability transition pore (; ), and several heat shock family members localized at the mitochondrial membranes, which are proposed to interact with the components of the mitochondrial permeability transition pore, such as VDAC (; ; ). Still, their exact role at the ER–mitochondria interface is not well known, although recently, weak links between chaperones were proposed to stabilize signaling and organellar cellular networks (; ).
Considering the central position of VDAC at the ER–mitochondrial interface outlined in the previous paragraphs, we used VDAC1 as a starting point for protein biochemical studies, to explore molecular interactions between the ER and mitochondrial networks. We found that through the OMM-associated fraction of the glucose-regulated protein 75 (grp75) chaperone (), VDAC1 interacts with the ER Ca-release channel IPR. Organellar Ca measurements, using targeted recombinant Ca probes, confirmed functional interaction between the IPR and the mitochondrial Ca uptake machinery, which was abolished by grp75 knockdown.
We performed yeast two-hybrid screens of human liver and kidney LexA-AD–fused libraries, using rat VDAC1-LexA-DNA- BD fusion protein as bait. Among the putative interactors we found cytoskeletal elements, which were previously thought to participate in sorting of VDAC or in mitochondrial dynamics (; ) and a group of chaperone proteins (). To investigate whether the chaperones participate in mediating organelle interactions, we focused our attention on the human heat shock 70 kD protein 9B/grp75 (nt 1,456–2,089 from GenBank/EMBL/DDBJ under accession no. ; aa 471–681). The yeast homologue of grp75 is part of the protein import motor associated with TIM23 in the mitochondrial matrix (), but it was also found in the cytosol and in OMM-associated high molecular weight protein complexes (; ; ). In addition, two further findings indicated that grp75 may be involved in ER–mitochondria Ca transfer: first, its C-terminal domain reduced the voltage dependence and cation selectivity of VDAC1 (), and second, grp75 overexpression was shown to promote cell proliferation and protect against Ca-mediated cell death (; ).
Based on these findings, we used further biochemical approaches to investigate the role of grp75 at the ER–mitochondria contact sites. We took advantage of a previously developed method to purify a mitochondria-associated ER subfraction (mitochondria-associated membrane (MAM) fraction []). The MAM was previously shown to be enriched in lipid synthases and transferases (), and it likely represents the membrane microdomain engaged in ER–mitochondrial Ca transfer (; ; ). Indeed, immunoblot screening of the MAM fraction, purified from rat liver and HeLa cells, revealed the presence of grp75, as well as Ca channels from both the OMM (VDAC1) and the ER (IPR; ). In liver cells, given the higher yield, the microsomal fraction and the different mitochondrial extracts (the crude mitochondrial pellet [Mito C], the low-density MAM, and the high-density mitochondrial fraction containing the matrix proteins [Mito P]) were separately analyzed. As expected, VDAC and grp75 are not enriched in the MAM, given that the former is highly expressed throughout the OM (because of its role in ion and metabolite diffusion) and the latter is mostly in the matrix (but the two pools show different macromolecular assembles; see following paragraph). Conversely, the IPR, besides the microsomes, is present in the crude mitochondria and the MAM fractions and is absent in the purified mitochondria ().
We next investigated whether grp75 and the ER and mitochondrial Ca channels are part of the macromolecular complex, and whether the grp75 pool involved is that present in the MAM fraction. For this purpose, we separately analyzed the MAM and the Mito P fractions by 2D Blue native (; first dimension, central part of the image) and SDS-PAGE protein separation. In the latter dimension, grp75, IPR, and VDAC were revealed by immunoblotting. Although VDAC1 was present in different amounts in complexes of a wide molecular weight range, we found a specific complex characterized by the presence of VDAC, the IPR, and grp75, suggesting their interaction in the native state. The specificity of the complex formation of VDAC, the IPR, and grp75 was corroborated by the finding that SERCA2b showed different localization in the 2D separation (Fig. S1, available at ). In Mito P, grp75 was found in lower molecular weight complexes (<400 kD), which is similar to previous data in yeast (), confirming that grp75 is involved in high molecular weight protein–protein interactions only in the MAM. To confirm the different location of the two grp75 pools, we verified that whereas the matrix-localized grp75 was resistant to proteinase K digestion (similar to the matrix enzyme MnSOD; ), grp75 of the MAM fraction was degraded by the enzyme, documenting its association with the cytosolic surface of mitochondria.
To further investigate the arrangement of the grp75–VDAC–IPR complex, we used coimmunoprecipitation studies of the involved proteins. Immunoprecipitation of both IPRs and VDAC led to the coprecipitation of grp75 (, respectively), but no IPR was found in the VDAC precipitate, and no VDAC was detectable in the IPR precipitate. However, immunoprecipitation of grp75 led to the copurification of both VDAC and the IPR (). These results strongly suggest that grp75 has a central role in setting up the protein complex with VDAC and the IPR. Moreover, the interactions were detected both in the presence and absence of Mg-ATP (unpublished data), further suggesting the scaffolding, rather than chaperoning, function of grp75 in the complex.
If the IPR is in a macromolecular assembly with VDAC, we assumed that the mitochondrial Ca uptake machinery might be regulated by the large cytoplasmic domain of the IPR. This scheme was also supported by previous studies showing that the ligand-binding domain of the IPR (aa 224–605; denoted as IPR-LBD), located on the surface of the cytoplasmic domain, participates in intramolecular interactions with other IPR domains (), as well as in linking the receptor with other protein partners (). To assess a direct role of the IPR in mitochondrial Ca uptake, we coexpressed in HeLa cells mRFP1-tagged IPR-LBD with cytosolic (cytAEQ) or mitochondrially targeted (mtAEQmut) aequorin-based Ca probes, and evaluated global and organellar Ca responses to agonist stimulation. After reconstitution with the aequorin cofactor coelenterazine, cells were challenged with histamine (in incremental doses from 1 to 100 μM), and luminescence was measured and converted to [Ca]. Recombinant expression of the IPR-LBD caused a marked increase in mitochondrial Ca uptake at each agonist concentration applied, in spite of reduced cytoplasmic Ca response ([Ca]), because of IP buffering and consequent reduction of IP-induced Ca release from the ER (see and Fig. S2 [available at ] for the lower agonist concentrations). The effect of the IPR-LBD was presumably exerted on the OMM because targeting the IPR-LBD to the OMM surface (by fusing to an N-terminal AKAP1 domain) augmented its stimulatory effect (see for intracellular localization of the mRFP1-tagged construct and for the effect on [Ca]). Morphological imaging and mitochondrial loading with the potential-sensitive dye teramethylrhodamine methyl ester showed that the effect was not caused by changes in mitochondrial morphology () or to the modification of mitochondrial membrane potential (not depicted).
To confirm that activation of mitochondrial Ca uptake can be exerted from the original site of the IPR (i.e., from the ER membrane), we expressed IPR-LBD fused to a C-terminal ER-targeting sequence derived from the yeast UBC6 protein (denoted as ER-IPR-LBD; ). Expression of this construct reduced the steady-state ER [Ca] ([Ca]) and IP-induced Ca release (Fig. S2 and , respectively), which were probably caused by direct activation of the IPR, as previously reported for COS-7 cells (), although store depletion was incomplete in HeLa cells at the expression levels of this study. Still, most importantly, expression of the ER-targeted IPR-LBD augmented mitochondrial Ca accumulation after cellular stimulation by histamine, similar to what was observed upon expression of the OMM-targeted IPR-LBD domain ( shows the intracellular localization of ER-IPR-LBD; shows the stimulatory effect of ER-IPR-LBD on [Ca]). These results strongly suggested that the IPR, acting from the ER surface, regulates mitochondrial Ca uptake at an OMM site, independent of its Ca-channeling function.
Based on our conclusions, we further investigated whether the effect of the N-terminal cytosolic domain of the IPR reflects specific protein–protein interactions at the ER–mitochondrial contacts. We first verified that the effect of IPR-LBD on mitochondrial Ca uptake is independent of IP buffering. For this purpose, we used a point-mutated (K508A) IPR-LBD, which is unable to bind IP. The K508 mutant increased the [Ca] rise in a manner similar to the wild-type (although slightly less efficient), but, as expected, did not modify the [Ca] response (). The capacity of an IP-insensitive IPR-LBD to enhance mitochondrial Ca uptake was also confirmed in digitonin-permeabilized HeLa cells. In this case, mitochondrial Ca uptake is exclusively dictated by the perfused [Ca], and it is totally independent of IPR activity. In permeabilized cells, Ca uptake was triggered by the perfusion of an intracellular buffer containing Ca buffered at 1 μM. Under those conditions, in which protein interactions might have been affected by the application of digitonin, both the wild-type and the K508A OMM-IPR-LBD increased the rate of mitochondrial Ca uptake, although also in this case the wild-type was more efficient (14.71 ± 4.66% increase, = 25, P < 0.01 vs. 6.58 ± 4.23% increase, = 25, P > 0.05).
The notion that IP binding cannot account for the mitochondrial effect was further confirmed by the demonstration that a structurally unrelated IP-binding protein domain, the PH domain of the PLC-like protein p130 (p130PH; ), targeted to the OMM, reduced both the [Ca] and [Ca] responses (). Interestingly, the reduction of the [Ca] response was more pronounced for the OMM-targeted PH domain than for the untargeted cytosolic version of the IP buffer, although the two proteins were equally effective on [Ca]. These data (Fig. S3, available at ) further stress the strict dependence of mitochondrial Ca homeostasis on the ER– mitochondrial contacts, and thus on the Ca release occurring in these microdomains.
The IPR-LBD was also shown to play an important role in the regulation of IPR channel activity by interacting with the N-terminal repressor domain (aa 1–223; ; ). Still, expressing the entire N-terminal surface domain of the IPR, targeted to the exterior of the OMM (OMM-IPR), augmented mitochondrial Ca uptake (). These results exclude that the stimulatory effect of the IPR-LBD was exerted through unmasking this intramolecular interaction in the endogenous IPR; instead, they support a model in which the entire N-terminal IPR exerts direct activation on the mitochondrial Ca uptake machinery.
Finally, we investigated the regulatory activity on mitochondria of the IPR-LBD when the [Ca] rise is elicited in the cell by the opening of plasma membrane channels. Under those conditions, not only the [Ca] rise is IPR-independent, but the [Ca] and ensuing [Ca] increases are markedly slower and smaller than upon ER Ca release. We thus measured [Ca] after emptying the ER Ca pool with the SERCA blocker tert-butyl-benzohydroquinone (tBHQ) in Ca-free medium and re-adding CaCl. This protocol induces capacitative Ca entry, causing a [Ca] rise and subsequent mitochondrial Ca uptake. As presented in , IPR-LBD–expressing cells showed an ∼60% increase in the influx-dependent [Ca] response (top), even if the [Ca] rise remained unaltered (bottom). This increase in [Ca] was almost doubled, as compared with the effect after histamine-/IP-induced Ca release from the ER (). Thus, we concluded that local IP buffering masks the stimulatory effect of the IPR-LBD upon ER Ca release, and, indeed, the effect of the IPR-LBD is established at the ER–mitochondrial contacts.
Because our proteomic studies suggested that the interaction of the VDAC and IPR channels is mediated by grp75, we investigated whether the stimulatory effect of the OMM-targeted IPR-LBD on mitochondrial Ca uptake requires the presence of grp75. A first series of experiments showed that strong inhibition of grp75 expression (48 h after transfection) in itself strongly reduced mitochondrial Ca uptake, most probably because of alterations of mitochondrial function through inhibition of protein import and Δψ loss (unpublished data). Thus, we opted for a lower silencing efficiency by conducting experiments 24 h after transfection (, inset). We expressed control and grp75 siRNAs in HeLa cells, cotransfecting them with the IPR-LBD construct and mtAEQmut. Under those conditions, grp75 siRNA had no effect on the [Ca] response to histamine stimulation (). However, the down-regulation of grp75 prevented the stimulatory effect on mitochondrial Ca uptake of the IPR-LBD, which was expressed both on the OMM and the ER surface (). Thus, we concluded that grp75 is not only physically associated with the IPR–VDAC1 complex, but is also necessary for functional coupling between these proteins. These results also show that although moderate knockdown of grp75 does not interfere with its function in the mitochondrial matrix, in accordance with previous results on mitochondrial protein import (), the low amount of grp75 at the ER–mitochondrial contacts is a limiting factor for the stimulatory effect of the IPR-LBD.
In the final set of experiments, we further investigated the role of grp75 in mitochondrial Ca uptake regulation by overexpressing the protein. Most likely caused by its differentially localized pools, grp75 appeared to modify mitochondrial Ca uptake after IP-induced Ca release through diverse mechanisms. Indeed, as shown in , overexpression of the wild-type protein led to reduced histamine-induced [Ca] response. However, at the same time, it also significantly decreased the steady-state [Ca] level (, right), thus, reducing the driving force for IP-induced Ca release, which in turn might be responsible for the dampened mitochondrial Ca accumulation. This parallel reduction of [Ca] and [Ca] may reflect two different effects of grp75: OMM-localized grp75, presumably through the interaction with the IPR or other members of the ER Ca-handling machinery, may increase the Ca leak from the ER through the IPR, as previously shown for Bcl-2 (; ); matrix-localized grp75 may modify mitochondrial parameters (e.g., pH) or import of Ca-handling proteins, leading to altered mitochondrial Ca uptake, as well as the ATP supply for ER Ca accumulation through the SERCA pumps. To dissect these effects, we again used the approach of measuring IP-independent mitochondrial Cauptake after capacitative Ca influx. In addition, to distinguish OMM-based effects from those in the mitochondrial matrix, we expressed a truncated grp75 lacking the N-terminal 51-aa mitochondrial-targeting sequence, and thus unable to enter the mitochondrial matrix. Ca influx was induced by depleting the ER Ca store with tBHQ in the absence of extracellular Ca, as described in the previous section (). This “cytosolic” form of grp75 (grp75) did not change the bulk cytosolic [Ca] response to readdition of Ca in the extracellular medium, but significantly increased mitochondrial Ca accumulation (). Moreover, grp75 further potentiated the stimulatory effect of IPR-LBD (), confirming the results obtained with siRNA grp75 and proving that the amount of grp75 present at the OMM in the VDAC–grp75–IPR complex is a limiting factor of the positive effect of the IPR-LBD on mitochondrial Ca uptake. Lastly, by coexpressing grp75 and IPR-LBD, we achieved a very high stimulation of mitochondrial Ca uptake rate during capacitative Cainflux (i.e., upon the increase of bulk [Ca] to ∼1 μM). Thus, we concluded that the VDAC–grp75–IPR complex renders mitochondria more sensitive at low extramitochondrial [Ca], as compared with higher local [Ca] increases during IP-induced Ca release (compare the effect of IPR-LBD on [Ca] in and or ). Indeed, by overexpression of grp75 we could not observe a significant increase in histamine-induced [Ca] responses even if the steady-state [Ca] remained unaltered (unpublished data).
Based on previous observations (; ; ; ), we used VDAC1 as the start point for proteomic search of interacting proteins and for unraveling the molecular basis of mitochondrial Ca homeostasis. An unexpected, but intriguing, finding of our biochemical studies was the central location of the chaperone grp75 in the interaction between ER and mitochondrial Ca channels. grp75, a conserved chaperone, has a well studied role in protein import through the IMM. Still, in yeast mitochondria, mtHsp70/Ssc1 was shown to be significantly more abundant than the translocase (TIM23 complex). Thus, only a small fraction of the protein appears to be involved directly in preprotein translocation (; ), suggesting the existence of different pools of the protein. Previous work also reported extramitochondrial localization of grp75 (), and its interaction with extramitochondrial proteins such as the cytosolic p53 or the ER luminal grp94 (; ), although the mechanisms that control the differential sorting of the protein are still completely unknown. According to our immunofluorescence and GFP-tagging studies in HeLa cells grp75 shows complete mitochondrial localization, but obviously cannot be discriminated from an OMM-associated pool. Biochemical studies, however, demonstrate that a matrix-localized pool participates in forming complexes in the 200–400-kD range and represents the major fraction of the total mitochondrial grp75 content, whereas a minor grp75 pool resides in the low-density (MAM) mitochondrial fraction, participating in complexes in the megaDalton range and comprising OMM and ER membrane proteins. To further support an independent function of the nonmatrix pool, we constructed a grp75 mutant lacking the mitochondrial presequence, and thus incompetent for import in the matrix. This protein retained the capacity to enhance mitochondrial Ca accumulation, strongly arguing for the notion that this role of grp75 is not only independent from its chaperone activity in the matrix but also depends on a physically separated protein pool.
How is the newly identified regulatory activity on mitochondrial Ca uptake exerted? In principle, two different mechanisms can be envisioned. In the first, grp75 could be involved in scaffolding the ER–mitochondria contacts, and thus determines the number of sites in which mitochondria are exposed to the high [Ca] microdomains generated at the mouth of IPRs. Fluorescent labeling studies of the ER and mitochondria revealed a partial (5–20%) colocalization, reflecting these interactions. However, no increase in colocalization has been observed by overexpression of grp75 (or of the IPR-LBD; unpublished data), suggesting that they do not directly function as structural determinants of the contacts. In a second scenario, grp75 could control the interaction of ER and mitochondrial proteins at the existing organelle contacts, and thus allow cross-talk between signaling partners, e.g., the ion channels of the two membranes. Indeed, grp75, as shown by its knockdown and overexpression models, was necessary and sufficient for the stimulatory effect of the IPR-LBD on mitochondrial Ca uptake. Moreover, the proteomic data also highlight the central role of grp75 in this interaction. VDAC and IPRs coprecipitate with grp75, and the chaperone is coimmunoprecipitated by both anti-IPR and -VDAC antibodies, indicating that it is the key assembling molecule in the loose interaction between the two ion channels.
Within the IPR–grp75–VDAC complex, potentiation of mitochondrial Ca accumulation by the IPR-LBD does not require IP binding, as demonstrated by the fact that it is retained by the K508A mutant, which is unable to bind IP (). Although the mutant shows the same stimulatory effect (), one should remember that wild-type IPR-LBD, because of IP buffering, reduces ER Ca release, and thus conclude that the wild type is somewhat more effective than the mutant. To further confirm independence from IP buffering, we measured mitochondrial Ca uptake after capacitative influx through the plasma membrane ( and ). Also, under those experimental conditions, the IPR-LBD potently stimulated mitochondrial Ca uptake.
As for the molecular mechanism of the effect on the mitochondrial Ca machinery, different scenarios could be envisioned. In the first, the recombinantly expressed IPR-LBD, both from the OMM and ER side, could interact with the endogenous IPR itself, and modify the probability of its interaction with grp75/VDAC. Indeed, it was previously shown that intramolecular interactions between different domains of the IPR, such as the 224–605 minimal IP-binding domain and the 1–223 N-terminal repression domain, regulate IPR channel opening upon IP binding. Thus, one could hypothesize that the high expression levels of IPR-LBD represses an interaction between the extreme N-terminal of the endogenous receptor and grp75/VDAC. To clarify this issue, we expressed the whole (aa 1–604) IPR-LBD, which is targeted to the OMM. The IPR-LBD had the same effect as IPR-LBD, thus, excluding competition of these two cytoplasmic, N-terminal domains of the IPR. In the second, simpler scenario, the IPR-LBD mimics the effect of the endogenous IPR. Thus, it directly enhances mitochondrial Ca uptake by maximizing, within the macromolecular complex, the interaction with the mitochondrial VDAC channel. Indeed, the density of the exogenous IPR-LBD, based on fluorescence labeling () and Scatchard plot analysis of IP binding (), can be assumed to be at least one order of magnitude higher than the endogenous receptor, and indeed, high expression levels were necessary for the effect of IPR on mitochondrial Ca uptake.
The central role of grp75 in the IPR-LBD–induced augmentation of Ca uptake was clearly shown by the siRNA-driven silencing of the protein, leading to the abolition of the effect. Conversely, high-level expression of grp75 induced a compound effect involving at least three different locations, as follows: the ER, decreasing the steady [Ca] level; the OMM, interacting with VDAC, whose permeability/ion selectivity was shown to be modified by grp-75 binding (); and the mitochondrial matrix, modifying mitochondrial parameters, such as pH or Ca buffering capacity. Expression of the cytosolic grp75 and measurement of Ca influx–induced mitochondrial Ca uptake allowed us to eliminate the intramitochondrial effect and changes of ER Ca handling. Importantly, mitochondrial Ca uptake in this approach was markedly increased, and grp75 potentiated the effect of OMM-IPR-LBD, clarifying the effect of the OMM-associated pool of grp75.
In conclusion, we demonstrated that the IPR is part of a signaling complex that directly controls Ca uptake into mitochondria. Much remains to be understood, but by these results the concept of macromolecular assembly of signaling elements, previously put forward for several plasma membrane channels, can be extended to defined microdomains at the ER–mitochondrial interface. Such an arrangement highlights novel routes for pharmacological intervention that may be used for the modulation of downstream events such as metabolism and apoptosis.
Yeast two-hybrid screening was carried out using the pLexA system according to the protocol of . For details see Supplemental materials and methods (available at ).
HeLa cells and rat liver were homogenized, and crude mitochondrial fraction (8,000-g pellet) was subjected to separation on a 30% self-generated Percoll gradient, as previously described (). A low-density band (denoted as the MAM fraction) and a high-density band (denoted as Mito P) were collected and analyzed by immunoblotting and Blue native/SDS-PAGE 2D separation, which are described in detail in the Supplemental materials and methods. Proteinase K (Sigma-Aldrich) digestion was performed with 50 μg enzyme in the presence of 50 μg proteins (10 min, on ice) in solution A used to resuspend subcellular fractions (250 mM mannitol, 5 mM Hepes, and 0.5 mM EGTA, pH 7.4). Hyposmotic shock (50 mM mannitol, 5 mM Hepes, and 0.1 mM EGTA, pH 7.4, for 30 min at room temperature) was applied to induce mitochondrial swelling.
Mouse grp75, cloned into the expression vector pTOPO (Invitrogen), was provided by R. Wadhwa (University of Tokyo, Tokyo, Japan; ). Full-length mouse IPR-1 was obtained from K. Mikoshiba (RIKEN Brain Science Institute, Wako City, Saitama, Japan). The constructs encoding the fusion proteins of the PH domain of the p130 protein (from GenBank/EMBL/DDBJ under accession no.; residues 95–233) and the IPR-LBD domain (residues 224–605) of the human IPR-1 with monomeric red fluorescent protein (mRFP1), GFP, or YFP, as well as the strategies for ER targeting, have been previously described (; ). For OMM tethering, the N-terminal mitochondrial localization sequence of the mouse AKAP1 protein (from GenBank/EMBL/DDBJ under accession no. ; residues 34–63) was fused to the N termini of the IPR-LBD and p130PH constructs through a short linker (DPTRSR). The OMM-IP3-LBD-mRFP1 construct was obtained by amplification of the 1–604 fragment of IPR-1 cDNA and insertion into the AKAP1/mRFP1 vector. The GRP75cyt cDNA was amplified from a human liver cDNA library (Origene) using the primers 5′-CCCAAGCTTATGAAGGGAGCAGTTGTTGGTATTG-3′ and 5′-CGCGGATCCTTACTGTTTTTCCTCCTTTTGATC-3′. After digestion with HindIII and BamHI, the product was ligated into the pcDNA3 plasmid (Invitrogen) digested with the same restriction enzymes. The construct was verified with bidirectional sequencing.
Transient transfection was done by the Ca-phosphate precipitation technique. Experiments were performed 24–36 h after transfection.
cytAEQ-, mtAEQmut-, or erAEQmut-expressing cells were reconstituted with coelenterazine and transferred to the perfusion chamber, and light signal was collected in a purpose-built luminometer and calibrated into [Ca] values, as previously described (). All aequorin measurements were performed in Krebs-Ringer bicarbonate (KRB) containing 1 mM CaCl (KRB/Ca; Krebs-Ringer modified buffer: 135 mM NaCl, 5 mM KCl, 1 mM MgSO, 0.4 mM KHPO, 1 mM CaCl, 5.5 mM glucose, and 20 mM Hepes, pH 7.4). [Ca] after capacitative Ca influx was measured by preincubating HeLa cells with the SERCA blocker tBHQ (100 μM) in a KRB solution containing no Ca and 100 μM EGTA. Cytoplasmic Ca signal and mitochondrial Ca uptake were evoked by adding 2 mM CaCl to the medium. For [Ca] measurements, erAEQmut-transfected cells were reconstituted with coelenterazine n, after ER Ca depletion in a solution containing 0 [Ca], 600 μM EGTA, and 1 μM ionomycin, as previously described (). Experiments in permeabilized HeLa cells were performed as previously described (), except that 25 μM digitonin was used to preserve ER–mitochondrial contacts.
For 3D morphological image acquisition, the cells were transfected with mRFP1-fused IPR-LBD constructs and loaded with 50 nM MitoTracker Green (Invitrogen) for 20 min at 37°C. For morphological studies, cells were placed in a thermostatted chamber at 37°C in KRB/Ca solution and imaged using an inverted microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) using a 63×/1.4 Plan-Apochromat objective, a CoolSNAP HQ interline charge-coupled device camera (Roper Scientific) and the MetaMorph 5.0 software (Universal Imaging Corp.). Z-series images were deconvolved using the PSF-based Exhaustive Photon Reassignment deconvolution software (; ), running on a Linux-based PC. For colocalization analysis, thresholded images were 3D rendered using the Data Analysis and Visualization Environment software (; ). To approximate real colocalization, and to exclude artificial ones produced by the noise of the signal, only the voxels with <50% difference in their normalized intensity were taken into account.
Table S1 shows [Ca] and [Ca] responses of HeLa cells expressing the constructs in this study. Fig. S1 shows the proteomic analysis of molecular components of the MAM fraction. Fig. S2 shows the effects of IPR-LBD on cytoplasmic Ca responses and ER Ca homeostasis. Fig. S3 shows the effect of cytosolic- and OMM-targeted p130-PH domain on mitochondrial Ca uptake. Online supplemental material is available at . |
Over 70 yr ago, Warburg discovered that cancer cells are more dependent on glycolysis for generation of ATP, even when abundant oxygen is present in the cellular environment (). During the past several decades, this metabolic alteration has been observed in many cancer types, including solid tumors and leukemia. It is now recognized that the Warburg effect represents a prominent metabolic characteristic of malignant cells. Although the exact mechanisms responsible for this metabolic alteration remain to be elucidated, malfunction of mitochondrial respiration or “respiration injury” due, in part, to mitochondrial DNA (mtDNA) mutations/deletions is thought to be an important contributing factor (; ; ; ). Recent studies revealed that cancer cells of various tissue origins exhibit frequent mutations in their mtDNA (; ; ). Because mtDNA encodes for 13 protein components of the mitochondrial respiratory chain, it is likely that certain mtDNA mutations may cause malfunction of the respiratory chain, forcing the cells to increase glycolysis to maintain their ATP supply. The active metabolism in cancer cells requires a constant supply of sufficient ATP. Paradoxically, generation of ATP through glycolysis (two ATPs per glucose) is far less efficient than ATP production through mitochondrial oxidative phosphorylation (36 ATPs per glucose). It is unclear how cancer cells, with this apparent disadvantage in energy metabolism, can survive the competition with other cells in vivo and develop as a malignant cell population with drug-resistant potential.
Environmental factors, especially hypoxic conditions in the tumor tissue microenvironment, may also force cancer cells to use glycolytic pathways to generate ATP to meet their energy supply. Hypoxia-induced metabolic adaptations cause the hypoxic cells to exhibit certain biochemical characteristics similar to those of mitochondrial respiratory-defective cells. Under hypoxic conditions, cancer cells have to use the “energy-inefficient” glycolytic pathway to generate ATP, leading to accumulation of a high level of NADH, which is normally channeled to the electron transport chain as the energy fuel in respiration-competent cells.
Cancer cells use multiple pathways to enhance their survival and prevent apoptosis under various conditions. Increased expression of Bcl-2 and -X antiapoptotic factors and activation of NFκB and phosphatidylinositol 3-kinase (PI3K)–Akt pathways are among the well-characterized mechanisms by which cancer cells promote their survival capacity. Overexpression of Bcl-2 and/or -X counteracts the proapoptotic effects of Bax and Bak and inhibits the mitochondria-mediated cell death pathway (). NFκB can be activated by various stimuli, such as cytokines, DNA-damaging agents, and reactive oxygen species (ROS). Its antiapoptotic effect is mediated mainly by promoting expression of a variety of cell survival factors through a transcriptional activation mechanism (). The PI3K–Akt pathway promotes cell survival and proliferation through a series of downstream events, including enhancing nutrient uptake and energy metabolism through activation of mTOR (), stimulating aerobic glycolysis (), and suppressing apoptosis by phosphorylation of the Bad protein (). The Akt survival pathway is positively and negatively regulated by PI3K and PTEN, respectively, through their opposing effects on phosphatidylinositol-3′-phosphate (PIP) generation (). Despite the fact that the regulatory mechanisms of these pathways have been characterized in great detail, their potential roles in promoting the survival of cancer cells with mitochondrial respiration defects remain largely unknown.
Because mitochondrial respiratory defects are frequently observed in human cancers, owing to genetic alterations of mtDNA and/or hypoxic conditions in tumor tissue environment, elucidation of the molecular and biochemical mechanisms contributing to the survival of cancer cells with such metabolic defects is obviously important in understanding the biology of the Warburg effect and in developing new strategies to overcome drug resistance. This study used experimental model systems to investigate the possible survival mechanisms in cells with mitochondrial genetic defects and in mitochondrial respiration-competent cells under conditions where the respiratory activity is compromised by specific pharmacological inhibitors or by hypoxia.
To investigate the molecular events contributing to the survival mechanisms in cancer cells with mitochondrial respiration defects, we first derived multiple clones of respiration-deficient cells (ρ) with altered mtDNA from two different human cancer cell lines, HL-60 (leukemia), and Raji (lymphoma), and then examined potential molecular alterations that promote cell survival. mtDNA was preferentially damaged using an established method, in which the respiration-competent parental cells were chronically exposed to a low concentration of ethidium bromide (50 ng/ml), and the resulting ρ cells were subcloned by a serial dilution method (; ). One subclone of ρ cells derived from HL-60 cells was designated as HL60-C6F cells, which have been previously characterized (). Four other ρ subclones (C2, C6, C7, and C8) were derived from Raji cells, and their alterations in mtDNA and mitochondrial functions are shown in . All ρ cells exhibited defects in respiration, as indicated by a lack of oxygen consumption (). However, the ρ cells still retained their mitochondrial mass with apparently normal transmembrane potential (). PCR analysis of mtDNA revealed multiple deletions of mtDNA at and the D-loop regions (). The PCR products for , and genes were reduced to various degrees in different ρ clones, suggesting possible reduction of these gene copy numbers or mutations in the mtDNA regions for the PCR primers, and thus retarded the PCR reactions. Interestingly, all ρ cells retained the gene sequence at levels comparable to the parental cells. The nuclear gene remained unchanged in all clones. The respiration-deficient phenotype of the ρ cells has been stable in long-term culture (over 20 mo) in medium without ethidium bromide.
All ρ cell clones derived from HL-60 and Raji cells were strictly dependent on glycolysis and required supplements of high glucose, pyruvate, and uridine for growth. Importantly, the ρ cells exhibited a better ability to grow in hypoxic conditions, a common situation cancer cells encounter in vivo. As illustrated in , under normoxia conditions, the parental Raji cells showed better colony-forming efficiency (CFE; 76%) than the respiration-deficient Raji-C8 clone (CFE 56%). However, under reduced oxygen conditions (5% O), the respiration-deficient clone C8 showed a higher CFE compared with the parental Raji cells (). We also noticed that the parental Raji cells formed larger colonies in normoxic conditions than in hypoxia, whereas the colony sizes of the Raji-C8 cells were similarly small both in hypoxia and normoxia. The colony formation ratio under hypoxia/normoxia conditions showed that the Raji-C8 cells had a significant survival advantage under hypoxia compared with the parental cells (; P = 0.009). Similarly, when the cells were cultured under a more hypoxic condition (1.5% O), the ρ C8 cells exhibited a higher colony formation rate than the parental Raji cells ().
Despite mitochondrial metabolic defects, ρ cells were not prone to drug-induced cell death, and exhibited reduced sensitivity to common anticancer agents. As shown in , HL60-C6F cells were significantly less sensitive to arsenic trioxide (AsO) and taxol than the parental HL-60 cells, as measured by annexin-V reactivity. Similarly, the Raji ρ clones also showed less sensitivity to AsO (), doxorubicin (), and vincristine (). These findings suggest that the ρ cells appear to have some survival advantage. Because cell cycle and cell growing rates may also affect drug sensitivity, we compared the cell cycle profiles of the ρ clones with their respective parental cells (HL-60 and Raji). Analysis of cellular DNA contents by flow cytometry revealed no substantial difference in cell cycle distribution between the ρ clones and their parental cells (Fig. S1 A, available at ). Although the ρ clones exhibited a moderate decrease in cell growth rate, they retained an active apoptotic response and showed massive cell death when incubated with the glycolytic inhibitor 3-bromopyruvate (). Interestingly, the proapoptotic factor cytochrome in the ρ cells was comparable to the parental cells or even slightly increased (Fig. S1 B). Analysis of proteins extracts from mitochondrial and cytosolic fractions showed that cytochrome was detected only in the mitochondrial fraction in both cell types (Fig. S1, C and D). Thus, the increased expression of this proapoptotic factor in ρ cells is confined in the mitochondria, which is consistent with a previous observation ().
All four clones of ρ cells were then used for examination of changes in gene expression in comparison with their parental Raji cells, using oligonucleotide microarray produced at the University of Texas MD Anderson Cancer Center Genomic Core Facility. Among the genes that consistently exhibited changes in all four ρ clones, we identified two key molecules, PIK3CA and PTEN, involved in regulation of the Akt survival pathway. This suggested a possibility that Akt might be involved in promoting the survival of ρ cells. To determine whether Akt was indeed activated in the ρ cells, the expression of Akt protein level and its phosphorylation status were first assessed by immunoblotting. As shown in , Akt phosphorylation at both Ser-473 and Thr-308 increased in all ρ cells. The total Akt protein did not increase in HL60-C6F cells, and moderately increased in the Raji ρ cells. Among the ρ clones, HL60-C6F, Raji-C2, -C6, and -C8 showed a significant increase in Akt phosphorylation, whereas only a slight increase was detected in Raji-C7 cells. Direct analysis of Akt enzyme activity in vitro using glycogen synthase kinase 3 (GSK-3) as the Akt substrate further confirmed the increase of Akt kinase activity in the ρ cells, albeit with some individual variation (). The reason for such variation among individual clones is likely caused by other changes induced by ethidium. To confirm that Akt activation is a general phenomenon in ρ cells, we analyzed an additional three ρ clones derived from HL-60 cells, and revealed that these ρ cells also exhibited a consistent increase of Akt activation ().
We then tested if the increase of Akt activation in ρ cells might be caused by an increased expression of PI3 kinase, which is a positive regulator of Akt. Western blot analysis showed that the protein expression of PI3Kp110α (catalytic subunit) and PI3Kp85 (regulatory subunit), as well as the tyrosine phosphorylation levels, were similar in the parental and ρ cells (unpublished data). Thus, Akt activation was unlikely because of an increase in PI3K. Because the phosphatase PTEN is a negative regulator of Akt pathway and the loss of PTEN activity has been correlated with increased Akt activity in cancer cells (), we compared the PTEN protein levels and its phosphorylation in ρ cells and their parental cells. PTEN protein and its phosphorylation at Ser-380 and Thr-382/-383 were reduced in most ρ clones (). The PTEN protein and its phosphorylation state were inversely correlated with the degree of Akt phosphorylation (), which is consistent with the negative regulatory role of PTEN in Akt signaling.
We reasoned that if deficiency in mitochondrial respiration was a key event that suppresses PTEN and causes Akt activation, inhibition of mitochondrial respiratory function in the parental cells should also cause PTEN suppression and Akt activation. To test this cause–effect relationship, we used rotenone, a specific inhibitor of the mitochondrial electron transport complex I, to block respiration in Raji cells. At 100 nM, rotenone effectively blocked respiration as early as 5 min after drug exposure (), but did not cause significant cell death during a 24-h incubation, as assessed by annexin-V/propidium iodide (PI) staining (). Inhibition of respiration by rotenone caused a time-dependent activation of Akt, indicated by an increase in Akt phosphorylation, which elevated significantly at 3 h and remained active for at least 24 h (). Concurrently, there was a time-dependent decrease in phospho-PTEN revealed by Western blotting using phospho-PTEN antibodies (). These observations suggest that the inhibition of respiration by rotenone may either cause dephosphorylation of PTEN, or mask the phosphorylated epitope. The overall PTEN protein appeared unchanged when respiration was acutely blocked. Consistently, other inhibitors of mitochondrial respiratory complexes, including antimycin A, cyanide, and oligomycin, also caused Akt activation (Fig. S2, available at ).
Because ROS are known to cause Akt activation, we tested the potential involvement of ROS in Akt activation when mitochondrial respiration was inhibited by rotenone. The addition of the antioxidant -acetylcysteine (NAC) did not affect rotenone-induced Akt activation, suggesting that ROS was not a major factor contributing to Akt activation in respiration-suppressed cells (). Furthermore, analysis of superoxide in ρ cells showed that they contained lower basal levels of superoxide than the parental cells (). Thus, these data suggest that ROS may not be a critical mediator to activate Akt in cells lacking mitochondrial respiration. It should be noted, however, that exogenous ROS could induce Akt activation in Raji cells, as indicated by a significant Akt phosphorylation in the presence of 0.5 mM HO (). As expected, the HO-induced Akt activation could be suppressed by NAC (). This was different from the Akt activation in respiration-deficient cells.
Because NADH is an essential substrate (electron donor) for the mitochondrial electron transport chain, defects in the respiratory chain function could lead to an accumulation of NADH. Indeed, chemical inhibition of respiration by rotenone caused a substantial increase in cellular NADH, from 81 to 130 arbitrary units (). Consistent with this observation, mitochondrial genetic defects also led to a significant increase of NADH in HL60-C6F cell and all ρ clones from Raji cells (). Interestingly, incubation of cell extracts from sonicated Raji cells with NADH (0.01–1 mM) in the presence of 1 mM ATP led to a significant increase in Akt phosphorylation without changing Akt protein levels (). Collectively, these observations suggest the possibility that the accumulation of NADH might be directly involved in Akt activation in the ρ cells.
Analysis of lactate (product of glycolysis) in culture medium showed that ρ cells produced significantly more lactate (Fig. S3, available at ). The lack of oxygen consumption and increased lactate accumulation in ρ cells indicate that the glycolytic pathway is highly active in these cells. We speculated that the increase in glycolysis for ATP generation in ρ cells might divert the glucose metabolic flow away from the pentose phosphate pathway (PPP) and lead to a decrease of NADPH because PPP is the major pathway for NADPH generation. To test this possibility, HPLC analysis was performed to determine the ratios of NADH/NADPH in ρ cells and parental cells. Under normal culture conditions, the ratios of NADH/NADPH in parental HL-60 and Raji cells were 0.50 and 0.36, respectively, suggesting an active NADPH generation in parental cells (Fig. S4, available at ). In contrast, all ρ clones exhibited reversed NADH/NADPH ratios (1.41–2.46), which is consistent with the accumulation of NADH and the decrease in NADPH generation (Fig. S4 B).
To investigate the mechanism by which changes in cellular NADH and NADPH promote Akt activation, we tested the possibility that the increase in NADH/NADPH ratio might modify the redox status and function of PTEN, which is known to be redox-sensitive and dependent on NADPH/thioredoxin (Trx) to maintain its enzyme activity (). Protein extracts from Raji cells were incubated with NADH, NADPH, Trx, or a combination of the three. The reaction products were divided into two portions for Western blot analysis under reducing and nonreducing conditions. As illustrated in , similar phospho-PTEN signals were detected in all samples under reducing condition, suggesting that there was no significant difference in PTEN phosphorylation. However, when the same samples were analyzed under nonreducing condition, striking differences in phospho-PTEN signals were observed. NADH caused a significant decrease of the phospho-PTEN signal detectable under nonreducing condition (, lane 1), suggesting that PTEN was likely in an oxidized state, which masked the epitope for antibody binding caused by disulfide bond formation. Oxidation of PTEN by hydrogen peroxide has been shown to cause such a conformational change (). Incubation of cell extracts with NADPH/Trx kept the PTEN protein in a reduced state, and the phosphoepitopes were readily detected (, lanes 2–3). Combination of NADPH and Trx also kept PTEN in a reduced state (, lane 5). Surprisingly, NADH + Trx did not reduce PTEN, and thus the accessibility to epitopes at Ser-380 and Thr-382/-383 was limited (, lane 4). These data suggest that NADH may compete with NADPH/Trx to affect the redox state of PTEN. To test this possibility, protein extracts were incubated with NADPH/Trx in the presence of various concentrations of NADH, and the accessibility to epitopes at Ser-380 and Thr-382/-383 was determined under nonreducing conditions. As shown in , NADH competed with NADPH, and caused a concentration-dependent decrease of phospho-PTEN signal (lanes 3–5).
Fluorescent confocal microscopy analysis was then used to evaluate the localization of PTEN protein in Raji cells under various conditions. As illustrated in , a portion of the PTEN protein was localized in cellular membrane region under normal culture conditions. Incubation of Raji cells with rotenone decreased the membrane localization of PTEN, which exhibited defused intracellular distribution. The decrease in PTEN membrane localization was also observed in the respiration-deficient Raji-C8 cells (). These data suggest that mitochondrial defects or respiration inhibition could decrease PTEN membrane localization.
To further evaluate the role of PTEN inactivation in mediating NADH-induced Akt activation, we used four cancer cell lines with either wild-type PTEN (HCT116 and LN-229 cells) or PTEN-null (Jurkat and U87-MG cells) and tested their Akt activation in response to modulation of mitochondrial respiration. Incubation of HCT116 and LN-229 cells (wt PTEN) with rotenone led to a time-dependent Akt activation, which was indicated by an increase in Ser-473 phosphorylation (). In contrast, rotenone failed to further activate Akt in the PTEN-null cell lines (Jurkat and U87-MG, ). It should be noted that although Akt phosphorylation was constitutively high in the PTEN-null cells, Akt could still be further activated in these cells by other stimuli, such as TRAIL () and T cell antigen receptor stimulation with anti-CD3 ().
Hypoxia is frequently seen in the tumor microenvironment. Cancer cells under hypoxic conditions mainly use glycolysis to generate ATP and are metabolically similar to ρ cells. We further evaluate the role of PTEN inactivation in mediating Akt activation under hypoxic conditions. When HCT116 and LN-229 cells (wild-type PTEN) were incubated under hypoxic conditions, there was a significant increase in Akt activation as early as 2 h; this increase was indicated by increased phosphorylation at Ser-473 (). In contrast, the PTEN-null cells (Jurkat and U87-GM) showed little change in their Akt phosphorylation (), further supporting the important role of PTEN in hypoxia-mediated Akt activation.
To evaluate the contribution of Akt activation to drug resistance in ρ cells, we tested the effect of inhibiting Akt pathway on drug-induced apoptosis in the ρ cells. Wortmannin, which is an inhibitor of the PI3K–Akt pathway, caused a concentration-dependent decrease in Akt phosphorylation at Ser-473 and Thr-308 in HL60-C6F cells (). Wortmannin also caused dephosphorylation of Akt in all Raji ρ clones (). Another PI3K–Akt inhibitor, LY249002, produced similar results (unpublished data). Inhibition of PI3K/Akt by wortmannin significantly increased the sensitivity of ρ cells to apoptosis induction by AsO, as assessed by annexin-V reactivity () and cleavage of procaspase-3 (). We also used SH-6, which is a selective Akt inhibitor that does not affect the upstream kinases (), to further test sensitization of ρ cells. SH-6 was effective in enhancing apoptotic response to AsO in ρ cells (). Thus, inhibition of the Akt pathway seems effective in enhancing the activity of anticancer agents against cancer cells with respiration defects. In contrast, inhibition of the mitochondrial ATP synthase (complex V) by oligomycin decreased cellular sensitivity to AsO and Taxol in parental HL-60 cells (Fig. S5, available at ). Consistently, oligomycin did not significantly affect drug sensitivity in ρ cells (HL60-C6F), which exhibited Akt activation and resistance to drug-induced apoptosis (Fig. S5). The insensitivity of ρ cells to oligomycin is consistent with previous observations that the sensitivity of ATP-synthase (complex V) to oligomycin depends on the mitochondrial DNA-encoded subunit 6 of F1-ATPase (), and that cells lacking mtDNA possess normal levels of the nuclear DNA- encoded α and β subunits of F1-ATPase, which is functional and sensitive to azide or aurovertin, but insensitive to oligomycin ().
Mitochondrial respiration malfunction and increased glycolysis are frequently observed in cancer cells. This metabolic alteration, known as the Warburg effect, is caused by complex biochemical and molecular mechanisms. mtDNA mutations and tissue hypoxia represent genetic and environmental factors contributing to the Warburg effect. Cells deficient in respiration because of mtDNA alterations or hypoxic conditions are forced to produce ATP through glycolysis, which is much less efficient than oxidative phosphorylation. Nevertheless, cancer cells manage to overcome such an apparent metabolic disadvantage, survive in vivo, and eventually emerge as a malignant cell population resistant to anticancer agents at the late stages of the disease progression. Thus, understanding the mechanisms underlying the increased survival capacity in cancer cells with compromised mitochondrial respiratory function is an important research area.
Our study suggests that mitochondrial respiration deficiency leads to activation of the Akt survival pathway through NADH-mediated inactivation of PTEN. This is a novel mechanism contributing to increased survival and drug resistance in cancer cells with compromised mitochondrial respiration. Several lines of evidences support this conclusion, as follows: (a) Cells that lack mitochondrial respiration because of mtDNA deletion, chemical inhibition of the electron transport chain, or exposure to hypoxia all exhibited significant Akt activation. (b) The cellular NADH/NADPH ratio abnormally increased when mitochondrial respiration was suppressed, and this was associated with a decrease in plasma membrane–associated PTEN. (c) Exogenous NADH led to inactivation of PTEN and activation of Akt in vitro. The inactivation of PTEN seems to be caused by redox modulation because NADH competed with NADPH/Trx to keep PTEN in an oxidized (inactive) state. These findings are consistent with previous studies showing inactivation of PTEN by oxidation using hydrogen peroxide (; ). (d) Cells lacking functional PTEN did not respond to respiratory inhibition or hypoxia, and exhibited no further Akt activation, indicating the important role of PTEN in this process.
Under physiological conditions, NADH is generated through glycolysis and the tricarboxylic acid cycle, whereas NADPH is produced mainly via the PPP (shunt). The proportion of glucose directed to each pathway is regulated by the cellular energy metabolic state. Mitochondrial defects render cancer cells dependent on glycolysis for ATP supply, and the NADH generated from the tricarboxylic acid cycle is not used effectively because of the decrease in oxidative phosphorylation. These metabolic alterations lead to an accumulation of NADH. At the same time, NADPH production from the PPP decreases because of increased utilization of glucose for glycolysis. Indeed, we consistently observed that the NADH/NADPH ratio was significantly increased in all eight clones of ρ cells (Fig. S4). Because NADH competes with NADPH and compromises the ability of NADPH/Trx to keep PTEN in a reduced state, the metabolic changes in cancer cells with mitochondrial defects would lead to inactivation of PTEN and activation of Akt. Interestingly, NAC suppressed Akt activation induced by HO, but did not decrease rotenone-induced Akt phosphorylation. The likely explanation is that the redox-sensitive PTEN was inactivated when the ratio of NADH/NADPH was significantly increased. This elevated ratio could not be modulated by NAC when cells were treated with rotenone. In contrast, the antioxidant NAC effectively decreased HO and reduced its direct effect on PTEN.
The PI3K–Akt pathway is critical for cell survival (; ). Activation of PI3K results in generation of PIP, which leads to activation of phosphoinositide-dependent kinase-1 (PDK-1) and phosphorylation of Akt. In contrast, the lipid phosphatase PTEN removes a phosphate from PIP, and thus acts as a negative regulator of Akt. Loss of PTEN leads to Akt activation in cancer cells (). Thus, it is likely that oxidation of PTEN suppresses its phosphatase activity and subsequently leads to Akt activation. Indeed, PTEN is sensitive to oxidative inactivation by HO (). The demonstration that respiration defects lead to activation of the Akt pathway caused by the accumulation of NADH and inactivation of PTEN reveals a novel mechanism by which cancer cells survive under respiration-compromised conditions. illustrates a model of this cell survival mechanism.
The degree of Akt activation among the ρ clones appeared somewhat heterogeneous. It is possible that during the process of establishing the ρ clones, the use of ethidium bromide to deplete mtDNA might also cause nuclear DNA mutations, which might affect PTEN function and/or Akt activation. This could also explain the heterogeneous colony formation efficiencies observed among the ρ clones. Although this heterogeneity reflects the complexity of the experimental systems, the conclusion that mitochondrial respiration defects lead to NADH-mediated PTEN inactivation and Akt activation remains valid. This argument is supported by the observations that Akt activation was observed in all eight ρ clones, in cells treated with respiratory chain inhibitor rotenone, and in cells under hypoxia in a PTEN-dependent manner.
Because mitochondrial DNA mutations and hypoxia with subsequently increased glycolysis are prevalent in cancer cells (; ; ; ; ), activation of the Akt pathway through NADH-mediated PTEN inactivation is likely an important survival mechanism for cancer cells with such metabolic alterations. Additionally, the ability of Akt to promote glucose uptake may also contribute to cell survival (). Interestingly, a recent study showed that Akt activation stimulates cells to use the glycolytic pathway to generate ATP (). The observations that hypoxia caused Akt activation in both HCT116 and LN-229 cells and that respiratory-deficient cells exhibited certain growth advantage in hypoxia conditions further illustrate the clinical relevance of this mechanism in cancer cell survival and growth in vivo. Furthermore, if Akt activation is an important mechanism contributing to decreased drug sensitivity associated with the Warburg effect, it is possible to overcome such drug resistance by inhibition of Akt activation. In fact, our data suggest that this is possible. Further investigation is warranted to evaluate the clinical implications of this therapeutic strategy.
Human leukemia cells lines (HL-60 and Jurkat) and lymphoma cell line (Raji) were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated FBS and 2 mM -glutamine at 37°C with 5% CO. Human malignant glioma cells (U87-MG and LN-229; S. Kondo, MD Anderson Cancer Center, Houston, TX) were cultured in DMEM medium supplemented with 10% FBS and 2 mM -glutamine. HCT116 human colon cancer cells from B. Vogelstein (Johns Hopkins University, Baltimore, MD) were cultured in McCoy's 5A medium supplemented with 10% FBS and 2 mM -glutamine. The respiration-deficient ρ cells were established using an established method with ethidium bromide (), and maintained in RPMI 1640 medium containing 0.47% glucose, 10% FBS, 2 mM -glutamine, 50 μg/ml uridine, and 1 mM pyruvate without ethidium bromide as previously described (). To compare colony formation capacity under normoxia and hypoxia conditions, the parental Raji cells and the respiration-deficient Raji-C8 cells were seeded onto 6-well plates (300 cells/well for normoxia and 400 cells/well for hypoxia) in semisolid medium (MethoCult; StemCell Technologies, Inc.) containing 0.47% glucose, 10% FBS, 2 mM -glutamine, 50 μg/ml uridine, and 1 mM pyruvate, and incubated at 37°C under normoxia (21% oxygen) or hypoxia (1.5–5% oxygen) conditions for 2 wk. Colonies were stained, photographed, and counted.
The following antibodies were used for immunoblotting analyses using standard Western blotting procedures: phospho-Akt antibodies, anti–GSK-3, anti–phospho-GSK-3α/β, anti-PI3Kp110α, anti-PI3Kp85, anti–phospho-Tyr of PI3Kp85, anti-PTEN, and phospho-PTEN antibodies (all purchased from Cell Signaling Technology); anti–β-actin (Sigma-Aldrich); anti-Akt1/2 antibodies (Santa Cruz Biotechnology, Inc); anti–pro-caspase-3 (BD Biosciences); LY294002, wortmannin, rotenone, NADH, NADPH, doxorubicin, taxol, AsO, and vincristine (Sigma-Aldrich); SH-6 (AG Scientific, Inc.); Trx (Promega); MitoTracker Green, dihydroethidium, and rhodamine-123 (Invitrogen); and Methocult H4230 (StemCell Technologies, Inc.).
Cells were stained with 60 nM MitoTracker Green (60 min) to measure the mitochondrial mass, with 100 ng/ml dihydroethidium (60 min) to detect superoxide, or with 1 μM rhodamine-123 (60 min) to evaluate the mitochondrial transmembrane potential, as previously described (). Analysis was performed using a FACScan flow cytometer (Becton Dickinson). A minimum of 10,000 cells per sample was analyzed. Cellular NADH was measured by quantifying its intrinsic fluorescence under ultraviolet excitation, using a flow cytometer (LSR II; Becton Dickinson) equipped with a 325-nm excitation laser and a 440-nm centered band-pass filter, as previously described (). To analyze drug-induced apoptosis, cells were stained with annexin-V–FITC and PI, or with a monoclonal FITC-conjugated antiactive caspase-3 antibody according to the manufacturer's instruction (BD Biosciences). Data acquisition and analysis were performed using a FACScan flow cytometer with the CellQuest software (BD Biosciences). Cells that were positively stained by annexin- V–FITC only (early apoptosis) and positive for both annexin-V–FITC and PI (late apoptosis) were quantitated, and both subpopulations were considered as overall death cells.
The Akt enzyme activity was assayed using an Akt kinase kit according to the manufacturer's directions (Cell Signaling Technology). To analyze the effect of NADH and NADPH on Akt activation in vitro, cell lysates were prepared by sonication. Raji cells were suspended in PBS containing a cocktail of protease inhibitors (Roche), sonicated on ice bath, and centrifuged in a refrigerated centrifuge (model 5415R; Eppendorf) at 10,000 rpm for 15 min to remove cell debris. Endogenous NADH and NADPH in the cell lysates were removed by dialysis in cold PBS containing protease inhibitors for 60 min. The cell lysates containing proteins and sonicated plasma lipid membranes were incubated for 20 min with NADH, NADPH, and ATP as specified in the figure legends. Akt phosphorylation at Ser-473 and total Akt protein were assayed by Western blotting.
Oxygen consumption in intact cells was measured as an indication of mitochondrial respiration activity. Cells (5 × 10) were suspended in 1 ml of culture medium preequilibrated with 21% oxygen; then they were placed in a sealed respiration chamber to monitor oxygen consumption, as previously described ().
Total DNA containing nuclear and mitochondrial DNA was isolated from 3 × 10 cells, as previously described (). The nucleotide sequences of the PCR primers for mitochondrial D loop (15–484), NDI (3,304–3,836), COXII (7,645–8,215), ATPase 6 (8,539–9,059), ND4 (11,403–11,927), Cytochrome b (15,260–15,774), and GAPDH, and the PCR reaction conditions were previously described (). The PCR products were analyzed by electrophoresis on a 1.2% Agarose gel, stained with ethidium bromide, and photographed.
The parental cells (HL-60 and Raji) and ρ cells in exponentially growing phase were washed twice with PBS, and NADH and NADPH were extracted using 0.4 N HClO, followed by neutralization with concentrated KOH. The neutralized cell extracts were immediately analyzed for NADH and NADPH, using a HPLC method adapted from previously described procedures (). In brief, NADH/NADPH standards or cell extracts (550 μl; equivalent to 5.5 × 10 cells) were applied to an anion-exchange column (Partisil-10 SAX; Whatman) and run at a flow rate of 1.5 ml/min using a concave gradient (curve #9) from 100% buffer A (5 mM NaHPO, pH 4.0) to 100% buffer B (250 mM NaHPO + 0.5 M NaCl, pH 4.75) over 15 min, followed by another 15-min isocratic 100% buffer B, using a HPLC system (Waters Alliance) equipped with Empower Software. NADH and NADPH were detected by their UV absorbance at 340 nm, with retention times of 17.1 and 20.1 min, respectively, under these conditions. Because auto-oxidation of NADH and NADPH in the samples can decrease the UV absorbance at 340 nm, HPLC analysis was performed immediately after cell extracts were prepared. To minimize the potential effect of auto-oxidation on quantitation, the ratio of NADH/NADPH was calculated from the peak areas of NADH and NADPH of the same HPLC chromatograms. This ratio did not change when the same sample was injected at different times. This normalization allows a quantitative comparison of NADH/NADPH ratios between the parental cells and the ρ clones.
Cells were cytospun onto poly--lysine–coated glass slides using a cytospin (Shandon-Elliot) and were immediately fixed with 100% acetone for 5 min. After blocking in PBS containing 1% BSA, the samples were stained with anti-PTEN mouse monoclonal antibody for 1 h (clone PTEN-18; Sigma-Aldrich), followed by a 1-h incubation with Texas red–conjugated anti–mouse secondary antibody (Vector Laboratories; 1:300). The cells were visualized using a laser scanning confocal microscope (FluoView 500; Olympus). Images were captured using a 60× objective with proper filter sets (model IX71FVSF-2; Olympus).
A -test was used to evaluate the statistical differences of the experimental values between two samples to be compared.
Fig. S1 shows the comparison of cell cycle profiles and cytochrome expression in ρ clones and their parental cells. Fig. S2 shows the activation of Akt in Raji cells treated with respiratory chain inhibitors. Fig. S3 shows the lactate production in ρ cells in comparison with their parental cells (HL-60 and Raji). Fig. S4 shows the comparison of NADH/NADPH ratios in parental HL-60 and Raji cells and respiration-deficient cell clones. Fig. S5 shows the effect of oligomycin on cellular sensitivity to taxol and AsO in HL-60 cells and HL-60–C6F cells. Online supplemental material is available at . |
As one of the major degradative mechanisms conserved among eukaryotic cells, autophagy mediates the turnover and recycling of long-lived cytosolic proteins and excess or damaged organelles (). The cargo destined for autophagic degradation is sequestered in a double-membrane vesicle called an autophagosome, which fuses with the lysosome in mammalian cells or the vacuole in yeast. Eventually, the cargo is degraded by lysosomal/vacuolar resident hydrolases. Autophagy occurs in response to physiological stress or developmental signals (). Recently, autophagy has been implicated in a variety of human diseases, including cancer, neurodegeneration, and pathogen infection (). The initial identification of >20 autophagy-related () genes in the budding yeast has highlighted this single-cell organism as a perfect model to study the molecular mechanism of autophagy, although orthologues of some yeast genes have been found in higher eukaryotes.
In yeast, autophagy can be induced under starvation conditions to reuse nutrients for essential cellular activities and proper cellular remodeling; this starvation-induced bulk autophagy is considered nonspecific. Studies in yeast have also revealed that has selective autophagic pathways that target specific cargos (; ). These pathways mechanistically and genetically resemble bulk autophagy. One such route is the cytoplasm to vacuole targeting (Cvt) pathway (). In this pathway, two vacuolar hydrolases, the precursor form of aminopeptidase I (Ape1 [prApe1]) and α-mannosidase, are transported to the vacuole in a double-membrane vesicle called a Cvt vesicle, with the former subsequently being processed into mature Ape1. Compared with starvation-induced autophagy, the Cvt pathway occurs constitutively in growing conditions. Although some Atg proteins appear to be pathway specific, most are involved in both the specific and nonspecific pathways. However, it is not fully understood how these proteins coordinate and function at the molecular level in either bulk or selective autophagy.
Most yeast Atg components localize at a perivacuolar punctate structure called the preautophagosomal structure (PAS) or phagophore assembly site, which is proposed to be the site of autophagosome and Cvt vesicle formation (; ). In most endomembrane trafficking systems, such as the early secretory pathway, vesicles form by budding from the surface of a preexisting organelle. However, in autophagy-related processes, the double-membrane sequestering vesicles appear to form de novo; that is, they expand by membrane addition during the formation process rather than being generated from a single piece of contiguous membrane (). One of the major current challenges is to unveil where the membrane materials for autophagosomes or Cvt vesicles come from and how the lipids are transported to the assembly site. Among all Atg proteins, Atg9 is the best candidate that can help us understand this pivotal issue. Atg9 is the only characterized integral membrane protein required for both autophagosome and Cvt vesicle formation (). However, this protein is absent from the completed vesicles, suggesting that it is retrieved before the vesicle sealing/completion step. Atg9 localizes to multiple punctate sites, with one of them corresponding to the PAS and others to mitochondria in addition to unidentified structures (). Recent studies reveal that Atg9 cycles between mitochondria and the PAS vesicle assembly site (, ). These characteristics make Atg9 a potential membrane carrier for vesicle formation.
We decided to investigate the molecular regulatory mechanisms underlying Atg9 cycling and, in particular, what factors regulate the anterograde transport of Atg9 to the site of vesicle formation. In this study, we discovered that a peripheral membrane protein, Atg11 (), is an interaction partner of Atg9. The interaction requires the second coiled-coil (CC) domain of Atg11 and the Atg9 N-terminal cytosolic domain. A missense mutation (H192L) in the Atg9 N-terminal domain that disrupts its interaction with Atg11 results in the impaired cycling of Atg9 and a defect in selective autophagy. In addition, we found that in actin mutant cells, Atg11 colocalized with Atg9 and was retained on mitochondria, indicating that Atg11 is not able to direct Atg9 to the PAS in the absence of an intact cytoskeletal network. These data support a model in which a pool of Atg11 links Atg9 to the PAS along the actin cable under vegetative growth conditions.
Atg9 is the only known transmembrane protein required for both bulk autophagy and selective autophagic processes (e.g., the Cvt pathway; ). Unlike most other Atg proteins, which are restricted to the perivacuolar PAS, Atg9 localizes to several punctate structures; one of them is at the PAS, whereas the others are primarily confined to mitochondria (). Atg9 cycles between the two compartments, suggesting that it plays a role in providing lipids to the forming autophagosomes or Cvt vesicles (). However, other than actin (), the factors that regulate the anterograde transport of Atg9 to the PAS have not been identified. Therefore, we performed a yeast two-hybrid–based screen of Atg proteins to identify potential Atg9-interacting proteins. Yeast two-hybrid cells harboring Atg9 and a peripheral membrane protein, Atg11, showed robust growth on plates lacking histidine, indicating that Atg11 could interact with Atg9 (). The same result was obtained on plates lacking adenine (unpublished data).
Atg11 functions in selective types of autophagy (i.e., the Cvt pathway and pexophagy) but is not essential for bulk autophagy (). Atg11 plays a role in organizing the PAS and linking cargo to the vesicle-forming machinery at the PAS (; ). This protein is a component of at least two complexes in yeast. One is the Atg1–Atg11 complex, which is involved in the induction of bulk and selective autophagy (). The other is the Atg19–Atg11 complex, which recognizes and delivers prApe1 and α-mannosidase to the PAS (; ). We decided to determine whether these complexes are involved in Atg11 and Atg9 interaction. To address this issue, we used a biochemical approach to examine whether Atg11 was able to form a complex with Atg9. We tagged Atg11 with GFP and tagged Atg9 with protein A (PA) at the chromosomal locus. Wild-type, Δ, and Δ cells expressing the integrated Atg9-PA and Atg11-GFP fusions were lysed, and the PA-tagged protein was isolated with IgG–Sepharose beads. Atg11 was coprecipitated with Atg9-PA in all three strains (), which verifies that these two proteins are present in a complex, although we do not know whether they interact directly. Thus, the absence of either Atg1 or Atg19 did not affect the formation of a complex between Atg9 and Atg11. This finding suggests that there might be multiple populations of Atg11 within the cell that interact with different sets of Atg proteins.
Our recent data show that two other Atg proteins, Atg23 and Atg27, interact with Atg9 and are required for Atg9 cycling. The interaction between Atg9 and either Atg23 or Atg27 is not mediated through Atg11 (; and unpublished data). Thus, we extended the analysis by determining whether the Atg9–Atg11 interaction was dependent on these other Atg9-interacting proteins. We found that Atg11 was coprecipitated with Atg9 in Δ and Δ Δ cells despite a lower overall efficiency of recovery (). This suggests that Atg9 and Atg11 were able to form a complex in the absence of Atg23 and Atg27, although these two proteins may facilitate the interaction. The interaction between Atg9 and Atg11 in the Δ or Δ strains was confirmed by yeast two-hybrid analyses; the Δ or Δ two-hybrid cells expressing Atg9 and Atg11 were able to grow on −histidine (−His) selective plates, which is comparable with the wild-type cells (unpublished data).
Atg11 is predicted to contain four CC domains (; ). Each CC domain mediates interactions of Atg11 with different Atg proteins. To test whether these CC domains are responsible for the interaction between Atg11 and Atg9, we used CC domain deletion mutants in a series of yeast two-hybrid assays. As shown in , the two-hybrid mutant activation domain (AD)–Atg11 C-terminal truncation (11N; Δ627–1,178 and lacking CC3-4) allowed the cells to grow in the presence of binding domain (BD)–Atg9 as well as the full-length AD-Atg11. In contrast, cells expressing an AD-Atg11 N-terminal truncation (11C; Δ1–817 and lacking CC1-3) could not grow on selective −His plates (). Thus, the Atg11 N terminus is sufficient for Atg9–Atg11 interaction. In addition, Atg11ΔCC2 (Δ536–576) abolished the interaction with Atg9, whereas cells containing mutants Atg11ΔCC1 (Δ272–321), Atg11ΔCC3 (Δ627–858), or Atg11ΔCC4 (Δ859–1,178) grew well on selective plates (). These data indicated that Atg11 CC domain 2 is required for the Atg9–Atg11 interaction.
From a hydropathy plot analysis, Atg9 contains hydrophilic N and C termini flanking six to eight transmembrane domains (). To define the topology of Atg9, we investigated the protease sensitivity of the Atg9 N and C termini. Spheroplasts derived from Δ cells expressing either the N-terminal–tagged PA-Atg9 fusion or the C-terminal–tagged Atg9-PA fusion were osmotically lysed and centrifuged at 13,000 . In agreement with previous studies, approximately two thirds of the total Atg9 was present in the S13 supernatant fraction that contained the PAS, and one third was present in the P13 pellet fraction (; ; ). When the P13 fraction was treated with exogenous proteinase K, both the N- and C-terminal PA tags were cleaved in the absence or presence of detergent, and no bands of smaller molecular mass were detected (), indicating that both the N and C termini of Atg9 were accessible to protease on the cytosolic side of the membrane. To verify that intracellular membranous structures were intact after osmotic lysis, we simultaneously monitored the protease sensitivity of an endogenous vacuole membrane protein, Pho8. The precursor form of Pho8 that accumulated in the background contains a small cytosolic tail and a lumenally oriented propeptide (). In the absence of detergent, only the Pho8 cytosolic tail but not the lumenal propeptide was accessible to proteinase K. Upon the addition of both detergent and proteinase K, the Pho8 lumenal propeptide was removed as a result of the disruption of all membranous compartments, which is shown as a further shift of the molecular mass (). Thus, these data verified the integrity of the relatively fragile vacuole and presumably other intracellular membranous compartments after osmotic lysis, suggesting that both the N and C termini of Atg9 are exposed to the cytosol (). Accordingly, the topology of Atg9 appears to be conserved between yeast and mammalian cells ().
So far, no known functional domains have been identified in the Atg9 N- or C- terminal regions. To further analyze the Atg9–Atg11 interaction, we generated truncated Atg9 mutants containing the N terminus, C terminus, or transmembrane region. As shown in , in the presence of Atg11, the Atg9 N-terminal domain supported the growth of two-hybrid cells as well as the full-length Atg9, whereas neither the C-terminal nor the transmembrane domains of Atg9 were able to do so (). This result showed that Atg11 interacts with the N-terminal region of Atg9. We further constructed a series of Atg9 N-terminal truncation mutants and analyzed them for interaction with Atg11 by yeast two-hybrid analysis. Atg9 N-terminal amino acids 159–255 appeared to be the minimal region that mediates the interaction with Atg11 (). Collectively, we concluded that Atg11 and Atg9 interact through the Atg11 CC domain 2 and the Atg9 N terminus.
In wild-type cells, Atg11 localizes at the PAS (), and Atg9 cycles between the PAS and mitochondria (, ). To examine the role of the interaction between Atg9 and Atg11, we analyzed the cycling of Atg9 in the presence of overexpressed Atg11 or in the absence of this protein. The chromosomally tagged Atg9-YFP chimera displayed a multiple punctate distribution in wild-type cells, with one of the puncta colocalizing with the PAS marker blue fluorescent protein (BFP)–Ape1 (). In contrast, in the Δ background, Atg9 was restricted to the PAS, which is in agreement with previous observations that indicated a role for Atg1 in the retrograde transport of Atg9 from the PAS to mitochondria (, ,b). In 91% (109/120) of the cells overexpressing Atg11, Atg9-YFP localized solely to the perivacuolar PAS (represented by CFP-Atg11), which is similar to the situation observed in Δ cells (; ), and did not localize to mitochondria (not depicted). Thus, excess Atg11 was able to restrict Atg9 to the PAS. Overexpressed Atg11 displays a dominant-negative phenotype, interfering with the vacuolar import of prApe1 through the Cvt pathway (unpublished data). The dominant-negative phenotype presumably reflects the defect in Atg9 cycling.
To further examine the role of Atg11 in recruiting Atg9 to the PAS, we used the TAKA (transport of Atg9 after knocking out ) assay (). This assay examines the epistasis of a second mutation relative to Δ with regard to Atg9 localization at the PAS. We visualized chromosomally tagged Atg9-YFP in Δ single and Δ Δ double deletion cells and simultaneously labeled the vacuolar membrane with FM 4-64. As shown previously, in Δ cells, Atg9-YFP localized to a single perivacuolar punctum, which corresponds to the PAS (). In contrast, in nearly 90% of the Δ or Δ Δ mutants, Atg9-YFP fluorescence showed multiple puncta and did not localize at a single perivacuolar structure, suggesting that its anterograde transport was blocked because of the deletion (). These results suggest that Atg11 is involved in the anterograde transport of Atg9.
Our mapping of the Atg9 and Atg11 interaction domains indicated that Atg11ΔCC2 but not Atg11 lacking its other CC domains was defective in forming a complex with Atg9 (). Accordingly, we examined the effect of Atg11 CC deletions on Atg9 subcellular distribution. As shown in , in cells expressing Atg11ΔCC1 in the Δ background, Atg9 localized to the PAS, which is similar to the result seen in Δ cells expressing wild-type Atg11. In contrast, in cells expressing Atg11ΔCC2, which lacks the Atg9-interacting domain, Atg9 displayed a multiple punctate distribution () resembling that observed in Δ Δ cells even though the mutant protein was expressed at a level similar to the wild-type protein (). Thus, we concluded that the interaction of Atg9 with Atg11 CC domain 2 is required to direct Atg9 to the PAS.
To further clarify the physiological functions of the Atg9–Atg11 interaction, we decided to test whether autophagic processes were affected by its disruption. It has been reported that Atg11 CC domain 2, the Atg9-interacting domain, interacts with multiple Atg proteins, including at least one (Atg1) that is involved in Atg9 retrograde transport (). Thus, Atg11ΔCC2 may cause pleiotropic effects when used in functional studies. To bypass this problem, we decided to use native Atg11 and instead to isolate mutations within the Atg9 N-terminal region that disrupt the interaction with Atg11. We performed a PCR-based random mutagenesis on the N-terminal region of Atg9 followed by a yeast two-hybrid screen for loss-of-interaction mutants. A missense mutation was identified with a single histidine to leucine substitution at position 192 (H192L), which is located in the minimal region (amino acids 159–255) needed for the Atg9–Atg11 interaction (). Two-hybrid cells harboring this mutation completely lost the capacity for growth on selective −adenine (−Ade) plates in the presence of Atg11 (), whereas the expression level of the mutant protein was comparable with that of the wild type (not depicted). This indicates that the interaction between Atg9 and Atg11 was abolished by the H192L mutation. The loss of Atg9–Atg11 interaction was confirmed by a coimmunoprecipitation assay. As shown in , endogenous Atg11 was recovered only with wild-type Atg9 but not with the Atg9 mutant. In contrast, yeast two-hybrid data indicated that the interaction between Atg9 and either Atg2 or Atg18, two proteins involved in the retrograde transport of Atg9, was unaffected (unpublished data).
Because both and are essential genes for the Cvt pathway, which is a type of selective autophagy, we monitored the processing of prApe1 as a marker protein for this transport route. As shown in , cells expressing Atg9 did not generate the mature form of Ape1 compared with cells expressing wild-type Atg9, indicating that the Cvt pathway was impaired by the Atg9 H192L mutation.
We were interested in determining whether the impairment of the Cvt pathway seen with Atg9 () was caused by defects in Atg9 cycling, particularly anterograde movement to the PAS. Accordingly, we used the TAKA assay to visualize the cycling of Atg9-GFP in growing conditions and concurrently stained mitochondria with the dye MitoFluor red. In the Δ background, wild-type Atg9-GFP was restricted to the PAS, as marked with BFP-Ape1 in 94% (116/123) of the cells examined (). In contrast, Atg9-GFP distributed to multiple punctate structures in 84% (131/156) of the Δ cells, which colocalized, in part, with the mitochondrial labeling but not with the PAS. Thus, the Atg9 mutation acted epistatically to Δ, suggesting that the anterograde movement of Atg9 from mitochondria to the PAS was defective.
Atg9 is essential for both bulk autophagy and the Cvt pathway, whereas Atg11 is required solely in the Cvt pathway (; ). To clarify whether Atg9 anterograde transport via Atg11 is involved in bulk autophagy, we used the Atg9 mutant to analyze the progression of bulk autophagy by several established assays.
Atg8 conjugated to phosphatidylethanolamine remains associated with the completed autophagosome and is a marker for autophagic delivery to the vacuole (, ; ). After delivery of the GFP-tagged Atg8 chimera, the GFP moiety is cleaved and remains relatively stable in the vacuole, whereas Atg8 is rapidly degraded. Thus, the accumulation of free GFP reflects the progression of bulk autophagy, which can be readily detected by Western blotting (; ). In Δ cells, essentially no free GFP was detected, indicating that bulk autophagy was blocked by deletion (). In cells expressing wild-type Atg9 or Atg9, free GFP was detectable starting 2 h after cells were shifted to starvation conditions (synthetic medium lacking nitrogen [SD-N]) to induce bulk autophagy, although GFP-Atg8 processing showed a delay in Atg9 cells compared with wild-type cells. This result demonstrated that bulk autophagy retained similar activity even when Atg9 failed to interact with the Cvt-specific component Atg11.
To confirm the aforementioned result, we quantitatively measured bulk autophagy activity using another marker protein, Pho8Δ60, which encodes an altered form of alkaline phosphatase that is only delivered to the vacuole via autophagy (). The Pho8Δ60 enzymatic activity was measured in wild-type and Δ cells and in Δ cells transformed with a plasmid expressing wild-type Atg9 or Atg9 or an empty vector in growing (synthetic minimal medium [SMD]) and starvation (SD-N) conditions (). The Δ or Δ cells transformed with an empty vector showed only the basal level of Pho8Δ60 activity after autophagy induction (SD-N), indicating that bulk autophagy was defective after deleting either gene. In contrast, there was an increase of Pho8Δ60 activity in Δ cells expressing either wild-type Atg9 or Atg9, indicating that Atg9 rescued the autophagy defect in Δ cells, which is comparable with the wild-type Atg9 protein. Collectively, these data suggested that bulk autophagy does not depend on the interaction between Atg9 and an Atg protein that is needed for anterograde movement during specific autophagy (Atg11).
Because bulk autophagy activity was not affected by the Atg9 H192L mutation, it was tempting to speculate that Atg9 cycled normally under starvation conditions even though it was not capable of forming a complex with Atg11. To test this hypothesis, we visualized the movement of Atg9 by the TAKA assay after autophagy induction. Cells were treated with the drug rapamycin, which mimics starvation conditions and induces bulk autophagy. As shown in , in the Δ background without rapamycin treatment (−rap), wild-type Atg9-GFP was restricted to the PAS, whereas Atg9-GFP could not move to the PAS and displayed a multiple punctate localization. After treatment with rapamycin (, +rap), in 85% (41/48) of the cells, Atg9-GFP colocalized with the PAS marker BFP-Ape1 similarly to wild-type Atg9, indicating that Atg9 recruitment to the PAS was normal. Therefore, this result demonstrated that during bulk autophagy, the Atg9 mutation did not interfere with the cycling of Atg9.
Recently, we have shown that Atg9 anterograde traffic to the PAS is blocked when the actin cytoskeleton is disrupted by either treatment with the drug latrunculin A or point mutations in , the gene encoding actin (). In particular, the impairment of actin function leads to a defect in the Cvt pathway, whereas bulk autophagy is normal; a similar phenotype was also observed with the Atg9 mutant. Because Atg9 anterograde transport was also dependent on its interaction with Atg11 (), we wondered whether there was a functional connection between actin and Atg11 in Atg9 cycling and autophagic processes. Previous data show that Atg11 is needed to recruit prApe1 to the PAS (). The -null mutant does not affect the recruitment of prApe1 to the PAS or the localization of Atg11 (unpublished data), suggesting that Atg11 localization at the PAS is not dependent on the Atg9–Atg11 interaction. Thus, we propose that Atg11 mediates actin-dependent Atg9 cycling in the Cvt pathway.
To test this hypothesis, we used an actin mutant, , which has defects in actin cable depolymerization, Atg9 cycling, and the Cvt pathway (). As shown in , in wild-type cells, GFP-Atg11 localized to the PAS as a single punctum; chromosomally tagged Atg9-RFP displayed a multiple punctate localization pattern. In the mutant, however, GFP-Atg11 redistributed to cytoplasmic patches, which partially colocalized with Atg9-RFP. To reveal the identity of these multiple compartments, we stained cells with the mitochondrial dye MitoFluor red and the vacuolar membrane dye FM 4-64. As shown in , in 81% (81/100) of the cells, the GFP-Atg11 patches were absent from the perivacuolar PAS position; instead, they colocalized with the mitochondrial labeling, indicating that Atg11 localized at mitochondria when the actin network was defective. Therefore, the proper localization of Atg11 to the PAS is dependent on the actin cytoskeleton and may underlie the actin-dependent Atg9 cycling during the Cvt pathway and other types of specific autophagy.
Atg9 is the only characterized transmembrane protein involved in the formation of the sequestering vesicles that form during the Cvt pathway, pexophagy, and autophagy. Accordingly, it is the best candidate to mark the source of the vesicle membrane. Recently, we have shown that Atg9 localizes to mitochondria in addition to the PAS, implicating this organelle in supplying membrane during autophagy-related processes (). Only Atg19, which is a receptor for biosynthetic cargos, and Atg8 remain associated with the completed vesicles; most of the soluble Atg proteins involved in vesicle formation presumably dissociate from the membrane before or upon vesicle completion. In contrast, a specific retrieval mechanism operates in the cycling of Atg9 and the associated protein Atg23 (). The retrograde movement of Atg9 from the PAS to mitochondria requires Atg1–Atg13, Atg2, Atg18, and the PtdIns3–kinase complex. The transit of Atg9 to the PAS involves Atg23, Atg27, and actin (; ; and unpublished data); however, the mechanism by which actin mediates Atg9 movement is not known.
To identify other Atg components involved in the anterograde movement of Atg9, we performed a yeast two-hybrid screen for proteins that interact with Atg9 and identified Atg11 (). This result is intriguing because Atg11 has previously been shown to interact with Atg19 and Atg1, which are components involved in distinct steps of a specific autophagic process. In conjunction with the present result, we propose that Atg11 acts as a scaffold to coordinate the delivery of multiple components, including the cargo–receptor complex, components involved in vesicle formation, and proteins involved in supplying membrane, to the site of vesicle formation, the PAS (). The Atg11 CC domain2 interacts with the N terminus of Atg9 (), and the interaction occurs in the absence of Atg1 or Atg19 (), suggesting that there are distinct and multiple populations of Atg11 within the cell. Atg11 self-interaction () may then allow these various populations of Atg proteins to be delivered to the PAS in a coordinated manner. Consistent with this model, the overexpression of Atg11 restricted Atg9 to the PAS, presumably as a result of enhanced delivery (). In contrast, a mutation of H192L that disrupts interaction between Atg9 and the specific autophagy component Atg11 resulted in a defect in transporting Atg9 to the PAS (). Furthermore, the absence of Atg9 at the PAS caused by this point mutation led to a block in the Cvt pathway ().
It is known that Atg11 is needed for specific types of autophagy such as the Cvt pathway but is not essential for nonspecific autophagy (). We found that the Atg9 point mutant that disrupts the interaction with Atg11 imposed little effect on the bulk autophagy induced during starvation (). The essentially normal autophagy function was correlated with the normal localization/transport of the binding-defective Atg9 mutant in the presence of the autophagy inducer rapamycin (). This finding suggests that the anterograde transport of Atg9 during bulk autophagy may be mediated by a different mechanism that is at least relatively independent of Atg11. Thus, other proteins that may interact with Atg9 deserve further investigation to reveal the Atg9 cycling machinery that operates during bulk autophagy; however, we cannot rule out the possibility that a low level of interaction between Atg9 and Atg11 allows the anterograde transport of Atg9 after autophagy induction.
Finally, we found that actin is required for the localization of Atg11 to the PAS (). This observation, coupled with the role of Atg11 in Atg9 anterograde movement but not vice versa, suggests that Atg11 mediates the connection between actin and Atg9 delivery from the mitochondria to the PAS. Actin is not needed for bulk autophagy in yeast (), which is in agreement with our findings in the present paper that the Atg9 mutant is not defective for nonspecific autophagy. It is not known how Atg9 might move along actin cables. The third Atg11 CC domain displays some similarity with that of Myo2; however, Atg11 lacks the N-terminal motor domain that functions in Myo2 movement (). Thus, it is not clear how Atg11 might actually mediate the anterograde movement of Atg9. Continued analysis of Atg9 cycling and its interactions with Atg11 and other Atg proteins may provide insight into the underlying mechanisms of membrane delivery during Cvt vesicle and autophagosome formation.
The strains used in this study are listed in . For gene disruption, the entire coding region was replaced by the , the or , the , or the gene using PCR primers containing ∼50 bases of identity to the regions flanking the open reading frame. For PCR-based integrations of the PA and GFP tags at the 3′ end of the and genes, pHAB102 and pFA6a-GFP-HIS3 were used as templates to generate strains expressing fusion proteins under the control of their native promoters (Longtine et al., 1998; ).
Plasmids expressing Atg11 truncations (), GFP-Atg8 (pGFP-AUT7; ), Atg9 (pAPG9; essentially constructed the same as pAPG9; ), Atg9-GFP (pAPG9GFP; ), and CFP-Atg11 (pCuHACFPCVT9; ) have been described previously. BFP-Ape1 (pBFPApe1) was constructed by introducing BFP (Qbiogene) on a BamHI PCR-generated fragment into a unique BglII site situated after the start codon of (). GFP-Atg11 (pTS495) was constructed in a similar manner starting with cloned into the pRS416 vector. PA-Atg9 (pCuPAAtg9) was generated by cloning into the pRS416-CuProtA vector (). Full-length Atg9 (pBD-Atg9) and the Atg9 N-terminal (pBD-Atg9N), C-terminal (pBD-Atg9C), and transmembrane (pBD-Atg9TM) regions were amplified by PCR from pAPG9 and introduced into the two-hybrid vector pGBDU-C1 using BamHI and SalI sites. A total of 19 other genes were amplified and cloned into the two-hybrid vector pGAD-C1 using BamHI and SalI sites. The serial deletion mutants of the Atg9 N terminus were generated by PCR amplification from pBD-Atg9N and ligated into the EcoRI and BamHI sites of pGBDU-C1. The plasmid expressing the Atg9 mutant (pAtg9) was constructed by releasing the fragment containing H192L from pAtg9GFP (see Random mutagenesis screen…mutant) and was cloned into pAPG9 using SacI and AgeI. pAtg9PA and pAtg9PA were constructed by PCR amplifying and together with its native promoter from pAPG9GFP and pAtg9GFP and incorporating them into pNopPA using XhoI and XmaI sites. pNopPA was a gift from F. Reggiori (University of Utrecht, Utrecht, Netherlands).
Cells were grown to OD = 0.8 in SMD, and 100 ml were harvested and resuspended in lysis buffer (PBS, 200 mM sorbitol, 1 mM MgCl, 0.1% Tween 20, 1 mM PMSF, and protease inhibitor cocktail). The detergent extracts were incubated with IgG–Sepharose beads overnight at 4°C. The beads were washed with lysis buffer eight times and eluted in SDS-PAGE sample buffer by incubating at 55°C for 15 min. The eluates were resolved by SDS-PAGE and immunoblotted with anti-YFP antibody.
The gap repair PCR mutagenesis method () was used to generate the Atg11 binding–defective Atg9 mutant. An AvrII site at 942 bp was introduced into the pAPG9GFP plasmid by site-directed mutagenesis. The Atg9 N-terminal region containing the AvrII site was amplified and introduced into pGBDU-C1 with EcoRI and BamHI to generate pBD-Atg9N(AvrII). A gapped pGBDU-C1 plasmid was generated by digesting with EcoRI and BamHI. The PCR reaction was performed using Taq polymerase (New England Biolabs, Inc.) with pBD-Atg9N(AvrII) as the template. The dATP concentration was lowered to three fifths that of the other three deoxynucleoside triphosphates. The resulting mutagenized PCR product shares overlapping sequences of ∼100 bp at both 5′ and 3′ ends with the gapped pGBDU-C1 plasmid. The PCR product and the gapped plasmid were cotransformed into the two-hybrid strain PJ69-4A. The transformants were replicated on both −His and −Ade plates. The transformants that were not able to grow on either plate were selected and analyzed by Western blotting for protein stability using anti-Atg9 antiserum (). Stable mutants were sequenced, and pBD-Atg9N was identified. The Atg9 N-terminal fragment containing the H192L mutation was then released from pBD-Atg9N by digestion at the introduced AvrII site and a natural NruI site. This fragment was cloned into pAPG9GFP digested with the same restriction enzymes to generate pAtg9GFP.
Cells expressing fusion proteins with fluorescence tags were grown in SMD media to OD = 0.8. For vacuolar membrane labeling, cells were pelleted, resuspended in fresh media at OD = 1.0, incubated with 2 μM FM 4-64 at 30°C for 15 min, and pelleted and cultured in the same media without FM 4-64 at 30°C for 30 min. For mitochondrial fluorescent labeling, MitoFluor red 589 (Invitrogen) was added to the growing culture at a final concentration of 1 μM, and the culture was incubated at 30°C for 30 min. Cells were then washed with the same culture medium before imaging to remove the excess dye. For rapamycin treatment, cells were cultured with 0.2 μg/ml rapamycin at 30°C for 1.5 h. When necessary, a mild fixation procedure was applied to visualize Atg9 without destroying various fluorescent proteins: cells were harvested, resuspended in a half volume of fixation buffer (50 mM KHPO, pH 8.0, 1.5% formaldehyde, and 1 μM MgCl), and incubated at room temperature for 30 min with gentle shaking. Cells were then washed once with an equal amount of wash buffer (50 mM potassium phosphate, pH 8.0, and 1 μM MgCl) and resuspended in 50 μl of wash buffer. Fluorescence signals were visualized on a fluorescence microscope (IX71; Olympus). The images were captured by a camera (Photometrics CoolSNAP HQ; Roper Scientific) and deconvolved using DeltaVision software (Applied Precision).
The GFP-Atg8 processing assay, the Pho8Δ60 activity assay, and the protease protection assay were performed as previously described (; ; ; ; ). |
The signal transducer and activator of transcription (STAT) family consists of seven members (STAT1–4, -5A, -5B, and -6). STATs are phosphorylated by cytokine stimulation, form homo- or heterodimers, and enter the nucleus, where they regulate expression of their target genes (; ). Although STATs have a variety of functions under physiological conditions, the pathological importance of STAT functions has also been reported in many studies. STAT3 and -5 were activated in a broad spectrum of human hematological malignancies as well as in solid tumors (). A constitutively active form of STAT5 and -3 transformed IL-3–dependent Ba/F3 cells and fibroblasts, respectively (; ; ). An internal tandem duplication (ITD) mutant of receptor tyrosine kinase Flt3 (ITD-Flt3), a causative mutation of acute myeloid leukemia (; ), induced phosphorylation of STAT5 on its tyrosine residues, thereby playing critical roles in cell transformation (; ; ).
The mechanisms by which STATs are phosphorylated by cytokines and the activated STATs regulate the expression of the target genes have been well characterized. How activated STATs are transported to the nucleus has also been investigated; activated STAT1 and -3 were reported to bind importin α5 and several importin αs, respectively, which mediated the nuclear transport of STATs (; ; ; ; ). However, molecules other than importins could also participate in the regulation of the nuclear translocation of STATs.
We have recently described the interactions among STAT3, Rac1, and a Rac/Cdc42 GTPase-activating protein (GAP), MgcRacGAP (male germ cell Rac-GAP), and have shown that MgcRacGAP is required for transcriptional activation of STAT3 (). However, the mechanisms by which Rac and MgcRacGAP regulate transcriptional activation of STAT3 remained unclear. In the present work, we investigated the molecular mechanisms of nuclear transport of a tyrosine-phosphorylated form of STAT5A, a close relative of STAT3, and found that GTP-bound Rac1 and MgcRacGAP were required for transport of activated STATs to the nucleus, indicating a novel function of Rac1 GTPase.
To test whether Rac1 and MgcRacGAP bind STAT5A, as was the case for STAT3 (), we did coimmunoprecipitation. STAT5A and MgcRacGAP were coimmunoprecipitated with Rac1 () and Rac2 in Ba/F3 cells (unpublished data). In addition, STAT5A was coimmunoprecipitated with MgcRacGAP in Ba/F3 cells and in several other human and mouse cell lines, as well as in human primary T cells (unpublished data). These data show that Rac, STAT5A, and MgcRacGAP form a complex in vivo.
A considerable amount of STAT5A protein was coimmunoprecipitated with MgcRacGAP in IL-3–starved Ba/F3 cells, and this association was enhanced by IL-3 stimulation (, left). Vice versa, a small amount of MgcRacGAP protein was coimmunoprecipitated with STAT5A in the starved cells, and this association was enhanced by IL-3 (, middle). In Ba/F3 cells expressing a constitutively active form of STAT5A (CA-STAT5A), which is more stable in the phosphorylated form than the wild-type STAT5A (), a considerable amount of STAT5A protein bound MgcRacGAP, even in unstimulated cells. This binding was also enhanced by IL-3 (). Thus, the association between MgcRacGAP and STAT5A does not require phosphorylation of STAT5A, but is enhanced by phosphorylation.
To map the interacting domains between MgcRacGAP and STAT5A, we prepared a series of truncated mutants of MgcRacGAP and STAT5A fused with maltose binding protein (MBP; Fig. S1, a, b, d, and e, available at ). It was found that STAT5A and Rac1 interacted with the Cys-rich and GAP domains of MgcRacGAP, whereas MgcRacGAP interacted with the DNA-binding domain (DBD) of STAT5A (Fig. S1, c and f). The binding domains between STAT5A and MgcRacGAP were similar to those between STAT3 and MgcRacGAP ().
We next investigated the stoichiometry of STAT5A/MgcRacGAP binding in the cytoplasm or nucleus. IL-3–starved Ba/F3 cells were stimulated with IL-3 for 0, 15, or 90 min, and the cell lysates were fractionated. The cytosol and nuclear fractions were then immunodepleted with the anti-MgcRacGAP or anti-STAT5A antibody. The amounts of total STAT5A and tyrosine-phosphorylated STAT5A (p-STAT5A) in the nuclear fraction increased 15 min after IL-3 stimulation and decreased 90 min after IL-3 stimulation (, lanes for the control antibody). Notably, most of p-STAT5A in the cytosolic fractions was immunodepleted with the anti-MgcRacGAP antibody as well as with the anti-STAT5A antibody (). On the other hand, a considerable part of p-STAT5A was left in the nuclear extracts of IL-3–stimulated cells after the immunodepletion with the anti-MgcRacGAP antibody (). These results suggested that most of p-STAT5A was bound by MgcRacGAP in the cytoplasm of IL-3–stimulated cells and was released from MgcRacGAP in the nucleus.
The amount of cytoplasmic STAT5A immunoprecipitated with the anti-MgcRacGAP antibody gradually increased after IL-3 stimulation (), and concomitantly the amount of cytoplasmic STAT5A immunodepleted with the anti-MgcRacGAP antibody gradually decreased (), implicating that MgcRacGAP maintained interaction with STAT5A in the cytoplasm of IL-3–stimulated cells even after the dephosphorylation of STAT5A. The fractionation was confirmed by Western blotting with the anti-HDAC (for nuclear fraction) or RhoA (for cytosol fraction) antibody (unpublished data).
Next, we visualized STAT5A and MgcRacGAP by immunostaining using adherent 293T cells. To enhance phosphorylation and nuclear translocation of STAT5, we used a constitutively active tyrosine kinase receptor, ITD-Flt3 (). In the absence of ITD-Flt3, ectopically expressed STAT5A-Flag localized to the cytoplasm and colocalized in part with the endogenous MgcRacGAP. Expression of ITD-Flt3 led to translocation and colocalization of STAT5A and MgcRacGAP in the nucleus (). These results indicated that MgcRacGAP translocated to the nucleus concurrently with STAT5A in response to IL-3 and ITD-Flt3 stimulation. Intriguingly, a dominant-negative form of Rac1, N17Rac1, completely inhibited the ITD-Flt3–induced nuclear translocation of STAT5A (Fig. S2, available at ). This result suggested that the GTP-bound form of Rac1 was required for the nuclear accumulation of activated STAT5A. However, N17Rac1 was recently reported to inhibit not only Rac1 but also other Rho-GTPases (). To confirm that the N17Rac1 inhibition of nuclear translocation of p-STAT5A was indeed due to the inhibition of Rac1, we used mouse embryonic fibroblasts derived from gene-targeted conditional Rac1-flox mice in the Rac2-null background ().
ITD-Flt3 induced the nuclear localization of STAT5A-Flag in the presence of Rac1 (). However, when Rac1 was depleted by Cre recombinase in Rac2Rac1 fibroblasts (), the nuclear translocation of STAT5A was severely impaired (, c and d), and even p-STAT5 mostly remained in the cytoplasm (, e and f). In addition, CA-STAT5A did not enter the nucleus in the absence of Rac1 (unpublished data). We performed a similar analysis using Rac1Rac2 fibroblasts and obtained identical results. These results demonstrate that Rac1 plays an essential role in the nuclear translocation of p-STAT5A.
We next used siRNA to knock down Rac1 or MgcRacGAP expression in Ba/F3 cells where STAT5 activation is required for cell growth. The siRNA treatment for Rac1 or MgcRacGAP resulted in severe growth retardation of Ba/F3 cells and caused apoptosis in some cells. The total cell number was only one tenth or one fifth 48 h after siRNA treatment for Rac1 or MgcRacGAP, respectively (unpublished data). The siRNA treatment for MgcRacGAP led to the formation of multinucleated cells, as reported previously (), but no more than 20% of the cells, indicating the failure of cytokinesis by MgcRacGAP depletion is not the major cause of the growth inhibition.
We then did semiquantitative RT-PCR analysis to test whether transcriptional activation of STAT5 is affected by the knock down of Rac1 or MgcRacGAP and found that expression of bcl-xL, one of the STAT5 target genes, was severely impaired by the siRNA treatment (). We also confirmed that siRNA treatments specifically decreased the expression levels of Rac1 or MgcRacGAP protein but not those of RhoA and HDAC, similar to the results shown in (not depicted).
We also investigated whether knock down of Rac1 or MgcRacGAP affects the subcellular distribution of STAT5A and p-STAT5A in Ba/F3 cells before and after IL-3 stimulation. The siRNA-treated Ba/F3 cells were starved for 6 h after the isolation of live cells using Ficoll and stimulated with IL-3 (15 min), and the cell lysates were fractionated. The siRNA treatments specifically decreased expression levels of Rac1 or MgcRacGAP protein () but not those of RhoA and HDAC (). The IL-3–induced nuclear accumulation of STAT5A and p-STAT5A was almost completely blocked in Ba/F3 cells treated with either Rac1 or MgcRacGAP siRNA when compared with those treated with the control siRNA (). The same treatment moderately decreased the amounts of p-STAT5A and total STAT5A in the cytoplasmic fraction (), suggesting that Rac1 and MgcRacGAP enhance the IL-3–induced phosphorylation of STAT5A and somehow stabilize STAT5A in the cytoplasm.
We previously found that STAT3 bound MgcRacGAP through its DBD (). To examine whether MgcRacGAP regulated transcriptional activity of STAT3 and -5A through direct interaction, we attempted to produce mutant STATs lacking the MgcRacGAP binding site. To this end, we narrowed down the binding site in DBD-STAT3 to a 25-amino-acid stretch, using MBP-fused DBD-STAT3 truncations (DB1-DB6; Fig. S3 A, available at ). We found that only DB2 (aa 338–362) of DBD-STAT3 interacted with MgcRacGAP (Fig. S3 B). Conversely, the mutant of DBD-STAT3 lacking DB2 (DBD-STAT3-dDB2) did not bind MgcRacGAP (Fig. S3 C). These results clearly demonstrated that the DB2 region (25 amino acid) of STAT3 bound MgcRacGAP. This region is well conserved among STAT family proteins and harbors a β-sheet structure, which is thought to mediate protein–protein interaction. Purified MgcRacGAP was pulled down by the MBP-DB2 of STAT3, and the corresponding region of STAT5 (aa 341–365) fused with MBP but not by MBP alone, demonstrating that MgcRacGAP directly bound DB2 of STAT3 and -5 (Fig. S3 D and ). Both of the STAT3 and -5A mutants lacking DB2 (STAT3- and STAT5A-dDB2) did not bind MgcRacGAP and the extent of tyrosine phosphorylation of these mutants was less prominent after IL-6 or ITD-Flt3 stimulation ( and not depicted). In addition, STAT3- and STAT5A-dDB2 lacked their transcriptional activities (Fig. S3 E and ). These results indicated that the interaction of MgcRacGAP/Rac1 with STAT3 and -5A facilitates cytokine receptor–induced tyrosine phosphorylation of both STAT3 and -5A. Considerable decrease in the tyrosine phosphorylation of STAT5A was also observed when Rac1 or MgcRacGAP was knocked down (). Intriguingly, MgcRacGAP also interacted with JAK2 (), suggesting that MgcRacGAP/Rac1 also mediated the tyrosine phosphorylation of STATs through the interaction with JAK2. Importantly, STAT3- and STAT5A-dDB2 that do not bind MgcRacGAP did not enter the nucleus even after tyrosine phosphorylation by IL-6 or ITD-Flt3 ( and not depicted), suggesting that MgcRacGAP/Rac1 is required not only for nuclear translocation of p-STATs but also for efficient tyrosine phosphorylation of STATs.
We established a nuclear transport assay using semi-intact permeabilized cells (), which enables us to biochemically analyze the roles of Rac1 and MgcRacGAP in the nuclear import of p-STAT5A. We confirmed the purities of STAT5A, MgcRacGAP, V12Rac1, N17Rac1, importin α1, importin α5, importin β1, Ran, and NTF2 produced by Sf-9 cells, and the tyrosine phosphorylation of STAT5A induced by coexpression with the kinase domain of JAK2 in Sf-9 cells (). It was confirmed that the purified p-STAT5A bound DNA in electrophoretic mobility shift analysis (EMSA) in a similar fashion with GM-CSF–activated STAT5 in TF-1 cells (Fig. S4 A, available at ), indicating that the recombinant in vivo phosphorylated STAT5A formed a proper dimer. Permeabilized HeLa cells were incubated with the indicated combinations of purified proteins in transport buffer (TB) plus an energy regenerating system. After the import reaction in the cells incubated with purified unphosphorylated STAT5A, a considerable amount of unphosphorylated STAT5A was detected at the cytoplasm in most cells (). The addition of purified MgcRacGAP, V12Rac1, importin α1, and importin β1 did not affect localization of unphosphorylated STAT5A (). Although rabbit reticulocyte lysate reduced cytoplasmic localization of unphosphorylated STAT5A (), it induced both the nuclear and plasma membrane localization of p-STAT5A (). These results suggested that rabbit reticulocyte lysate contained cofactors that are required for the nuclear translocation of p-STAT5A in this transport assay. Interestingly, p-STAT5A accumulated at the nuclear membrane, with some migrating into the nucleus in the presence of purified MgcRacGAP and V12Rac1, but the nuclear translocation of p-STAT5A was inhibited in the presence of purified MgcRacGAP and N17Rac1 (). These results indicate that the GTP-bound form of Rac1 and MgcRacGAP facilitate the nuclear translocation of p-STAT5A. Given that purified importin β1 also accumulated mostly in the nuclear envelope and only partially migrated to the nucleus in our assay system (unpublished data) as reported previously (), the accumulation of p-STAT5 and importin β1 in the nucleus might have been caused by residual amounts of nuclear transporters left in the assay system. Thus, it is likely that the GTP-bound form of Rac1 and MgcRacGAP play critical roles in targeting p-STAT5A to the nuclear envelope and that cofactors are required for the efficient nuclear import of p-STAT5A from the nuclear envelope. In fact, nuclear translocation of p-STAT5A was enhanced by further addition of the purified nuclear transporters, including importin α1, importin β1, Ran, and NTF2 to the assay (). This nuclear translocation of p-STAT5A was not observed in the absence of MgcRacGAP even in the presence of cofactors ().
To confirm whether the unphosphorylated recombinant STAT5A conserved a native folded state, we did nuclear transport assay using the in vitro phosphorylated STAT5A. The recombinant full-length JAK2 efficiently phosphorylated the recombinant STAT5A in the kinase reaction buffer (Fig. S4 B). This in vitro phosphorylated STAT5A behaved in the nuclear transport assay like the in vivo phosphorylated STAT5A (Fig. S5, a–i, available at ). The nuclear transport of p-STAT5A requires both MgcRacGAP and V12Rac1. The nuclear import of the in vitro phosphorylated recombinant STAT5A was also achieved by the presence of the cytosol fraction of HeLa cells (HeLa-CS), which had been prepared as described previously (). In addition, immunodepletion of MgcRacGAP or Rac1 considerably inhibited the nuclear import of the in vitro phosphorylated recombinant STAT5A (Fig. S5, j–m). This inhibition was restored by add-back of the purified recombinant MgcRacGAP or Rac1 (Fig. S5, n and o).
To determine whether the Rac1 activation or the presence of MgcRacGAP is required for the interaction of p-STAT5A with importin αs, an in vitro binding assay was done using purified proteins. Intriguingly, p-STAT5A formed complexes with importin α1 and α5 only in the presence of both MgcRacGAP and V12Rac1 or another constitutively active mutant L61Rac1, but not N17Rac1 (). These results demonstrated that GTP-bound Rac1 and MgcRacGAP functions as p-STAT5A nuclear chaperone, facilitating p-STAT5A to form protein complexes with importin αs.
In the present work, we demonstrate that Rac1 and MgcRacGAP are essential for the nuclear translocation of STAT5A, based on the following observations. First, Rac1 and MgcRacGAP directly bound STAT5A, and the interaction between MgcRacGAP and STAT5A was enhanced by IL-3 stimulation. Second, STAT5A and MgcRacGAP simultaneously entered the nucleus upon IL-3 and ITD-Flt3 stimulation. Third, knock down of Rac1 or MgcRacGAP profoundly inhibited both the IL-3–induced transcriptional activation of STAT5A and the nuclear accumulation of p-STAT5A in IL-3–dependent Ba/F3 cells. Fourth, depletion of Rac1 in fibroblasts, as well as expression of N17Rac1 in 293T cells, prevented p-STAT5A from entering the nucleus. Fifth, p-STAT5A lacking the MgcRacGAP binding site (p-STAT5A–dDB2) did not accumulate in the nucleus. Last, in a nuclear transport assay, purified V12Rac1 and MgcRacGAP induced accumulation of purified p-STAT5A on the nuclear envelope, with some p-STAT5A migrating into the nucleus, and the further addition of nuclear transporters, including importin α1, importin β1, Ran, and NTF2, achieved the efficient nuclear translocation of p-STAT5A. Moreover, either the absence of MgcRacGAP or the presence of N17Rac1 inhibited this nuclear translocation of p-STAT5A.
suggested that an active form but not an inactive form of Rac1 bound STAT3 and played important roles in EGF-induced STAT3 activation. These authors did not, however, specifically examine the nuclear transport of STAT3. Interestingly, EGF receptor–mediated endocytosis is required for cytoplasmic transport of STAT3 (), and MgcRacGAP is recruited to the EGF receptor complex after EGF stimulation (). We also found that STAT3 bound Rac1 and Rac2, which was enhanced by IL-6 stimulation. In addition, STAT3 bound MgcRacGAP, which was required for the transcriptional activation of STAT3, and some population of MgcRacGAP entered the nucleus together with STAT3 (). Although these results suggested a role of Rac1/MgcRacGAP in STAT3 activation, the underlying molecular mechanisms remained elusive. We studied the functional interactions using a nuclear transport assay and found that the nuclear translocation of p-STAT3 as well as p-STAT5A was induced in the presence of a combination of purified proteins, including V12Rac1, MgcRacGAP, importin α1, importin β1, Ran, and NTF2 (, Fig. S5, and not depicted). These results demonstrate a novel Rac1 function in the nuclear transport of p-STAT3 as well as p-STAT5A.
Although we showed the results for STAT5A, we obtained identical results in experiments so far performed for closely related STAT5B (unpublished data). In addition, the phenotypes of STAT3- and STAT5A-dDB2 were nearly identical ( and Fig. S3), and the region of STAT3 that binds to MgcRacGAP (STAT3-DBD-DB2) is well conserved among STAT family proteins, suggesting a general role for MgcRacGAP and Rac1 in the nuclear transport of p-STAT proteins.
Ba/F3 cells were maintained in RPMI1640 medium (Invitrogen) containing 10% FCS and 1 ng/ml mIL-3 (R&D Systems). An ecotropic retrovirus packaging cell line PLAT-E was maintained as described previously (). An anti-STAT5A antibody and anti-STAT5B antibody were obtained from R&D Systems. Affinity-purified anti-MgcRacGAP antibody was produced as described previously (). An anti-Rac1 mAb and anti–importin α1 mAb were purchased from BD Biosciences. The rabbit polyclonal anti-Rac1, anti-RhoA, anti-JAK2, and goat polyclonal anti-HDAC or anti–importin α5 antibodies were obtained from Santa Cruz Biotechnology, Inc.
Immunoprecipitation, gel electrophoresis, and immunoblotting were done as described previously (), with minor modifications. Cell lysates (2 × 10 cells/ml) were incubated at 4°C for 2 h with the indicated antibodies and protein A–Sepharose. The immunoprecipitates were subjected to Western blot analysis with an anti–p-STAT5 mAb (Upstate Biotechnology), anti-MgcRacGAP, or anti-STAT5A antibody. The loading amounts were verified with the anti-STAT5A or anti-MgcRacGAP antibody after stripping the filters. The filter-bound antibody was detected using the ECL system (GE Healthcare). Cytosol and nuclear fractions were prepared as described previously ().
MBP fusion proteins (0.5 μg) bound to amylose resin beads were incubated with cell lysates (10 μg) from IL-3–stimulated Ba/F3 cells as described previously ().
The 293T cells were transfected with 1.0 μg pME/STAT5A-Flag together with 0.5 μg pMKIT (MOCK) or pMKIT/ITD-Flt3, and in some experiments cells were transfected with 0.5 μg pME/STAT5A-HA and 0.5 μg pMKIT (MOCK) or pMKIT/ITD-Flt3 together with 1.0 μg pCMV5/N17Rac1-Flag, using Lipofectamine Plus reagents (Life Technologies). After 24 h, cells were plated on glass coverslips, and the next day the cells were immunostained as described previously ().
Fluorescence images were analyzed on a confocal microscope (Fluoview FV300 Scanning Laser Biological Microscope IX 70 system; Olympus) equipped with two lasers (Ar 488 and HeNe 543) using a 60× oil objective (PlanApo; Olympus). Fluoview version 4.3 software (Olympus) was used for image acquisition from confocal microscopy. Photoshop 7.0 or Photoshop Elements 2.0 software (Adobe) was used for processing of images.
For the silencing of Rac1 or MgcRacGAP, SMARTpool Rac1 or MgcRacGAP siRNA (L-041170 or ; Dharmacon) was used. A control siRNA was used as a nonsilencing control (). 5 μl of 40 μM double-stranded siRNA were introduced in to 2 × 10 cells of Ba/3F cells with Nucleofector II (Amaxa) set at program T-16 using a Cell Line Nucleofector kit V (Amaxa) according to the manufacturer's instruction. A control vector carrying GFP was introduced to >80% of Ba/3F cells under this condition. 24 h after transfection, live cells were isolated using Ficoll-Paque PLUS (GE Healthcare), and gene expression was examined by semiquantitative RT-PCR analysis as described previously (). The primers used are as follows: 5′bcl-x, 5′-GAAAGAATTCACCATGTCTCAGAGCAACCGG-3′; 3′bcl-x, 5′-GAAAGCGGCCGCTCACTTCCGACTGAAGAGTG-3′; 5′GAPDH, 5′-ACCACAGTCCATGCCATCAC-3′; 3′GAPDH, 5′-TCCACCACCCTGTTGCTGTA-3′.
High-titer retroviruses harboring Cre recombinase were produced in a transient retrovirus packaging cell line PLAT-E () and were used to deplete Rac1 in Rac2Rac1 fibroblasts ().
To construct baculovirus vectors, the cDNAs encoding STAT5A, MgcRacGAP, V12Rac1, L61Rac1, N17Rac1, importin αs, importin β1, Ran, and NTF2 with the C-terminal Flag epitope tag, and a kinase domain of JAK2 (JH1; ) were subcloned into pBacPAK (BD Biosciences). The resulting constructs were used to obtain recombinant baculoviruses by cotransfection with Bsu36 I–digested BacPAK viral DNA (BD Biosciences) into Sf-9 cells according to the manufacturer's protocol. For protein expression, Sf-9 cells were infected with high-titer viral stocks for 96 h and lysed. The lysate was clarified by centrifugation, and the supernatant was immunoprecipitated with the anti-Flag M2-agarose affinity gel (Sigma-Aldrich) for 2 h at 4 °C. The recombinant Flag-tagged proteins were eluted with 3× Flag peptide (Sigma-Aldrich).
To determine whether purified p-STAT5A formed a proper dimer, EMSA was performed using consensus sequence of STAT5A as a probe, as described previously ().
An in vitro kinase reaction of purified STAT5A was performed as described previously with minor modifications (). In vitro phosphorylated STAT5A was immunoprecipitated with the anti-Flag M2-agarose affinity gel and reeluted with a 3× Flag peptide. The purified in vitro phosphorylated STAT5A was dialyzed against TB, and the final concentrations of STAT5A protein were determined for use in SDS-PAGE and in the nuclear transport assay.
FITC-labeled BSA (Sigma-Aldrich) conjugated with a synthetic peptide containing the SV40 large T antigen (CGGGPKKKRKVED; NLS-conjugated FITC-BSA) was prepared as described previously (), as a control protein harboring an NLS. We confirmed that NLS-conjugated FITC-BSA was imported to the nucleus in our experimental conditions as reported previously (), which was not inhibited by immunodepletion of MgcRacGAP or Rac1 (Fig. S5, p–t).
HeLa cells were grown on poly--lysine–coated coverslips and permeabilized with 40 μg/ml digitonin (Roche) in TB (20 mM Hepes, pH 7.3, 110 mM KOAC, 2 mM Mg(OAC), 1 mM EGTA, 2 mM DTT, 0.4 mM PMSF, 3 μg/ml aprotinin, 2 μg/ml pepstatin A, 1 μg/ml leupeptin, and 20 mg/ml BSA) for 10 min at RT. Subsequently, the cells were washed twice in TB. Incubation with 50 μl import mix was performed at 37°C for 30 min. The import mix contained TB, an energy regenerating system (0.5 mM ATP, 0.5 mM GTP, 10 mM creatine phosphate, and 30 U/ml creatine phosphokinase), and 1 μM of purified unphosphorylated or phosphorylated STAT5A alone, or STAT5A plus the 1 μM of other purified cofactor proteins as indicated in . After the import reaction, the cells were washed with ice-cold TB and immunostained with the anti-STAT5A antibody and anti–p-STAT5 mAb as described previously (). Fixed cells were examined using a Fluoview FV300 confocal microscope (Olympus).
Fig. S1 depicts the binding domains of MgcRacGAP with STAT5A and that of STAT5A with MgcRacGAP. Fig. S2 shows that N17Rac1 expression inhibits the nuclear translocation of p-STAT5A. Fig. S3 shows that the DB2 region is required for the interaction of STAT3 with MgcRacGAP and the transcriptional activation of STAT3. Fig. S4 shows that purified p-STAT5A forms a dimer and binds DNA containing the STAT5 consensus sequence and that the purified STAT5A can be phosphorylated in vitro. Fig. S5 shows that the in vitro phosphorylated recombinant STAT5A can be imported to the nucleus in the nuclear transport assay and that immunodepletion of Rac1 or MgcRacGAP specifically inhibits the nuclear import of p-STAT5A using HeLa cytosol extract. Online supplemental material is available at . |
Migrating cells form a dynamic, thin, veil-like structure called lamellipodia at the leading edge. The dendritic nucleation model describes the reorganization processes of actin arrays in lamellipodia (; ). Within the lamellipodia, actin filaments display a branched network (). The Arp2/3 complex nucleates actin filaments off the sides of preexisting filaments (; ; ; ). Nucleated filaments elongate toward the leading edge (), and their growth is terminated by capping protein (CP; ; ). Actin then depolymerizes to replenish the monomeric actin pool for the next round of polymerization. Quantitative modeling of the dendritic nucleation actin array is a logical goal in cell migration research, but it will require precise knowledge on the mechanisms governing actin filament turnover in vivo.
Our previous study using single-molecule speckle microscopy investigated actin filament lifetime distribution in lamellipodia of spreading fibroblasts (). The filament lifetime spanned a wide range of time with a maximum of 148 s. Notably, one third of the filaments had a short lifetime of <10 s. This raised the question of what mechanisms may account for the observed lifetime of actin filaments.
In this study, we further extend our single-molecule speckle analysis to major actin end-binding proteins to elucidate the filament turnover mechanism. Single-molecule observations allow us to precisely measure the dissociation kinetics of a molecule in cells given that the molecule binds cellular structures on the order of seconds. Surprisingly, the dissociation of CP from actin was found to occur three orders of magnitude faster in cells than in vitro and even much faster than the actin disassembly rate. We demonstrate that the CP dissociation rate is prolonged under the several conditions in which actin filaments are stabilized. Our data indicate that cofilin-mediated actin disassembly is required for the fast CP dissociation in lamellipodia. Based on the marked difference in the dissociation rate between actin (0.03 s) and CP (0.58 s), we predict that fairly frequent filament severing and end- to-end annealing might take place in the dendritic nucleation actin arrays.
CP has been attributed to most of the actin capping activity in cell lysates (; ) and has been found as an essential protein in the regulation of lamellipodium morphology (). Therefore, we analyzed the dynamics of CP to elucidate the state of barbed ends in the dendritic nucleation network. We generated three expression constructs by tagging either the α or β subunit of CP with an EGFP. Of the three, two probes, EGFP-CPβ1 and CPβ1-EGFP, distributed to lamellipodia, which is consistent with previous studies (; ). EGFP-CPβ1 and CPβ1-EGFP speckles moved along with the retrograde actin flow, which suggested their association with actin structures ( and Videos 1 and 2, available at ). We selected these two probes for single-molecule speckle analysis of CP.
We measured the lifetime distribution of single-molecule CP speckles in lamellipodia of XTC cells spreading on poly- -lysine (PLL)–coated glass coverslips. EGFP-CPβ1 and CPβ1-EGFP showed a similar lifetime distribution (). Our measurement revealed that 53–60% of CP dissociated from actin within 1 s. When the CP lifetime distribution was fit with a single exponential curve, the half-life was 1.20 and 1.23 s for EGFP-CPβ1 and CPβ1-EGFP, respectively, which is in marked contrast to the slow dissociation of CP from barbed ends in vitro (
= ∼28 min; ). CP displayed similar lifetime distributions throughout lamellipodia ().
We next examined the spatial location of CP recruitment. We recorded positions of newly appearing CP speckles and measured the distance from the cell edge. Although CP speckles appeared most frequently in the tip region, appearance was widely distributed throughout the lamellipodia (). This bias toward the lamellipodium tip region was weaker than that observed for the Arp2/3 complex (see ).
Given this unexpectedly fast dissociation, we needed to test whether our EGFP-tagged CP probes retained high affinity to barbed ends. We characterized biochemical properties of the recombinant CPα2/EGFP-CPβ1 heterodimer expressed in (). First, we determined the on-rate constant (k) of CPα2/EGFP-CPβ1 using a pyrene-actin polymerization assay (). Low concentrations of CPα2/EGFP-CPβ1 efficiently inhibited barbed end elongation from filamentous actin (F-actin) seeds. The on-rate constant was calculated to be 3.9 μMs, which was comparable with that of native CP (3.5–5.7 μMs; ). Second, we measured the dissociation constant (K) using a pyrene-actin steady-state assay (). The K of CPα2/EGFP-CPβ1 with barbed ends was 1.3 nM, which was comparable with the K of native CP (0.06–1 nM; ; ; ). The K and on-rate constant predict an off-rate constant of 5.1 3 10 s, which is ∼100-fold smaller than the dissociation rate of our CP probes in lamellipodia (). We also noted that recombinant EGFP-CPβ1, which was expressed and purified in the absence of CPα subunits, did not interfere with actin elongation (unpublished data). This is consistent with previous findings (). Therefore, CP speckles observed in cells should correspond to EGFP-CPβ1 coupled with endogenous CPα subunits but not deficient EGFP- CPβ1 monomers.
To further evaluate the fidelity of our CP probes, we permeabilized cells expressing EGFP-CPβ1 using Triton X-100 and followed the decay of EGFP fluorescence. The decay rate of fluorescence intensity was 1.7 × 10 s in lamellipodia (). Cells retained most of the expressed CP speckles during the permeabilization procedure (), eliminating the possibility that our permeabilization procedure selected a specific fraction of CP tightly bound to actin. These results indicate that our CP probes bind barbed ends with a high affinity, similar to that measured for unmodified protein in vitro (; ; ). All together, our CP probes can be used as a reliable marker to monitor the dynamics of CP. We conclude that fast dissociation kinetics of CP speckles represent the dynamics of endogenous CP.
To test whether other barbed end interactors might cap actin tightly, we visualized the molecular dynamics of Eps8, VASP, and gelsolin (Fig. S1 and Videos 3–7, available at ; ; ; ; ; ). None of the three molecules displayed long-term association with the actin network. These results suggest that the majority of actin filaments may release barbed end cappers and revert to the growing phase quickly in lamellipodia.
We next examined the molecular dynamics of the Arp2/3 complex in lamellipodia. We generated expression constructs by tagging each of seven Arp2/3 subunits with EGFP, testing N- and C-terminal fusions in several cases. Among them, four probes— EGFP-p40, p40-EGFP, EGFP-p21, and p21-EGFP—were localized to lamellipodia and cytoplasmic punctate actin structures, which is consistent with the reported Arp2/3 localization (; ; ; ). Using an anti-p40 antibody, we confirmed that distributions of EGFP-p21 and p21-EGFP were identical to that of the endogenous p40 subunit of Arp2/3 (Fig. S2, available at ). We also compared localization of the four probes with each other by generating probes tagged with monomeric red fluorescent protein 1 (mRFP1; ). All four probes showed identical localization (Fig. S2) and moved inward at the same rate as the retrograde actin flow. Based on these observations, we selected these four probes for further analysis.
We analyzed the kinetics of Arp2/3 dissociation from the actin network in lamellipodia. We followed each newly emerged single-molecule Arp2/3 speckle and measured the duration between its appearance and disappearance (; and Video 8, available at ). Arp2/3 interacts with the side of actin filaments with micromolar affinity (; ). Thus, Arp2/3 is expected to dissociate from the side of the filament on a subsecond timescale after leaving the pointed end, and rapidly dissociating side binding would not be detected using our current experimental settings. Therefore, we interpret that the lifetime of Arp2/3 speckles represents the duration of Arp2/3 association with the pointed end, although we do not know whether Arp2/3 dissociation occurs simultaneously with debranching. The lifetime distribution of Arp2/3 speckles could be approximated by a single exponential curve, which suggests that the dissociation of Arp2/3 is governed by a single rate-limiting step. Half-life determined for four Arp2/3 probes yielded similar values that spanned 12.5 to 16.9 s. The half-life of Arp2/3 speckles is about half of that of actin filaments in lamellipodia of XTC cells ().
Next, we examined the spatial location of Arp2/3 recruitment by recording positions of newly appearing speckles. Although Arp2/3 speckles appeared throughout lamellipodia, appearance was heavily biased to a zone within 0.5 μm from the cell edge (). This bias was greater than that previously observed for actin speckles () and suggests that most Arp2/3-driven nucleation occurs in the tip region of the leading edge, where known Arp2/3 activators are localized (). The mean lifetime of Arp2/3 did not change throughout lamellipodia, although Arp2/3 with prolonged lifetime was observed more frequently in the cell edge region than in the rest of lamellipodia ().
Under our normal observation conditions, up to 2 h after cells were seeded on PLL-coated coverslips, our CP probes were predominantly colocalized with Arp2/3 in lamellipodia and cytoplasmic punctate actin structures (). When cells were allowed to grow for another several hours, CP probes were also associated with actin stress fibers (SFs; ). The similar localization of CP to myofibrillar structures in cultured cardiomyocytes has been reported previously (). We noticed that a fraction of CP speckles stayed stably associated with SFs, whereas CP speckles that did not localize with SFs in the lamella region emerged and disappeared as quickly as in lamellipodia. When subjected to speckle regression analysis (), the decay rate of persistent CP speckles located on SFs was slower than that of CP in the other area of lamella (). These results suggest that CP dissociation is differentially regulated in distinct actin structures and that the dissociation rate of CP may be correlated with the actin filament turnover rate.
Next, we examined the effect of stabilization of actin filaments on the dissociation kinetics of CP. We tested jasplakinolide (Jas), an actin depolymerization inhibitor () that induces a three- to fourfold increase in the half-life of actin filaments within 1 min in lamellipodia (). We observed a rapid, marked prolongation of CP speckle lifetime upon treatment with 1 μM Jas ( and Video 9, available at ). This finding is striking, as the half-life of actin filament turnover (∼30 s; ) greatly (∼20-fold) exceeds that of CP, and it is a counterintuitive phenomenon that inhibiting the disassembly of a network system has a strong impact on the kinetics of its associated molecule, which displays much faster binding-dissociation kinetics.
To confirm this, we tested whether the inactivation of cofilin, a major actin depolymerizing factor implicated in lamellipodium formation (; ; ), may also have similar effects. LIM kinase (LIMK) phosphorylates cofilin and prevents the interaction between cofilin and actin (). We generated an expression construct for mRFP1-tagged human LIMK (mRFP1– hLIMK-1) to examine the effect of various levels of LIMK-1 expression on CP. It was confirmed that cells expressing mRFP1–hLIMK-1 displayed slower actin filament turnover (Fig. S3 and Video 10, available at ), which is consistent with previous findings (). The decay rate of persistent CP speckles in lamellipodia was also markedly prolonged in cells expressing mRFP1–hLIMK-1, and this effect was correlated with the expression level of mRFP1–hLIMK-1 (). These results indicate that cofilin-mediated actin disassembly is responsible for the fast dissociation kinetics of CP in lamellipodia. Because three different conditions of actin filament stabilization led to the prolongation of CP speckle lifetime (–
), filament severing may frequently take place in lamellipodia, which could lead to the fast dissociation of CP from the actin network.
This study reveals the astoundingly fast dissociation kinetics of CP in lamellipodia. In vitro, native CP binds barbed ends with subnanomolar affinities, and CP dissociates from barbed ends at a very slow rate (∼0.0004 s; ; ; ). In contrast, our CP probes displayed three orders of magnitude faster dissociation kinetics in cells. Several studies have already suggested the weak interaction of CP with barbed ends in cell lysates. Nearly 1 μM of free barbed ends are present despite ∼1–2 μM CP being present in lysates obtained from cells () and neutrophils (). The dissociation constant of CP in cell lysates was estimated to be ∼0.1 μM (). The weak affinity of CP to barbed ends has now been confirmed by our data obtained using intact cells. Moreover, our results conclude that the fast dissociation of CP, but not its slow association as postulated (), is responsible for the weak capping activity. Attenuation of the fast dissociation of CP probes upon permeabilization can be explained by the removal of CP dissociation promoters such as phosphatidylinositol bisphosphate () and CARMIL (). An unidentified anti-CP factor that differs from VASP and CARMIL has also been reported previously (). Twinfilin, a barbed end–interacting protein with a G-actin sequestering activity (), interacts directly with CP without affecting the interaction of CP with barbed ends (). These molecules may contribute to the fast dissociation of CP. However, based on our findings of the strong dependency of the CP dissociation rate on actin filament turnover (–
), we propose that cofilin-mediated filament severing may trigger the dissociation of CP from the dendritic nucleation actin arrays.
Possible mechanisms for actin filament turnover–dependent CP dissociation are depicted in ). Jas might strengthen CP–actin interaction by affecting filament conformation (). However, because CP alone can bind F-actin with a very high affinity, further stabilization of CP–barbed end interaction by Jas seems questionable.
Alternatively, filament severing induced by cofilin may possibly trigger the dissociation of CP attached to a small actin oligomer. Jas, which binds actin filaments competitively with phalloidin (), probably exerts antisevering effects in a similar manner to phalloidin. If filament severing by cofilin triggers fast CP dissociation, there must be a mechanism that prevents the actin network from equally fast breakdown. Disassembly of actin filaments (
= 30 s; ) is ∼20-fold slower than the dissociation of CP in lamellipodia of XTC cells. One possible explanation is that cofilin may sever actin filaments preferentially near the barbed end (). Another possibility is that cofilin-mediated severing occurs frequently throughout the actin arrays, and rapid end-to-end annealing may prevent severed actin oligomers from leaving the filament network (). In vitro, rapid reannealing of sheared actin filaments has long been recognized (; ; ). Annealing appears to be a favorable reaction given the high density of actin filaments, which is ∼1,000 μM in lamellipodia (). To gain evidence for filament annealing, we examined the distribution of free filament ends using recombinant EGFP-tagged CP and tropomodulin () in permeabilized XTC cells. From the binding density of these fluorescent probes to lamellipodia, the concentrations of free barbed ends and pointed ends were estimated to be 0.99 and 4.6 μM, respectively (Fig. S4, available at ). The coexistence of high amounts of free barbed ends and pointed ends may support our hypothesis of frequent end-to-end filament annealing ().
The current dendritic nucleation models did not incorporate end-to-end filament annealing as a dominant reaction because CP has been believed to terminate barbed end growth within 1–2 s after the nucleation of filaments (; ). Now, our finding of fast CP dissociation suggests that CP is not sufficient to strictly block barbed ends in lamellipodia. The present study also tested whether other molecules exist that are capable of tightly capping a major fraction of filaments. Gelsolin fibroblasts display a reduced membrane ruffling response to EGF (; ). Eps8 fibroblasts are defective in membrane ruffling formation induced by PDGF (), and lacking Eps8 displays severe defects in the organization of intestinal brush border microvilli (). However, our data show that none of these barbed end factors, including gelsolin, Eps8, and VASP, cap actin tightly in lamellipodia. Notably, the dissociation constant of Eps8 with barbed ends determined in vitro () was also considerably lower than that observed in vivo. Collectively, these results suggest that free barbed ends are abundant in lamellipodia despite these cappers, whose affinities to barbed ends are within the nanomolar range in vitro. Our findings of the fast dissociation kinetics of these cappers from the actin network have opened up a possibility that end-to-end filament annealing may ubiquitously occur in lamellipodia.
summarizes kinetics in regulation of the dendritic nucleation actin arrays. Our single-molecule speckle analysis revealed 0.048 and 0.58 s for the dissociation rates of Arp2/3 and CP, respectively. Another important notion is the fast actin elongation in living cells as revealed by the actin polymerization–driven movement of mDia1 (). Based on the speed of mDia1F2 (Fig. S5, available at ), we estimate that free barbed ends may elongate at ∼66 subunits/s in lamellipodia. Together with the fast dissociation kinetics of CP, these results suggest that barbed end growth is not strictly restricted as previously postulated. Given the relatively free and fast barbed end growth, we also predict the requirement of filament severing in catalyzing the fast actin disassembly in lamellipodia based on kinetic modeling (unpublished data). Filament severing is also relevant, as it seems indispensable to facilitate fast actin turnover (; ) in the presence of actin filaments of several microns in length found in lamellipodia ().
Collectively, we predict that filament severing and its counteracting reaction, end-to-end annealing, may take place frequently in lamellipodia. Filament severing and end-to-end annealing could play a pivotal role in leading edge dynamics by rapidly changing the direction of growth, length distribution, disassembly, and polarity of the actin filament arrays. Currently, it is difficult to estimate the frequency of these two inverse reactions in cells. However, our finding of the fast CP dissociation kinetics may imply a severing frequency of up to one per second for an actin filament of several tens of subunits in length. Further studies will be required to test this high frequency severing-annealing hypothesis.
EST clones encoding subunits of Arp2/3, CP, and other actin-related proteins were obtained from the IMAGE consortium. GenBank/EMBL/DDBJ accession no. for sequence data are as follows: (Arp3), (Arp2), (p40), (p34), (p21), (p20), (p16), (CPα1), (CPα2), (CPβ1), (VASP), (gelsolin), and (tropomodulin). Each cDNA was subcloned into the pEGFP-C1 (CLONTECH Laboratories, Inc.)–derived vector harboring the defective cytomegalovirus (CMV) promoter () delCMV- EGFP-C1. mRFP1 fusion constructs were generated by replacing EGFP sequences with mRFP1 cDNA (gift from R.Y. Tsien, University of California, San Diego, San Diego, CA; ). Full-length pEGFP-Eps8 was described previously (). pmRFP1–hLIMK-1 was constructed by introducing wild-type human LIMK-1 cDNA (gift from K. Mizuno, Tohoku University, Sendai, Japan; ) into the mRFP1 expression vector. Jas was purchased from Calbiochem.
The guinea pig anti-p40 antibody was raised against the 6xHis-tagged p40 subunit using pET-30a (Novagen). For immunocytochemistry, cells were extracted with 0.1% Triton X-100 in Cytoskeleton buffer (CB; 10 mM MES, pH 6.1, 90 mM KCl, 3 mM MgCl, 2 mM EGTA, and 0.16 M sucrose) for 10 s and fixed with 3.7% PFA in CB for 20 min at room temperature. Texas red–X phalloidin (Invitrogen) was used for staining F-actin. AlexaFluor488 and -594 anti–guinea pig IgG (Invitrogen) were used as secondary antibodies.
Speckle imaging was performed as described previously (). In brief, cells were allowed to spread on a PLL-coated glass coverslip attached to a flow cell in 70% Leibovitz's L15 medium (Invitrogen) without serum. The flow cell was placed on the stage of a microscope (BX52 or IX71; Olympus) equipped with either 100-W mercury or 75-W Xenon illumination and with a cooled CCD camera (MMX1300-YHS, CoolSNAP HQ, or Cascade II:512; Roper Scientific; or UIC-QE; Molecular Devices). Time-lapse imaging was performed at 21–23°C using MetaMorph software (Universal Imaging Corp.) up to 120 min after cells were seeded. Fluorescent speckle microscopy was performed by observing cells expressing a low amount of EGFP-tagged proteins using a planApo 100× NA 1.40 oil objective (Olympus). A restricted area near the cell edge was illuminated. Speckle lifetime measurement was performed by tracking individual speckles manually. The photobleaching rate of EGFP probes was measured by illuminating entire cell areas under an identical condition on the day of each experiment. Normalization for photobleaching was described previously ().
XTC cells expressing EGFP-CPβ1 or CPβ1-EGFP were permeabilized with 0.1% Triton X-100 in CB for 10 s, and the medium was replaced with a buffer (10 mM Hepes, pH 7.4, 100 mM KCl, 2 mM MgCl, 0.2 mM EGTA, and 1 mM DTT). Time-lapse images were acquired at intervals of 5 min. Illumination was attenuated using neutral density filters so that fluorescence decay caused by photobleaching was negligible. EGFP fluorescence intensity was measured in a cell peripheral region after the subtraction of background values obtained in an area outside of the cell.
Cells were grown at 37°C in Luria-Bertani medium overnight. After subculturing into fresh media, chaperon proteins were induced, and, at OD = 0.6, 1 mM IPTG was added. Cells were cultured at 30°C for an additional 6 h and collected by centrifugation. Cells were then sonicated in buffer C (10 mM Tris-Cl, pH 7.5, 300 mM NaCl, 1 mM MgCl, 1 mM DTT, and 1 mM ATP) supplemented with 0.5% Triton X-100, 200 μM PMSF, 2 μg/ml leupeptin, and 1 μg/ml pepstatin A. The sonicate was clarified by centrifugation. The supernatant was incubated with glutathione–Sepharose 4B (GE Healthcare), and the beads were washed by repeated centrifugation with buffer C. The beads were incubated with thrombin (Sigma-Aldrich) in buffer C, and CPα2/EGFP-CPβ1 cleaved from GST was collected in the supernatant. Thrombin activity was neutralized with hirudin (Sigma-Aldrich).
Pyrene-labeled actin and rabbit skeletal muscle actin were purchased from Cytoskeleton, Inc. Before use, the protein concentrations were verified using densitometry on the SDS-PAGE gel stained with Coomassie Brilliant blue. Unlabeled actin was polymerized at 5 μM in buffer F (10 mM Tris-Cl, pH 7.5, 100 mM KCl, 2 mM MgCl, 1 mM ATP, 0.2 mM EGTA, and 0.2 mM DTT) for 1 h. For assembly from F-actin seeds, a 12.5-μl aliquot of unlabeled, preassembled actin filaments was inserted into each well. To start the reaction, 6.67 μl of a 15-μM pyrene-actin droplet in buffer G containing 0.2 mM ATP was washed into the F-actin seeds with 71 μl of 1.2× buffer F and 10 μl of buffer C containing a range of concentrations of CPα2/EGFP-CPβ1. Pyrene fluorescence was monitored at room temperature using Fluoroskan Ascent FL (Labsystems).
The rate constants for capping barbed ends were determined from kinetic parameter optimization using data collected in actin polymerization assays. The concentration of F-actin seeds, [S], was first determined by fitting the fluorescence data without CPα2/EGFP-CPβ1 with the equationwhere [A] is the concentration of polymerized actin, [A] is the initial concentration of free actin monomers, B and B are on and off rates of barbed ends, and P and P are on and off rates of pointed ends.
Polymerization in the presence of CPα2/EGFP-CPβ1 is described as B + A = B, B + C = BC, and P + A = P, where A is free actin monomers, B is free barbed ends, P is free pointed ends, C is free CP, and BC is capped barbed ends. Using [S], actin polymerization curves were calculated numerically by Euler's method. The on and off rates of CPα2/EGFP-CPβ1 were varied to fit calculated polymerization curves with pyrene-actin fluorescence data.
2 μM pyrene-labeled actin was polymerized for 2 h in buffer F. Various amounts of CPα2/EGFP-CPβ1 in 10 μl of buffer C was added to 90 μl of the polymerized pyrene-actin solutions. Samples were incubated for 24 h at 22°C, and pyrene fluorescence was measured. At steady state,where [A] is the concentration of free actin monomers, [B] is the concentration of free barbed ends, [P] is the concentration of free pointed ends, [C] is the concentration of free CP, and [BC] is the concentration of capped barbed ends.
At CP concentrations of ≥1 nM, the concentration of free CP, [C], is almost equal to its total concentration, [C]. The concentration of polymerized actin is calculated by the following equation:
The K for CPα2/EGFP-CPβ1 was determined by optimizing this equation to fluorescence data.
Fig. S1 shows the fast dissociation of barbed end–interacting proteins Eps8 and VASP from the actin network in lamellipodia. Fig. S2 shows a comparison of the localization of our Arp2/3 constructs and endogenous Arp 2/3 in XTC cells. Fig. S3 shows reduced actin filament turnover in cells overexpressing LIMK. Fig. S4 shows the distribution of free barbed ends and pointed ends in permeabilized XTC cells visualized by recombinant CPα2/EGFP-CPβ1 and GST-EGFP-tropomodulin, respectively. Fig. S5 shows the speed of processive movement of the FH2 mutant of mDia1 in the cell peripheral area. Videos 1–10 show time-lapse images of EGFP- tagged probes in XTC cells. Online supplemental material is available at . |
Retrograde, or centripetal, flow was originally identified as the net transport of surface receptors and the underlying actin cytoskeleton from the cell cortex toward the center of the cell (; ). This process has been studied extensively in motile cells, including neuronal growth cones, cultured fibroblasts, and fish keratocytes (; ; ), where it is required for the generation of traction forces on the substratum for cell movement and necessary for the maintenance of polarized, directional growth. Retrograde actin flow also occurs in nonmotile cells, such as sea urchin coelomocytes (). In both motile and nonmotile cells, retrograde flow occurs as a result of two forces; the “push” of actin polymerization caused by the addition of G-actin at the barbed ends of elongating actin filaments at the cell cortex, and the “pull” of myosin molecules further back in the network. Numerous studies, including those using inhibitors for myosin or myosin light chain kinase, have implicated conventional type II myosin in retrograde actin flow (; ; ; ; ). Other studies have also implicated tropomyosin in the regulation of actin flow, through the isoform-specific regulation of both the length of the actin network and the rate of retrograde flow ().
Retrograde actin flow similar to that observed in cortical actin networks in motile and nonmotile cells also occurs in polarized actin bundles within microvilli, filopodia, and stereocilia (for review see ). In each of these cortical protrusions, actin bundles elongate by the addition of new material to the tip of the bundle, which comprises F-actin barbed ends and a “tip complex” detected by ultrastructural analysis (; ; ). Imaging of actin dynamics revealed that addition of material to the tip of actin bundles results in retrograde, or centripetal, flow of F-actin within the bundle from the tip toward the base of the bundle ().
Several lines of evidence indicate that actin cables in budding yeast are conserved structures, which exhibit retrograde flow. Actin cables are parallel bundles of F-actin that align along the mother-bud axis and contain cross-linking and stabilizing proteins, including fimbrin (Sac6p), Abp140p, and two tropomyosin isoforms (Tpm1p and Tpm2p; ; ; ; ; ; ; ; ; ). Actin cables are required for the localization of vacuoles, secretory vesicles, mitochondria, late-Golgi components, spindle alignment elements, and mRNA to the developing bud (; ; ; ; ; ; ; ). With the exception of mitochondria, which use Arp2/3 complex–mediated actin polymerization for anterograde movement (), all of the other particles that rely on actin cables for distribution require type V myosin for localization in the bud. In the case of secretory vesicles and spindle-alignment elements, bud- directed movement is linear and occurs at a velocity that is directly proportional to the length of the type V myosin lever arm (; ). Thus, there is evidence that actin cables serve as tracks for myosin-driven, bud-directed particle movement. Moreover, because type V myosins are barbed-end–directed motors, it is likely that actin cables, like actin bundles in other eukaryotes, consist of actin filaments that are physically or functionally polarized with their barbed end oriented toward their assembly site in the bud.
Recent studies indicate that actin cables undergo retrograde flow. Previously, we showed that actin cables are labeled in living cells using an Abp140p-GFP fusion protein (). Time-lapse imaging of Abp140p-GFP revealed that actin cables are dynamic structures that assemble in the bud and bud neck and exhibit assembly-dependent retrograde flow away from assembly sites along the mother-bud axis. Fluorescence loss in photobleaching experiments and analysis of the movement of fiduciary marks of Abp140p-GFP on motile actin cables revealed that insertion of material at the tip of the cable located within the bud results in retrograde movement of the entire cable toward the mother cell. Consistent with these findings, other laboratories showed that the yeast formins Bni1p and Bnr1p are required for actin cable stability, have the capacity to stimulate actin polymerization, and localize to the sites of actin cable assembly detected using Abp140p-GFP (; ; ; ; ,; ).
Our previous results also showed that actin cable movement could be mediated by assembly-independent mechanisms. Specifically, we found that fixed-length fragments of actin cables, which are produced by treatment with low levels of the actin monomer–sequestering factor latrunculin A, move along the cell cortex. This movement did not depend on actin cable assembly and occurred parallel to the long axis of the actin cable (). These findings raised the possibility that myosin molecules that are anchored to the cortex could provide a pulling force for actin cable retrograde flow. has five myosin genes: two type I myosin genes ( and ), one type II myosin gene (), and two type V myosin genes ( and ; ; ; ; ; ). Type I myosins localize to actin patches (actin-coated endosomes), where they contribute to the control of actin organization, endocytosis, and activation of the actin nucleation activity of the Arp2/3 complex (; ; ; ; ). The type II myosin of yeast localizes to the bud neck, and has been implicated in cytokinesis (; ; ). Finally, type V myosins are motors for bud-directed cargo movement along actin cables, and accumulate in the bud tip (; ; ). We studied the role of yeast myosins in actin cable dynamics.
In complementary studies, we evaluated the role of yeast tropomyosins in retrograde actin cable flow. The yeast tropomyosins share 64% homology, and show similar affinities for binding to F-actin. Deletion of both Tpm1p and Tpm2p results in cell death in budding yeast. These findings indicate that Tpm1p and Tpm2p together support an essential function in budding yeast. Because both proteins localize to actin cables, and actin cables are essential for cell viability in budding yeast, it is likely that the essential function supported by Tpm1p and Tpm2p is stabilization of actin cables. On the other hand, overexpression of in yeast bearing a deletion in does not suppress the defects in growth or chitin deposition associated with the deletion. Moreover, actin cables are destabilized in , but not in mutants, (). These findings indicate that the two tropomyosin isoforms of budding yeast carry out distinct functions, and that the primary function of Tpm1p is control of actin cable stability, along with its function in chitin deposition and bud growth.
We report novel roles for a type II myosin (Myo1p) and a specific tropomyosin isoform (Tpm2p) in retrograde actin cable flow in budding yeast. Our findings indicate that retrograde actin flow is conserved from yeast to humans, and raise the possibility that budding yeast may be useful as a model system for studying this fundamental process. Our studies also support an isoform-specific function for Tpm1p in promoting actin cable activity as a track for myosin-driven anterograde cargo movement.
In wild-type cells, actin cables extend from the bud tip or bud neck along the mother-bud axis. In yeast bearing deletions in the type I myosins, actin cables do not align along the mother-bud axis and are less robust than those observed in wild-type cells; however, actin cables are present in these cells. To analyze the rate of retrograde actin cable flow, we monitored as a function of time the change in position of either the tips of Abp140p-GFP–labeled actin cables, or of fiduciary marks of Abp140p-GFP on actin cables (). The velocity of retrograde actin cable flow in various myosin mutants and their corresponding wild-type controls are shown (). Because the velocity of retrograde actin cable flow was similar in all wild-type stains analyzed at 23 and 37°C (0.36–0.38 μm/s), the velocity of actin cable movement in various myosin mutants was compared with the mean velocity of all wild-type strains evaluated ().
The retrograde flow rate of actin cables in strains bearing a deletion in the two type I myosins of yeast (
) was not significantly different from rates observed in corresponding wild-type cells at 23°C. Because the type V myosin gene is essential, we studied the effect of loss of type V myosin function using a yeast strain bearing a temperature-sensitive mutation in the gene and a deletion in which is the other, nonessential, type V myosin gene. We found that the rate of retrograde actin cable flow in this strain () was similar to that observed in wild-type cells upon incubation at permissive (23°C) or restrictive (37°C) temperatures. In contrast, in the strain lacking the type II myosin (), the rate of retrograde actin cable flow of 0.19 μm/s was significantly lower than that observed in the wild-type control (P < 0.01). These findings indicate that type II myosin of budding yeast is required for normal rates of retrograde actin cable flow ( and ).
Type II myosin localizes to the bud neck from the time of commitment to a round of cell division (G) to the end of the cell division cycle. We tested whether this localization of type II myosin is critical for retrograde actin cable flow by taking advantage of the finding that overexpression of the tail of type II myosin acts in a dominant-negative fashion and causes delocalization of the endogenous, full-length protein from the bud neck (). We inserted the C-terminal 868 amino acids of the type II myosin into a multicopy plasmid under control of the galactose-inducible promoter. The resulting plasmid, pGal-MYO1-tail, was used to evaluate the effect of regulated overexpression of the type II myosin tail on the localization of endogenous, GFP-tagged type II myosin and on the dynamics of actin cables.
We confirmed that the addition of GFP to the C terminus of the endogenous type II myosin gene () had no obvious effect on cell viability, polarity, or growth rates (unpublished data). As expected, we found that full-length, GFP-tagged type II myosin (Myo1p-GFP) forms a ring at the bud neck in untransformed cells, as well as in cells that were transformed with pGal-MYO1-tail and incubated under noninducing conditions (). Thus, endogenous type II myosin exhibits normal localization when the pGal-MYO1-tail expression is not induced. In contrast, galactose-induced overexpression of pGal-MYO1-tail caused mislocalization of endogenous, GFP-tagged type II myosin from the bud neck to the cytosol ().
Using Abp140p-GFP as a marker for actin cables, we found that wild-type cells bearing pGal-MYO1-tail showed normal rates of actin cable retrograde flow when grown under noninducing conditions. However, galactose-induced overexpression of pGal-MYO1-tail resulted in a significant decrease in the rate of retrograde actin cable flow (). Indeed, the retrograde flow rate in cells expressing the dominant-negative pGal-MYO1-tail construct was similar to that observed upon deletion of . This decrease in the rate of retrograde flow was not the result of incubation in galactose, as wild-type cells grown in galactose showed no change in actin cable retrograde flow rate. Instead, these findings indicate that localization of type II myosin at the bud neck is important for its function in retrograde actin cable flow.
Two approaches were taken in our study of the role of the myosin motor in retrograde actin cable flow. The first approach was based on recent findings that the Myo1p tail supports cytokinesis when expressed in lieu of full-length Myo1p and at a level similar to that of endogenous, wild-type Myo1p (). The Myo1p tail and full-length Myo1p were expressed in cells using the plasmids constructed by , and retrograde actin cable flow was examined in these cells using Abp140p-GFP. Full-length Myo1p and the Myo1p tail localize to the bud neck, as previously reported (). The velocity of retrograde actin cable flow observed upon expression of full-length, plasmid-borne Myo1p (0.419 μm/s; = 32) was similar to that observed in yeast expressing endogenous, wild-type . In contrast, expression of the Myo1p tail alone resulted in a reduced rate of retrograde actin cable flow (0.227 μm/s; = 33) that was similar to that observed in untransformed cells and in cells that were transformed with the vector used for Myo1p and Myo1p tail expression. Thus, the motor domain of is required for normal rates of retrograde actin cable flow.
As a second approach to study the role for the type II myosin motor activity in retrograde actin cable flow, we constructed a mutant that carries a single amino acid substitution in the Myo1p motor domain, which inhibits the ATP-sensitive actin-binding activity of the protein. The site was chosen for mutagenesis based on the observation that the corresponding mutation in the essential type V myosin of budding yeast, the mutation, produces a temperature-dependent loss of cell viability (). The mutation is a substitution of lysine for glutamic acid 511 (), a residue conserved in all yeast myosin proteins () that corresponds to Glu527 in chicken skeletal muscle myosin and Glu528 in yeast type II myosin. X-ray crystallography of chicken skeletal muscle myosin mapped this residue to the actin-binding face of the lower 50-kD component of the motor domain, and showed that this glutamic acid forms a salt bridge with Lys486, which is another highly conserved residue in the motor domain ().
To generate the mutant, the type II myosin gene (), which was tagged with GFP at its C terminus, was inserted into a low-copy (CEN) plasmid under control of the endogenous type II myosin promoter. Glu528 was than replaced with Lys in the plasmid-borne, GFP-tagged gene. We refer to the mutated gene as to indicate that it shares the same mutation as . The GFP-tagged mutant was expressed in lieu of wild-type myosin II, i.e., in a yeast strain bearing a deletion in the gene (p). Using Western blots probed with anti-GFP antibody, we found that the steady-state level of myo1-66p was 70% of the wild-type type II myosin level upon growth at permissive temperature (23°C) or after incubation at restrictive temperature (37°C) for 30 min (). Moreover, cells exhibited elevated levels of multibudded cells; the level of multibudded cells in these cultures increased upon incubation at 37°C (unpublished data). Nonetheless, the mutant protein localized normally to the bud neck throughout the cell cycle (). Moreover, the growth rate of these cells at 23°C was similar to that observed for the corresponding wild-type strain at 23°C. Finally, cells did not show temperature-dependent lethality, as expected, because type II myosin is not essential in the yeast genetic background used for these studies ().
The effect of the mutation on the mechanochemical properties of type V myosin of yeast is not known. Therefore, we evaluated the rigor-binding activity of wild-type Myo1p and mutant myo1-66p to actin (). Wild-type myosin was immobilized on magnetic beads and incubated with yeast F-actin in the presence of ATP or ADP. In the presence of ATP, F-actin did not bind to wild-type type II myosin. However, in the presence of ADP, there was a significant increase in the amount of F-actin binding. Thus, we concluded that type II myosin exhibits ATP-sensitive actin-binding activity. Mutant protein also exhibited ATP-sensitive actin- binding activity under these experimental conditions. However, actin-binding activity of the mutant protein was lower than that observed in the wild-type protein (). This decrease in actin binding to mutant type II myosin was evident at all concentrations of actin tested (). Because wild-type protein lost activity after short-term incubation at elevated temperatures, we could not test whether the mutation confers temperature-dependent loss of actin-binding activity in vitro. Nonetheless, our results indicate that substitution of Glu528 with Lys in the actin-binding site of the type II myosin motor domain reduces the ATP-sensitive actin-binding activity of the protein.
Given a clear understanding of the effect of the mutation on the type II myosin motor domain, we examined the retrograde flow rates of Abp140p-GFP–labeled actin cables in the mutant, relative to the wild-type (). At 23°C, the rate of retrograde actin cable flow in yeast expressing the mutant was similar to that observed in yeast expressing wild-type myosin II protein (). Thus, the mutant protein supports normal rates of retrograde flow at reduced temperatures, despite the fact that it is present at lower levels compared with wild-type cells. The rate of retrograde actin cable flow in wild-type cells was not affected by incubation at 37°C. In contrast, the rate of actin cable retrograde flow is significantly reduced in yeast expressing the mutant at 37°C. (). However, the rate of retrograde actin cable flow observed in the mutant at restrictive temperatures is similar to that observed in yeast bearing a deletion in and in yeast expressing the Myo1p tail at wild-type levels in lieu of full-length Myo1p (). Thus, the mutant exhibits temperature- dependent reduction in the rate of retrograde actin cable flow. These data suggest that normal rates of actin cable retrograde flow require a functional type II myosin motor domain.
Earlier studies suggested that deletion of results in loss of actin cables (). However, with the advent of more sensitive optical imaging technology, it is now clear that deletion of reduces the length and/or abundance of actin cables, but does not result in the complete loss of these structures. Time-lapse videos illustrating retrograde flow of actin cables in wild-type cells and mutants are shown (Videos 1 and 2, available at ). Because actin cables are present in both and cells, we studied the effect of loss of the yeast tropomyosins on actin cable dynamics. We found that the rate of actin cable retrograde flow in strains was not significantly different from that observed in wild-type cells (). In the strain, however, the rate of actin cable retrograde flow was 0.78 μm/s, which is approximately twice that observed in wild-type cells (). This finding indicates that actin cable retrograde flow can occur at rates that are at least twice the mean rate of flow observed in wild-type cells. Equally important, this finding indicates that Tpm2p, but not Tpm1p, reduces the rate of retrograde actin cable flow.
To further explore the role of Tpm1p in retrograde actin cable flow, we studied Abp140p-GFP dynamics in yeast bearing a deletion in . Mdm20p was originally identified in a screen for mutations affecting mitochondrial inheritance (). Deletion of results in reduced actin cable length, but has no effect on alignment of actin cables along the mother-bud axis. Recent evidence indicates that Mdm20p is associated with Nat3p, a protein that catalyzes acetylation of Tpm1p (). We find that deletion of has no significant effect on the rate of retrograde actin cable flow ().
We found that the type II myosin Myo1p stimulates the rate of actin cable retrograde flow, whereas the tropomyosin Tpm2p has a negative effect on this process. Because both proteins bind to the lateral surface of F-actin, it is possible that Tpm2p regulates actin cable dynamics by competing with type II myosin for binding to F-actin within actin cables. One approach to test this model is to study actin cable dynamics in a cell bearing a deletion of and . In the strains used for these studies, deletion of and produced synthetic lethality. Because Myo1p and Tpm2p have functions that are not related to their function in retrograde actin cable flow, and all strains used to generate the double mutant also carried a GFP tag on the gene, the basis for the synthetic lethality is not readily apparent.
double mutant.
As an alternative approach to test this model, we performed in vitro assays to determine whether the tropomyosins exhibit isoform-specific effects on type II myosin motor–driven microfilament sliding and on the binding of type II myosin to F-actin. We found that type II myosin of budding yeast supports microfilament sliding in vitro. The assay used was a modification of a previously described assay (). GFP-tagged type II myosin from whole-cell extracts was immobilized in a microscope flow cell using anti-GFP antibody. The flow cell was treated with unlabeled F-actin and ATP to saturate ATP-insensitive, rigor-binding sites on immobilized type II myosin. Rhodamine-phalloidin–stained and stabilized yeast F-actin was then added in ATP- free buffer, and ATP was then added after washes to remove unbound material. Type II myosin–driven microfilament sliding was then monitored through time-lapse imaging.
Microfilament sliding driven by the yeast type II myosin was similar to that observed for other actin-based motors (). Microfilament sliding was unidirectional, continuous, and parallel to the long axis of the filament. The rate of movement of microfilament sliding was 0.63 ± 0.10 μm/s. This rate is similar to the rate of 0.5 μm/s observed for microfilament sliding by type II myosin isolated from the fission yeast (). To our knowledge, this is the first direct evidence that the type II myosin of budding yeast has motor activity on actin.
To determine whether tropomyosins exert isoform-specific effects on this process, we used established protocols to purify Tpm1p from yeast carrying a deletion in , and Tpm2p from yeast carrying a deletion in . Pull-down assays, which measured binding of the yeast tropomyosins to yeast actin, revealed that purified Tpm1p and Tpm2p retained biological activity (). The affinity of purified Tpm1p and Tpm2p for yeast actin is similar to that reported previously for rabbit skeletal muscle actin ().
Finally, we tested the effects of Tpm1p and Tpm2p on type II myosin–driven microfilament sliding and on binding of F-actin to type II myosin. To do so, we compared the rates of type II myosin–driven sliding of F-actin that was either untreated or pretreated with saturating amounts of Tpm1p or Tpm2p (). We found that the rate of microfilament gliding of F-actin containing bound Tpm1p (0.64 ± 0.10 μm/s) was similar to that observed for untreated F-actin without tropomyosin (0.63 ± 0.10 μm/s). In contrast, the rate of sliding of actin filaments containing bound Tpm2p (0.41 ± 0.12 μm/s) was significantly lower than that observed for either untreated or Tpm1p-treated F-actin. Interestingly, Tpm2p exhibits isoform-specific effects on type II myosin–driven microfilament sliding in vitro that is equal and opposite to the observed effect of deletion of on type II myosin–mediated retrograde actin cable flow in vivo. This finding is consistent with the interpretation that Tpm2p exerts an isoform-specific effect on inhibition of binding of F-actin to yeast type II myosin. To test this hypothesis directly, phalloidin-stabilized yeast F-actin was pretreated with equal amounts of purified Tpm1p or Tpm2p, and tested for binding to immobilized type II myosin (). At the concentrations tested, we found that Tpm2p, but not Tpm1p, inhibited the binding of actin filaments to type II myosin of budding yeast.
Retrograde actin flow occurs in cortical actin networks in both motile and nonmotile cells, as well as in actin bundles in apical protrusions including microvilli, filopodia, and stereocilia. Several studies support a role for type II myosin and a tropomyosin in this process. First, both proteins localize to the lamella, an actin network that is proximal to the lamellipodium and exhibits retrograde flow in migrating cells (). Second, retrograde flow of actin lamella and of actin bundles in neuronal filopodia is perturbed by 2,3-butanedione monoxime and blebbistatin, agents that inhibit type II myosin ATPase activity (; ; ; ; ; ). Third, microinjection of the skeletal muscle tropomyosin, but not endogenous tropomyosin, into PtK1 cells produces an altered rate of retrograde actin flow ().
Although type II myosins and specific tropomyosin isoforms have been implicated in retrograde actin flow in animal cells, there was no evidence for a role of these proteins in actin cable dynamics in budding yeast. Moreover, the precise function of tropomyosin in retrograde actin flow in animal cells has been unclear. We provide evidence that type II myosin and a specific tropomyosin contribute to the retrograde flow of actin cables in budding yeast. Our findings are consistent with a role for type II myosin in providing a pulling force on actin cables as they undergo retrograde flow. Our findings also indicate that type II myosin and Tpm2p may have antagonistic effects during retrograde actin cable flow. Tpm2p may exert its isoform-specific negative regulatory effect by inhibiting the binding of type II myosin with F-actin within actin cables. These findings indicate that retrograde actin flow is conserved from yeast to animal cells, and extend our understanding of the mechanism of action of type II myosin and tropomyosin in this process.
Yeast strains used in this study are listed in .
strain, which is in the W303 genetic background, all strains used are S288C. Yeast cell growth and manipulations were performed according to established protocols (). To visualize actin cables in living cells, the C terminus of Abp140p was tagged with GFP (S65T) using PCR-based insertion of the GFP gene () into the chromosomal copy of as previously described (). The cassette used to insert GFP into the gene of wild-type and myosin mutant cells was amplified from pFA6a:GFP (S65T): with the following primers: forward 5′- AAATATTGATAGTAACAATGCACAGAGTAAAATTTTCAGTCGGATCCCCGGGTTATTAA-3′ and reverse 5′-AAAGGATATAAAGTCTTCCAAATTTTTAAAAAAAAGTTCGGAATTCGAGCTCGTTAAAC-3′. Amplified DNA was transformed into the strains of interest, and cells that had integrated the GFP cassette into the gene were selected by their ability to grow on synthetic media lacking histidine.
The strain containing a deletion in the gene and a GFP tag on Abp140p (THY114-2d) was constructed by mating a haploid cell (RG11767) with a wild-type cell expressing Abp140p-GFP (YCY027). The resulting diploids were sporulated in medium containing 1% KOAc + 0.025% glucose, and haploids produced from the diploids were selected using a visual screen for GFP-labeled actin cables and PCR to verify deletion of the gene of interest.
Plasmid pTH12 is a two-micrometer plasmid that encodes the 868 C-terminal amino acids of Myo1p under the control of the promoter (similar to that described by ). For the construction, DNA encoding the tail region of was amplified from wild-type yeast genomic DNA using the following forward and reverse primers: 5′-TCATGACATAAAATTGGTCACTTTAGA-3′ and 5′-AGTTAA
TTAACTGAAAATTTTACTCTGTGCATT-3′, respectively. The forward primer contains a BamHI restriction site and the reverse primer contains a HindIII restriction site. Restriction sites are underlined. The product was purified and ligated into the BamHI and HindIII sites of the plasmid pGAL68. The resulting vector contains an in-frame ATG from codon 1060 of downstream of the promoter. Induction of Myo1p tail overexpression was performed by shifting cells from synthetic media supplemented with raffinose into synthetic media containing both raffinose and galactose.
Plasmid pTH14 is a CEN-based plasmid that contains the promoter within 555 bases of the upstream untranslated region of the gene followed by a GFP (S65T)-tagged gene that carries an E528K mutation (). To construct this plasmid, site-directed mutagenesis was performed on the plasmid p846 using the QuikChange Site-Directed Mutagenesis kit (Stratagene), which is a PCR-based method for mutagenesis. Primers were designed to mutate glutamate 528 to a lysine by a single nucleotide substitution (GAA→AAA). After PCR amplification, the mutated fragment was cleaved using AgeI and BclI and reinserted into p846. This new plasmid was sequenced to confirm the presence of the mutation. The plasma used to construct pTH14, which contains wild-type MYO1, was provided by E. Bi (University of Pennsylvania, Philadelphia, PA).
Plasmids pTL2 and pTL3 are centromere-based plasmids that contain full-length and the tail (aa 842–1,928) under control of the promoter. These plasmids were produced from pGFP-MYO1 and pGFP-MYO1-, which were generously provided by M. Lord (University of Vermont, Burlington, VT) and T. Pollard (Yale University, New Haven, CT; ). To allow for transformation into strains of interest, the marker on pGFP-MYO1 and pGFP-MYO1- was replaced with . To do so, the marker was excised from pGFP-MYO1 and pGFP-MYO1- by digestion with PvuI. The marker, which was amplified from pRS415, was then inserted into the plasmid at the PvuI site.
For time-lapse fluorescence imaging of Abp140p-GFP, cells were grown in synthetic complete or lactate medium () until early log phase at 25°C. 3 μl of cell suspension was applied to a microscope slide and covered with a coverslip. Microscopy was performed using a microscope (E600; Nikon) equipped with a Plan Apo ×100/1.4 NA objective and a cooled charge-coupled device camera (Orca-ER; Hamamatsu). Illumination with a 100 W mercury arc lamp was controlled with a shutter (Uniblitz D122; Vincent Associates). The temperature of the objective lens was controlled using an objective heater (Bioptechs). Images were collected and analyzed using Openlab 3.0.8 software (Improvision) and ImageJ 1.28 (National Institutes of Health), respectively. The exposure time and time interval between successive image acquisitions were 400 and ∼500 ms, respectively. The total imaging time for the time-lapse imaging was 20 s. For determination of the velocity of elongating actin cables, the change in position of fluorescent fiduciary marks on elongating cables was measured as a function of time, as previously described ().
Time-lapse imaging of fluorescently labeled actin filaments in actin- gliding assays was performed as described in the previous paragraph. The exposure time and time interval between successive image acquisitions were 400 and 500 ms, respectively. The total imaging time for the time-lapse imaging was 50 s. For determination of the velocity of microfilament gliding, the change in position of the tips of gliding actin filaments was measured as a function of time, as previously described ().
For analysis of the localization of GFP-tagged wild-type and mutant proteins, 25 z sections were obtained at 0.2-μm intervals through the entire cell. Z sectioning for 3D imaging was performed using a piezoelectric focus motor mounted on the objective lens of the microscope (Polytech PI). Out-of-focus light was removed by digital deconvolution, and each series of deconvolved images was projected and rendered with Volocity software (Improvision, Inc.).
Yeast actin was isolated from as previously described (). In brief, actin was purified from cell lysate by DNase affinity chromatography, followed by ion-exchange chromatography using DE52. Peak fractions were pooled and subjected to multiple rounds of polymerization/depolymerization to ensure polymerization-competent actin. Yeast tropomyosins (Tpm1p and Tpm2p) were purified as previously described (). Tpm1p was purified from a strain that carries a deletion in the gene and is transformed with a plasmid bearing under control of the galactose-inducible promoter (THY196). Tpm2p was purified from a strain that carries a deletion in the gene and is transformed with a plasmid bearing under control of the galactose-inducible promoter (THY197). Tropomyosin constructs were provided by A. Bretscher and D. Pruyne (Cornell University, Ithaca, NY). Protein concentrations were determined using the bicinchoninic acid protein assay reagent (Pierce Chemical Co.) according to the manufacturer's protocol.
For purification of actin and the GFP-tagged wild-type and mutant Myo1p proteins from yeast, cells were grown to mid-log phase, concentrated by centrifugation (1,500 for 10 min), and suspended in cell lysate buffer (25 mM imidazole-HCl, pH 7.4, 25 mM KCl, 4 mM MgCl, 200 μM ATP, 2 mM DTT, 1 mM EGTA, 1 mM PMSF, and a previously described protease inhibitor cocktail []). The cell suspension was then added drop-wise into liquid nitrogen. The resulting frozen pellets were ground in a liquid nitrogen–cooled coffee grinder for 5–10 s, and the ground powder was stored at –80°C. The ground powder was thawed at 4°C and clarified by centrifugation at 1,500 for 4 min at 4°C. The pellet was discarded and the supernatant was used as in whole-cell lysate for microfilament gliding and myosin–actin–binding assays.
For quantitation of the steady-state level of Myo1p and myo1-66p, cells were grown to mid-log phase (OD = 0.7) in 10 ml of synthetic complete–ura liquid medium. Cells were concentrated by centrifugation, washed with distilled water, and resuspended in 300 μl lysis buffer (50 mM imidazole, pH 7.4, 1% Triton X-100, 2 mM PMSF, and protein inhibitor cocktails 1 and 2). 1 ml of 0.5-mm glass beads (Biospec Products, Inc.) were washed in lysis buffer and added to the cell suspension. The mixture was vortexed vigorously for 6 min at 4°C. Thereafter, 200 μl of the lysate was removed and mixed with 66.6 μl of 4× SDS sample buffer. The mixture was incubated at 100°C for 3–4 min. Proteins in the solubilized lysate were analyzed by SDS-PAGE, followed by Western blots.
Purified yeast tropomyosins at the indicated concentrations were added to 0.4 mg/ml F-actin in tropomyosin-binding buffer (150 mM KCl, 5 mM MgCl, 10 mM Tris-HCl, pH 7.5, 1 mM EGTA, and 1 mM DTT). Samples were incubated for 20 min at 4°C. Samples were then subjected to ultracentrifugation for 20 min at 100,000 at 4°C in a fixed angle rotor (TA 100.3) using a ultracentrifuge (TL-100; both Beckman Coulter). Supernatant was removed, and the pellet was resuspended in an equal volume of SDS sample buffer. Equal volumes were loaded in wells of a polyacrylamide gel and subjected to electrophoresis, followed by Coomassie blue staining. Quantitative densitometry was performed to assess the cosedimentation of tropomyosin with actin (bound vs. unbound fractions).
h
a
n
o
l
-
w
a
s
h
e
d
c
o
v
e
r
s
l
i
p
s
w
e
r
e
c
o
a
t
e
d
w
i
t
h
0
.
5
%
n
i
t
r
o
c
e
l
l
u
l
o
s
e
a
n
d
a
l
l
o
w
e
d
t
o
a
i
r
-
d
r
y
.
T
h
e
c
o
v
e
r
s
l
i
p
w
a
s
t
h
e
n
a
p
p
l
i
e
d
o
n
t
o
a
m
i
c
r
o
s
c
o
p
e
s
l
i
d
e
w
i
t
h
t
h
e
c
o
a
t
e
d
s
i
d
e
f
a
c
i
n
g
d
o
w
n
,
u
s
i
n
g
d
o
u
b
l
e
-
s
t
i
c
k
t
a
p
e
a
s
a
d
h
e
r
e
n
t
s
p
a
c
e
r
s
.
T
h
e
r
e
s
u
l
t
a
n
t
f
l
o
w
c
h
a
m
b
e
r
v
o
l
u
m
e
w
a
s
∼
3
0
μ
l
.
#text
T
o
b
l
o
c
k
a
l
l
r
i
g
o
r
-
b
i
n
d
i
n
g
s
i
t
e
s
o
f
i
m
m
o
b
i
l
i
z
e
d
t
y
p
e
I
I
m
y
o
s
i
n
,
u
n
l
a
b
e
l
e
d
p
h
a
l
l
o
i
d
i
n
-
s
t
a
b
i
l
i
z
e
d
y
e
a
s
t
a
c
t
i
n
f
i
l
a
m
e
n
t
s
(
2
5
μ
g
/
m
l
i
n
m
o
t
i
l
i
t
y
b
u
f
f
e
r
)
w
e
r
e
a
d
d
e
d
t
o
t
h
e
f
l
o
w
c
h
a
m
b
e
r
a
n
d
a
l
l
o
w
e
d
t
o
i
n
c
u
b
a
t
e
f
o
r
2
m
i
n
a
t
4
°
C
.
U
n
b
o
u
n
d
m
a
t
e
r
i
a
l
w
a
s
t
h
e
n
r
e
m
o
v
e
d
b
y
t
w
o
w
a
s
h
e
s
w
i
t
h
m
o
t
i
l
i
t
y
b
u
f
f
e
r
s
u
p
p
l
e
m
e
n
t
e
d
w
i
t
h
1
0
μ
M
A
T
P
a
n
d
a
n
A
T
P
-
r
e
g
e
n
e
r
a
t
i
n
g
s
y
s
t
e
m
(
1
0
m
M
c
r
e
a
t
i
n
e
p
h
o
s
p
h
a
t
e
a
n
d
1
0
0
μ
g
/
m
l
c
r
e
a
t
i
n
e
p
h
o
s
p
h
o
k
i
n
a
s
e
)
.
U
n
b
o
u
n
d
m
a
t
e
r
i
a
l
w
a
s
r
e
m
o
v
e
d
b
y
t
h
r
e
e
w
a
s
h
e
s
w
i
t
h
m
o
t
i
l
i
t
y
b
u
f
f
e
r
.
a
s
t
F
-
a
c
t
i
n
t
h
a
t
w
a
s
f
l
u
o
r
e
s
c
e
n
t
l
y
l
a
b
e
l
e
d
a
n
d
s
t
a
b
i
l
i
z
e
d
w
i
t
h
r
h
o
d
a
m
i
n
e
p
h
a
l
l
o
i
d
i
n
w
a
s
s
u
s
p
e
n
d
e
d
i
n
A
T
P
-
f
r
e
e
m
o
t
i
l
i
t
y
b
u
f
f
e
r
t
o
a
f
i
n
a
l
c
o
n
c
e
n
t
r
a
t
i
o
n
o
f
1
0
μ
g
/
m
l
.
T
h
i
s
s
o
l
u
t
i
o
n
w
a
s
a
d
d
e
d
t
o
t
h
e
f
l
o
w
c
h
a
m
b
e
r
c
o
n
t
a
i
n
i
n
g
i
m
m
o
b
i
l
i
z
e
d
t
y
p
e
I
I
m
y
o
s
i
n
.
A
f
t
e
r
i
n
c
u
b
a
t
i
o
n
a
t
R
T
f
o
r
2
m
i
n
,
u
n
b
o
u
n
d
m
a
t
e
r
i
a
l
w
a
s
r
e
m
o
v
e
d
b
y
t
w
o
w
a
s
h
e
s
w
i
t
h
m
o
t
i
l
i
t
y
b
u
f
f
e
r
.
T
h
e
m
i
c
r
o
s
c
o
p
e
s
l
i
d
e
f
l
o
w
c
h
a
m
b
e
r
w
a
s
p
l
a
c
e
d
o
n
t
h
e
s
t
a
g
e
o
f
a
n
e
p
i
f
l
u
o
r
e
s
c
e
n
c
e
m
i
c
r
o
s
c
o
p
e
(
s
e
e
a
b
o
v
e
)
.
M
o
t
i
l
i
t
y
b
u
f
f
e
r
c
o
n
t
a
i
n
i
n
g
A
T
P
a
n
d
a
n
A
T
P
-
r
e
g
e
n
e
r
a
t
i
n
g
s
y
s
t
e
m
w
a
s
a
d
d
e
d
a
n
d
m
i
c
r
o
f
i
l
a
m
e
n
t
g
l
i
d
i
n
g
w
a
s
m
o
n
i
t
o
r
e
d
b
y
t
i
m
e
-
l
a
p
s
e
f
l
u
o
r
e
s
c
e
n
c
e
i
m
a
g
i
n
g
a
s
d
e
s
c
r
i
b
e
d
i
n
M
i
c
r
o
s
c
o
p
y
a
n
d
i
m
a
g
e
a
n
a
l
y
s
i
s
.
Goat anti–mouse IgG magnetic beads (New England Biolabs; 10 μl per reaction) were washed three times with sterile-filtered TBS, pH 8.0, and incubated with 5 μg of mouse anti-GFP IgGs (Roche) for 1 h at 4°C. Beads were washed two times in sterile-filtered TBS and two times in lysate buffer, and then incubated with 100 μl of either 1 mg/ml BSA, lysate from cells expressing GFP-tagged Myo1p, or lysate from cells expressing GFP-tagged p for 2 h at 4°C. Beads were washed twice in lysate buffer and twice with motility buffer to remove unbound material. Phalloidin-stabilized yeast F-actin was added at the concentrations indicated in the presence of 1 mM ATP or ADP, pH 7.4. After incubation at 4°C for 90 s, beads were separated from the supernatant. Bound material (beads) and unbound material were solubilized in SDS sample buffer and subjected to PAGE and Western blot analysis using antibodies that recognized GFP and actin. Quantitative densitometry was performed to assess the fraction of actin that was bound to myosin or released to the supernatant.
Video 1 shows retrograde actin cable flow in wild-type cells. Video 2 shows retrograde actin cable flow in myo1Δ cells. Online supplemental material is available at . |
Among the landmarks for apicobasal cell polarity, microvilli are actin-based protrusions on the apical surfaces of epithelial and sensory cells. In epithelia of the small intestine and kidney proximal tubule, closely packed microvilli known as the brush border are observed on their apical plasma membranes. They participate in a variety of functions such as nutrient absorption, mechanosensory transduction, and phototransduction. The core structure of microvilli is comprised of parallel actin filaments, and numerous actin-binding proteins, including villin, espin, fimbrin, fascin, and myosins as well as ezrin/ radixin/moesin (ERM) proteins, have been identified as components of microvilli (for reviews see ; ; ). Some transmembrane proteins such as Cad99C, the orthologue of the vertebrate procadherin 15 (; ), are also known to be constituents of microvilli.
ERM proteins (; ; ) bind not only to actin filaments at their C-terminal domains but also to various transmembrane proteins at their N-terminal 4.1 ERM domains, thereby acting as cross-linkers between the cytoskeleton and the plasma membrane (for reviews see ; ; ). Their N-terminal halves are also indirectly associated with membrane proteins through Na/H exchanger regulatory factor 1/ERM-binding phosphoprotein 50 (EBP50; ; ; ; ; for reviews see ; ). However, ERM proteins are known to exist in a dormant state in terms of their cross-linking activity, when the 4.1 ERM domains interact with their own C-terminal tail (; ). Phosphorylation on the conserved threonine residues in the C terminus (T567/T564/T568 in ezrin/radixin/moesin, respectively) is considered to cause conformational changes in ERM proteins, resulting in their activation to interact with other molecules (). The suppressed expression of ERM proteins in cultured cells by antisense oligonucleotides leads to the loss of microvilli, suggesting their functional significance in microvillus formation (). Although moesin-deficient mice show no obvious abnormalities (), the absence of radixin causes the disappearance of microvilli in bile canalicular membranes of hepatocytes in mice (). In addition, ezrin knockout and knockdown mice as well as EBP50-deficient mice exhibit shortened and irregular microvilli in enterocytes with differences in their extent (; ; ), indicating their important roles in microvillus development. Thus, although molecular constituents of microvilli and their interactions have been disclosed, it remains obscure which signals cue microvillus morphogenesis.
Hepatocyte nuclear factor 4α (HNF4α), a member of the nuclear receptor superfamily, transcriptionally regulates the expression of many target genes involved in glucose, fatty acid, amino acid, ammonia, cholesterol, steroid, and drug metabolism as well as in hematopoiesis and blood coagulation (; ; ; ; for reviews see ; ). During early development, it is initially detected in primitive endoderm cells and afterward is expressed in visceral endoderm cells (), which cover the fetal components and possess various properties similar to those in hepatocytes (for review see ). In the adult, HNF4α is expressed in several types of epithelial cells, such as hepatocytes, enterocytes, β cells of the pancreas, and proximal tubular epithelia in the kidney, and is reported to contribute to the differentiation of these cells. Several lines of evidence have made it clear that HNF4α also plays an essential role in activation of the expression of genes encoding cell junction molecules (; ; ; ; ). In addition, recent studies using four distinct cell lines have shown that HNF4α is implicated in the control of cell proliferation (; ; ). Moreover, chromatin immunoprecipitation combined with promoter microarrays has revealed that HNF4α occupies an exceptionally huge number of promoters in human hepatocytes and pancreatic islets, implying its broad range of physiological functions.
Mouse F9 embryonal carcinoma cells show no or little spontaneous differentiation when cultured in the absence of retinoic acid. Conversely, those grown as monolayers and aggregates differentiate upon retinoic acid treatment into primitive endoderm– and visceral endoderm–like cells, respectively, both of which represent polarized epithelial cells bearing junctional complexes (; ). Judging from these properties, F9 cells are regarded as an attractive system to study the molecular mechanisms not only of early embryonic development, cell differentiation, and retinoid signaling but also of the organization of cell junctions and epithelial polarity. To facilitate the study of gene functions of interest in F9 cells, we previously established the cell line F9:rtTA:Cre-ER L32T2 (also called F9 L32T2), which allows sophisticated genetic manipulation such as the sequential inactivation of -flanked genes and strict regulation of gene expression without altering its general characteristics ().
We subsequently generated F9 cells expressing doxycycline (Dox)-inducible HNF4α (F9 L32T2:HNF4α) and found that HNF4α triggered the expression of several tight-junction molecules as well as the establishment of cell–cell junctions and epithelial polarity (; ). Therefore, we hypothesized that HNF4α might also be involved in microvillus morphogenesis. To test this assumption, we used F9 L32T2:HNF4α and the rat lung endothelial (RLE) cell line RLE:rtTA:HNF4α, in which HNF4α expression is also conditionally induced by Dox (). Our study highlights that HNF4α provokes microvillus biogenesis as a morphogen via the induction of EBP50 expression. We also compared the phenotype in retinoid X receptor α (RXRα)/retinoic acid receptor γ (RARγ)–deficient F9 cells (; ) with that in wild-type (WT) cells and determined the functional relevance of retinoid receptors in microvillus formation. Furthermore, we examined whether the up-regulation of EBP50 was sufficient for the morphogenesis of microvilli in F9 cells.
By scanning and transmission electron microscopy, we first determined whether microvillus formation was also induced in the cells. In undifferentiated cells grown in the absence of Dox, filopodia- and lamellipodia-like structures were observed, but there were few microvilli (, A and B; top). On the other hand, in F9 L32T2:HNF4α cells treated for 72 h with 1 μg/ml Dox, the number of microvilli, which contained parallel actin filaments, was strikingly increased on the apical cell surfaces and along the cell borders (, A and B; bottom). Note that the majority of the Dox-exposed cells exhibited microvilli with variations in number and length and that tightly packed arrays of microvilli were frequently detected.
Because HNF4α provoked microvillus biogenesis in F9 cells, we hypothesized that HNF4α may induce the gene expression of microvillus components such as ERM proteins and EBP50. As shown in , the expression of EBP50 mRNA was activated in F9 cell lines expressing Dox-induced HNF4α. The levels of EBP50 transcripts in F9 L32T2:HNF4α cells were elevated by Dox in time- and dose-dependent manners in terms of the amount of HNF4α (for HNF4α expression, see , ) and were three- to fourfold higher in the cells treated for 48 h with 1 μg/ml Dox than in the cells grown without Dox exposure. In contrast, the expression of ezrin, radixin, and moesin mRNAs was marginally altered in the cells after Dox treatment.
By Western blot analysis, we subsequently checked the expression levels of these four microvillus proteins as well as that of villin in F9 L32T2:HNF4α cells. EBP50 protein was abundantly induced in the cells after 24 and 48 h of 1 μg/ml Dox treatment (approximately 6- and 11–15-fold increases, respectively; ). In addition, the level of EBP50 protein appeared to be increased by Dox and, thereby, HNF4α in a dose-dependent fashion (). The expression of villin was moderately induced in the cells (Fig. S1, A and B; available at ), as reported for human embryonic kidney cells expressing HNF4α (). In contrast, little or no change in the expression levels of ezrin or moesin was observed in the cells after Dox exposure, and radixin expression was only weakly increased in cells treated for 48 h with 1 μg/ml Dox (1.5–2-fold increase). Together with the data of RT-PCR analysis, these results indicated that HNF4α extensively activated the expression of EBP50.
ERM proteins are reported to be activated after phosphorylation at the conserved C-terminal threonine residues (T567/T564/T558 residues in ezrin/radixin/moesin, respectively) in order to interact with other molecules (; for review see ). Therefore, by immunoblotting with an antibody against the phospho–ezrin (T567)/radixin (T564)/moesin (T558), we next analyzed the phosphorylation states of ERM proteins in F9 L32T2:HNF4α cells (). The threonine-phosphorylated ERM proteins were barely detected in the cells grown without Dox, whereas their levels were extremely increased in the cells exposed for 72 h to 1 μg/ml Dox (about a 70-fold increase). On the other hand, the amounts of ezrin, radixin, and moesin were not largely altered after Dox treatment. Thus, ERM proteins were heavily phosphorylated on the conserved C-terminal threonine residues in F9 cells expressing Dox-induced HNF4α.
We then examined the distribution of ezrin, radixin, moesin, phospho-ERM, EBP50, and villin in F9 L32T2:HNF4α cells by immunostaining. In the cells exposed for 72 h to 1 μg/ml Dox, these proteins were concentrated on the apical cell surfaces and cell boundaries in a dotlike manner, which was consistent with the localization pattern of microvilli (). Ezrin, radixin, moesin, and EBP50 proteins were also colocalized, at least in part, with filamentous actin (). Conversely, in the cells grown without Dox, they showed neither apical enrichment nor any specific subcellular distribution except moesin exhibiting weak positive signals along lateral and apical plasma membranes ().
In villin-deficient mice, no gross abnormalities are observed in microvillus formation (; ). Therefore, the important question that arises from the aforementioned results is whether HNF4α-induced EBP50 expression is required for the phosphorylation and apical concentration of ERM proteins as well as microvillus formation. To address this issue, we used an RNAi approach to knock down the expression of EBP50. F9 L32T2:HNF4α cells were transfected with the siRNAs against EBP50 or negative control siRNA and were incubated for 6 h after transfection followed by treatment for 72 h with 1 μg/ml Dox. Western blot analysis revealed that the three distinct EBP50 siRNAs effectively reduced the EBP50 expression (). Importantly, when the expression of EBP50 was suppressed in the cells, the levels of threonine-phosphorylated ERM proteins were strongly decreased, whereas total amounts of ezrin, radixin, and moesin were basically not affected (). In addition, the reduction of EBP50 expression in the cells resulted in impairment of the apical enrichment of ezrin and radixin (). Moreover, the knockdown of EBP50 expression led to a remarkable decrease in both the number and length of microvilli on the apical cell surface (). Altogether, these results indicated that HNF4α triggered the phosphorylation and apical concentration of ERM proteins as well as microvillus morphogenesis through the up-regulation of EBP50 expression.
To elucidate whether the induction of HNF4α causes microvillus biogenesis in another type of cell, we subsequently used the vascular endothelial cell line RLE:rtTA:HNF4α, in which HNF4α expression was induced by Dox treatment as described previously (). On scanning electron microscopy, only a small number of fingerlike protrusions were detected on apical surfaces of the cells grown in the absence of Dox (, top). In contrast, in the cells exposed for 72 h to 1 μg/ml Dox, the number and length of microvilli-like structures were extremely increased (, bottom) as in the Dox-treated F9 L32T2:HNF4α cells. On transmission electron microscopy, however, parallel actin bundles were not apparent in fingerlike extensions of the Dox-exposed cells (). Furthermore, the expression of EBP50 and villin was induced in the cells after 24 and 48 h of 1 μg/ml Dox treatment, and its levels were elevated by Dox in a dose-dependent manner (; for HNF4α expression, see ). On the other hand, the expression of ezrin, radixin, and moesin was not grossly changed in the cells after Dox exposure. Immunostaining showed that ezrin, phospho-ERM, EBP50, and villin were apically enriched in the cells treated for 72 h with 1 μg/ml Dox but not in the vehicle-treated cells (Fig. S2, available at ).
We previously reported that retinoid receptors like HNF4α induced the gene expression of the same tight-junction molecules (occludin, claudin-6, and claudin-7) as well as the formation of functional tight junctions and epithelial polarity (; ; ). Therefore, we investigated whether retinoid receptors contributed to microvillus formation. To this end, we used WT and RXRα/RARγ F9 cells, the latter of which are defective for differentiation into epithelial cells (; ). Scanning electron microscopy showed that in WT cells treated for 96 h with 10 M all-trans retinoic acid (tRA), a small number of short microvilli developed on the apical-free surfaces (, top; arrows). On the other hand, in the tRA-exposed RXRα/RARγ cells, microvilli were hardly observed (, bottom). Note that the number of microvilli was increased in the tRA-treated WT F9 cells compared with that of the vehicle-treated WT cells and the tRA-exposed RXRα/RARγ cells (, b and d; and not depicted), although it was much smaller than on F9 L32T2:HNF4α cells treated with Dox ().
We also examined the expression of ezrin, radixin, moesin, EBP50, and villin in WT and RXRα/RARγ F9 cells. As shown in (B and C) and Fig. S1 C, the expression of ezrin but not radixin, moesin, EBP50, or villin was induced in WT cells after 48 h of 10 M tRA treatment and was dose-dependently increased in the cells by tRA. Interestingly, no induction of ezrin expression was detected in the tRA-exposed RXRα/RARγ cells (), indicating that retinoid signals for the induction of ezrin expression were mediated by these retinoid receptors. Moreover, ERM proteins were barely phosphorylated in WT F9 cells exposed for 96 h to 10 M tRA compared with the Dox-treated F9 L32T2:HNF4α cells ().
We subsequently determined whether EBP50 overexpression was sufficient for microvillus formation in F9 cells. When F9 cells were transfected with EBP50 cDNA, little or no phospho-ERM protein was detected (). Interestingly, ERM proteins were phosphorylated when the EBP50 transfectants were exposed to 10 M tRA for 72 h, although their signals were relatively weak compared with those in the Dox-treated F9 L32T2:HNF4α cells, most probably because partial fractions of F9 cells were transfected with EBP50. In addition, ezrin, villin, and phospho-ERM proteins as well as EBP50 appeared to be apically concentrated in EBP50-overexpressed cells that were grown in the presence of tRA, whereas they were only sparsely detected on the apical cell surfaces in the tRA-treated nontransfectants ( and Fig. S3 A, available at ). In the vehicle-exposed transfectants, EBP50 and ezrin but not villin or phospho-ERM were observed in a dotlike pattern at a different height of multilayer of the cells (, Fig. S3 A, and not depicted). Furthermore, on scanning electron microscopy, well- developed microvilli were observed on apical cell surfaces of EBP50 transfectants exposed to 10 M tRA for 96 h but not in the vehicle-treated nontransfectants (). The number and length of microvilli were fewer and shorter in both vehicle-treated transfectants and tRA-exposed nontransfectants than those observed in the tRA-treated transfectants (Fig. S3). Collectively, these results strongly suggest that the up-regulation of not only EBP50 but also other molecules that could be induced by retinoid receptors are required for HNF4α-triggered microvillus morphogenesis.
We and others previously showed that HNF4α activated the expression of cell adhesion molecules, resulting in the acquisition of epithelial cell polarity and junctional complexes (; ; ; ). In the present study, we have demonstrated by scanning and transmission electron microscopy that HNF4α also provokes the biogenesis of microvilli in F9 cells. In addition, the ectopic expression of HNF4α in the vascular endothelial cell line RLE initiated the development of apical fingerlike protrusions, although these did not bear visible parallel actin bundles. Because HNF4α is known to be expressed in epithelial cells such as enterocytes, proximal tubular epithelia, and hepatocytes, in which well-developed microvilli are observed, it may also be involved in microvillus organization in these cells. The importance of HNF4α in microvillus formation is supported by the observation that microvilli are absent in the bile canaliculi of hepatocytes in mice with conditional knockout of the gene (). Thus, we concluded that HNF4α played a fundamental role not only in cell junction formation but also in microvillus morphogenesis, establishing at least two aspects of apicobasal cell polarity. It should also be noted that HNF4α-induced microvilli in F9 cells possessed some variations in number and length compared with the brush border-type microvilli in vivo. This might be explained by differences in the differentiation stages of epithelial cells derived from F9 stem cells.
One of the best-known mechanisms that activates ERM proteins as cross-linkers is phosphorylation on the conserved threonine residues in the C terminus (; ; ; for reviews see ; ). We have indeed shown that ERM proteins appear to be heavily phosphorylated at the threonine residues and apically concentrated during the HNF4α-initiated formation of microvilli in F9 cells. This finding reinforces the notion that ERM proteins become activated and exhibit asymmetrical subcellular localization via their phosphorylation to perform their functions.
Our results also revealed that the expression of EBP50 was induced by HNF4α in F9 L32T2:HNF4α and RLE:rtTA:HNF4α cells, whereas that of ezrin, radixin, and moesin was marginally altered in these cells. We examined the published gene array data and found that EBP50 expression is reduced in the HNF4α-null embryonic liver (). More importantly, we found that the suppression of EBP50 expression by RNAi in F9 L32T2:HNF4α cells hindered threonine phosphorylation and apical enrichment of ERM proteins as well as microvillus biogenesis. These findings clearly indicated that HNF4α activated ERM proteins and drove microvillus morphogenesis via the induction of EBP50. In other words, the expression of EBP50 is required for both the activation of ERM proteins and the formation of microvilli by HNF4α. EBP50-deficient mice possess not only disorganized and shortened microvilli in the intestine but also concomitant reduction in phosphorylated ERM proteins at the apical membranes of polarized epithelia (), further supporting our conclusion.
Furthermore, we have shown that tRA induces the expression of ezrin but not radixin, moesin, or EBP50 in WT F9 cells. Up-regulation of ezrin by tRA in F9 cells has also been recently reported (). Nevertheless, tRA faintly caused the phosphorylation of ERM proteins and microvillus formation in WT F9 cells compared with those in Dox-treated F9 L32T2:HNF4α cells. It is also noteworthy that neither the induction of ezrin expression nor the weak biogenesis of microvilli was observed in the tRA-exposed RXRα/RARγ cells, revealing that retinoid-induced events were definitely mediated by these retinoid receptors. Importantly, phosphorylation and apical enrichment of ERM proteins as well as microvillus formation were much more strikingly initiated when EBP50-overexpressed F9 cells were treated with tRA than those in the vehicle-treated transfectants and the tRA-exposed nontransfectants. Thus, in addition to EBP50 up-regulation, other molecules that are also induced by tRA are most likely necessary for the HNF4α-provoked activation of ERM proteins and microvillus biogenesis. It will be important in future studies to identify additional factors whose expression is induced by both HNF4α and retinoid receptors that trigger microvillus morphogenesis.
In summary, we have provided evidence showing that HNF4α functions as a robust morphogen to activate ERM proteins and provoke microvillus formation via the induction of EBP50 expression. The F9 L32T2:HNF4α cell line, like F9 L32T2, allows various experimental manipulations, such as tamoxifen-dependent Cre-mediated recombination, inducible gene expression (), and RNA interference (this study). In addition, it can be differentiated simply upon the addition of Dox from nonepithelial cells into polarized epithelial cells harboring prominent microvilli and mature junctional complexes (; ; and this study). Thus, the F9 L32T2:HNF4α cell line is definitely a powerful system that can be used to investigate molecular mechanisms underlying the process of epithelial polarization.
Rat mAbs against ezrin, radixin, and moesin were purchased from Sanko Junyaku. Rabbit pAbs against EBP50, phospho–ezrin (Thr567)/radixin (Thr564)/moesin (Thr558), and actin were obtained from Affinity BioReagents, Inc., Cell Signaling Technology, and Sigma-Aldrich, respectively. A goat pAb against villin was purchased from Santa Cruz Biotechnology, Inc. The secondary antibodies used were as follows: HRP-conjugated anti–rat, anti–rabbit, or anti–goat IgG (DakoCytomation), AlexaFluor488 (green)-labeled anti–rabbit or anti–goat IgG (Invitrogen), AlexaFluor594 (red)-labeled anti–rabbit IgG (Invitrogen), and FITC-conjugated anti–rat IgG (DakoCytomation).
The F9 mouse embryonal carcinoma cell line F9 L32T2, which exhibits both Dox-inducible gene expression (Tet-on; ) and tamoxifen-dependent Cre-mediated recombination () systems, was generated as described previously (). To establish F9 cells expressing Dox-induced HNF4α (F9 L32T2:HNF4α), F9 L32T2 cells were electroporated with the expression vector pUHD10-3-rHNF4α, in which the expression of rHNF4α1 is under the control of the Tet operator along with the puromycin-resistant gene expression vector pHRLpuro1 () as described previously (, ). RXRα/RARγ F9 cells were generated as reported previously ().
A Tet-on system in RLE cells was established as for F9 cells (), and it was designated the RLE:rtTA L20 cell line as described previously (). Cells showing the Dox-inducible expression of HNF4α were generated as described previously ().
For scanning electron microscopy, cells grown on coverslips were fixed with 2.5% glutaraldehyde in PBS overnight at 4°C. After several rinses with PBS, they were postfixed in 1% OsO at 4°C for 3 h and washed with distilled water followed by being dehydrated through a graded series of ethanol and freeze drying. Samples were subsequently coated with platinum and examined under a scanning electron microscope (S-4300; Hitachi).
For transmission electron microscopy, cells were fixed with 2.5% glutaraldehyde and 0.1 M cacodylate buffer, pH 7.3, overnight at 4°C. After washing with 0.1 M cacodylate buffer, pH 7.3, they were postfixed in 1% OsO and 1.5% potassium ferrocyanide in 0.1 M cacodylate buffer for 2 h. Samples were subsequently stained with uranyl acetate for 2 h at room temperature, washed, and dehydrated followed by embedding in Epon 812. Ultrathin sections were cut with a diamond knife, stained with lead citrate, and examined with an electron microscope (1200Ex; JEOL) at an acceleration voltage of 100 kV.
For analysis of gene expression, total RNA was isolated from cells using TRIzol reagent (Invitrogen), and RT-PCR was performed as previously described (, ; ). The PCR primers for mouse cDNAs were as follows: ezrin (GenBank/EMBL/DDBJ accession no. ), 5′-AGAGTACACGGCCAAGATC-3′ (nt 1,361–1,379) and 5′-TCTACATGGCCTCGAACTC-3′ (nt 1,839–1,857); radixin (GenBank/EMBL/DDBJ accession no. ), 5′-AAGCCAAGTCTGCAATCGC-3′ (nt 1,358–1,376) and 5′-TTGCCTTGTCGAATCTGCC-3′ (nt 1,862–1,880); moesin (GenBank/EMBL/DDBJ accession no. ), 5′-GAACTTGAGCAGGAACGGA-3′ (nt 1,087–1,105) and 5′-CAGTCGCATGTTCTCAGCA-3′ (nt 1,602–1,620); and EBP50 (GenBank/EMBL/DDBJ accession no. ), 5′-AGGTCAATGGTGTCTGCA-3′ (nt 724–741) and 5′-CTTTAGCCACAGCCAAGGA-3′ (nt 1,104–1,122). The primers for 36B4 were described previously ().
Cells were grown on 60-mm tissue culture plates, washed twice with ice-cold PBS, and scraped with 150 μl of ice-cold NaHCO buffer (1 mM NaHCO and 1 mM PMSF, pH 7.5). They were subsequently collected into a microcentrifuge tube, sonicated for 10 s, and put on ice for 30 min. Total cell lysates were resolved by one-dimensional SDS-PAGE and electrophoretically transferred onto a polyvinylidene difluoride membrane (Immobilon; Millipore). The membrane was saturated with PBS containing 4% skim milk and incubated for 1 h at room temperature with primary antibodies in PBS. After rinsing in PBS containing 0.1% Tween 20, the membrane was incubated for 1 h at room temperature with HRP-conjugated anti–rat or anti–rabbit IgG (diluted 1:1,000) in PBS. For the detection of phospho-ERM proteins, TBS was used instead of PBS, and 5% BSA with 0.1% Tween 20 in TBS was used as a blocking buffer. It was then rinsed again and finally reacted using an ECL Western blotting detection system (GE Healthcare). The blots were stripped with Restore Western blot stripping buffer (Pierce Chemical Co.) according to the manufacturer's instructions and immunoprobed with an antiactin antibody. Signals in immunoblots were quantified using Image 1.62c software (Scion).
Cells grown on coverslips were fixed in 1% formaldehyde in PBS for 10 min. After being washed three times with PBS, they were treated with 0.2% Triton X-100 in PBS for 10 min, rinsed again with PBS, and preincubated in PBS containing 5% skim milk. They were subsequently incubated for 1 h at room temperature with primary antibodies and/or rhodamine phalloidin and rinsed again with PBS followed by a reaction for 1 h at room temperature with appropriate secondary antibodies. For immunohistochemistry of phospho-ERM proteins, TBS was used instead of PBS, and 5% BSA in TBS was used as a blocking solution. All samples were examined using a laser-scanning confocal microscope (MRC 1024; Bio-Rad Laboratories) and a planApo 60× NA 1.40 oil immersion objective (Nikon). Photographs were recorded with a computer (PowerEdge 2200; Dell) and OS/2 Warp software (IBM) and were processed with Photoshop 6.0 (Adobe). Observations were made at room temperature.
Stealth siRNA duplex oligonucleotides against mouse EBP50 were synthesized by Invitrogen. The sequences were as follows: EBP50 RNAi #1, sense (5′-GGACCGAAUUGUGGAGGUCAAUGGU-3′) and antisense (5′-ACCAUUGACCUCCACAAUUCGGUCC-3′); EBP50 RNAi #2, sense (5′-CCAGCGAUACCAGUGAGGAGCUAAA-3′) and antisense (5′-UUUAGCUCCUCACUGGUAUCGCUGG-3′); and EBP50 RNAi #3, sense (5′-UACCAGUGAGGAGCUAAAUUCCCAA-3′) and antisense (5′-UUGGGAAUUUAGCUCCUCACUGGUA-3′). Cells were transfected with 100 pmol siRNAs or Stealth RNAi negative control by using LipofectAMINE 2000 reagent (Invitrogen) according to the manufacturer's protocols and were treated 6 h after transfection followed by exposure to Dox. For transient transfection of EBP50 cDNA, cells were transfected with the empty vector or pEGFP-EBP50 (provided by T. Shibata, National Cancer Research Institute, Tokyo, Japan; ) by using LipofectAMINE 2000 reagent and were treated for 6 h after transfection followed by culture in the presence or absence of tRA.
Fig. S1 shows the effects of HNF4α and retinoic acid on villin expression in F9 monolayers. Fig. S2 shows the staining pattern of ERM protein, EBP50, and villin in RLE cells expressing Dox-induced HNF4α. Fig. S3 shows the staining pattern of ezrin and villin and scanning electron microscope images in EBP50-transformed F9 cells and nontransfectants that were treated with either the vehicle or tRA. Online supplemental material is available at . |
The vascular system is the first functionally developed organ system during ontogeny. The development of blood vessels starts with differentiation of mesodermal precursor cells toward endothelial cells (ECs), hematopoietic cells, and vascular smooth muscle cells. The first vascular plexus in embryos is formed by de novo aggregation of hemangioblasts (vasculogenesis). Thereafter, vessel growth mainly occurs by the sprouting of capillaries from preexisting vessels (angiogenesis; ; ). Endothelial sprouting is a complex process that involves the increase of vascular permeability, vessel wall disassembly, degradation of the basement membrane, migration and proliferation of ECs, and, finally, the formation of a capillary lumen. The newly formed sprouts are subsequently stabilized by formation of cellular junctions, extracellular matrix deposition, and recruitment of perivascular cells (; ). Among other proteases, matrix metalloproteinases (MMPs) are required for proteolytic remodeling of the extracellular matrix. Degradation of the EC basement membrane, as well as cleavage of helical interstitial collagens, is rate limiting for endothelial sprouting (; ; ).
Although angiogenic signaling through growth factors like VEGF or basic FGF and their receptors has been studied in great detail, the role of endothelial transcription factors orchestrating the angiogenic response is less well understood. Genetic approaches suggested that transcription factors of the hypoxia-inducible factor, Ets, Gata, Hox, Hey, SCL/Tal, and Smad families are implicated in different stages of vascular development (; ). Recently, an increasing body of evidence associated members of the Activator Protein-1 (AP-1) family of transcription factors either directly or by cooperation with aforementioned factors with angiogenic responses and/or programs (; ; ). AP-1 consists of homo- or heterodimers of Jun (c-Jun, JunB, and JunD), Fos (c-Fos, FosB, Fra-1, and -2), and ATF (ATF-2, -3, -4, and ATFa) family members (; ; ). Although different Jun factors have been reported to regulate genes implicated in angiogenesis, like VEGF and MMPs (; ; ), a direct link between endothelial Jun proteins and angiogenesis has thus far remained questionable. When the different Jun members were deleted in mice, only JunB-deficient embryos displayed vascular abnormalities and died between embryonic day (E)8.5 and E10 because of placental failure (). Conditional gene targeting provided novel insight into the physiological processes regulated by JunB. Bone development (; ) and adaptive and innate immunity (; ), as well as wound healing and epidermal proliferation (; ), are affected by the loss of JunB. Furthermore, JunB-deficient mice develop a myeloproliferative disease similar to human chronic myeloid leukemia ().
We investigated the cell-autonomous role of JunB in EC function in vivo and in vitro. EC-specific ablation of JunB resulted in early embryonic lethality, underscoring an essential role for JunB in vessel development in vivo. We found that capillary sprouting of JunB-deficient aortic explants was strongly diminished, underlining the crucial role of JunB as a regulator of angiogenic programs. Isolated ECs lacking JunB expression failed to form capillary-like structures when cultured on Matrigel.
() as a novel JunB target gene. Importantly, reintroduction of CBFβ alone rescued the tube formation defect of JunB-deficient ECs, implying a critical role for CBFβ in EC morphogenesis. In line with these findings, expression of the common AP-1 and CBF target metalloproteinase MMP-13 was impaired. Consequently, ECs isolated from MMP-13–deficient mice failed to form capillary-like networks on Matrigel, thus, recapitulating the phenotype of JunB-deficient ECs.
JunB is expressed in ECs during vascular development (). To investigate the cell-autonomous function of JunB in ECs, we generated mice that lack JunB expression specifically in their ECs. -Cre transgenic mice were crossed to −/flox mice to generate a −/Δ allele in ECs ( −/Δ). To better visualize ECs, the mice were crossed to - lacZ reporter mice. The Cre-mediated deletion of the locus was detected by PCR analysis (unpublished data). Coexpression of JunB and the EC marker CD31 was confirmed by immunofluorescence staining of control embryos (). In contrast, −/Δ embryos showed no JunB expression in blood vessels, whereas other cell types (e.g., neurons) retained JunB expression. These data confirm the efficient deletion of the floxed allele, specifically in ECs. Embryonic lethality was observed around E10 in −/Δ mice. The embryos were already severely retarded at E9.5 compared with controls (). Embryos had not yet turned and displayed enlarged pericardial sacs. At E10.5, control mice developed a highly organized vascular tree in the head region, whereas −/Δ embryos showed a disorganized vasculature with aberrantly branching and dilated vessels (). Additionally, the yolk sacs of −/Δ embryos showed marked defects in vascular remodeling, as vessels formed only a primitive vascular plexus. In contrast, a hierarchically organized vessel structure developed in control mice (). Interestingly, this phenotype closely resembled the vascular defects observed in the embryo proper of the complete knockout mice (). However, vascularization of the placenta was not impaired in −/Δ mice (Fig. S1, available at ). These data define an important role for JunB in the endothelial lineage in vivo.
Because of the early embryonic death, it was difficult to investigate altered EC function in −/Δ mice. Therefore, we used a second transgenic mouse model with conditional inactivation of JunB. In -Cre/ −/Δ mice ( −/Δ), efficient recombination of the locus occurred in many cell types, as reported previously (). In contrast to completely JunB-deficient mice, these animals were viable () because of the fact that the deletion of occurred at a later stage of embryonic development (around E14; unpublished data). Analysis of genomic DNA revealed complete deletion in aortic tissue and isolated primary ECs of −/Δ mice (). To prove loss of JunB expression on protein level, primary ECs were isolated from these mice. Expression of endothelial marker proteins was confirmed by FACS analysis and immunofluorescence staining for CD31, endoglin (CD105), VEGFR-2, or VE-cadherin ( and not depicted), demonstrating the isolation of a 94–98% pure population of ECs. Immunoblotting revealed that the JunB protein was indeed undetectable in −/Δ ECs, whereas it was expressed in control ECs ().
To investigate whether −/Δ mice have a defect in blood vessel growth, mouse aortic ring assays were performed with explanted tissue of control or −/Δ aortae (). Within 7 d of explant culture, strong outgrowth of capillary sprouts from the aortic rings was observed in control () or -lacZ reporter mice. Staining for LacZ activity or immunofluorescence staining for CD31 confirmed that these sprouts originate from ECs and that they form a vessel lumen (Fig. S2, A–E, available at ). In contrast, −/Δ rings showed nearly no capillary outgrowth (). To exclude that the effect is solely caused by paracrine VEGF acting on ECs, aortic ring assays were performed in the presence of exogenously added recombinant VEGF. In control aortae, the addition of VEGF to the medium did not further increase the already strong sprouting activity (), indicating that the VEGF levels in the medium are already in a saturation range. In −/Δ explants, additional VEGF had no effect on sprout formation (). Thus, VEGF is not the limiting factor for sprouting in this system, and JunB most likely induces capillary growth via other mechanisms.
The capability of primary ECs to form tube-like structures was investigated upon cultivation on Matrigel (). Control ECs formed capillary-like structures that connected to anastomosing networks within 18 h in culture. In sharp contrast, −/Δ ECs largely failed to generate tubes, but instead formed cellular aggregates, even if cultured for longer time periods. Like in the aortic ring assay, addition of recombinant VEGF did not rescue the tube formation defect (unpublished data). Our finding was supported by the use of a second model involving JunB-deficient transformed ECs, so-called endothelioma (END) cells that provide a model for activated embryonic endothelium (). Control END cells formed tubelike structures and networks of EC cords very rapidly (within 6 h) on Matrigel (). As observed for −/Δ ECs, cord formation of JunB-deficient END cells was drastically decreased (13% of the control cells), which was quantified by counting interendothelial spaces (). A similar phenotype was observed with END cells derived from flox/flox mice, in which the floxed allele was deleted by transduction of a retroviral Cre expression vector ( Δ/Δ; , bottom). The floxed allele was efficiently deleted (∼90%), as revealed by FACS analysis for IRES-eGFP expression (unpublished data) and PCR amplification of the recombined locus (). Defective tube formation could be caused by a failure in proliferation; thus, the proliferation rate of control and −/Δ ECs was determined by BrdU incorporation and subsequent FACS analysis, which revealed no considerable difference (). Lateral migration was investigated by scratching EC monolayers and by monitoring wound closure after 24 and 48 h. A slight delay in wound closure was observed in JunB-deficient cells after 48 h (); however, these small differences in migration alone cannot account for the severe defects in tube formation (). The expression levels of genes typically involved in angiogenesis were analyzed by RT-PCR. No differences in expression of and or their receptors and were observed. Decreased levels were found, but the main signaling receptor, , was unaffected. In addition, similar levels of the AP-1 family members c-Jun, c-Fos, and ATF-2 were detected in control and JunB-deficient ECs (unpublished data). Collectively, analysis of proliferation, migration, and expression of known angiogenic candidate genes did not reveal first signs for a mechanism that could explain the phenotype of JunB-deficient cells. Therefore, a systematic approach was chosen to compare the gene expression profiles of JunB-deficient and control cells.
To search for target genes affected by the loss of JunB in the endothelium, large-scale comparative gene expression analysis was performed using cDNA microarray technology. JunB expression is highly up-regulated by hypoxia (); thus, a combinatorial analysis of gene expression was performed in control and JunB-deficient ECs cultured under normoxia or hypoxia. Under hypoxic conditions, 729 genes were up-regulated in control cells, whereas the expression of 956 genes was repressed. However, by comparing the expression profile of control with JunB-deficient cells under hypoxic conditions, only 23 genes were up-regulated and 16 genes were repressed in a strictly JunB-dependent manner. The newly identified JunB-dependent genes () included transcription factors (, , , , , and ), factors involved in protein synthesis (, , and ), genes involved in intracellular trafficking (, , and ), genes involved in cytoskeletal architecture (, , and ), and genes regulating apoptosis ( and ) and metabolism (, , , and ).
Of the newly identified JunB target genes, our attention was attracted by CBFβ. AP-1 and CBF factors physically interact in a transcriptional complex and transactivate common target genes (; ). CBFβ is the common partner of the Runt-related transcription factors (Runx). Real-time RT-PCR analysis confirmed the results from the microarray analysis, showing a 3.3-fold induction of CBFβ under hypoxia, which was lost in −/− END cells, as well as in Cre-transduced Δ/Δ cells ( and not depicted). Induction of CBFβ correlated with that of JunB by hypoxia in END cells (). To confirm that CBFβ is a JunB target in vivo, bone marrow cells were isolated from control and JunB-deficient mice. Indeed, CBFβ expression was considerably reduced (by 60%) in the absence of JunB, as determined by real-time RT-PCR and Western blot analysis (). Moreover, immunofluorescence staining was performed on embryo sections of −/Δ and control embryos. CBFβ expression colocalized with CD31-positive vessels in control embryos, whereas no CBFβ signal was detected in vessels of −/Δ embryos (). To investigate whether CBFβ is a direct JunB target, the murine promoter region was cloned and different fragments were fused to a luciferase reporter gene. Promoter analysis revealed three potential AP-1 binding sites (12--tetradecanoylphorbol-13-acetate–responsive elements [TREs]) in the distal promoter region (nt −1,496 to −1,061) and one TRE and two cAMP-responsive elements (CREs) in the proximal promoter region (nt −1,127 to −251; , top). Coexpression of either the complete or proximal reporter constructs (nt −1,496 to −251 or −1,127 to −251) with a JunB expression vector in F9 teratocarcinoma cells already increased promoter activity by threefold (, top). Coexpression of JunB with the dimerization partners ATF-2 or c-Fos further increased CBFβ promoter activity four- or eightfold induction, respectively, whereas expression of ATF-2 or c-Fos alone had no effect (, top). The region containing the three distal TREs alone (nt −1,496 to −1,061) showed hardly any activity in this assay (, bottom); therefore, we hypothesized that JunB binding to the CBFβ promoter should take place at the proximal CREs and TRE. To finally prove the physical interaction of JunB with the CBFβ promoter, chromatin immunoprecipitation (ChIP) analysis was performed in ECs in the absence or presence of CoCl. Primers were designed to amplify specific regions of the CBFβ promoter (P1, P2, and P3) that contain the TRE and CREs for potential binding of JunB. We found JunB binding at the proximal CREs and TRE (, P1 and P2). This binding was strongly enhanced in CoCl-treated cells. In contrast, no interaction was observed at the three distal TREs (P3), confirming the data of the luciferase reporter assay (). To find the dimerization partner of JunB regulating CBFβ expression, ChIP analysis was performed using antibodies against ATF-2, c-Fos, and Fra-1 that are expressed in ECs. We only detected a constitutive interaction of ATF-2 with the promoter in uninduced and CoCl-treated ECs (, bottom). The binding occurred at the proximal CREs and TRE (P1), in the same region where JunB binding was detected. No ATF-2 binding was found at the three distal TREs. We detected no interaction of c-Fos or Fra-1 with the promoter (unpublished data). These data strongly suggest that either a JunB–JunB homodimer or a JunB–ATF-2 heterodimer regulates CBFβ expression in ECs in vivo.
To determine whether CBFβ is functionally implicated in the tube-forming process of ECs, we attempted to rescue the phenotype. By retroviral gene transfer CBFβ, JunB, or empty control vector (Fig. S3 A, available at ) were reintroduced into JunB-deficient END cells. After culture in selective medium, >96% of the cells integrated the retroviral vector into their genome, as determined by FACS analysis for the coexpressed IRES-GFP reporter (Fig. S3 B). Immunoblotting confirmed strong expression of CBFβ or JunB in transduced cells (Fig. S3 C). As expected, JunB-deficient END cells transduced with empty vector did not form capillary-like structures when cultured on Matrigel (). Tube formation was efficiently rescued by reexpression of JunB in END cells (). Most notably, CBFβ expression also rescued tube formation in JunB-deficient cells (). To quantify vascular organization in these experiments, the interendothelial spaces were counted, which revealed a significant difference between the END cells transduced with empty vector and those transduced with CBFβ or JunB, respectively (). These data underscore a functional implication of CBFβ in the morphogenic process downstream of JunB.
To mechanistically explain the observed defects in sprouting and tube formation of JunB-deficient cells, we studied the invasive capacity of control and JunB-deficient ECs. The ECs were plated on type I collagen gel, and after 10–14 d control cells changed from polygonal to spindlelike morphology and invaded the gel underlying the monolayer. By focusing through the optical planes, cells underneath the monolayer were identified that displayed long cellular protrusions and filopodia, indicating an invasive/angiogenic phenotype (). In contrast, JunB-deficient cells formed a monolayer on top of the collagen, but failed to invade deeper into the gel (). To investigate the molecular basis of impaired invasion, supernatants of control and −/Δ ECs were analyzed for collagenolytic activity. In this assay, native helical collagen served as a substrate, which is cleaved only by MMP-8, -13, and -14. Supernatant of JunB-deficient ECs revealed a 40% decrease in overall collagenase activity (), suggesting that reduced collagenase activity accounts for the impaired EC sprouting. RT-PCR analysis revealed strongly reduced expression of not only but also , a known JunB target gene (), in JunB-deficient ECs in comparison to control cells. In contrast, and mRNA levels were not affected (), whereas expression was not detectable in cells of both genotypes (unpublished data). The reduced expression of and was confirmed by real-time RT-PCR analysis (). MMP-13 represents a well-documented common target of AP-1 and CBF transcription factors that can physically interact and cooperate in the transactivation of MMP-13 in osteoblastic cells (; ). We were able to confirm these findings in ECs (Fig. S4, available at ).
To unequivocally determine the function of MMP-13 in this process, primary ECs were isolated from −/− mice (; ) and applied in a tube formation assay. Indeed, MMP-13–deficient ECs largely failed to form capillary-like networks on Matrigel (14% compared with control cells), confirming an essential role of MMP-13 in tube formation in our model system (). Finally, we investigated MMP-13 expression in JunB-deficient END cells transduced with CBFβ. Although −/− END cells transduced with empty vector showed no MMP-13 expression, introduction of CBFβ strongly up-regulated MMP-13 expression (). Together, these findings suggest that the defects in tube formation and sprouting observed in JunB-deficient ECs are a consequence of impaired CBFβ expression, resulting in reduced expression of proteases.
Previously, we have reported critical functions for JunB in controlling cytokine-regulated mesenchymal–epidermal interactions in skin by regulating keratinocyte proliferation and differentiation in both a paracrine and autocrine manner (, ). Very recently, we could demonstrate an essential role for JunB in basal and hypoxia-mediated VEGF expression and tumor angiogenesis, implying a paracrine mechanism for fibroblast or tumor cell–derived VEGF acting on the endothelium (unpublished data). As JunB is also highly expressed in ECs, in this study we specifically addressed the EC-intrinsic requirement of JunB and JunB-dependent genetic programs. Indeed, JunB deletion in the endothelial lineage using -Cre mice resulted in severe vascularization defects in the yolk sac and embryo proper and early embryonic death, suggesting an EC-intrinsic role for JunB. Accordingly, sprout formation of JunB-deficient aortic explants and capability of primary ECs lacking JunB to form capillary-like cords when cultivated on Matrigel was strongly impaired, even in the presence of excessive recombinant VEGF, excluding that limiting VEGF levels account for this phenotype.
In the search for JunB-dependent target genes implicated in EC morphogenesis, we focused on genes differentially expressed in JunB-deficient ECs kept under normoxic or hypoxic conditions compared with control cells to mimic a physiological condition for angiogenesis induction. Applying stringent criteria for expression profiling, 23 up- and 16 down-regulated genes were identified to be dependent on JunB and hypoxia. These genes can functionally be grouped in cytoskeletal components, transcription factors, translational factors, and genes implicated in apoptosis, cellular trafficking, or metabolism.
For example, the group of cytoskeletal genes includes , which is an intermediate filament protein that is highly expressed in neuronal stem cells, but also in activated ECs. As expression is down-regulated in quiescent ECs, it was recently designated as a specific marker for angiogenic ECs and probably a modulator of cytoskeletal dynamics in activated ECs (). A second cytoskeletal gene found in our screen is , which encodes a Rho-GTPase–binding protein that modulates Rho activity, and thereby influences rearrangement of the actin cytoskeleton (). Previously, it has been shown that Rho-mediated actin organization is essential for cord formation of angiogenic ECs (; ). A second group of JunB target genes is comprised of transcriptional regulators, such as , which is a mammalian homologue of the transcription factor family originally found in and implicated in BMP-4 signaling. Interestingly, LBP-1a–deficient mice display a phenotype very similar to that of JunB knockout mice, as the embryos die at E10 and have severe defects in the vascularization of placenta and yolk sac (). In summary, the physiological features of these identified target genes imply that JunB is required for the transition from quiescent to angiogenic endothelium.
Yet, we concentrated our analysis on another transcriptional regulator found among the JunB-dependent genes, namely the , which is the common heterodimerization partner of the Runx comprising the three α-subunits Runx-1 (CBFα-2 and AML-1), -2 (CBFα-1), and -3 (CBFα-3). The Runx–CBFβ dimer binds to a specific DNA sequence and, depending on the recruitment of coactivators or co-repressors, activates or suppresses the transcription of target genes, respectively ().
Inactivation of the genes in mice revealed essential functions in definitive hematopoiesis (Runx-1), in bone formation (Runx-2), or in epithelial and neuronal development (Runx-3). Runx-1–deficient embryos had a lower number of small capillaries in the hindbrain; moreover, vessels in the hindbrain, pericardium, and yolk sac showed less branching ().
Deletion of the CBFβ subunit disrupted normal hematopoiesis and caused embryonic lethality between E12.5 and E13.5. Transgenic CBFβ expression in the hematopoietic system rescued embryonic lethality and revealed impaired bone formation in CBFβ-deficient mice (; ). Hemorrhages were observed in the cephalic and lumbar region of CBFβ-deficient embryos starting at E10.5. This was associated with perivascular edema and cell death in areas of actively growing capillaries. It was hypothesized that CBFβ may play a role in certain aspects of vessel development, although this was not tested directly (). Moreover, Runx-1 or CBFβ are frequent targets of chromosomal translocation in humans, which account for 25% of adult acute myeloid leukemia. In −/Δ mice, CBFβ expression was decreased in bone marrow, which is the major site of endogenous CBFβ expression in vivo. Interestingly, phenotypes of CBFβ- and JunB-deficient mice exhibit striking similarities. Mice with reduced JunB expression develop a leukemia-like disease () and have defects in bone formation (; ), providing further physiological evidence for a role of JunB in regulating CBFβ expression.
Our data point to an additional function of CBFβ in the hypoxia-response of ECs. ChIP analysis revealed that is a direct target gene of JunB in ECs. We demonstrate that CBFβ is functionally important for EC morphogenesis, as reexpression of CBFβ in JunB-deficient ECs rescued tube formation on Matrigel. CBFβ, Runx-1, and -2 expression was reported in ECs (; ; ). In line with our findings, a function of CBFβ in ECs has been proposed previously, as overexpression of a dominant-negative mutant of CBFβ inhibited EC tube formation in type I collagen gels (). Similarly, loss of Runx-1, or a truncated Runx-2 protein, perturbs tube formation on Matrigel (; ).
CBF and AP-1 are known to cooperate in the transactivation of the metalloproteinase MMP-13 (; ; ). MMP-13 is an interstitial collagenase able to degrade native fibrillar collagens in the triple helical domain in vivo that is required for endochondral bone formation and homeostasis (; ). Recent evidence points to a critical and specific function of MMP-13 derived from activated stromal cells and inflammation-responsive hematopoietic cells at sites of tumor tissue (; ) and in the chick chorioallantoic membrane in response to angiogenic factors (; ). We provide experimental evidence that loss of EC-intrinsic MMP-13 expression caused by ablation and, consequently, impaired CBFβ expression results in diminished collagenolytic activity of JunB-deficient ECs and failure in sprouting and tube formation. Accordingly, −/− ECs also exhibited strongly reduced tube formation on Matrigel, suggesting that loss of MMP-13 may account for the angiogenic defects in JunB-deficient ECs. Reexpression of CBFβ in JunB-deficient cells led them to regain MMP-13 expression and tube formation. Thus, our data provide strong evidence for a regulatory JunB–CBFβ–protease axis in the EC sprouting process. Interestingly, knockout mice display delayed enchondral ossification and reduced vessel ingrowth into the primary ossification center (; ), presumably caused by the reduced bioavailability of VEGF. It was concluded that MMP-13 produced by chondrocytes might be critical for the release of matrix-bound VEGF. So far there exists only one study stressing an EC-intrinsic role for MMP-13 as an important effector of nitric oxide-activated EC migration (). Thus, our findings underscore the importance of MMP-13 in EC function, which may also contribute to the phenotype of MMP-13–deficient mice.
In conclusion, we demonstrate an essential cell-autonomous function of JunB in ECs and identify CBFβ as a novel JunB target. The fact that a common target of CBF and AP-1, MMP-13, is crucial for capillary-like tube formation suggests important tasks for CBFβ in diseased neovascularization such as cancer and retinopathy. Henceforth, it will be interesting to dissect not only the individual contribution of autocrine and paracrine pathways addressed by JunB but also JunB/CBFβ converging on commonly known, as well as yet to be defined, targets. It is conceivable that these pathways may prove to be a promising target for antiangiogenic therapy in the future.
The generation of +/− (), flox (), -Cre (), -Cre (, ), and −/− mice () has been previously described. -lacZ () transgenic mice were provided by U. Deutsch (University of Berne, Berne, Switzerland). The -Cre or -Cre mice were crossed to +/− mice and their offspring were mated to flox/flox mice to obtain −/Δ mice. +/flox, +/Δ, −/flox, or +/+ mice were used as controls. For genotyping, genomic DNA was isolated and PCR was performed for the locus as previously described (). Embryos were fixed and stained for LacZ activity, as previously described (). For image acquisition, a dissecting microscope (M 10; Leica) equipped with a PlanApo 1.0×/0.04 NA objective (Leica) and a digital camera (DXM1200; Nikon) were used. Images were processed with Photoshop CS software (Adobe).
END cells were established by infection of ECs derived from midgestation embryos with the N-TKmT retrovirus, as previously described (). END cells expressed endothelial marker genes (VE-cadherin, CD31, CD105, and VEGFR-2), as determined by immunofluorescence staining or FACS analysis, and were able to endocytose DiI-acetyl-LDL particles (CellSystems). For hypoxia treatment, END cells were placed in an incubator with 1.5% O partial pressure (CB; Binder) for 16 h.
The murine CBFβ promoter (nt −1,496 to −251) was amplified by PCR using the primers CBFβ1 and CBFβ2 (Table S1, available at ). The promoter was ligated to a tata-luciferase reporter plasmid (). F9 teratocarcinoma cells were cultured and transiently transfected as previously described (). 1 μg CBFβ promoter fused to the luciferase reporter gene was cotransfected with 0.5 μg AP-1 expression plasmids (JunB, c-Fos, and ATF-2 under control of the Rous sarcoma virus promoter) and 0.05 μg -luciferase for normalization. Cells were lysed and luciferase activity was determined using the Dual-Luciferase reporter system (Promega) and a Sirius luminometer (Berthold).
Microvascular ECs were isolated from mouse lungs using a magnetic cell separation method, as previously described (). In brief, magnetic beads (Dynabeads; Dynal) were coated with anti-CD31 antibodies (Mec13.3; BD Biosciences). Two mouse lungs of similar genotype were pooled, minced, and digested with collagenase A solution (Roche). Cells were incubated with anti-CD31–coated Dynabeads and separated in a magnetic field. After washing away unbound cells, the beads were released by trypsin/EDTA treatment. Isolated ECs were analyzed for the expression of endothelial marker genes as described in Cell culture and luciferase reporter gene assay.
Aortic explants were prepared as previously described (). Mouse aortae were dissected free from connective tissue, cut in 1–2-mm pieces using a scalpel, and washed in ice-cold DME. For each tissue explant, a drop of 40 μl rat type I collagen (Serva) was placed into a culture dish. Aortic rings were placed into the droplet with fine forceps. After polymerization of the gel, the dishes were filled with culture medium (see Isolation of primary ECs) supplemented with 40 or 100 ng/ml recombinant VEGF-A164 (R&D Systems). Capillary sprouting was quantified in a semiquantitative manner by classification of aortic rings from 1 (no capillary sprouting) to 4 (strong outgrowth reaching the margins of the collagen matrix). In each experiment, at least nine aortic rings of control and −/Δ mice were scored, and the experiment was repeated four times.
Matrigel (BD Biosciences) was mixed with an equal volume ice-cold DME. 150 μl Matrigel solution was poured into 48-well plates and incubated for at least 6 h at 37°C. 3.5 × 10 cells were seeded on top of the gel and incubated for 6–24 h. Capillary-like tube formation was quantified by counting interendothelial spaces in three randomly chosen optical fields. Each assay was performed in triplicate.
Proliferation was assessed by BrdU incorporation and subsequent FACS analysis using the BrdU Flow kit (BD Biosciences) according to the instructions of the manufacturer. In brief, 10 ECs were seeded on 6-well plates and cultured for 24 h. Cells were incubated with medium containing BrdU for 2 h and subjected to the assay protocol.
ECs were plated in 6-well dishes and grown to confluency. The monolayer was wounded by scratching with a pipette tip and photographed. After 24 and 48 h, the wounded area was photographed again and the wound closure was measured using ImageJ software (National Institutes of Health).
JunB or CBFβ coding sequences were amplified by PCR using the primers junB1, junB2, CBFβ3, and CBFβ4 (Table S1). The fragments were inserted into the pMXpie vector. JunB-deficient END cells were transduced with retroviral vectors as previously described (). In brief, cellular supernatant containing viral particles was transferred to 8 × 10 END cells in 24-well plates. Cells were transduced by spin infection (3 h; 2,000 rpm) and thereafter cultured in selective media containing 3 μg/ml puromycin (Sigma-Aldrich). Efficiency of retroviral infection was determined by FACS analysis of GFP reporter gene expression.
Total RNA was extracted from END cells cultured under normoxic or hypoxic conditions (1.5% O for 16 h) using TRIZOL reagent as suggested by the manufacturer (Invitrogen). A collection of 15,247 sequences from embryonic cDNA libraries was used to generate the 15K NIA cDNA microarray. Details on probe labeling, hybridization, data acquisition, and analysis were previously described ().
Total RNA was isolated from mouse tissue or cultured ECs using peqGold RNApure reagent (PEQLAB) according to the instructions of the manufacturer. RNA was reverse transcribed using AMV reverse transcriptase (Promega). cDNA was used for semiquantitative RT-PCR or quantitative real-time RT-PCR using Absolute QPCR SYBR Green mix (ABgene) and a thermal cycler (iCycler; BioRad Laboratories) controlled by MyiQ software (BioRad Laboratories). Primer sequences for MMPs were previously described (). For primer sequences for β-tubulin, HPRT, CBFβ5 and -6, and JunB 3 and 4 see Table S1.
Whole-cell extracts or nuclear extracts were prepared as previously described (). Proteins were separated by SDS-PAGE and transferred to nitrocellulose membrane. Immunodetection was performed using an enhanced chemiluminescence system (Perkin Elmer Life Science) and the following primary antibodies: anti-JunB (1:500; N17; Santa Cruz Biotechnology, Inc.), anti-RCC1 (1:500; BD Biosciences), anti-CBFβ (1:500; E20; Santa Cruz Biotechnology, Inc.), and anti-HSC70 (1:10,000; Noventa).
Collagenolytic activity in supernatants of EC cultures was determined using the Collagenase Activity Assay kit (CHEMICON International, Inc.) according to the instructions of the manufacturer. Equal numbers of control and JunB-deficient cells were seeded and cultured for 24–48 h. 100 μl of the supernatant were then subjected to the assay protocol.
Cryosections were fixed with acetone at −20°C and stained at RT, as previously described (). An epifluorescence microscope (DMLB; Leica) equipped with PlanApo objectives (10×/0.4 NA, 20×/0.7 NA, 40×/0.85 NA, and 63× oil/1.3 NA) was used. A DXM1200 digital camera was used for documentation, and images were processed with Photoshop CS software. The primary antibodies used were CD31 (BD Biosciences), JunB (210; Santa Cruz Biotechnology, Inc.), and CBFβ (E20; Santa Cruz Biotechnology, Inc.). The secondary antibodies used were goat anti–rat–Alexa Fluor 488 (Invitrogen) and goat anti–rabbit–Cy3 (Dianova).
In brief, 2.5 × 10 END cells were incubated with 200 μM CoCl for 4 h. Cells were cross-linked in 1% formaldehyde for 20 min at RT. Cells were washed and sonicated using a Bioruptor device (Diagenode). Immunoprecipitation was performed using protein A–Agarose. After washing, protein–DNA complexes were eluted and cross-links were reversed. Proteins were digested by proteinase K and DNA was extracted using QIAquick spin columns (QIAGEN).
The SD is indicated by error bars. Unpaired two-tailed tests were performed using Sigma Plot 8.0 software. Significance was assumed for P values (P < 0.05; indicated by asterisks).
Table S1 shows details on PCR primer sequences. Fig. S1 shows the investigation of placental development, Fig. S2 shows the microscopic evaluation of aortic ring assays, Fig. S3 shows the validation of retroviral gene transfer, and Fig. S4 shows promoter analysis in ECs. |
The β4 integrin is a laminin receptor that mediates the stable adhesion of epithelial cells to the basal membrane (when incorporated in hemidesmosomes) and movement of carcinoma cells through the extracellular matrix (when redistributed at actin-rich protrusions; ; ). These roles of β4 in modulating static or dynamic adhesion and in translating positional cues into different cellular fates are consistent with a conventional mechanochemical activity that has been traditionally ascribed to integrins. But, in addition to these properties, β4 is also endowed with an unorthodox function: it can be a positive regulator of the anchorage-independent growth of neoplastic cells (; ; ). This function is counterintuitive given the established notion that transformed cells are able to grow and survive in the absence of adhesion and, thus, without demanding integrin-derived signals (). Nonetheless, this atypical behavior could explain the frequently observed up-regulation of β4 in human carcinomas () and suggests that β4 overexpression could afford cells with a selective advantage for proliferation and survival in neoplastic contexts where tissue architecture and canonical cell–matrix interactions are compromised ().
A few modes of action have been proposed to explain this feature. In some instances, carcinoma cells may secrete abundant quantities of laminin-5, which can ligate β4 and sustain the expansion of nonadherent colonies through activation of a Rac–nuclear factor κB antiapoptotic pathway (). Alternatively, the expression of β4 is accompanied by the translational up-regulation of VEGF, which can act as a survival factor favoring colony formation (). Finally, we have recently demonstrated that β4 can promote anchorage-independent growth in response to activation of the Met receptor for hepatocyte growth factor (HGF; ). This activity does not rely on laminin engagement, as it is retained by a β4 truncated variant devoid of most of the extracellular domain and requires Met-dependent tyrosine phosphorylation of the β4 cytoplasmic domain. However, the signaling pathway regulating this laminin-independent activity remains to be elucidated. Using biochemical, pharmacological, and genetic approaches, we found that the Met-mediated phosphorylation of β4 initiates a previously unexplored transduction pathway that involves Shp2 recruitment, Src activation, and Gab1 phosphorylation and ultimately leads to dedicated stimulation of the MAPK/extracellular signal-regulated kinase (ERK) cascade. These results substantiate the crucial role of β4 in epithelial tumorigenesis and describe a novel integrin-dependent signaling pathway that could be exploited by growth factor receptors for the implementation of oncogenic responses.
We previously demonstrated that the β4 cytodomain can be tyrosine phosphorylated by Met. This event potentiates HGF-mediated cellular invasion through the stimulation of Ras and PI3K () and enhances anchorage-independent growth through still uncharacterized mechanisms (). Although the signals implicated in Met/β4-dependent cell invasion also play a crucial role in cellular proliferation and are likely to contribute to β4-mediated anchorage-independent growth, the involvement of additional transduction pathways cannot be excluded.
To begin to explore new signaling circuits potentially controlled by β4 activity, we performed a progressive tyrosine/phenylalanine mutagenesis along the β4 tail and mapped three tyrosines (Tyr1257, Tyr1440, and Tyr1494) that, when sequentially mutated into phenylalanines, cause a gradual reduction in β4 phosphotyrosine content up to complete abrogation (). Interestingly, an in silico analysis of the sequences surrounding these tyrosines revealed common structural features (): Tyr1257 and Tyr1494 are both located within immune T cell inhibitory motifs that have been characterized as canonical binding regions for protein and lipid phosphatases, including the SH2-containing tyrosine phosphatase Shp2; similarly, Tyr1440 is embedded in a degenerated consensus for Shp2. Based on these observations, we hypothesized that Tyr1257, Tyr1440, and Tyr1494 might be collectively involved in central signaling functions whose trait of union is multiple binding for Shp2.
To investigate this issue, we initially tested the potential association between β4 and Shp2 using an overexpression system (). COS-7 cells were transiently transfected with the following: the wild-type forms of Met and β4, a condition in which Met overexpression results in elevated tyrosine phosphorylation of the integrin; wild-type β4 and a kinase-inactive variant of Met (Met), which is unable to phosphorylate the integrin; wild-type Met and β4 mutants bearing single (β4-ΔShp2), double (β4-ΔShp2), or triple (β4-ΔShp2) phenylalanine substitutions of the three critical tyrosines, with the consequent progressive reduction of β4 tyrosine phosphorylation levels (); and wild-type β4 alone or wild-type Met alone as negative controls.
First, we examined direct interaction in vitro between β4 and Shp2 in far-Western analysis using a GST fusion protein of the Shp2 N-terminal SH2-containing domain (GST-Shp2). This probe recognized β4 immunoprecipitates when the integrin was fully phosphorylated by activated Met but not when Met-dependent phosphorylation was inhibited (). Association between GST-Shp2 and β4 was gradually weaker when using the single and double β4 mutants and was totally prevented upon expression of the β4 triple mutant (), indicating that all of the putative consensus sites for Shp2 are involved in this interaction. Notably, the efficacy of Shp2 binding paralleled the extent of β4 phosphorylation, suggesting that the tyrosines involved in Shp2 recruitment contain the bulk of the integrin substrate capacity for Met (). Tyrosine phosphorylation of β4 as well as Shp2 binding could not be observed in the absence of transfected Met (). Similarly, the transfection of Met alone was ineffective because COS-7 cells do not express endogenous α6β4 (). When Tyr1257 and Tyr1494 were mutated into phenylalanines, the overall phosphotyrosine content of β4 partially decreased, and β4–Shp2 interaction was less efficient; however, the complete abolition of β4 phosphorylation and Shp2 association was obtained only upon the phenylalanine permutation of Tyr1440 (). Thus, in principle, Tyr1440 could autonomously account for all of the observed binding events. To rule out this possibility, we repeated the far-Western experiment using a β4 mutant bearing a single phenylalanine substitution of Tyr1440 (β4-Y1440F). In COS-7 cells overexpressing Met and β4-Y1440F, the tyrosine phosphorylation of β4 and interaction between β4 and Shp2 were reduced but not abolished (). This indicates that Tyr1440 is necessary but not sufficient for Shp2 association and that Tyr1257 and Tyr1494 display a residual binding activity for the phosphatase.
Next, we analyzed association in vivo by coimmunoprecipitation experiments in COS-7 cells overexpressing Shp2 and the various forms of Met and β4. Met has been reported to associate with both Shp2 () and β4 (). Therefore, it may act as a bridging molecule between the integrin and the phosphatase. To exclude this possibility and assess a direct interaction between β4 and Shp2, we exploited a Met variant (Met) that retains full catalytic activity (therefore, it is still able to phosphorylate β4) but contains phenylalanine substitutions of Tyr1349 and Tyr1356, which represent the docking residues for several signal transducers, including Shp2 (; ; ). This mutant was validated for its inability to bind Shp2 () and for its capacity to phosphorylate β4 () upon transient transfection in COS-7 cells. Under these conditions, Shp2 could be efficiently recovered in β4 immunoprecipitates. Conversely, the interaction was abolished in the absence of β4 phosphorylation or in the presence of the β4 triple mutant displaying complete substitution of the critical tyrosines and was progressively diminished in the single and double mutants (). Again, the efficiency of the β4–Shp2 interaction paralleled the extent of β4 phosphorylation; no tyrosine phosphorylation of β4 nor association with Shp2 could be detected when β4 or Met were transfected individually ().
Finally, we assessed whether tyrosine-phosphorylated β4 can also recruit Shp2 in cells expressing endogenous proteins. Accordingly, we set up coimmunoprecipitation experiments in GTL16 cells, which express a constitutively active form of Met, and in FG2 cells, in which Met phosphorylation is inducible by HGF. Indeed, the association of Shp2 with β4 could be detected basally in GTL16 () and only after ligand stimulation in FG2 cells (). The binding between tyrosine-phosphorylated β4 and Shp2 is specific, as no coimmunoprecipitation of Shp2 was observed with antibodies against an unrelated myc antigen (). As expected, β4 tyrosine phosphorylation was constitutive in GTL16 cells and was induced by HGF in FG2 cells ().
Translocation of Shp2 at the plasma membrane is essential for its function (); therefore, the interaction of Shp2 with the intracellular domain of β4 is likely to stimulate Shp2-dependent transduction pathways. One of the best-documented activities of Shp2 relies on its ability to stimulate Src catalysis by controlling Csk recruitment (; ). This notion, together with the established observation that Src can promote anchorage-independent growth when aberrantly activated (), prompted us to investigate whether the signaling contribution of β4 to this process could entail the stimulation of Src.
To explore this subject, we examined the activation of Src in MDA-MB-435 breast carcinoma cells, which do express Met but are devoid of β4, and their counterpart cells stably transfected with a human β4 cDNA (MDA-MB-435–β4 cells; ). Under basal conditions, Src enzymatic function was already higher in β4 cells compared with mock cells, possibly as a result of residual Met-based β4 signaling or to adhesion-triggered Src stimulation as previously demonstrated (; ). This priming favored the further activation of Src in response to HGF; indeed, ligand stimulation led to a rapid boost of Src activity in β4 cells, whereas the HGF-dependent activation of Src in mock cells displayed slower kinetics and weaker intensity (). To evaluate whether the stronger and accelerated activation of Src in β4-expressing cells is caused by the integrin ability to bind Shp2, we generated MDA-MB-435 transfectants expressing the β4 triple variant that is unable to bind Shp2 (hereafter referred to as β4). This mutant was exposed at the cell surface as efficiently as wild-type β4 together with the endogenous α6 subunit (). In line with our hypothesis, disruption of the association between β4 and Shp2 almost abolished the HGF-dependent activation of Src (). Interestingly, the β4 mutant affected the stimulation of Src more potently than the simple absence of β4, suggesting that this variant could not only behave as a signaling-dead molecule but also as a dominant repressor of other endogenous β4 signaling partners. Accordingly, β4 has been shown to impinge on signals emanating from the Ron tyrosine kinase receptor (), the EGF receptor (; ), Erb-B2 (; ), and tetraspanins ().
Next, we searched for a Src substrate that could be a potential target of this Met–β4–Shp2 signaling pathway. The Gab1 multiadaptor protein is the major Met substrate () and can also be phosphorylated by Src (). To verify whether the activation of Src in β4-expressing cells supports a more efficient phosphorylation of Gab1 in response to HGF, we analyzed Gab1 phosphorylation levels upon HGF stimulation in mock and β4 cells. In time-course experiments, the HGF-induced phosphorylation of Gab1 was more intense and durable in β4 than in mock cells (). The increased phosphorylation of Gab1 in β4 cells likely depends on Shp2 recruitment; indeed, the HGF treatment of cells expressing the β4 mutant led to a less efficient phosphorylation of Gab1 compared with cells expressing wild-type β4 (). Likewise, stimulation with increasing amounts of HGF produced a weak dose-dependent curve of Gab1 phosphorylation in mock cells. Conversely, Gab1 phosphorylation in β4 cells reached a plateau at a very low concentration of ligand, and the overall phosphorylation levels were more robust than in mock transfectants (). Again, the response of β4 cells was similar or even lower than that showed by mock cells ().
In a complementary approach, we abated β4 levels by lentiviral delivery of siRNA in MDA-MB-231 breast carcinoma cells, which endogenously synthesize both Met and the integrin (). We then reestablished β4 expression in β4-deficient cells using a siRNA-resistant variant that lacks most of the extracellular portion, including the siRNA target region, but retains the ability to transduce Met-dependent responses (β4; ; ). In addition to wild-type β4, which displays an intact cytoplasmic domain, we also expressed a β4 mutant with phenylalanine substitutions of the tyrosines involved in Shp2 binding (). In all cases, Met expression was unaffected (). Consistent with that observed in MDA-MB-435 cells, β4 knockdown in MDA- MB-231 resulted in the reduced HGF-dependent phosphorylation of Gab1, whereas the rescue of β4 expression was accompanied by the restoration of higher Gab1 phosphorylation levels in the presence of a signaling-competent cytoplasmic portion but not when Shp2 recruitment was abolished ().
If Shp2 and Src specifically contribute to Gab1 phosphotyrosine content in a β4-dependent manner, the attenuation of Shp2 and/or Src activity should affect the HGF-triggered phosphorylation of Gab1 only in cells expressing β4. In fact, Shp2 inhibition by expression of a catalytically inactive dominant interfering isoform (Shp2) resulted in the decreased phosphorylation of Gab1 in MDA-MB-435–β4 cells but not in mock cells (). Similar results were obtained when mock and β4 cells were treated with the Src pharmacological inhibitor PP2 () or upon transfection of a kinase-dead variant of Src (Src; ).
Gab1 is a scaffolding adaptor that associates with a variety of signal transducers after phosphorylation on multiple sites (). We wondered whether the Src-dependent increase of Gab1 phosphorylation in β4 cells leads to a generic, quantitative enhancement of the overall Gab1 phosphotyrosine content or to a selective, qualitative phosphorylation of defined tyrosine residues with consequent specific binding to a given transducer.
To investigate this issue, we overexpressed Gab1 in the various MDA-MB-435 transfectants and assessed the ability of Gab1 to form supramolecular complexes with a representative panel of SH2-containing signaling molecules, including Grb2, Shc, PI3K, and Shp2 itself. Among these signaling effectors, Grb2 appeared to bind Gab1 more efficiently in β4 versus mock cells after HGF stimulation (). This binding was reduced in cells expressing the β4 mutant (); similarly, the PP2-mediated inhibition of Src, which decreases HGF-induced Gab1 phosphorylation only in β4 cells, impaired the association of Grb2 with Gab1 in the presence but not in the absence of β4 ().
Grb2 can bind tyrosine-phosphorylated Gab1 using the SH2 domain, but it also combines with Gab1 in a phosphorylation-independent manner using the SH3 domain (). To verify that the increased association between Gab1 and Grb2 in β4 cells is in fact caused by the enhanced HGF-dependent phosphorylation of the Gab1 tyrosines responsible for SH2-mediated Grb2 binding, we performed a far-Western analysis using a GST fusion protein of the Grb2 SH2 domain (GST-SH2-Grb2). In accordance with the coimmunoprecipitation data, the interaction between GST-SH2-Grb2 and Gab1 after HGF stimulation was much more efficient in β4 cells. Again, binding was decreased in cells expressing the β4 mutant, and treatment with PP2 impaired Gab1–Grb2 association in β4 transfectants but not in mock cells (). Together, these results indicate that the Shp2-mediated activation of Src in β4 cells leads to the dedicated phosphorylation of Gab1 on tyrosine residues specifically responsible for Grb2 binding, a preferential association that can be abrogated by uncoupling β4 from Shp2 or by hindering Src function. Privileged interaction between Gab1 and Grb2 in β4 cells results in the increased stimulation of Ras-dependent effectors: indeed, the amplitude and persistence of ERK/MAPK activation after HGF stimulation were higher in β4 compared with mock and β4 transfectants ().
To verify whether the aforementioned signaling pathway is in fact responsible for the ability of β4 to promote anchorage-independent growth, we performed soft agar assays on several different MDA-MB-435 transfectants ( and ). As expected, the ectopic expression of β4 considerably enhanced the basal and HGF-stimulated formation of suspended colonies compared with mock cells both in absolute numbers (, graph) and in size (, images). Analogous results were obtained upon transfection of the β4 variant, further validating the role of this nonadhesive mutant as an efficient substitute of wild-type β4 in HGF-driven responses and confirming the notion that the Met-dependent signaling activity of β4 does not require laminin ligation (; ; ). In contrast, expression of the β4 signaling-dead mutant was less effective, indicating that the interaction between β4 and Shp2 is critical for this process (). Transfection of dominant-negative isoforms of Shp2 or Src in β4-expressing cells () as well as treatment of β4 cells with the Src inhibitor PP2 or with the ERK inhibitor PD98059 () potently impaired HGF-dependent clonogenic activity in soft agar. Together, these results reinforce the observation that Shp2, Src, and ERKs are central downstream transducers of β4-driven anchorage-independent growth.
In a complementary way, siRNA-mediated reduction of β4 expression in MDA-MB-231 cells led to an almost complete abolition of basal and HGF-stimulated anchorage-independent growth. Colony-forming ability was partially reestablished upon the expression of the β4 siRNA-resistant variant displaying a wild-type intracellular domain but not upon expression of the ΔShp2 mutant ().
When a normal cell of epithelial origin happens to lose contact with the basement membrane, the surrounding stromal compartment impedes its multiplication and, in the meantime, stimulates its apoptotic elimination (). To avoid this and to ensure successful metastatic dissemination, malignant elements undergo an adaptive process that relies mainly on two strategies. One is the abnormal and unpolarized secretion of basement membrane molecules, which aberrantly reconstitute the original histological niche, thus imparting surrogate proliferative and survival signals. Parallel (or alternative) to this is the hyperactivation of oncogenic pathways, which produce a global amplification of the cell signaling activity with consequent elusion of the external adhesive consensus (). Whatever the mechanism used, the phenotypic outcome of this process is that neoplastic cells acquire the ability to grow in the absence of proper anchorage.
In this study, we show that the β4 integrin conspires with the tyrosine kinase Met for efficient execution of anchorage-independent growth by channeling Met signals toward activation of the Ras-ERK oncogenic cascade. This function relies on Met-triggered phosphorylation of the β4 cytoplasmic domain and on the ensuing recruitment of the tyrosine phosphatase Shp2 to the integrin tail. β4–Shp2 association results in the stimulation of Src, which, in turn, favors a more efficient phosphorylation of the Gab1 multiadaptor protein, mainly on tyrosines involved in Grb2 binding. Ultimately, privileged interaction between Gab1 and Grb2 leads to dedicated stimulation of the MAPK–ERK pathway.
We used the following reagents: anti–human Met, anti-Shp2, anti-Src, antiactin, GST-Shp2, and GST-SH2-Grb2 (Santa Cruz Biotechnology, Inc.); anti-β4 integrin (BD Biosciences and Chemicon); anti-Gab1, anti-Shc, anti-PI3K, antiphosphotyrosine, and anti-GST (Upstate Biotechnology); anti-Grb2 (Transduction Laboratories); antiactive and total ERK/MAPK (Promega); antiactive Src (nonphospho-Tyr529; Biosource International); PP2 (Calbiochem); wild-type Shp2 and catalytically inactive C/S Shp2 in pJ3H vector (obtained from B.G. Neel, Harvard Medical School, Boston, MA); and wild-type Src and kinase-dead K/R Src in pCMV5 vector (obtained from J. Brugge, Harvard Medical School). The β4 construct containing phenylalanine mutations of Y1257, Y1440, and Y1494 was generated by PCR amplification of the BssHII–NotI fragment of a β4 template already containing the Y1257F and Y1494F substitutions (obtained from L.M. Shaw, Harvard Medical School). To create the β4 siRNA expression vector, oligonucleotides used by were annealed and ligated into pSUPER between the BglII and HindIII sites. BamHI- and XhoI-digested inserts were then subcloned into the pRLL5 lentiviral vector. A scrambled β4 oligonucleotide was used as a control. The constructs encoding for wild-type Met, kinase-inactive Met, wild-type β4, β4-1440F, and β4 have been described previously (). The β4 cDNA used in this study corresponds to PubMed accession no. and matches with the sequence originally cloned by .
COS-7, MDA-MB-435, and MDA-MB-231 cells were cultured in DME supplemented with 10% FBS (Invitrogen). The expression of exogenous proteins was obtained with LipofectAMINE- or LipofectAMINE 2000 (Invitrogen)–mediated transfection according to the manufacturer's protocol or with retroviral or lentiviral infection. Viral hybrid vectors were produced by the transient transfection of 293T cells. Viral supernatants were filtered through a 0.22-μm filter, and infections were performed in the presence of 4 μg/ml polybrene (Sigma-Aldrich).
For immunoprecipitations, 5 × 10 cells were lysed for 20 min at 4°C with 1 ml of a buffer containing 50 mM Hepes, pH 7.4, 5 mM EDTA, 2 mM EGTA, 150 mM NaCl, 10% glycerol, and 1% Triton X-100 in the presence of protease and phosphatase inhibitors. For β4–Shp2 coimmunoprecipitations in FG2 cells, 1% Brij58 was used instead of Triton X-100. Extracts were clarified at 12,000 rpm for 15 min, normalized with the BCA Protein Assay Reagent kit (Pierce Chemical Co.), and incubated with different mAbs for 2 h at 4°C. Immune complexes were collected with either protein G– or protein A–Sepharose, washed in lysis buffer in the presence of 1 M LiCl, and eluted. Total cellular proteins were extracted by solubilizing the cells in boiling SDS buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 1% SDS). Extracts were electrophoresed on SDS-polyacrylamide gels and transferred onto nitrocellulose membranes (Hybond; GE Healthcare). Nitrocellulose-bound antibodies were detected by the ECL system (GE Healthcare). Unless otherwise indicated, cells were stimulated with 50 ng/ml HGF for 30 min. PP2 was generally used at a 10-μM concentration and applied to cells for 30 min. In experiments aimed at analyzing the association of Gab1 with signal transducers and the HGF-dependent activation of ERKs, mock cells and cells expressing wild-type β4 or β4 were transiently transfected with a Gab1 cDNA.
Src kinase assays were performed on Src immunoprecipitates using a commercial kit (Upstate Biotechnology) based on the phosphorylation of a specific substrate peptide (KVEKIGEGTYGVVYK) using the transfer of the γ-phosphate of γ-[P]ATP by Src. The phosphorylated substrate was then separated from the residual γ-[P]ATP using phosphocellulose paper and quantified with a scintillation counter.
3,000 cells were resuspended in complete medium containing 0.5% Seaplaque agar. Cells were seeded in 24-well plates containing a 1% agar underlay and supplemented twice a week with complete medium. In some experiments, MDA-MB-435–β4 cells were treated twice a week with 5 μM PP2 or 10 μM PD98059. Colonies were stained by the incorporation of tetrazolium salts 2 (for MDA-MB-435) or 3 wk (for MDA-MB-231) after seeding. Colonies were coded and scored in a blinded fashion by a second observer. Colony numbers were obtained using a phase-contrast light microscope (DMIL; Leica) fitted with a 32-grid eyepiece at a total magnification of 20×. Images were captured with ImageReady software (Adobe) using a microscope (DMIL; Leica) and a 20 × 0.30 objective (Leica) equipped with a digital camera (DFC320; Leica).
Results are means ± SEM. Comparisons were made using the two-tailed test. P-values <0.05 were considered to be statistically significant. Blot images were captured using a molecular imager (ChemiDoc XRS; Bio-Rad Laboratories). Densitometric analysis was performed with analysis software (Quantity One 1-D; Bio-Rad Laboratories) installed on the imager. Images were arranged and labeled using Illustrator (Adobe). |
Embryonic, fetal, and adult tissues are used as sources to investigate the developmental and therapeutic potential of stem cells. Because of their accessibility and the possibility that the patient could act as a stem cell donor, adult stem cells from the skin have received particular attention (). Apart from multipotent epithelial stem cells that form hair follicles, sebaceous glands, and epidermis (; ; ; ) and so-called melanocyte stem cells that generate pigmented cells (), a multipotent cell dubbed skin-derived precursor cell (SKP) has been isolated from both the murine and human skin (). SKPs have the potential to produce in vitro cell types normally not found in the skin, such as neuronal cells. Subsequently, several laboratories reported the existence of self-renewing cells present in the skin of mice, pigs, and humans and able to differentiate in vitro into cells expressing neuronal, glial, osteoblast, chondrocyte, smooth muscle, melanocyte, and adipocyte lineage markers (; ; ; ; ; ; ).
The formation of cells normally not present in skin might be due to transdifferentiation, which describes the conversion of a cell type of a specific tissue lineage into a cell type of another lineage (). Alternatively, cells from a given lineage might dedifferentiate into a more naive state that allows the cell to redifferentiate along new lineages. Finally, multipotent cells with stem cell features might persist until adulthood, able to generate a broad variety of cells, depending on their environment. To distinguish among these possibilities, the origin and nature of the cell in question has to be determined and its developmental potential has to be analyzed at the single cell level ().
The developmental origin and exact localization of skin cells giving rise to neural and nonneural progeny is unclear in many of the reported cases. Multipotent skin-derived cells have been enriched by means of markers found on hematopoietic stem cells () or have been isolated from transgenic animals expressing GFP from promoter elements of (), a gene also expressed in neural progenitor cells. One source that has been associated with sphere-forming SKPs is the dermal papilla from whisker follicles (). Whisker follicles are large hair follicles of the face that serve as sensory organs for a wide range of mammals, excluding humans. Genetic in vivo cell fate mapping revealed that the dermal papilla of these follicles is of neural crest origin (). Similarly, culturing explants of bulge and dermal sheath of whisker follicles allowed the identification of neural crest–derived multipotent cells in the upper part of the whisker follicle (; ).
A neural crest origin might explain the multipotency of at least some stem and progenitor cells in the skin. Indeed, the neural crest contributes during vertebrate development to a variety of tissues, including the peripheral nervous system and nonneural cell types such as melanocytes in the skin (). Clonal analysis revealed that multipotent, self-renewing neural crest stem cells (NCSCs) cannot only be isolated from migratory neural crest but also from different tissues at later stages and even from the adult organism (; ; ). Thus, it is conceivable that apart from the whisker follicle, other neural crest–derived compartments in the skin might contain multipotent neural crest–derived cells.
Floating sphere cultures have previously been used to identify self-renewing cells in both murine and human skin (, ; ; ; ). To further characterize sphere-forming cells derived from the trunk skin of adult mice, dorsal and ventral skin biopsies comprising both dermis and epidermis were dissociated and cultured, and formation of spheres was observed within 4–7 d of culture. These spheres could be passaged for several months without overt morphological changes (), pointing to the self-renewing capacity of cells present in the spheres. Intriguingly, unlike SKPs enriched by marker selection () or cultured in slightly different conditions than used here (), 100% of all primary, secondary, and later passage spheres generated from mouse trunk skin ( > 50 spheres) contained cells expressing the low-affinity neurotrophin receptor p75 and the transcription factor Sox10, both markers for NCSCs (; ; ). In spheres passaged >20 times, 67.0 ± 10.5% of all cells expressed p75, 76.6 ± 4.5% of all cells expressed Sox10, and 58.6 ± 10.5% of all cells were double positive for p75 and Sox10. 15.0 ± 6.2% of all cells were negative for these markers, pointing to a cellular heterogeneity within skin-derived spheres, as also observed in sphere cultures from other tissues (). Thus, skin-derived cells expressing NCSC markers can be propagated in culture for prolonged time periods.
Similarly, spheres readily formed from dissociated surgical samples of adult human thigh and face skin (). These spheres could be expanded by passaging, such that after 3 mo >10 cells had been generated from a 16-cm skin sample used as starting material. Similar to mouse cultures, all spheres contained p75/Sox10-positive cells, which accounted for >60% of all cells (). However, other markers for premigratory or migratory NCSCs, such as Sox9 and HNK-1, were not expressed.
As p75 and Sox10 are markers for NCSCs (; ), we next examined whether the mouse trunk skin–derived spheres originate from the neural crest. The fate of neural crest cells was mapped in vivo by mating Cre reporter () mice, which express β-galactosidase upon Cre-mediated recombination, with mice expressing Cre recombinase under the control of the promoter (; ). In double-transgenic mice, virtually all NCSCs express β-galactosidase (; ). Importantly, despite the transient expression of Cre recombinase, the progeny of neural crest cells continue to express β-galactosidase because of the genomic recombination event. Anti–β-galactosidase antibody staining revealed that all primary and late passage spheres generated from the back skin of adult double-transgenic mice were composed of neural crest–derived cells ( and not depicted). In particular, 100% of all p75-positive cells coexpressed β-galactosidase, as revealed by a typical punctuated staining pattern (). Because 87.3 ± 6.0% of all p75-positive cells also expressed Sox10 (three independent experiments with spheres obtained after 20–35 passages), the data demonstrate that sphere-forming, p75/Sox10-expressing cells from the adult mouse skin are neural crest derivatives.
To test the developmental potential of sphere cells derived from murine and human skin, spheres containing p75/Sox10-positive neural crest cells were allowed to differentiate at high cellular density. The formation of glia expressing glial fibrillary acidic protein (GFAP), βIII tubulin (TuJ1)–positive neuronal cells, and smooth muscle actin (SMA)–expressing nonneural cells was readily detectable in both mouse and human cell cultures (), although the number of neuronal cells generated was highly variable and low in comparison to that of glia and smooth muscle cells. Upon addition of ascorbic acid and bone morphogenic protein (BMP) 2, the generation of chondrocytes was observed (), whereas treatment with stem cell factor and endothelin-3 resulted in formation of a few melanocytes (). Finally, occasional adipocytes were detected (). However, we never observed the generation of keratinocytes as assessed by staining with a pan-keratin antibody (unpublished data), demonstrating that neural crest–derived sphere-forming cells are distinct from epithelial stem cells of the skin.
The aforementioned data are consistent with the idea that skin-derived spheres contain multipotent cells capable of generating neural and nonneural cell types. In analogy to NCSCs isolated from other stages and locations, it is likely that this broad potential is inherent to the p75/Sox10-expressing neural crest–derived cells found in the spheres. To address this hypothesis, we plated cells from mouse trunk skin–derived spheres at clonal density and prospectively identified and mapped single undifferentiated, unpigmented p75-positive clone founder cells (; ; ; ). The clone founder cells were then incubated in culture conditions permissive for neurogenesis, gliogenesis, and nonneural cell formation (). 57.9% of all p75-positive founder cells were at least tripotent, giving rise to clones consisting of neural and nonneural cell types (). Virtually no p75-positive cell was restricted to a single cell lineage. Thus, p75/Sox10-positive neural crest–derived cells prepared from the adult trunk skin are multipotent and can be expanded in culture. Upon isolation, these cells therefore exhibit properties of NCSCs.
Several instructive growth factors, including Wnt, BMP, neuregulin (NRG), and TGFβ, have been shown to promote specific fate decisions in NCSCs at the expense of other possible fates. Although Wnt responsiveness is lost at later developmental stages (), postmigratory NCSCs isolated from various structures maintain their responsiveness to BMP2, NRG1, and TGFβ, although the biological activity of these factors changes with time and location (; ). Similarly, single prospectively identified p75-positive neural crest cells isolated from the adult back skin were sensitive to BMP2, NRG1, and TGFβ (). All three instructive growth factors suppressed multipotency without affecting survival of founder cells and promoted the generation of clones containing nonneural cells that were mostly SMA positive. However, we were unable to identify growth factors inducing exclusively neuro- or gliogenesis in skin-derived neural crest cells, whereas NCSCs isolated from other sources give rise to neurons and glia, respectively, in response to BMP2 and NRG1 (; ). Hence, adult skin–derived neural crest cells, although displaying NCSC features, are intrinsically different from other types of NCSCs and show altered factor responsiveness.
Apart from back skin–derived p75/Sox10-positive multipotent cells ( and ), the neural crest origin of sphere-forming cells in the adult skin has been demonstrated for whisker follicle–derived SKPs, which, however, are negative for the NCSC markers p75 and Sox10 (). This could either reflect differential regulation of NCSC markers in the same cell type because of varying culture conditions or indicate sphere-forming capacity of skin cells from different neural crest derivatives. To address this issue, we first mapped neural crest derivatives in the adult skin and investigated which of these neural crest derivatives express the NCSC marker Sox10 in vivo. We initially focused on the whisker follicle because this structure has been identified before as a source of multipotent neural crest–derived cells (; ). In the head, the neural crest contributes to many mesenchymal structures (). Thus, many mesenchymal structures in whisker follicles isolated from double-transgenic mice expressed β-galactosidase (). In particular, the capsula, the ringwulst, the dermal sheath, and, as previously published (; ), the dermal papilla turned out to be neural crest derived. The neural crest origin of all these structures was confirmed by fate mapping experiments performed in () mice, in which Cre recombinase is expressed in neural crest cells independently from promoter activity (; ). As revealed by X-gal staining of whisker follicles isolated from mice (that express β-galactosidase from the locus; ), capsula, ringwulst, and dermal papilla did not express Sox10 in vivo, whereas the dermal sheath, glial cells in nerve endings, and melanocytes were Sox10 positive (). Thus, the whisker follicle comprises various Sox10-positive and -negative tissues of neural crest origin.
To investigate which of these neural crest derivatives contain cells with sphere-forming potential, dermal papilla, capsula, the upper part of the dermal sheath (without the bulge), and the lower part of the dermal sheath were isolated from whiskers of adult double-transgenic mice by microdissection, dissociated, and cultured in the same conditions as used before for trunk skin–derived multipotent neural crest cells. In addition, rat whiskers were used to dissect the ringwulst, which in mice was too small to be isolated without contamination from other tissues. Strikingly, all these whisker follicle structures appear to harbor cells with the capacity to generate spheres (). X-gal staining of mouse cell cultures confirmed that the spheres were neural crest derived. Therefore, neural crest cells with sphere-forming potential are not confined to a particular niche in the whisker follicle.
Unlike in the head, the mesenchyme in the trunk is not derived from the neural crest (), and β-galactosidase expression in back skin of mice was thus restricted to a few locations (). The same structures were also labeled in the back skin of mice (). In particular, both in the anagen and telogen stage, X-gal staining was found in the permanent part of the pelage follicle, including the bulge region below the sebaceous gland (; and Fig. S1, available at ). This area comprises the location of melanocyte stem cells () and glial cells in nerve endings (). In addition, pigmented melanocytes in the bulb region (the lower part of the hair follicles; ) and nerves expressed β-galactosidase. In contrast, other hair follicle structures such as the dermal papilla, dermal sheath, and the outer and inner root sheaths were X-gal negative and, in the trunk skin, do not originate from the neural crest ().
To determine the potential origin of Sox10-positive sphere-forming cells in the skin (), we assessed Sox10 expression by virtue of β-galactosidase activity in the back skin of mice in vivo. Interestingly, Sox10-expressing cells were confined to exactly the same areas as were X-gal–positive cells in and mice, including nerves, melanocytes, and a domain consistently found below the sebaceous gland in anagen and telogen stage that encompasses the hair follicle bulge with the niche for melanocyte stem cells and nerve endings (Fig. S1). Importantly, in both and mice, X-gal–positive cells in the region below the sebaceous gland coexpressed Sox10 and p75 protein (). Thus, p75/Sox10-positive multipotent neural crest–derived cells from the trunk skin (–
) are connected to the glial or the melanocyte lineage or to both of these lineages.
To elucidate whether p75/Sox10 expression and the capacity to form spheres are associated with glial cells from skin, we made use of () mice that express Cre recombinase in the peripheral glial lineage from early stages onward, but not in migrating neural crest cells or in neural crest–derived cells of other than glial lineages (). β-Galactosidase activity was detectable in nerves and nerve endings in the back skin of adult mice (). As predicted from the proposed location of glial cells associated with nerve endings in the hair follicle (), X-gal staining in pelage follicles of mice was confined to a region around the bulge (), corresponding to the area that also contains β-galactosidase–expressing cells in , , and mice (; and Fig. S1). In mice, X-gal–labeled cells of the bulge region were also labeled with anti-Sox10 antibody () and anti-p75 antibody (). Pigmented melanocytes in the hair follicle bulb were X-gal negative, however, indicating that cells labeled in mice do not give rise to melanocyte s and thus are not related to the melanocyte lineage ().
To directly demonstrate that cells from the glial lineage tracked by promoter activity possess sphere-forming potential, these cells have to be prospectively identified and freshly isolated. One possibility to achieve this would be by using specific surface antigen markers. However, such markers for the early glial lineage are currently unavailable. Furthermore, nerves present in the skin cannot be isolated by microdissection. Therefore, we used a genetic strategy to prospectively identify and directly isolate cells associated with the glial lineage. mice were mated with mice that express EYFP upon Cre-mediated recombination (). Cells expressing EYFP in the trunk skin of double-transgenic mice were isolated by FACS and transferred into medium permissive for sphere formation (). Although from unselected skin samples >10 cells were used to generate ∼50 spheres (; see Materials and methods), <10,000 cells from both the EYFP-positive and -negative cell fraction were seeded in these experiments, to assess a possible enrichment in the spherogenic potential of FACS-selected cells. In two independent experiments, the EYFP-positive (, green frame), but not the EYFP-negative (, blue frame), cell population gave rise to spheres. Moreover, acutely fixed primary spheres of EYFP-positive cells were composed of cells expressing both p75 and Sox10 (). Thus, p75/Sox10-positive cells related to the glial lineage can be isolated from the skin and form spheres.
We next asked whether sphere-forming potential is a common feature of peripheral glia. Therefore, we investigated whether sphere cultures can also be established from adult peripheral nerves. In agreement with others (), we were unable to obtain spheres from cultures of dissociated sciatic and trigeminal nerves from adult mice (unpublished data). Thus, nerves or nerve endings in skin, but not peripheral nerves in general, contain cells with sphere-forming potential.
In cell preparations from the trunk skin of mice, only a fraction of all p75/Sox10-positive cells also expressed β-galactosidase (unpublished data). This could point to inefficient Cre-mediated recombination in mice. Alternatively, sources in the skin other than the glial lineage might yield sphere-forming neural crest–related cells. To address whether spherogenic neural crest–derived cells might be connected to the melanocyte lineage, we traced the fate of trunk skin cells in mice (). codes for the enzyme dopachrome tautomerase (also called Trp-2), which is required for melanin synthesis and already expressed in melanocyte stem cells (). As expected, β-galactosidase activity in the back skin of mice was detectable in melanocytes () and in the hair follicle bulge region corresponding to the location of melanocyte stem cells (; ). Moreover, some X-gal–positive cells in the bulge region also expressed Sox10 () and p75 ().
To investigate whether, in addition to cells of the glial lineage, the early melanocyte lineage also comprises undifferentiated neural crest–derived cells with the capacity to generate spheres, we isolated EYFP-expressing cells prospectively identified in the skin of mice. Intriguingly, in two independent experiments, FACS isolation and culturing of <10,000 cells revealed that only EYFP-expressing (, green frame), but not EYFP-negative, cells (, blue frame) were able to form spheres (). Analysis of acutely fixed primary spheres revealed many cells coexpressing p75 and Sox10, whereas pigmented differentiated melanocytes were absent (). These data indicate that the early melanocyte lineage comprises p75/Sox10-positive cells that can be propagated as spheres. Thus, as in whisker follicles of the face, the trunk skin contains more than one source of sphere-forming neural crest–derived cells, namely, cells of the glial and melanocyte lineages.
t
h
e
p
r
e
s
e
n
t
s
t
u
d
y
,
w
e
s
h
o
w
t
h
a
t
c
e
l
l
s
w
i
t
h
N
C
S
C
f
e
a
t
u
r
e
s
c
a
n
b
e
i
s
o
l
a
t
e
d
f
r
o
m
t
h
e
a
d
u
l
t
t
r
u
n
k
s
k
i
n
o
f
b
o
t
h
m
o
u
s
e
a
n
d
h
u
m
a
n
.
L
i
k
e
N
C
S
C
s
f
r
o
m
o
t
h
e
r
e
m
b
r
y
o
n
i
c
a
n
d
p
o
s
t
n
a
t
a
l
s
o
u
r
c
e
s
,
t
h
e
s
e
n
e
u
r
a
l
c
r
e
s
t
–
d
e
r
i
v
e
d
c
e
l
l
s
i
n
t
h
e
s
k
i
n
e
x
p
r
e
s
s
p
7
5
a
n
d
S
o
x
1
0
a
n
d
a
r
e
m
u
l
t
i
p
o
t
e
n
t
,
a
b
l
e
t
o
g
e
n
e
r
a
t
e
s
e
v
e
r
a
l
n
e
u
r
a
l
a
n
d
n
o
n
n
e
u
r
a
l
l
i
n
e
a
g
e
s
.
M
o
r
e
o
v
e
r
,
m
u
l
t
i
p
o
t
e
n
t
n
e
u
r
a
l
c
r
e
s
t
–
d
e
r
i
v
e
d
c
e
l
l
s
f
r
o
m
t
h
e
a
d
u
l
t
s
k
i
n
d
i
s
p
l
a
y
a
s
e
l
f
-
r
e
n
e
w
i
n
g
c
a
p
a
c
i
t
y
,
i
n
t
h
a
t
m
o
u
s
e
a
n
d
h
u
m
a
n
s
k
i
n
–
d
e
r
i
v
e
d
c
e
l
l
s
c
a
n
b
e
g
r
o
w
n
a
n
d
e
x
p
a
n
d
e
d
f
o
r
m
o
n
t
h
s
i
n
f
l
o
a
t
i
n
g
s
p
h
e
r
e
c
u
l
t
u
r
e
s
.
I
n
t
r
i
g
u
i
n
g
l
y
,
i
n
w
h
i
s
k
e
r
f
o
l
l
i
c
l
e
s
o
f
f
a
c
i
a
l
s
k
i
n
,
s
e
v
e
r
a
l
s
t
r
u
c
t
u
r
e
s
o
f
n
e
u
r
a
l
c
r
e
s
t
o
r
i
g
i
n
a
p
p
e
a
r
t
o
c
o
m
p
r
i
s
e
c
e
l
l
s
w
i
t
h
s
p
h
e
r
e
-
f
o
r
m
i
n
g
c
a
p
a
c
i
t
y
.
I
n
t
h
e
t
r
u
n
k
s
k
i
n
,
h
o
w
e
v
e
r
,
g
e
n
e
t
i
c
c
e
l
l
f
a
t
e
m
a
p
p
i
n
g
,
p
7
5
/
S
o
x
1
0
e
x
p
r
e
s
s
i
o
n
a
n
a
l
y
s
i
s
i
n
v
i
v
o
,
a
n
d
,
i
m
p
o
r
t
a
n
t
l
y
,
p
r
o
s
p
e
c
t
i
v
e
i
d
e
n
t
i
f
i
c
a
t
i
o
n
a
n
d
d
i
r
e
c
t
i
s
o
l
a
t
i
o
n
d
e
m
o
n
s
t
r
a
t
e
t
h
a
t
c
e
l
l
s
d
i
s
p
l
a
y
i
n
g
N
C
S
C
p
r
o
p
e
r
t
i
e
s
d
o
n
o
t
r
e
s
i
d
e
i
n
m
e
s
e
n
c
h
y
m
a
l
s
t
r
u
c
t
u
r
e
s
o
f
h
a
i
r
f
o
l
l
i
c
l
e
s
b
u
t
,
r
a
t
h
e
r
,
a
r
e
a
s
s
o
c
i
a
t
e
d
w
i
t
h
t
h
e
m
e
l
a
n
o
c
y
t
e
a
n
d
g
l
i
a
l
l
i
n
e
a
g
e
s
.
T
h
u
s
,
s
e
l
f
-
r
e
n
e
w
i
n
g
n
e
u
r
a
l
c
r
e
s
t
–
d
e
r
i
v
e
d
c
e
l
l
s
f
r
o
m
t
h
e
s
k
i
n
a
r
e
n
o
t
c
o
n
f
i
n
e
d
t
o
a
p
a
r
t
i
c
u
l
a
r
n
i
c
h
e
b
u
t
c
a
n
b
e
a
t
t
r
i
b
u
t
e
d
t
o
d
i
s
t
i
n
c
t
l
o
c
a
t
i
o
n
s
i
n
f
a
c
e
a
n
d
t
r
u
n
k
s
k
i
n
.
Human thigh skin from an adult man (∼45 yr of age) and face skin from an adult woman (∼57 yr of age; provided by G. Beer, University Hospital of Zurich, Zurich, Switzerland) were obtained in the frame of cosmetic surgery according to the guidelines of the University Hospital of Zurich. Murine skin was taken from adult C57/BL6 mice of at least 8 wk of age. Skin samples (composed of both dermis and epidermis) were dissected, cut into small pieces, and digested in 0.1% Trypsin-EDTA (Invitrogen) in HBSS without Ca and Mg (Animed) and digested for 50 min at 37°C. Partially digested skin pieces were dissociated mechanically and filtered through a 40-μm cell strainer (BD Biosciences). The cell suspension was centrifuged and washed with medium, and the cell pellet was resuspended in growth medium (GM) consisting of ME-F12 1:1 containing 1× B-27 supplement (Invitrogen), 20 ng/ml FGF2 (PeproTech), 10 ng/ml EGF (PeproTech), penicillin/streptomycin (P/S), and Fungizone. GM for human cells also contained 10 ng/ml leukemia inhibitory factor (Sigma-Aldrich). 2.5–4 million cells were seeded in GM into an uncoated T-25 cell culture flask (BD Biosciences). After 4–7 d in culture, sphere formation was observed. For FACS analysis, skin was taken from mice between 10 and 16 d of age. Samples were incubated in 0.5 mg/ml Dispase (Roche) in HBSS for 30 min at 4°C. Fat tissue was removed with forceps, and the rest of the skin was cut into small pieces and digested in 1 mg/ml collagenase (Worthington) in HBSS for 45 min at 37°C. After a final digestion with 0.1% Trypsin-EDTA in HBSS for 5 min at 37°C, the partially digested skin pieces were dissociated mechanically and treated as described above. FACS was performed with a FACS Aria (Becton Dickinson).
Once a week, the sphere suspension was transferred into a 15-ml Falcon tube. Cells adhering to the flask bottom were discarded. Spheres were centrifuged, and one third of the supernatant was transferred as conditioned medium into a new T-25 flask. Spheres were incubated with 300 μl Trypsin-EDTA solution (0.25%) for 3–5 min at RT. 400 μl of Ovomucoid solution (1 mg/ml Trypsin inhibitor [Sigma-Aldrich] and 10 mg DNase [Roche] in 25 ml medium) were added, and spheres were dissociated mechanically, centrifuged, resuspended in fresh GM, and seeded into a new flask containing one third conditioned medium. After some passages, spheres were cultured in flasks coated with Poly(2-hydroxyethylmethacrylate) (Poly-Hema; Sigma-Aldrich). Coating was performed at RT with a solution of 16 mg/ml Poly-Hema in 95% ethanol.
Anagen-phase whisker follicles from mouse and rat were dissected out of the whisker pad and microdissected as described previously (). Follicle structures were incubated in 0.05% Trypsin-EDTA in ME for 1.5–2 h at 37°C. Trypsin activity was stopped with ME containing 10% FCS. After two washing steps with GM, cells of each structure were plated in a well of a 24-well dish. Sphere formation was observed within 1–2 wk. Passaging was performed as described for skin spheres using 0.05% Trypsin-EDTA for 2 min at RT.
All differentiation assays were performed using spheres plated on dishes coated with fibronectin (FN) or poly--lysine/FN as described previously (). Neurogenesis, gliogenesis, and smooth muscle formation were observed after 3–7 d in GM. Chondrocyte formation was obtained after 9 d in ME containing 10% FCS, 50 μg/ml ascorbic acid 2-phosphate (Sigma-Aldrich), 10 ng/ml FGF2, and P/S, followed by ME containing 10% FCS, 50 μg/ml ascorbic acid 2-phosphate, and 10 ng/ml BMP2 (PeproTech) for another 3 d. Adipocytes were occasionally observed when spheres were cultured in DME-F12 containing B-27 supplement and 10 ng/ml BMP2. Melanocytes were observed when cultured in MEM containing 10% FCS, 50 ng/ml murine stem cell factor (PeproTech), 100 nM endothelin-3 (Sigma-Aldrich), and P/S for at least 10 d.
Murine skin–derived spheres passaged >17 times were dissociated with Trypsin-EDTA as described and plated at clonal density on pDL/FN-coated 35-mm dishes (Corning) in standard medium prepared as reported previously (). Single p75-positive cells were labeled and mapped as described previously () and incubated in standard medium alone or supplemented with 100 ng/ml BMP2, 1 nM NRG1 (R&D Systems), or 0.1 ng/ml TGFβ (R&D Systems). After 10 d, the cells were fixed and analyzed immunocytochemically.
Anti-p75, anti-Sox10, anti-SMA, and anti-GFAP antibody stainings on cells were done as described by . Anti-TuJ1 antibody (1:200; Sigma-Aldrich) and anti-Keratin antibody (1:500; Abcam) were used for 2 h at RT, whereas anti–β-galactosidase (1:100; Roche) and anti-NG2 (1:200; Chemicon) antibodies were used with incubation overnight at 4°C. The following secondary antibodies were used for 1 h at RT: Cy3-conjugated goat anti-mouse (1:200; Jackson ImmunoResearch Laboratories), Cy3-conjugated goat anti-rabbit (1:200; Jackson ImmunoResearch Laboratories), Alexa 488–conjugated goat anti-mouse (1:100; Invitrogen), and Alexa 488–conjugated goat anti-rabbit (1:100; Invitrogen). Cell nuclei were stained with DAPI. Paraffin sections of X-gal–treated skin were stained for Sox10 as described previously () using a controlled antigen-retrieval device (FSG 120-T/T; Milestone). Heat unmasking for the p75 staining was done in 10 mM trisodium citrate, pH 6.0, using the same antigen-retrieval device. The antibody (Chemicon) was used at a dilution of 1:5,000. Alexa 594–conjugated goat anti-mouse and Alexa 488–conjugated goat anti-rabbit (1:200; Invitrogen) were used as secondary antibodies. Immunofluorescence of cells was analyzed using a microscope (Axiovert 100; Carl Zeiss MicroImaging, Inc.) and magnifications of 32×. Pictures were made with a camera (AxioCam MRm) and Axiovision 4.3 software (Carl Zeiss MicroImaging, Inc.). Immunofluorescence of sections was analyzed using a microscope (Axioskop 2; Carl Zeiss MicroImaging, Inc.) with 20×, 63×, and 100× magnifications. Pictures were made with a camera (AxioCam HRc) and Axiovision 4.2 software.
For Alcian blue staining, cells fixed with 4% formaldehyde were incubated with a 3% solution of glacial acetic acid in distilled water for 3 min at RT followed by a 1% Alcian blue solution (Chroma Gesellschaft) in 3% acetic acid for 5 min at RT. For oil red O staining, fixed cells were incubated in 60% isopropanol in water for 15 min at RT followed by another 15 min in an oil red O mixture (0.35 g oil red O [Sigma-Aldrich] in 50 ml isopropanol and water in a 3:2 dilution). For DOPA reaction, cells were incubated in a 0.1% solution of 3-(3,4-Dihydroxyphenyl)--alanine (L-DOPA) in PBS for 5 h at 37°C.
Skin pieces from 12-d-old mice for anagen or 8-wk-old mice for telogen stage and single whiskers from , , and mice were fixed and incubated in X-gal solution (). Samples were washed three times with PBS, incubated for 24 h at RT in Bouin's fixative, washed twice for 15 min in HO, and dehydrated in 80%, 95%, and three times 100% ethanol. Samples were transferred into a 1:1 mixture of 100% ethanol and glycomethacrylate (Leica), left for infiltration at 4°C for 1–2 wk, and embedded in glycomethacrylate resin. 4–5-μm sections were cut on a microtome using a glass knife. Counterstainings were performed with Fast red (Merck) for 30 min at 60°C.
Skin from 12-d-old mice from and mice was stained with X-gal, embedded in paraffin, and sectioned as described previously ().
Fig. S1 shows localization of neural crest–derived cells expressing p75/Sox10 in the adult skin. Online supplemental material is available at . |
To date, a physiological function for Rhamm (receptor for hyaluronan [HA]-mediated motility) has remained elusive. Analyses of animal models demonstrated instructive roles of Rhamm in tumorigenesis and inflammatory diseases (; ; ), consistent with evidence for a role of Rhamm in motility and proliferation/apoptosis in culture (; ). However, given that migration and proliferation/apoptosis are essential functions for morphogenesis and tissue homeostasis, it is surprising that genetic deletion of Rhamm does not affect embryogenesis or adult homeostasis.
Rhamm was originally isolated from subconfluent fibroblasts in culture () and subsequently cloned from mesenchymal cells (). Antibodies prepared against a shed form of Rhamm blocked HA-stimulated fibroblast motility, suggesting that Rhamm is a cell surface protein able to transduce motogenic signaling in culture (). Rhamm-bound HA was detected in cancer cell lines and shown also to occur in intracellular compartments/structures, including the cytoskeleton, nucleus, and cytoplasm (; ). These results suggest that Rhamm has both extracellular and intracellular functions. However, the role of Rhamm as a cell surface HA receptor became controversial partly because cloning of the human (; ; ) and mouse genes () revealed an absence of both a signal peptide required for export through the Golgi/ER and membrane spanning domains common to most cell surface receptors. We now know that Rhamm resembles a group of intracellular proteins that also lack these signature characteristics of classical cell surface proteins but that are nevertheless found at the cell surface and regulate multiple functions by transmitting signals across the cell membrane. Examples include epimorphin/syntaxin-2 and autocrine motility factor/phosphoglucose isomerase (; ). However, neither the physiological functions of proteins such as Rhamm nor the mechanisms by which these proteins regulate signaling pathways are known.
We isolated cell surface Rhamm as a motogenic factor required for rapid fibroblast motility, and we and others also provided evidence for a role of both cell surface and intracellular Rhamm in GM progression in culture (; ). We showed that Rhamm expression is high in aggressive human fibromatoses (desmoid) tumors () and demonstrated that genetic deletion of Rhamm strongly reduced desmoid tumor initiation and invasion in a mutant adenomatous polyposis coli and βcatenin–driven mouse model of this mesenchymal tumor. Fibroproliferative processes such as aggressive fibromatosis resemble proliferative/migratory stages of wound healing (). The expression of Rhamm is modulated during wounding () and by fibrogenic cytokines such as TGF-β (). Because factors that regulate fibroblast function play dual roles in wound repair and tumorigenesis (; ), we have assessed whether Rhamm is involved in response to injury using models of “wounds” with Rhamm (Rh) fibroblasts in vitro as well as excisional skin wounds in vivo with Rh mice. We show that Rhamm loss results in defective migration in culture as a result of aberrant CD44–ERK1,2 (extracellular-regulated kinase 1,2) signaling. We also show that ERK1,2 activity is defective in early phases of skin repair in Rh mice and that this defect is associated with aberrant mesenchymal cell migration and differentiation.
We and others have previously shown that Rhamm mediates cell migration on tissue culture plastic (). To determine whether Rhamm is required for fibroblast migration under more physiological settings and whether other proteins compensate for loss of Rhamm expression, the motogenic behavior of Rh and wild-type (Wt) fibroblasts were compared using scratch wounds and 3D collagen gel assays, which are designed to mimic aspects of migration in vivo (). Significantly fewer Rh than Wt fibroblasts migrated across 3-mm scratch wounds in culture (). Time lapse of wounds revealed that motility speed of Rh fibroblasts was less than Wt (). To confirm that Rhamm expression is sufficient to restore migration to Wt levels, we expressed full-length Rhamm (Rh) cDNA in Rh fibroblasts. This rescued migration defects (Fig. S1 a, available at ). Thus, loss of Rhamm expression results in an inherent migration defect related to a reduced ability of fibroblasts to orient and locomote toward haptotactic cues rapidly.
The invasive properties of Rh versus Wt fibroblasts were compared in 3D collagen type I gels (). Migration of primary Rh dermal fibroblasts into the PDGF/HA/collagen gel plug was reduced by 90% compared with Wt fibroblasts (), indicating an intrinsic defect in haptotaxis and invasion of Rh fibroblasts. Rh expression rescued these defects (Fig. S1 b).
We previously showed that Rhamm is required for activation of ERK1,2 in culture (). Fibroblast migration requires appropriate temporal regulation of signaling pathways such as ERK1,2, which provide cues for promoting and sustaining migration/invasion (). To determine whether ERK1,2 activity and/or subcellular targeting is aberrant in Rh fibroblasts and whether the loss of Rhamm is the direct cause of these defects, we quantified serum induction of ERK1,2 activity in Rh versus Rh-rescued Rh fibroblasts using ELISA, Western blots, and confocal microscopy (). ELISA analysis showed activation of ERK1,2 in both Rh-rescued and Rh fibroblasts (), but activity was less and declined more rapidly in Rh cells (). Western blots confirmed these results (). Confocal analysis showed that active ERK1,2 are targeted to the nucleus in both Rh-rescued and Rh fibroblasts, but activity was significantly less in Rh cells (). A similar reduction in activated ERK1,2 was observed in other subcellular compartments of Rh cells, including lamellae. These results suggest that Rhamm is required for sustaining ERK1,2 activity in multiple subcellular compartments, a process that can affect fibroblast motility and differentiation ().
CD44 is a commonly expressed, integral membrane adhesion and HA receptor that activates motogenic signaling cascades, including ERK1,2 (). CD44 coassociates with ERK1,2 via adaptor proteins that bind to its cytoplasmic tail (; ). To begin to identify molecular mechanisms that are deficient in Rh fibroblasts, we first assessed whether Rhamm can coassociate with CD44 and whether both Rhamm and CD44 are required for activation of ERK1,2. CD44s protein was expressed in equivalent amounts in Rhamm-expressing and Rh fibroblasts (). The ability of Rhamm, CD44, and ERK1,2 to associate with each other in Rh-rescued fibroblasts was demonstrated in pull-down assays (). Confocal analyses confirmed coassociation of these proteins in cell processes and CD44-positive perinuclear vesicles (). Both anti-CD44 and anti-Rhamm antibody significantly reduced the intensity of nuclear phospho-ERK1,2 (), raising the possibility that CD44–Rhamm complexes are required for maximal activation of ERK1,2. To test this, the consequences of blocking Rhamm, CD44, and ERK1,2 motogenic functions on fibroblast migration were measured.
Anti-Rhamm, anti-CD44 antibody, and a mitogen-activated kinase kinase 1 (Mek1) inhibitor (UO126) significantly blocked serum-induced motility of Rh-rescued but not Rh fibroblasts (), confirming that Rhamm, CD44, and ERK1,2 activity are required for optimal motility.
Signaling through CD44 and ERK1,2 are required for motility in response to HA (). We therefore next measured the motility of Rh-rescued versus Rh fibroblasts stimulated with HA. To render nontransformed cells sensitive to HA, fibroblasts can be pretreated with PMA; this activates PKC-dependent processes, permitting motogenic responses to HA (). Compared with PMA treatment alone, a mixture of high molecular weight HA and oligosaccharides significantly promoted random motility of Rh-rescued but not Rh fibroblasts (). Motogenic stimulation by HA required cell surface Rhamm, as anti-Rhamm antibodies blocked a response to HA. The inability of HA to increase Rh fibroblast motility was puzzling because these cells express similar levels of CD44s as Wt (). We suspected that the localization of CD44 to the cell surface might be altered in Rh cells, as a reduced number of perinuclear CD44-positive vesicles, which can traffic to the cell surface (), were observed in Rh versus Rh-rescued fibroblasts ().
Adherent Rh fibroblasts exhibited reduced CD44 surface display compared with Rh-rescued fibroblasts (). Both Mek1 inhibition with UO126 () and anti-Rhamm antibody () reduced CD44 display. Rh-rescued fibroblasts also showed increased colocalization of CD44 with active ERK1,2, particularly in the cell nucleus and perinuclear vesicles, compared with Rh fibroblasts (). Anti-Rhamm antibody reduced colocalization of these proteins (). These results suggest a role for cell surface Rhamm in ERK1,2 activation and, consequently, surface display of CD44. We next examined whether activation of ERK1,2 in the absence of Rhamm would have a similar effect.
Expression of active Mek1 restored sustained serum-induced ERK1,2 activity in Rh fibroblasts (). This effect was not enhanced by coexpression of Rh (). Mek1 also restored migration of Rh fibroblasts () and promoted CD44 display (). These results suggest that Rhamm and Mek1/ERK1,2 act on the same CD44-regulated motogenic signaling pathway because Mek1 can compensate for Rhamm in these functions. Both cell surface and intracellular Rhamm have been proposed to regulate functions associated with motility and to affect signaling cascades (; ). The ability of anti-Rhamm antibody to mimic defective motogenic signaling of Rh fibroblasts suggests a role for cell surface Rhamm in these functions but does not exclude an involvement of intracellular Rhamm. Blocking antibodies do not permit a direct assessment of the relative roles of cell surface versus intracellular Rhamm in cell functions. To directly assess the importance of cell surface Rhamm to motogenic signaling, we exposed Rh fibroblasts to recombinant Rhamm linked to Sepharose beads to restrict it to the extracellular compartment.
Rhamm beads significantly stimulated motility of Rh fibroblasts that contacted beads, whereas GST beads had no effect (). Both anti-CD44 and anti-Rhamm antibody (unpublished data) significantly blocked Rhamm bead–stimulated motility. CD44 surface display was dramatically increased in Rh fibroblasts contacting Rhamm beads (). Importantly, fibroblasts deficient in both Rhamm and CD44 (Rh:CD44) did not increase motility in response to Rhamm beads (). Collectively, these results indicate that cell surface Rhamm is required for CD44 display and motility but that intracellular Rhamm proteins are not required for these functions. Furthermore, these results show that CD44 is required for the motogenic effect of cell surface Rhamm.
To understand the molecular mechanisms by which cell surface Rhamm promotes CD44 surface display and increased motility, we determined whether Rhamm bead–induced motility of Rh fibroblasts requires ERK1,2 activity. Inhibition of Mek1 significantly reduced motility of Rh fibroblasts in contact with Rhamm beads (). Conversely, Rhamm beads significantly stimulated serum-induced ERK1,2 activity and translocation to the nucleus (). ERK1,2 activity was not as high as that of Rh-rescued fibroblasts, suggesting a possible role for intracellular Rhamm forms in maintaining maximal activity of these kinases. A role for CD44 in Rhamm bead–promoted ERK1,2 activity was shown by the ability of CD44 antibodies to block Rhamm bead effects and by the lack of Rhamm bead–induced ERK1,2 activity in Rh:CD44 fibroblasts. Because expression of Rhamm and CD44 are increased during excisional skin wound repair () and fibroblast migration and differentiation are essential components of successful excisional wound repair, we asked whether Rhamm loss alters repair of excisional skin wounds in vivo.
We have shown that Rh fibroblasts exhibit defective activation and subcellular targeting of ERK1,2 via CD44. To prove that Rhamm is indeed a player in the physiological regulation of ERK1,2, their activity was quantified in fibroblasts of Rh versus Wt wound granulation tissue. Wt granulation tissue fibroblasts exhibited strong staining for active ERK1,2 3 d after wounding (). Staining intensity increased sixfold by day 7 and did not drop significantly until day 13. Rh granulation tissue fibroblasts also exhibited strong activation of ERK1,2 at 3 d after wounding but dropped dramatically by day 7 and remained low at day 13 (). These differences in Rh versus Wt ERK1,2 activity were not due to decreases in total ERK1,2 protein levels because immunoblot analyses revealed that Rh and Wt granulation tissue expressed similar amounts of ERK1,2 protein (unpublished data). These results suggest that Rhamm expression is required for sustaining ERK1,2 activity in granulation tissue fibroblasts in vivo as well as in culture.
MAPKs have been implicated in processes relevant to wound repair, including contraction, cell migration, and mesenchymal differentiation (; ).
We first confirmed that Rhamm expression increases during repair of excisional wounds by analyzing wound Rhamm mRNA during the first 7 d after injury. RT-PCR analysis of uninjured skin confirmed low Rhamm expression (Fig. S2 a, available at ). A marked increase in Rhamm mRNA was obvious 1 d after injury, and expression was increased until day 3, when mRNA levels began to drop. By day 7, the mRNA levels were only slightly higher than those observed in uninjured skin. These results indicate that Rhamm is expressed during wound contraction, reepithelialization, and early granulation tissue formation. The consequence of Rhamm loss for the integrity of these early processes was recorded.
Wt and Rh wounds both contracted by 1–3 d after injury, but early contraction of Rh wounds was significantly reduced compared with Wt (Fig. S2 b). By day 14, Wt and Rh wounds appeared resolved at the macroscopic level (unpublished data). However, analysis of serial tissue sections of wound centers revealed significant reductions in contraction and obvious differences in the dermal structure of Rh wounds at all time points (Fig. S2 c). 21 d after injury, both the thickness and the cellularity of the remodeling dermis were significantly greater in Rh versus Wt wounds. In addition, the differentiation of structures such as hair shafts and muscle within the wound site was reduced in Rh versus Wt wounds (Fig. S3 b, available at ). A significant decrease in the thickness of Rh versus Wt dermis was also observed before injury (Fig. S3 a). These observations are consistent with an ERK1,2-regulated defect in Rh wound fibroblast function. We next analyzed the consequences of Rhamm loss on granulation tissue formation/resolution, a process that is dependent on fibroblasts.
A temporal defect in the formation and resolution of granulation tissue was confirmed in Rh versus Wt wounds (). Tenascin-positive granulation tissue was abundant in day 3 Wt wounds and began to decrease by day 7 (). At day 14, Wt wound granulation tissue was largely resolved (). In contrast, the area of tenascin-positive granulation tissue in day 3 and 7 Rh wounds was smaller than Wt. Rh day 14 wounds were still strongly tenascin-positive and highly variable (), as the pattern of staining was “patchy,” in contrast to Wt (). An additional difference in Rh wounds was the transient appearance of a thick layer of subcutaneous adipocytes in day 1–3 Rh wounds (; unpublished data). These results suggest that a prominent effect of Rhamm loss during wound repair may be a miscuing of signals required for a correct ratio of fibroblasts to adipocytes that arise from progenitor cells, which leads to aberrant regulation of granulation tissue formation and resolution.
Fibroplasia is a particularly prominent feature of granulation tissue in excisional skin wounds. The biological activities of fibroblasts are key factors in the formation of early granulation tissue architecture (). Robust fibroplasia, as quantified by the density/unit area of granulation tissue fibroblasts, was apparent in Wt wounds on day 3 and was increased by day 7 (). Myofibroblasts, detected by smooth muscle actin staining, appeared in Wt wounds by day 7 (). Fibroplasia was observed in day 3 and 7 Rh granulation tissue but was blunted and accompanied by a significant decrease in myofibroblasts at day 7 (). The presence of abundant wound-edge adipocytes, indicated by the presence of vacuolated cells (, arrow), was confirmed by staining with the lipophylic dye BODIPY493/503 (; unpublished data). Rh cells explanted from both uninjured skin (day 0) and day 7 wounds expressed less smooth muscle actin and accumulated more lipid than explanted Wt cells (). Thus, deletion of Rhamm results in lower fibroblast density as well as aberrant differentiation in Rh granulation tissue in vivo and in culture.
Our study identifies Rhamm as a fibrogenic factor that is required for temporal and spatial regulation of granulation tissue formation and resolution. An underlying signaling defect associated with Rh wounds is deregulated ERK1,2 activation, which promotes fibroblast migration as well as mesenchymal cell differentiation. This conclusion is supported by the demonstration that Rh fibroblasts are unable to appropriately activate ERK1,2 in culture and exhibit migration defects as measured by several locomotion assays, and that these defects are rescued by expression of mutant active Mek1. Our results further reveal an autocrine mechanism by which cell surface Rhamm promotes motility in culture. This form of Rhamm promotes ERK1,2 activation via an association with CD44, which in turn is required for maintaining cell surface display of CD44. ERK1,2 is acting “downstream” of cell surface Rhamm in this function, as expression of mutant active Mek1 is sufficient to maintain cell surface CD44, activate ERK1,2, and restore motility in the absence of Rhamm expression. This motogenic mechanism involves formation of cell surface Rhamm–CD44–ERK1,2 complexes and is apparently required for both growth factor– and HA-mediated motility. These findings identify for the first time a mechanism by which Rhamm, a nonintegral membrane protein, can activate intracellular signaling cascades and provide a novel mechanism by which ERK1,2 promotes motility. Furthermore, our results suggest a previously unidentified role for ERK1,2 activation kinetics in excisional skin wound granulation tissue formation/resolution.
Rhamm belongs to a group of proteins that are predominantly intracellular but can be exported to the cell surface via unconventional transport mechanisms that do not involve the export through the Golgi/ER (). We show that cell surface Rhamm is displayed in culture after injury, and our results have begun to clarify functions for cell surface Rhamm versus intracellular Rhamm forms. Although we did not set out to define a role for intracellular Rhamm, indirect evidence suggests that it plays a role in mitotic events, at least in culture, as cell surface Rhamm did not rescue the abnormal mitosis observed during time-lapse analysis of Rh fibroblasts (unpublished data). These results are consistent with evidence for intracellular Rhamm function during progression through GM of the cell cycle (; , ) and its presence on centrosomes and mitotic spindle microtubules (; ). Nevertheless, our current data do not provide support for an essential role of either intracellular or cell surface Rhamm in mitotic spindle formation and cell cycle regulation during wound repair in dermal fibroblasts in vivo as judged by the lack of detectable differences in proliferation or apoptotic indices within Rh versus Wt wound sites. The slightly disorganized migration of Rh fibroblasts from scratch wound assays on tissue culture plastic is consistent with a possible centrosome defect that could contribute to aberrant migration () and merits further experimentation. A role for Rhamm in collagen contraction has been controversial in culture (; ). Unexpectedly, therefore, our studies suggest that Rhamm is necessary for recruitment/differentiation of myofibroblasts and contraction of the wound bed. As is increasingly reported and recognized, both of these results emphasize the importance of context and the microenvironment in regulating tissue-specific signaling (). Thus, data obtained in culture, especially on 2D substrata, need to be confirmed in vivo.
ERK1 and -2 are closely related MAPK isoforms that are activated by Mek1 or -2 and regulate signaling pathways that control cell motility, invasion, and cytoskeleton remodeling during migration in culture (; ). Our results show that migration defects of Rh fibroblasts result from an inability to sustain and maximally activate ERK1,2 after growth factor stimulation. These results are consistent with our previous evidence that cell surface Rhamm is required for PDGF-stimulated ERK1,2 activity in mesenchymal cells and for promoting migration by regulating signaling through upstream activators of ERK1,2, including HA, Src, Ras, and FAK (, ; ). Others have also documented a role for cell surface Rhamm in activating signaling cascades that regulate motility and that directly or indirectly affect ERK1,2 activation (; ; ). Although cell surface Rhamm can promote ERK1,2 activity to levels sufficient to sustain motility in the absence of intracellular forms, levels were lower than in cells expressing both cell surface and intracellular Rhamm (e.g., Rh-rescued fibroblasts). Furthermore, although cell surface Rhamm–activated ERK1,2 translocated to the nucleus, activity did not persist here as it did when intracellular Rhamm was also present. These results indirectly implicate intracellular Rhamm forms in aspects of ERK1,2 activation/compartmentalization. These intracellular deficiencies did not affect the ability of cell surface Rhamm to promote motility during our experimental time frame but could affect other functions associated with ERK1,2 activity, such as invasion and mitosis, neither of which were rescued by cell surface Rhamm alone (unpublished data). The consequences of ERK1,2 signaling on cell differentiation, migration, and proliferation depends on activation kinetics and subcellular compartmentalization (; ; ). These factors are determined by receptor dimerization and internalization, cross talk with other receptors, association of ERK1,2 with adaptors, and activation of other kinases or phosphatases that modify ERK1,2 activity. Our study raises the possibility that cell surface and intracellular Rhamm may differentially affect the activation levels and subcellular targeting of ERK1,2, which have consequences to motility- and invasion-related gene expression and phosphorylation of intracellular substrates that are involved in cell migration/invasion ().
ERK1,2 regulate motility by both transcriptional and posttranslational mechanisms. For example, initiation and early phases of migration during wound repair do not require transcription () but, rather, involve phosphorylation of predominantly cytoskeleton-associated substrates required for motility, such as myosin (; ; ). Our results also identify a role for ERK1,2 activity in sustaining cell surface display of CD44, an integral membrane protein required for motility in response to growth factors and HA (). ERK1,2 promote recycling of clathrin-negative early endosomes back to the cell surface, a pathway associated with recycling of β1 integrins and E-cadherin (). A similar ERK1,2-regulated recycling event may be responsible for maintaining CD44 at the cell surface.
We have shown that CD44 and Rhamm have overlapping functions in regulating migration events and that Rhamm can compensate for loss of CD44 in aspects of splenocyte migration into arthritic joints, although the reverse may not be true (). These and other studies (; ) suggest that Rhamm can promote cell motility independently of CD44. Very likely, in these instances, cell surface Rhamm associates with other adhesion receptors involved in cell motility, and partnering may depend on expression and cell surface display levels of these receptors, which will vary with the nature of disease, cell type, and temporal stage of wound repair.
Medical-grade HA prepared from bacterial fermentation (a gift from SkyePharma) was free of detectable proteins, DNA, or endotoxins (). The average molecular mass and polydispersity was 276.7 and 1.221 kD, respectively. HA oligosaccharides (average molecular mass 10 kD; a gift from F. Winnik, University of Montreal, Montreal, Canada) were prepared by partial digestion with testicular hyaluronidase and purification by gel filtration. The following primary antibodies were used: ERK1 (immunohistochemistry/Western blot), actin (Western blot), vimentin (immunohistochemistry), α-smooth muscle actin (immunohistochemistry), nonimmune IgG (Santa Cruz Biotechnology, Inc.), Ki67 (immunohistochemistry; DakoCytomation), tenascin (immunohistochemistry; Chemicon), phospho-ERK1,2 (immunohistochemistry/Western blot/immunofluorescence; Cell Signaling), CD44 (Western blot/immunofluorescence/blocking; IM7; BD Biosciences), CD44 (blocking; Hermes-3; a gift from D. Naor, The Hebrew University of Jerusalem, Jerusalem, Israel), Rhamm (immunofluorescence; ProSci), and Rhamm (blocking; Zymed Laboratories). Specificity of Rhamm and CD44 antibodies were determined using Rh and CD44 cells, respectively. The following secondary antibodies were used: anti-rabbit Alexa 555 and anti-rat Alexa 433 (Invitrogen), HRP-anti-mouse (Bio-Rad Laboratories), anti-rabbit (BD Biosciences), and anti-rat (Santa Cruz Biotechnology, Inc.). ApopTag peroxidase in situ apoptosis detection kit (Chemicon) was used for quantification of apoptosis, and FACE ERK1/2 ELISA kit (Active Motif) was used to detect phospho-ERK1,2. Other reagents used were as follows: human plasma fibronectin (BD Biosciences), PDGF (PDGF-BB), and PMA (Sigma-Aldrich); 50 μM PD098059 and 10 μM U0126 (BD Biosciences); collagen (Vitrogen100; Cohesion); Matrigel (BD Biosciences); immunofluorescence mounting medium with DAPI (Vectashield; Vector Laboratories); Cytoseal (Richard-Allan Scientific); and ABC staining system (Santa Cruz Biotechnology, Inc.). All antibodies and reagents were used according to the manufacturer's instructions unless otherwise stated.
All animal experiments complied with the University of Western Ontario (London, Canada) animal use committee regulations. Preparation of Rh mice and mouse embryonic fibroblasts () and CD44 mice () have been described. Rhamm and CD44 heterozygous mouse mating generated Rh:CD44 double-knockout mice. Dermal fibroblasts were isolated from newborn skin explants and granulation tissue cells from wound punch explants (cultured dermal side down).
Cell culture was done as described previously (; ). 25 ng/ml PDGF, 500 ng/ml–1 mg/ml HA, or 10% serum (FCS) were added to 24-h serum-starved, 50% subconfluent fibroblasts plated on 25 μg/ml fibronectin-coated dishes (; ). For HA responsiveness, cells were pretreated with 5 nM PMA (). For antibody blocking experiments, serum-starved cells were preincubated for 30 min with anti-Rhamm antibody, anti-CD44 antibody, or control IgG (10 μg/ml) before the addition of 10% FCS. Immortalized Rh cells were transfected with murine Rh and/or mutant active Mek1 (a gift from N. Ahn, University of Colorado at Boulder, Boulder, CO) using Lipofectamine Plus (Invitrogen) as described previously () and were selected in 1–5 mg/ml G418 (Sigma-Aldrich).
Cells were plated and stimulated with HA or PDGF. For FCS stimulation, fibroblasts were plated onto serum-coated flasks. Cells were filmed as described previously (; ).
Confluent cell monolayers were starved overnight, scratch wounded (3 mm) with a sized cell scraper, and stimulated with FCS or PDGF for 24–48 h. Monolayers were fixed (3% paraformaldehyde), stained (0.1% methylene blue), and imaged. For 3D assays, collagen or Matrigel gels containing fibroblasts (5 × 10 cells/ml) were prepared with plastic inserts placed in the gel center. After 24–48 h, the inserts were removed and the cell free space was filled with collagen containing 25 ng/ml PDGF, 100 μg/ml HA, and 25 ng/ml fibronectin. Gels were fixed and analyzed 72 h later.
Western blots of CD44, phospho-ERK1,2, and total ERK1,2 proteins were performed as described previously (; ; ). Densitometry was performed using Image Quant 5.1 software (Molecular Dynamics).
Immunofluorescence of phospho-ERK1,2 was done as described previously (). Immunofluorescence of CD44 and Rhamm was done using the same methods with overnight incubation of the primary antibodies at 4°C. Live-cell CD44 immunofluorescence was done as described previously ().
For quantification of serum or PDGF-induced ERK1,2 activation, cells were plated, serum starved, and stimulated with 10% FCS or PDGF as described. Western blots and immunofluorescence were done as described. The ERK1,2 ELISA was done according to the manufacturer's instructions.
Rhamm-GST (72-kD murine isoform) and GST recombinant proteins were prepared as described previously (). For pull-down assays, recombinant Rhamm-GST or GST beads were incubated with 500 μg of Rh-rescued lysate overnight at 4°C. Beads were then washed with cold lysis buffer. Proteins associated with beads were boiled in SDS buffer and were detected by Western blot as described.
Wt and Rh mice were anaesthetized by Halothane inhalation. Two full-thickness 4-mm wounds were placed at the same location on denuded backs of age-matched male mice (9–18 mo), were harvested at the indicated times using an 8-mm metal punch, and were fixed and paraffin embedded as described previously (). To ensure that serial sections were cut starting at the wound center, samples were cut in half through the wound center before embedding. The first and last sections were stained with Masson's trichrome.
Rhamm mRNA was amplified from excised wound tissue, and PCR products were detected as described previously (). βActin was used as a loading control ().
Immunohistochemistry of paraffin processed sections was done as described previously (). Antigens were retrieved in 10 mM sodium citrate buffer, pH 6, heated to boiling, except for smooth muscle actin staining. Neutral lipids were detected in fixed cultured cells or frozen wound sections with 25 μg/ml BODIPY 493/503 (Invitrogen; ).
Masson's trichrome, hematoxylin and eosin, vimentin, tenascin, and phospho-ERK1,2 stained tissue section images were taken with air objectives (4×, NA 0.16; 20×, NA 0.7; Olympus) on a microscope (AX70 Provis; Olympus) with a color camera (Cooke SensiCam; CCD Imaging) and Image Pro Plus 4.5.1.2.9 (Media Cybernetics). Phospho-ERK1,2 staining was quantified with Photoshop 6.0 (Adobe). The area of blue (hematoxylin, total number of cells) was quantified. After deletion of the blue pixels, the peroxidase substrate (phospho-ERK1,2) was quantified. Tenascin staining was quantified using Simple PCI (Compix). Images in are composites of images taken with the 4× air objective. The colors were enhanced using Photoshop. Scratch wound and surface CD44 immunofluorescence images were taken with air objectives (4×, NA 0.1, and 20×, NA 0.4, respectively [Nikon], with Hoffman Modulation Contrast optics) using a microscope (Eclipse TE300; Nikon) with a digital camera (Hamamatsu) and Simple PCI software. In vivo wound images (Fig. S2 a) were taken with a digital camera (Dimage Z3; Minolta) with a 12× zoom. The wound area was quantified using Simple PCI. Confocal images were taken using an oil objective (63×, NA 1.4; Carl Zeiss MicroImaging, Inc.) with a confocal microscope (510 LSM Meta; Carl Zeiss MicroImaging, Inc.) using LSM 5 software (Carl Zeiss MicroImaging, Inc.). Fluorescence intensity of images was measured using LSM 5 software. Colocalization was identified and quantified using ImageJ software (). Cell surface CD44 images were deconvolved using Simple PCI's nearest-neighbor deconvolution. All images were acquired at room temperature. Unless otherwise indicated, comparisons between samples were assessed for statistical significance using a test; P < 0.05 was considered significant, and significant differences between values are marked with asterisks.
Fig. S1 shows that transfection of Rh into Rh fibroblasts rescues defective migration in scratch wounds and invasion in collagen gel assays. Fig. S2 shows that Rhamm mRNA expression is transiently increased after excisional skin wounding between days 1 and 7. Photographs of wounds show that Rh wounds contract more slowly than Wt. Tissue sections through excisional skin wounds confirm that loss of Rhamm results in significantly reduced wound contraction relative to Wt. Fig. S3 shows that the dermal structure of uninjured and repaired Rh skin is aberrant compared with Wt. This includes alterations in dermal thickness and in differentiation of dermal cell types. Online supplemental material is available at . |
Remarkably, neurons have the ability to undergo long-lasting changes in the strength and pattern of their synaptic connectivity in response to environmental stimuli. Long-term modification of synapses is thought to mediate many aspects of brain functions, particularly learning and memory (). In general, long-term synaptic plasticity has two distinct features. First, long-lasting changes in the structure of synapses accompany alterations in synaptic efficacy that include both the modification of existing synapses and the formation of new synapses. Second, long-term synaptic plasticity requires changes in gene expression and new protein synthesis (; ). The transcription factor cAMP response element (CRE)–binding protein (CREB) is thought to play a key role in the synthesis of new proteins required for long-term synaptic plasticity.
CREB, which is a member of basic leucine zipper (bZIP) family of dimeric transcription factors, controls the expression of many plasticity-related genes through binding to the CRE. Numerous studies have established that CREB plays a critical role in long-term synaptic plasticity and memory in a variety of model systems, such as flies, mollusks, and rodents (; ; ; ; ). Is CREB involved in both structural and functional modifications of synapses? Inhibition of CREB activity in blocks both long-term facilitation of synaptic efficacy and synaptic growth (). In , however, CREB controls only functional, and not structural, synaptic plasticity (). Gene knockout experiments in mice have yielded variable results regarding hippocampal synaptic plasticity. Some investigators have reported that disruption of CREB function impairs some forms of long-term potentiation (LTP; ; ), whereas others find no considerable LTP deficits (; ; ; ). There is some evidence that CREB is also involved in structural changes at hippocampal synapses ().
Recently, neurotrophins have emerged as major factors that regulate synaptic structure and function (; ; ). These factors are widely expressed in the brain, and can elicit both acute and long-term modifications of synapses (). In general, neurotrophins bind to their cognate Trk receptor tyrosine kinases and initiate their synaptic functions through the following three signaling pathways: MAPK, phosphatidylinositol 3-kinase, and PLC-γ (; ). An important task is to determine the signaling mechanisms for specific neurotrophins in specific aspects of synaptic modulation. Several lines of evidence suggest that long-term synaptic modulation by neurotrophins also involves synaptic growth and requires new protein synthesis (). Long-term treatment of hippocampal slices with brain-derived neurotrophic factor increases the number of synapse, spine density, and synaptic proteins in rodent CA1 pyramidal neurons (; ). In cultured neuromuscular synapses, long-term exposure to neurotrophin-3 (NT-3) induces a series of profound structural changes, specifically, an increase in the number and size of synaptic varicosities (). In a recent study, we demonstrated that the long-term structural changes at the neuromuscular synapses induced by NT-3 require protein synthesis in the presynaptic neurons ().
Does long-term synaptic modulation by neurotrophins require the transcription factor CREB? What are the relationships between structural and functional changes at synapses induced by neurotrophins? In this study, we used the neuromuscular synapses as a model system to study the signaling mechanisms underlying the structural and functional plasticity induced by NT-3. We demonstrated that although NT-3–induced enhancement in the efficacy at a single synapse requires CREB activation in presynaptic neurons, the long-term increase in the number of synaptic sites elicited by NT-3 is CREB-independent. Instead, the morphological changes in the presynaptic terminals require activation of the MAPK pathway. Thus, the structural and functional synaptic modifications by NT-3 are mediated by separate mechanisms. Using various imaging tools, we have also identified signaling events upstream of CREB and MAPK. Together, these data argue that a concomitant activation of CaMKIV–CREB–mediated transcription and sustained Rap1–MAPK signaling may be necessary for the long-term functional and structural changes necessary for neurotrophin-dependent synaptic modulation.
CREB activation is critical for several forms of sustained synaptic plasticity (, ; ). It is unclear, however, whether CREB mediates the long-term synaptic modulation by neurotrophins. To measure CREB activation, we used an antibody that detects the phosphorylation at serine 133 of CREB. spinal neurons exhibited rapid activation of CREB (<30 min) after NT-3 treatment (5 ng/ml), as indicated by dark nuclear staining (). To better quantify CREB activation, we used immunofluorescence, and measured the peak fluorescence intensity in the nuclear region (, and Fig. S1 A, available at ). Compared with untreated neurons, NT-3 treatment increased the level of fluorescence signals in the nucleus of spinal neurons by threefold ().
CREB can be phosphorylated on serine 133 by multiple kinases, including protein kinase A (PKA), calcium/calmodulin-dependent kinases (CaMKs), and MAPK. To identify the signaling pathways mediating the CREB activation, we examined phospho-CREB (pCREB) fluorescence in cultured spinal neurons in the presence of inhibitors specific for these pathways. Pretreatment of spinal neurons with the MAPK kinase (MEK) inhibitor PD098059 (10 μM; ) or PKA inhibitor Rp-cAMP (5 μM; ) for 1 h could not attenuate the increase in pCREB fluorescence after NT-3 application (). In contrast, pretreatment with the CaM kinase inhibitor KN93 (1 μM; ) completely prevented the increase in pCREB fluorescence induced by NT-3 (). These data suggest that CREB activation, at least within 1 h after NT-3 application, is mediated through the CaMKs.
It has been reported that in hippocampal neurons, CaMKIV is responsible for early activation of CREB, whereas the late, sustained phosphorylation of CREB depends on MAPK activity (). To examine whether CaMK or MAPK mediates the sustained CREB phosphorylation in developing spinal neurons, we monitored the time course of CREB activation in the presence or absence of specific inhibitors. NT-3 treatment significantly increased CREB phosphorylation as early as 30 min, and maximum activation of CREB was observed 1 h after NT-3 treatment (P < 0.05; ). This increase was maintained at 93% level until 24 h after NT-3 treatment, indicating that NT-3 induced a sustained phosphorylation of CREB. The sustained activation of CREB was abolished by KN93, but not PD098059 (). Thus, in developing spinal neurons, NT-3–induced CREB phosphorylation is mediated by CaMKs, not MAPK.
NT-3 is known to induce both acute and long-term changes in synaptic efficacy at the neuromuscular synapses (; ; ; ). To determine whether CREB activation is necessary for either the acute or the long-term effects of NT-3, we introduced dominant-negative CREB (DnCREB) mutants into spinal neurons by embryo injection. Two different DnCREB mutants were used; the K-CREB, a mutant that contains mutations in the DNA-binding domain of CREB; and the A-CREB, a mutant that contains basic-to-acidic residue mutations within the CREB bZIP domain (; ). These variants of CREB proteins act as dominant repressors for gene transcription by forming an inactive dimer with endogenous CREB and blocking its ability to bind cAMP response element (CRE). Therefore, expression of these DnCREB mutants cannot change the phosphorylation status of Ser133 in the endogenous CREB, but can still inhibit CREB-mediated gene transcription. The mRNAs of DnCREB mutants were coinjected with GFP mRNAs into one blastomere of embryos at the two-cell stage. Nerve-muscle cocultures were prepared from the injected embryos after 24 h. Expression of K-CREB in spinal neurons had no effect on the basal synaptic transmission. The frequency of spontaneous synaptic currents (SSCs) recorded from K-CREB–positive (N+) and K-CREB–negative (N−) neurons were 4.7 ± 0.5 events/min ( = 11) and 4.3 ± 0.6 events/min ( = 7), respectively. Moreover, identical to that observed in control synapses (N–), acute application of NT-3 induced an increase in frequency, but not amplitude, of SSCs (). Collectively, inhibition of CREB activation does not affect the acute modulation of transmission by NT-3.
In contrast, expression of K-CREB in presynaptic neurons completely blocked the enhancement of synaptic efficacy induced by long-term exposure to NT-3. In a culture treated with NT-3 for 48 h, the mean SSC frequency of the synapses made by K-CREB–expressing neurons was reduced to the level of NT-3–untreated neurons (). However, expression of K-CREB in the postsynaptic muscle cells had no effect on NT-3–induced long-term synaptic potentiation (), suggesting that CREB activation is only required in presynaptic neurons. Moreover, expression of K-CREB either pre- or postsynaptically had no effect on SSC amplitude, whether the synapses were treated with NT-3 or not (Fig. S2 A, available at ). Almost identical results were obtained from synapses expressing A-CREB presynaptically (). In addition to potentiating spontaneous synaptic activity, long-term treatment with NT-3 enhances functional synaptic transmission, as reflected by an increase in the amplitude of evoked synaptic currents (ESCs), which are elicited by stimulating presynaptic somata of spinal neurons (). Expression of K-CREB in presynaptic spinal neurons completely blocked the increase in ESC amplitude induced by NT-3 (). Collectively, these results suggest that presynaptic activation of CREB is required for NT-3– induced long-term, but not acute, synaptic transmission at the neuromuscular synapses.
Given that CaMK signaling was required for NT-3–induced CREB activation (), we next asked whether the CaMK pathway also mediates the long-term effects of NT-3. Treatment with NT-3 for 48 h increased the frequency of SSCs by ∼2.5-fold over untreated control cells (). In contrast, in cultures pretreated with 1 μM KN93, NT-3 elicited no increase in SSC frequency, indicating that NT-3–induced synaptic potentiation requires activation of the CaMK signaling pathway. Inhibition of either MAPK () or PKA (Fig. S2 D) did not exhibit statistically significant change in NT-3–induced long-term physiological effects. Therefore, these data suggest that the NT-3–induced long-term physiological effect is primarily mediated through a CaMK-dependent mechanism.
KN93 is a general inhibitor of CaMK family members, which include CaMKI, CaMKII, and CaMKIV, but CaMKI and CaMKII are enriched in neuronal processes and synapses, whereas CaMKIV is relatively enriched in the nucleus (; ; ). The nuclear localization of CaMKIV suggests its ability to link extracellular signals to CREB activation in the nucleus. Therefore, we tested whether CaMKIV mediates both CREB activation and the long-term synaptic effects elicited by NT-3 by using Dn-CaMKIV. This mutant of CaMKIV bears a T196A mutation, which disrupts CaMKIV from binding to CaMKIV kinases (). Expression of the Dn-CaMKIV mutant was reflected by GFP fluorescence in cultured spinal neurons (). NT-3 treatment for 30 min caused an approximately fourfold increase in pCREB immunoreactivity in control spinal neurons, but not in neurons expressing Dn-CaMKIV (). These findings are consistent with a recent study showing that expression of Dn-CaMKIV could attenuate CREB activation and CREB-mediated transcription in hippocampal neurons ().
Cells expressing the Dn-CaMKIV mutant did not display any defect in growth or morphology (unpublished data). To determine whether CaMKIV activation is required for long-term effects of NT-3 on synaptic transmission, we measured SSC frequency in spinal neurons expressing Dn-CaMKIV mutant. Expression of Dn-CaMKIV in spinal neurons completely blocked the enhancement of synaptic efficacy induced by long-term exposure to NT-3 (). In cultures treated with NT-3 for 48 h, the frequency of SSCs in the synapses made by Dn-CaMKIV–expressing neurons was significantly lower than that in synapses made by nonexpressing control neurons in the same dishes (). Dn-CaMKIV had no effect on SSC amplitude, suggesting a pure presynaptic effect (Fig. S2 B). Expression of Dn-CaMKIV in presynaptic neurons also completely prevented the potentiating effect of NT-3 on ESC amplitude (). Inhibition of CaMKII with a specific peptide inhibitor blocked the acute potentiation of synaptic transmission by NT-3 (). In neurons expressing the Dn-CaMKIV, however, acute application of NT-3 could still elicit a marked increase in SSC frequency (Fig. S3, available at ). Thus, the acute effect of NT-3 on synaptic transmission is mediated by CaMKII, whereas the long-term effect is mediated by CaMKIV. These results suggest that the long-term effect of NT-3 on synaptic transmission is mediated by the CaMKIV–CREB pathway.
CaMKIV activation requires the translocation of Ca-calmodulin (CaM) into the nucleus, and such nuclear translocation of CaM appears to be important for the activation of CaMKIV–CREB pathway (). To test whether NT-3 is capable of inducing nuclear localization of CaM, we expressed GFP-CaM fusion protein in spinal neurons (). Expression of GFP-CaM did not affect the normal growth or morphology of neurons (, left). Western blot showed the expression of GFP-CaM in neural tubes 2 d after embryonic DNA injection. In embryos injected with GFP-CaM cDNA, a prominent band of 47 kD was detected using an anticalmodulin antibody. This band was also stained positively with antibody against GFP, confirming the expression of GFP-CaM fusion protein (, right).
We next studied the NT-3–induced movement of GFP-CaM in spinal neurons. GFP-CaM fluorescence was monitored in the nucleus and axonal periphery of neurons. Before application of NT-3, culture medium was replaced with Ringer's solution for 2 h to reduce the background CaM movement caused by unspecified factors accumulated in the medium. Fluorescence intensity changes of selected areas in the nucleus and axons of a neuron (indicated by dotted lines in , top) upon NT-3 application were monitored by using time-lapse microscopy. Application of NT-3 evoked a rise in the fluorescence in the nucleus, paralleling a decrease in the fluorescence intensity in the axons (, top). The increase in nuclear CaM fluorescence was relatively fast; within 20 min, the nuclear fluorescence reached a plateau (, bottom left). In all experiments, the total averaged cellular fluorescence (corrected for photobleaching effects) remained unchanged, indicating that observed fluorescence intensity changes are caused by redistribution of GFP-CaM. Application of NT-3 for 30 min increased the intensity of nuclear fluorescence by 29%, but decreased that of the axonal fluorescence by 35%. These results indicate that application of NT-3 evoked active translocation of CaM from nonnuclear regions into nuclear region in neurons.
In addition to physiological changes, long-term treatment with NT-3 induces the morphological development of presynaptic terminals (; ). Using FM dye to label endocytosed synaptic vesicles, our previous study showed that chronic, but not acute, treatment with NT-3 promotes the formation of “synaptic varicosities,” which are enlargements along the axons that reflect clusters of synaptic vesicles and other presynaptic elements (). There are two potential problems associated with this approach. First, FM dye also labels postsynaptic membrane structures and organelles. To avoid this problem, we visualized the synaptic varicosity by expressing a fusion protein in which GFP is fused with the C terminus of synaptophysin (SYP-GFP), a major integral membrane protein of synaptic vesicles (; ). SYP-GFP was coinjected with rhodamine-dextran in embryos so that we could simultaneously label axonal morphology and synaptic varicosity (). Furthermore, to ensure that SYP-GFP labels functional synaptic vesicles, neurons transiently transfected with SYP-GFP were loaded with FM 4–64 dye. We found that that SYP-GFP fluorescence overlapped completely with FM dye–labeled structures at the presynaptic terminals (). The second problem is that labeling presynaptic vesicles alone may not faithfully reveal specific changes in the presynaptic terminals at the synapses. Therefore, we labeled postsynaptic acetylcholine receptors (AChRs) with a low concentration of Cy5-labeled α-bungarotoxin (, Cy5-α-BTX). The “synaptic site” was revealed by the juxtaposition of presynaptic SYP-GFP (green) and postsynaptic AChR (, red).
Using the double-labeling technique, we showed that chronic treatment (48 h) of NT-3 significantly increased the number of synaptic sites (). To test the role of CREB in NT-3–induced long-term morphological changes, we inhibited CREB activation by expressing K-CREB in spinal neurons and quantified the changes in synaptic sites. Treatment with NT-3 induced a twofold increase in the number of synaptic sites per myocyte. Surprisingly, although expression of K-CREB blocked the physiological changes of synapses triggered by NT-3, expression of K-CREB in spinal neurons could not inhibit a NT-3–induced increase in synaptic sites (). These results suggest that CREB activation is only required for functional, but not structural, changes of synapses induced by NT-3. Furthermore, these results raise the possibility of parallel signaling pathways mediating NT-3–induced functional and morphological changes.
What are the signaling pathways that mediate the increase in the number of synaptic sites induced by long-term exposure to NT-3? NT-3 is known to activate MAPK in spinal neurons (). Given that neurotrophins have been shown to promote neurite outgrowth via MAPK pathway in cortical neurons and PC12 cells (), we sought to test whether MAPK activation is required for NT-3–induced increase in synaptic varicosity. Treatment with the MEK inhibitor U0126, which is known to block NT-3–induced MAPK activation (), completely blocked changes in synaptic sites elicited by NT-3 (). However, neither the PKA inhibitor Rp-cAMP nor the CaMK inhibitor KN93 could block the NT-3 effect (). These results, together with the finding that MAPK is not involved in the long-term changes in synaptic efficacy induced by NT-3 (), reveal a complementary role of MAPK and CREB; MAPK mediates the morphological, whereas CREB mediates physiological, changes induced by NT-3.
If CaMKIV is the upstream activator for CREB that mediates the functional changes of synapses upon NT-3 treatment, what is the upstream activator for MAPK that mediates the structural change of synapse induced by NT-3? In PC12 cells, Ras and Rap1, two related small GTP-binding proteins of the Ras subfamily, have been shown to mediate transient and sustained activation of MAPK, respectively (). To monitor the kinetics of Ras and Rap1 activation by NT-3 in living spinal neuron, we used fluorescent resonance energy transfer (FRET) techniques using Raichu-Ras and -Rap1 constructs (). Raichu-Ras contains a pair of mutant fluorescence proteins (YFP and CFP) in its C terminus. Thus, activation of Ras caused by intramolecular binding of the Ras domain to the RafRBD domain brings CFP close to YFP, leading to FRET from CFP to YFP (). Raichu-Rap1 works very similarly, except Ras is replaced with Rap1. We found that the activation of Ras differed from that of Rap1 in two ways. First, Ras activation, as reflected by the color change, was faster and more transient than Rap1 activation (, A [top] and B). Ras activation reached its maximum in <10 min and was diminished in 90 min (). This may reflect the localization of Ras at the plasma membrane, where NT-3/TrkC signals were rapidly transduced (). In contrast, Rap1 activation was much slower and more sustained. Second, Ras activation was observed not only in the cell body, but also in the axonal processes. In contrast, Rap1 activation was concentrated primarily at the neuronal cell body and never extended to the axons (, A [bottom] and B). Thus, although Ras and Rap1 could both be activated by NT-3, they showed different spatial and temporal activation patterns.
Next, we asked whether the inhibition of Ras or Rap1 could block the long-term structural and functional change induced by NT-3. To test this, we expressed Dn-Ras or -Rap1 in spinal neurons by embryo injection, and the nerve-muscle cocultures were treated with NT-3 for 2 d. Expression of Dn-Ras and -Rap1 did not affect basal synaptic transmission (compare the white bars in ). Moreover, long-term treatment with NT-3 induced a similar increase in SSC frequency in control neurons and neurons expressing either Dn-Ras or -Rap1 ( and Fig. S2 C). However, expression of Dn-Rap1, but not Dn-Ras, completely blocked the long-term effect of NT-3 on synaptic sites (). These data, together with the spatiotemporal pattern of Rap1 activation induced by NT-3, suggested that the Rap1–MAPK pathway was responsible for NT-3–induced long-term structural, but not functional, changes at the neuromuscular synapses.
It is well known that neuronal activity elicits long-term changes in the structure and function of synapses. Neurotrophins have recently emerged as a class of important regulators for synapse development and function. Neurotrophin-induced long-term synaptic changes resemble activity-dependent long-term synaptic plasticity in two fundamental ways: synaptic growth and dependence on protein synthesis (). A question of general interests is whether neuronal activity and neurotrophins use the same or different molecular mechanisms to modulate synapses. In this study, we show that NT-3 induced two parallel molecular pathways. One involves the CaMKIV–CREB pathway, which is responsible for enhancement of synaptic efficacy, but not synaptic growth. The other involves Rap1–MAPK, which leads to increase in the number of synaptic sites, but not in synaptic transmission. To our knowledge, no study so far has demonstrated that neuronal activity triggers two parallel, but distinct, signaling pathways for structural and functional plasticity in vertebrate synapses. Thus, although similar molecules are used to mediate synaptic modulation induced by either neuronal activity or neurotrophins, the specific mechanisms underlying the two types of synaptic modulation are not the same.
The results obtained from this study, together with our previous findings and those of other laboratories, support a model in which neurotrophins, through activation of Trk receptor tyrosine kinases, induce long-term structural and functional changes at synapses through two parallel signaling pathways (). The Ca–CaMKIV–CREB pathway ensures maturation of transmitter-release machinery for efficient functional transmission. The endocytosed Trk receptors and activation of Rap1–MAPK pathway promotes synaptic growth, leading to the development of properly matched pre- and postsynaptic structures. For proper development of neuromuscular synapse in vivo both pathways have to be activated in concert. Having one ligand to trigger both pathways synchronously is an ideal way to coordinate the structural and functional development of synapses. Similar mechanisms may also apply to synapse regeneration after injury. Local secretion of neurotrophins may ensure simultaneous morphological and physiological recoveries of the injured synapses. Thus, concomitant activation of these two pathways via pharmacological means may help treatment of nerve injuries or other neurodegenerative diseases.
dnCREB cDNA constructs (K- and A-CREB) were purchased from CLONTECH Laboratories, Inc. Dn-CaMKIV was a gift from M. Ehlers (Duke University Medical Center, Durham, NC). GFP-CaM expression vector was a gift from D. Chang (Hong Kong University of Science and Technology, Hong Kong, China). Dn-Ras and -Rap1a were gifts from A. Imamoto (University of Chicago, Chicago, IL). Raichu-Ras and -Rap1a were obtained from M. Matsuda (Osaka University, Osaka, Japan). cDNAs were digested and subcloned into the expression vector pcDNA3.1(+), which contains a T7 promoter for in vitro transcription of sense mRNAs. Capped mRNAs were generated by using the mMessage mMachine kit (Ambion). egg laying was induced by injecting female with human chronic gonadotropin (Sigma-Aldrich). Resulting eggs were fertilized artificially with sperms derived from male testis. mRNAs for NT-3, Dn forms of CREBs, Ras and Rap1a, or CaMKIV were mixed with 1 μg/μl GFP mRNA at a 1:1 ratio. Approximately 6–12 nl of the solution was injected into one blastomere at the 2- or 4-cell embryonic stage using the Picospitzer pressure ejector (Parker Hannifin). 1 d after injection, the neural tube and associated myotomal tissues were dissected and used to prepare nerve-muscle cultures. Neural tube and associated myotomal tissue of embryos at stage 20 were dissociated in Ca-Mg–free medium (58.2 mM NaCl, 0.7 mM KCl, and 0.3 mM EDTA, pH 7.4) for 15–20 min. Cells were plated on clean glass coverslips and grown in the presence or absence of NT-3 (2 nM; a gift from Regeneron Pharmaceuticals, Tarrytown, NY) for 2 d at room temperature, as previously described (). The culture medium consisted (vol/vol) of 50% L-15 medium, 1% fetal calf serum, and 49% Ringer's solution (117.6 mM NaCl, 2 mM CaCl, 2.5 mM KCl, and 10 mM Hepes, pH 7.6). Various inhibitors and NT-3 were added to the cultures 6 h after plating, when cells were completely settled.
Synaptic currents were recorded from innervated muscle cells in 2- or 3-d-old cultures by the whole-cell recording method in culture medium at room temperature (). The internal pipette solution contained 150 mM KCl, 1 mM NaCl, 1 mM MgCl, and 10 mM Hepes buffer, pH 7.2. The membrane potentials of the muscle cells recorded were generally in the range of –55 to –75 mV and were voltage clamped at –70 mV. All data were collected by a patch clamp amplifier (Axonpatch 200B; Axon Instruments), with a current signal filter set at 3 kHz. SSC frequency is defined as the number of SSC events per minute. The amplitudes of SSCs were analyzed using SCAN software (Dagan). To elicit ESCs, square current pulses (0.5–1 ms; 0.5–5 V) were applied to the soma of spinal neurons with a patch electrode filled with Ringer's solution at the neuronal soma under loose seal conditions (). Pipette and membrane capacitance and seal resistance were compensated. For acute effect, the data from a 10-min recording before application of NT-3 was taken as “control,” and those from a 10-min recording after the increase has reached peak as “NT-3 treated.” For long-term effect, a 10-min mean volume of SSC frequency/amplitude taken from any parts of the recording from a synapse gave rise to very similar numbers. For the convenience of comparison, we averaged data from control synapses without NT-3 treatment (control), and normalized data from all other conditions to the control. The numbers in the graph bars represent the number of synapses examined (, , , , , S2, and S3).
To visualize synaptic varicosity, we coinjected a small volume of a mixture, consisting of 2 μl of SYP-GFP plasmid DNA (1 μg/ml) and 2 μl of rhodamine-conjugated dextran dye (molecular weight, 10,000 D), by embryo injection. SYP-GFP construct was a gift from J. Sullivan (University of Washington, Seattle, WA). To label postsynaptic AChRs, 100 nM Cy5-labeled α-bungarotoxin (α-BTX) was applied to culture medium for 20 min. Cells were then rinsed extensively with Ringer's solution to remove unbound α-BTX. Confocal imaging was performed using an inverted LSM 510 META laser scanning microscope using 1.3 NA oil-immersion objectives (both Carl Zeiss MicroImaging, Inc.). For triple-color imaging, excitation lines of a 488-nm argon laser and two 543- and 640-nm helium lasers were used. Fluorescence was detected using a 488-/543-/633-nm dichroic beam splitter and a 530–560-nm bandpass filter for SYP-GFP, a 580–620-nm bandpass filter for rhodamine, and a 650-nm long-pass filter for Cy5. With the narrow band-pass filters, any crossover or bleed-through of fluorescence was eliminated. Postacquisition images were analyzed by IPLab software (Scanalytics). The criteria for synaptic puncta were that their fluorescence intensity must be >100% above the background and their size must be >0.5 μm in diameter. By using the region of interest (ROI) tool in the IPLab program, the number of synaptic sites, SYP-GFP puncta that juxtaposed with Cy5-α-BTX puncta within a myocyte, were measured. The calculated number of synaptic sites in a single myocyte per neuron were pooled and averaged. tests were used for statistical analysis. The numbers in the bar graphs represent the number of cells used for morphological analysis (, , , , and S1 B). For FM dye imaging, the fluorescent styryl membrane dye FM 4–64 (Invitrogen) was loaded into spinal neurons by applying high K loading solution (60 mM KCl, 57.6 mM NaCl, 3.5 mM CaCl, 10 mM Hepes, pH 7.6, and 2 μM FM 4–64) to cultures for 2 min. Cells were then rinsed extensively with Ringer's solution to remove membrane-bound FM dye. Fluorescent images of FM-dye were acquired by a cool charge-coupled device camera (MicroMax 1300; Roper Scientific) mounted on an inverted epifluorescence microscope (IX70; Olympus) and analyzed. Fluorescence images were taken with 500-ms exposure time with a ×40, 0.55 NA, objective.
embryos at stage 20 were quickly homogenized in extraction buffer (100 mM NaCl, 50 mM Tris-HCl, pH 7.5, 1% NP-40, 2 mM PMSF, 1 μg/ml aprotinin, 1 μg/ml leupeptin, 1 μg/ml pepstatin A, and 2 μM NaVO) and subsequently sonicated. High-speed centrifugation produced an insoluble pellet that was discarded. Supernatants were transferred to a fresh tube containing 300 μl Freon (1,1,2-trichlorotrifluoroethane; Sigma-Aldrich), vortexed for 1 min, incubated on ice for 5 min, and centrifuged again to remove yolk protein. The protein concentrations of the supernatants were determined using the BioRad protein assay (BioRad Laboratories). The proteins were separated by SDS-polyacrylamide electrophoresis, and blotted onto Immobilon-P membrane (Millipore). The blots were probed with primary antibodies, anti-CaM (1:100), and anti-GFP (1:10,000; Cell Signaling Technology), followed by a secondary antibody conjugated with HRP. Signals were detected by a chemiluminescence kit (Pierce Chemical Co.).
nerve-muscle cultures were treated with or without pharmacological drugs for 30 min at room temperature before NT-3 application. The cultured were fixed 30 min after NT-3 application by using 4% paraformaldehyde (EM Science) for 30 min at room temperature, and washed three times with 0.1 M NaBH (sodium borohydrate) in PBS to reduce autofluorescence. Fixed cells were permeabilized with 0.1% Triton X-100 in PBS for 5 min. Fixed cells were blocked with 5% nonfat milk and incubated with primary antibodies against phospho-CREB 1:200 (Cell Signaling Technology) overnight at 4°C. Secondary antibody for immunofluorescence was Alexa Fluor 546 (Invitrogen). To visualize DAB staining, an ABC kit (Vector Laboratories) was used. The images of immunocytochemistry were captured with an Olympus IX70 microscope using a charge-coupled device camera, acquired with IPLab software and processed with Photoshop (Adobe).
Confocal imaging was performed using an inverted LSM 510 laser scanning microscope and 40×, 1.3 NA, or 63×, 1.3 NA, oil-immersion objectives (all from Carl Zeiss MicroImaging, Inc.). For imaging coexpression of YFP and GFP constructs, excitation lines of a blue diode laser of 405 nm for both CFP and YFP were used. Fluorescence was detected using a 458-/ 514-nm dichroic beam splitter and a 470–525-nm bandpass filter for CFP and a 530–590-nm bandpass filter for YFP. In this way, any cross-talk and bleed-through of fluorescence were eliminated. Time-lapse scanning was performed with LSM 510 imaging system software. Post acquisition image processing was performed with the LSM 5 Image Browser (Carl Zeiss MicroImaging, Inc.) and Photoshop 7.0 software (Adobe). spinal neurons plated on a 35-mm diam coverslip were expressing pRaichu-derived plasmids by embryonic injection. A phase-contrast image and fluorescent images for ECFP and EYFP were recorded every 20 s. Starting from 5 min, cells were stimulated with 10 ng/ml NT-3. We demonstrated cell images in intensity-modulated display (IMD) mode by MetaMorph software (Molecular Devices), according to the manufacturer's instructions. In brief, background-subtracted images of EYFP and ECFP were first used for calculating of the EYFP/ECFP ratio of each pixel. After determination of the upper and lower thresholds, the ratio value of each pixel was associated with one of six hues from black (low) to white (high). The intensity of each pixel was determined by the brightness of ECFP. Thus, in IMD mode, the hue and its intensity at each pixel represent the FRET efficiency and the probe concentration, respectively. The FRET images were collected every 20 s.
Fig. S1 shows that NT-3 activates CREB. Fig. S2 shows the SSC amplitude data for K-CREB, Dn-CaMKIV, and Dn-Ras, Dn-Rap1 (A–C), and SSC frequency data for Rp-cAMP. Fig. S3 shows that inhibition of CaMKIV activation does not prevent an acute effect of NT-3 on synaptic transmission. Online supplemental material is available at . |
It's in my family. My father was an engineer, my mother was a nurse, which is pretty much as close as a Chinese woman could get to science during WWII, and my oldest brother, Dick, is an eminent neurobiologist at Stanford.
I have lots of engineers amongst my relatives, on both my mother's and father's side. My father's cousin was quite a famous engineer (H.S. Tsien). He was the head of the ballistic missile program for the People's Republic. It's actually a strange tale. He was at Cal Tech as an aeronautical engineer, but then got in trouble in the 1950s because he had some communist links in his past. He was put under house arrest and then deported. Back in China, he started up the ballistics program, I guess partly because he was mighty pissed off at the US!
Yes, and I remember my first interest in science was actually in chemistry. My parents had bought me a chemistry set, but I found it boring because the experiments were so safe. But then I found a book in the school library that told you how to make a bright purple solution turn bright green, just by passing it through a funnel of filter paper. Amazingly, it worked when I tried it, which I thought was pretty cool. So I decided chemistry was more fun than chemistry sets let on. That was in elementary school. By high school, I had started to accumulate stuff in the basement at home. One time, my brother and I surreptitiously made gunpowder.
We thought we could set off a controlled explosion, but it was just slightly out of control—we set fire to the ping-pong table! We managed to put it out, but there was a good deal of smoke, and my parents got a little bit alarmed.
After that incident, I tended to shift to the outdoor patio, which was made of concrete. That was a safer place. It was effectively like working in a fume hood.
I thought I was going to be a chemist. I was good at it in high school and was awarded the Westinghouse Science Prize for an NSF-funded chemistry project I did in the summer of my junior year, 1967.
But I went off to Harvard and found that I hated the way chemistry was taught there. I resolved that chemistry was really boring. I couldn't see the point of organic chemistry. I couldn't stand it anymore. I started drifting around, looking for things, and eventually the thing that most attracted me was neurobiology.
No. When it was time to look for a graduate school, I received a Marshall Scholarship for Cambridge. I heard from the Marshall Commission that I had been assigned to a Dr. R.H. Adrian as my Ph.D. supervisor, but I had never met him.
As it turned out, my older brother, Dick, who had just come back from Oxford, knew Richard Adrian. I asked him, “Who's this Richard Adrian?” He said, “Oh, well, he's a very eminent muscle electrophysiologist.” I said, “Muscle? You're kidding me; I want to work on the brain. Muscle is a backwater!”
Dick said, “Don't worry, Richard Adrian is a true British gentleman. He'll let you work on whatever you want.” So, I thought, “I'll see how it goes. I won't immediately write a letter of protest.” That was maybe January or February '72 before I had finished at Harvard.
In October, I went to Cambridge University. I was sitting in Churchill College Refectory, and this distinguished gentleman in a tweed jacket sat down across from me and asked me in a very upper-class British accent, “Are you Roger Tsien, by any chance?”
I guessed correctly that he must be Richard Adrian, who had come looking for me. Within about five minutes, he asked, “Is it true that you think muscle is a backwater?”
That's how I got started with my Ph.D. supervisor!
Neurobiology at the time was dominated by single-neuron recordings. People would drill a hole in the skull, drop an electrode blindly into the cortex, and start recording. It's like ice fishing, and after you've caught a couple hundred fish over several years, you write a Ph.D. describing in detail the classification of the fish that you pulled up from beneath the ice.
People would write for their theses, “I recorded 300 neurons from this region of the brain, and this was their firing pattern, and this is how they responded to this, that, or the other stimulus.” I didn't like that. I wanted some means by which one could see the activity of the brain in a more parallel fashion. The brain is made up of billions of neurons, and listening to them one at a time has limitations. It would be better, I thought, to see networks of neurons signaling to each other. That immediately implied imaging.
To begin with, I wanted to image membrane potential. This was difficult because any dye you make should be just in the plasma membrane and not in the internal membranes, and also because it wasn't clear how a dye would sense the voltage. Larry Cohen and his collaborators at Yale were randomly testing lots of photographic dyes for sensitivity. They had some success, but I attempted to be clever and specifically design just a few. I was not successful at the time. I struggled with that for years and eventually realized that, for the time being, it was just too hard, though two decades later we found a way.
Targeting calcium, however, was an enormously simpler solution to the problem of following neuronal activity. Action potentials almost always lead to major calcium influxes somewhere in the neuron. Chemically speaking, a lot was known about calcium, but it was ignored by biologists. Meanwhile, chemists were unaware of the biological importance of calcium signals. That gave me an angle, because there was almost no competition. My crummy chemistry on a chemically easier problem produced a unique solution.
Strangely enough, this is when I decided chemistry wasn't so bad. I realized that chemistry could actually be quite interesting if you had the right motivation. That's what was missing before. Chemistry for its own sake didn't interest me, but biology has interesting problems, for which chemistry can provide solutions.
Most biologists hate or fear chemistry. If a biological problem needed a chemical solution that you couldn't just buy, nobody expected you to do it. It was assumed that chemistry is impossible for outsiders. That view is gradually changing. But back when we started, it gave us a tremendous opportunity, because being able to do any chemistry was like being the one-eyed man in the kingdom of the blind.
It has been very useful. For a long time, it kept the competition out. We could be very simple minded as chemists, but since hardly any chemists knew of the problems, and few biologists could do any chemistry, we had the problem pretty much to ourselves!
I've always liked pretty colors. I tell students it's valuable to get some degree of gut pleasure out of what you're doing, to keep you going. Because, yes, when experiments work, there's nothing like that thrill, but that happens once in a while, and you can go for long, long periods where nothing works.
Enjoying pretty colors is one of my ways to keep going. It may be why I wound up doing this type of science. When all you are doing is looking at colorless solutions being pipetted from one tube to another, and visualizing the results on a gel by radioactivity much later, I find that very dry.
I much prefer experiments where there are pretty colors, and real activity, where the cells can talk back to you while they are alive. It makes it possible to design the experiments based on the conversation you're having with them.
We're imaging neurons to look at protein turnover at synapses, and we're also imaging tumors inside animals.
Yes. For the neurons, we use a special protease from hepatitis C virus. We fuse an epitope tag to our protease and then fuse the protease to our protein of interest. As the protein comes off the ribosome, the protease immediately cuts itself and the epitope tag off, leaving the host protein unmarked.
That happens throughout the life of the animal until we deliver a drug that specifically blocks this protease. From the time you administer the drug onward, all new copies of the protein remain intact and tagged.
Immunofluorescence against the tag will then light up synapses with new protein turnover. Being able to distinguish between newly made and old proteins should give us a more direct attack on the question of where brains store, for example, short-term and long-term memory.
We have capitalized on peptide sequences that are known to be taken up into cells, called cell-penetrating peptides (CPPs). If you tie the positively charged CPPs to negative charges, you get something that's fairly biologically inert and that won't get taken into cells unless the linker between the plus and minus ends is cut. Cancers express particular enzymes that they use to cut through the extracellular matrix to be invasive. When our inert fluorescently labeled CPPs reach the tumor, the tumor cells cut through the linker and thus unleash and take up the CPP. We're making the tumor cells take up fluorescence, but it could be used for other types of imaging such as magnetic resonance imaging—and perhaps eventually even for putting in drugs to kill the cancer.
I'm getting fussier and fussier about the importance of the problem I want to work on. Imaging cancer is pretty important. My father died from cancer, and so did Richard Adrian, my Ph.D. supervisor. It would be nice to contribute something in that area.
I probably only have time for one more phase in my career. The neuronal and the cancer projects are my attempt to do something different. We're reinventing ourselves, starting again in fields where we're not the experts anymore. Who knows how this phase is going to work out?
I have to say, that's one thing that getting some prizes for the GFP work was good for. I'd rather not just use the prizes to congratulate myself. It feels more like they've given me a license to go and try something else, like, “You've been there, you've done that, now do something better.” |
Asymmetric cell division plays a crucial role in generating cell diversity. Proper positioning of the mitotic spindle is an essential step of an asymmetric division. In the early embryo of the nematode , asymmetric positioning of the mitotic spindle depends on an imbalance in cortical force generators that act on astral microtubules and pull on spindle poles (). Although the molecular nature of the force generators is not known, their spatial and temporal activation is controlled by heterotrimeric G protein signaling. Inactivation of two Gα subunits, GOA-1 and GPA-16, as well as the receptor-independent activators of G protein signaling GPR-1 and GPR-2 (two nearly identical proteins containing a Goloco domain, hereafter referred to as GPR-1/2) results in strongly reduced and symmetric pulling forces (). The coiled-coil protein LIN-5 also plays a crucial role in spindle positioning. LIN-5 interacts with GPR-1/2, and its inactivation results in a phenotype very similar to the phenotype of embryos lacking both Gα subunits or GPR-1/2 (; ). The role of heterotrimeric G proteins in spindle positioning is conserved in other organisms, including flies and mammals (; ). In neuroblasts, Gα and PINS, the functional homologue of GPR-1/2, are required for apical basal orientation of the mitotic spindle. In mammalian cells, Gα, the Goloco-containing protein LGN, and the microtubule-binding protein nuclear mitotic apparatus (NuMA) form a complex that has been suggested to regulate the interaction of astral microtubules with the cell cortex (). Interestingly, NuMA can bind to the dynein–dynactin complex (), and recent work has shown that Mud and LIN-5 are the homologues of NuMA (; ; ). These results suggest a model in which the interaction of cortically localized NuMA/LIN-5/Mud with dynein results in the activation of this minus end–directed motor, locally increasing pulling forces and, thereby, resulting in posterior displacement of the mitotic spindle. To date, there is no evi dence for a dynein requirement in spindle positioning in . Recent work in using temperature-sensitive mutants has shown that partial dynein inactivation does not abolish spindle displacement to the posterior (), but other studies have suggested a role for dynein in force generation during spindle positioning (; ).
In this study, we show that , which encodes a dynein light chain, plays a role in the regulation of spindle positioning. DYRB-1 is the only member of the highly conserved roadblock/LC7 family (). Mutations in the gene result in defects in mitosis, the accumulation of vesicles in axons, and larval or pupal lethality (). Studies in have suggested that LC7 light chains are required for both motor assembly and regulation (). However, the exact function of roadblock/LC7 is unknown. We demonstrate that the depletion of enhances the phenotype of and temperature-sensitive mutants. Furthermore, we show that the forces pulling on the astral microtubules are reduced after the inactivation of . Interestingly, we find that DYRB-1 coimmunoprecipitates with GPR-1 and LIN-5. Based on these results, we propose a model in which GPR-1/2 and LIN-5 function with dynein to control spindle positioning.
To identify additional genes that act with heterotrimeric G proteins in spindle positioning, we performed RNAi enhancer screens using and temperature-sensitive alleles (see Materials and methods and enhancer screen section; unpublished data). In qualitative assays, we identified T24H10.6 as a candidate gene whose disruption by feeding bacteria that express double-stranded RNA (dsRNA) had no effect in wild-type animals but resulted in embryos that failed to hatch in and mutant backgrounds. This gene, which is named (dynein light chain roadblock type-1), encodes a homologue of the roadbock/LC7 dynein light chain family (). The synthetic lethality was confirmed by feeding assays on solid media. In such assays, we found that the disruption of in wild-type animals resulted in >99% embryonic viability, whereas viability was decreased considerably in both and mutants (). Injection of dsRNA resulted in ∼50% embryonic viability in wild-type animals, whereas viability was substantially decreased in and mutants (). These results indicate that genetically interacts with components of the heterotrimeric G protein pathway.
Because the injection of dsRNA in wild-type animals results in reduced embryonic viability, we investigated by time-lapse differential interference contrast (DIC) microscopy whether embryos have early defects. The progression of events in wild-type embryos is illustrated in . In wild type, the oocyte pronucleus migrates to the posterior to meet the sperm pronucleus, which also moves slightly to the center of the embryo. As a result, the two pronuclei meet at ∼68% of embryo length (0% anterior-most and 100% posterior-most; ). The two pronuclei and associated centrosomes then migrate toward the cell center while undergoing a 90° rotation () that aligns the centrosomes along the anterior-posterior axis of the embryo. The spindle sets up in the center of the cell along this axis and is displaced toward the posterior by an imbalance of pulling forces at metaphase/anaphase. During this displacement, the posterior spindle pole undergoes transverse oscillations called rocking (). At telophase, the posterior spindle pole flattens, whereas the anterior remains round (). This asymmetric displacement of the spindle results in an asymmetry in cell size after cytokinesis, with a larger anterior cell (AB) and a smaller posterior cell (P1; ).
We found that many of these processes are affected in embryos. embryos display delayed migration of the oocyte pronucleus (Table S1, available at ), failure in pronuclear centration (), and delay in rotation of the pronucleus–centrosome complex (), although in all embryos, the mitotic spindle is aligned along the anterior-posterior axis at cytokinesis (). In addition, the first mitotic spindle is significantly shorter than in wild type (), spindle rocking is absent in all embryos, and the posterior aster does not flatten (). These phenotypes are highly penetrant (as observed in nearly 100% of the embryo) and, therefore, cannot solely be the cause of embryonic lethality because 50% of the embryos are viable in these conditions.
The phenotypes of –depleted embryos are similar to the phenotypes observed upon weak RNAi depletion of , which encodes a dynein heavy chain (), or in animals bearing temperature-sensitive alleles of (). Strong RNAi depletion of results in more severe phenotypes, including defects in centrosome separation, bipolar spindle formation, and cytokinesis (). We did not observe such phenotypes upon the RNAi depletion of . Although transgenically expressed GFP–DYRB-1 is strongly depleted under our RNAi conditions (), we cannot exclude that there is remaining endogenous DYRB-1 in the depleted animals, and attempts to obtain anti–DYRB-1 antibodies were unsuccessful. However, embryos from homozygote mutant mothers, which are predicted to produce a truncated protein that only contains the first 29 amino acids of DYRB-1, show phenotypes identical to ) embryos (Fig. S1, available at ). Because there are at least six other dynein light chains predicted in the genome, it is possible that one or more of these partially compensate for the loss of in the early embryo. Collectively, these results suggest that DYRB-1 is required for several dynein-dependent processes in the one-cell embryo, including proper spindle orientation, and are consistent with DYRB-1 regulating dynein activity.
To further investigate the genetic interaction between and genes in the heterotrimeric G protein pathway, we analyzed the phenotype resulting from depletion in or mutants. Codisruption of and or at permissive temperature for both mutants resulted in a complete loss of embryonic viability () and a strong synthetic phenotype in early events. The male pronuclear envelope broke down before the pronuclei met in 9/17 cases for and in 9/16 cases for embryos. During prophase in wild type, , , and one-cell embryos, the two centrosomes localized to opposite sides of the male pronucleus. In 7/17 embryos codisrupted for and in 6/10 embryos codisrupted for , this separation failed, and centrosomes remained in close proximity to each other (Videos 1–5, available at ). This phenotype is similar to embryos strongly depleted of DHC-1 (). Furthermore, although in , , and embryos, the mitotic spindle was oriented along the anterior-posterior axis at the onset of cytokinesis, rotation was further delayed in the double mutants, and the spindle was misaligned in 13/17 embryos and 3/16 embryos (). This phenotype is independent from the centrosome separation defect because we observed misaligned spindles in embryos in which the centrosomes had correctly separated.
These results indicate that and double mutant embryos have a stronger phenotype than any single mutant, which is consistent with GPR-1, LIN-5, and DYRB-1 regulating common processes in the one-cell embryo. Interestingly, the pronuclear migration defect was enhanced in both double mutant combinations, suggesting that LIN-5 and GPR-1/2 play a role in this process. Therefore, we quantified the position of pronuclear meeting in and embryos and found that they meet more anterior when compared with wild-type embryos (). This indicates a novel role for the heterotrimeric G protein pathway in the regulation of pronuclear migration.
Altogether, the phenotypes of embryos suggest that DYRB-1 may regulate dynein activity. To address this, we generated transgenic animals expressing DYRB-1 fused to GFP in the early embryo. We found that DYRB-1 localizes in a punctate manner in the cytoplasm and is found at the periphery of pronuclei and nuclei (). During anaphase, GFP–DYRB-1 localizes to the mitotic spindle, the two centrosomes, and around the chromosomes (). We also observed a weak localization to the cortex that is more apparent in two- and four-cell embryos as well as in older embryos (). The depletion of GPR-1/2 or LIN-5 did not affect this localization pattern at any stage (unpublished data). Therefore, DYRB-1 localization appears similar to that of DHC-1 (), which is consistent with a role in regulating dynein activity in the early embryo. To further test this possibility, we depleted in mutants, which exhibit hyperactive nuclear movements during centration (Fig. S2, available at ). This phenotype results from an excess of Gα signaling and dynein activity because it is suppressed in α embryos and in embryos in which has been weakly depleted by RNAi (, ). We found that the depletion of also suppresses the nuclear hyperactive movement of embryos (Fig. S2, M–P; and Video 6). These results indicate that DYRB-1 plays a positive role in the regulation of dynein activity.
Recent work in has shown that dynein activity may contribute to posterior spindle displacement, although its function is not strictly essential for this process (; ; ). Several observations indicate that the pulling forces that regulate spindle positioning are weaker in embryos: the spindle is significantly shorter than in wild type, spindle rocking does not occur, and the posterior aster does not flatten (). Therefore, we tested whether regulates spindle positioning by investigating whether pulling forces are compromised in embryos when the mitotic spindle is severed by a laser microbeam (). Because the mitotic spindle forms at the posterior of embryos, comparisons were made with wild-type embryos in which the spindle was also severed when it was more posterior at the onset of anaphase B (see Materials and methods RNAi, microscopy, and spindle severing section). In wild-type embryos, forces on each side of the spindle are asymmetric, and the mean peak velocity of the posterior centrosome is ∼1.6 times higher than the one of the anterior centrosome (; ). The mean peak velocity of both anterior and posterior centrosomes in embryos after spindle severing is significantly reduced compared with wild type (P < 0.001; test), and the asymmetry is lost ( and Video 7, available at ). The mean peak velocity of both centrosomes in embryos was faster than that in embryos, indicating that RNAi disruption of does not completely inactivate the force generators, as is the case for . These reduced forces are not a consequence of polarity defects because PAR proteins are properly localized in embryos (Fig. S3). Likewise, the depletion of did not affect the localization pattern of GPR-1/2 or LIN-5 (Fig. S3 and not depicted). These results indicate that plays an important role in regulating the asymmetry in forces that pull on astral microtubules and that it functions downstream or in parallel to polarity cues and LIN-5/GPR-1/2.
Our results suggest that DYRB-1 could function together with the heterotrimeric G protein pathway to regulate microtubule-dependent events. In mammalian cells, NuMA was shown to physically interact with both dynein and LGN (; ). This suggested the possibility that DYRB-1 could molecularly interact with components of the heterotrimeric G protein pathway. Thus, we investigated whether DYRB-1 can be recovered in a complex with LIN-5 and/or GPR-1/2 using embryonic extracts made from the transgenic strain expressing GFP–DYRB-1. We found that both LIN-5 and GPR-1/2 could be coimmunoprecipitated with anti-GFP antibodies (). Conversely, both GPR-1/2 and GFP–DYRB-1 could be coimmunoprecipitated with anti–LIN-5 antibodies. These interactions are specific because they are not detected in control immunoprecipitations in which anti-GFP antibodies were incubated with extracts made from -depleted animals. GFP–DYRB-1 could still be coimmunoprecipitated with LIN-5 and GPR-1/2 in the presence of 50 μM of the microtubule-depolymerizing drug nocodazole, indicating that this interaction is microtubule independent. Unfortunately, the depletion of or in the GFP–DYRB-1–expressing strain resulted in complete sterility, thus precluding us from assessing whether this interaction depends on LIN-5 or GPR-1/2 (see Materials and methods Preparation of extracts and Western blot analyses section). These results suggest that LIN-5, GPR-1/2, and DYRB-1 are parts of a common protein complex, which is consistent with these three proteins functioning in the same pathway.
In conclusion, we have shown that the embryonic loss of , which encodes a dynein light chain subunit of the roadblock family, phenocopies a weak loss of dynein activity. Furthermore, we have demonstrated that DYRB-1 genetically and physically interacts with LIN-5 and GPR-1/2, which are two positive regulators of the heterotrimeric G protein pathway. This suggests that DYRB-1 or another component of the dynein complex is an effector of the heterotrimeric G protein pathway. Interestingly, spindle-positioning forces are reduced, and the asymmetry in pulling forces between anterior and posterior poles is lost in –depleted embryos. Collectively, these results suggest a model in which heterotrimeric G protein signaling controls spindle positioning, at least in part, by regulating dynein activity. One possibility is that LIN-5 and GPR-1/2 could asymmetrically activate cortically anchored dynein, which would then promote pulling of the mitotic spindle toward the posterior pole of the embryo. Recent results by showed that a weak depletion of dynein results in a loss of pole oscillations during spindle positioning, which is consistent with our observation that such oscillations are lost in embryos. However, their results and results from predicted that reducing dynein activity leads to an overall reduction in pulling forces rather than a loss in asymmetry, which is inconsistent with our observation that forces are weaker and symmetric in embryos. One possibility to reconcile these results is to suggest that DYRB-1 regulates all asymmetries in pulling forces and that the remaining forces in embryos are not DYRB-1 dependent. These remaining forces could depend on a variety of other potential regulators of force generators (e.g., another dynein light chain, the dynactin complex, or microtubule–cortex interactions).
Spindle positioning was recently proposed to be mainly controlled by the regulation of microtubule dynamics () based on computer simulations. It is unclear at present how dynein regulates spindle positioning, but previous studies in other cell types have shown that its motor activity can regulate microtubule dynamics (; ). Therefore, spindle positioning in embryos could occur through a regulation of dynein motor activity by DYRB-1 and the heterotrimeric G protein signaling pathway, which could, in turn, modulate microtubule dynamics. These two events need not be mutually exclusive.
Bristol strain N2 was used as the standard wild-type strain. Nematode culturing was performed as described previously (). The alleles used in this study are , (), , and . Temperature-sensitive mutant strains were grown at a permissive temperature of 15°C and shifted to 22°C 24 h before the embryos were examined.
was isolated in a screen for temperature-sensitive embryonic lethal mutations () and was outcrossed six times with N2 males. The mutation mapped between and , close to and its paralogous locus . These two closely linked paralogues are 97% identical in coding DNA and cannot be distinguished by expression analysis or RNAi. Because embryos from homozygotes show a mild form of the cytological defects resulting from RNAi inactivation of , we sequenced both loci in the mutant. Genomic DNA fragments encompassing the F22B7.13 and C38C10.4 loci were amplified from the mutant by PCR using primer sequences unique to each locus. Pooled PCR products were sequenced using a sequencer (CEQ 800; Beckman Coulter) at the University of Oregon Sequencing Facility. was found to carry a missense mutation in that was predicted to convert glycine 289 to arginine. This mutation was not present in the background strain used for mutagenesis (CB1309), and no other mutations were found in within 0.5-kb or coding regions. Eggs laid by ts homozygotes were >99% viable at 23°C, 22% viable at 25°C, and <1% viable at 26°C. Eggs laid by heterozygotes were >99% viable at 26°C, indicating that is fully recessive. However, trans-heterozygotes for and nDf20, a deficiency predicted to uncover but not , produced 33% viable eggs at 26°C, suggesting that the mutant GPR-1 protein acts in a dominant-negative fashion for both GPR-1 and -2.
For the GFP–DYRB-1 strain, full-length DYRB-1 cDNA was cloned into pID3.01 () using Gateway technology (Invitrogen). GFP lines were created using the microparticle bombardment technique as described previously (). The transgene analyzed fully rescues the embryonic lethality of the allele, indicating that the GFP fusion is functional.
The screen was performed as described previously () with some modifications. In brief, and mutant animals were grown in large quantities on solid media at 15°C and were bleached to collect embryos. These embryos were incubated at 15°C with rocking in M9 buffer to allow hatching of L1 larvae. Assays were performed in plates containing 96 wells. RNAi clones from the available collection () were seeded in individual wells and grown overnight at 37°C in Luria Broth medium containing 100 μg/ml carbenicillin. Each well of fresh 96-well plates was then filled with 75 μl 3x nematode growth medium (NGM; regular NGM with a triple amount of peptone), and 2 μl of overnight bacterial culture was added. The plates were incubated at 37°C for 2.5 h without shaking. 25 μl 3xNGM containing 24 mM IPTG was then added to each well (6 mM IPTG final), and the plates were incubated at 37°C for 5 h without shaking. 5–10 L1 worms of each genotype were then added to each well (15 μl from a suspension containing approximately eight worms per 15 μl M9 buffer). The plates were incubated for 6–7 d at the semirestrictive temperature of 17°C without agitation until food was depleted and F1 progeny had hatched. Enhancement of lethality was estimated by visual inspection under a dissecting scope, and the relative decrease of swimming L1 larvae compared with the control was scored as positive. Liquid handling was performed using a robotic system (Biomek FX; Beckman Coulter) equipped with a 96-channel pipetting head.
dsRNA was produced as described previously (). For live imaging of embryos, dsRNA was injected into young adult hermaphrodites. Animals were dissected, and embryos were analyzed 24 h after injection.
For DIC analysis of living embryos, embryos were mounted as described previously (). The first cell cycle of embryos was visualized with a camera (Orca ER; Hamamatsu) mounted on an inverted microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.), and the acquisition system was controlled by Openlab software (Improvision). Images were captured at 5-s intervals using a plan Apochromat 63× 1.4 NA objective. For immunofluorescence experiments, a microscope system (DeltaVision 3000; Olympus) was used for capturing and deconvolving images of embryos.
Spindle-severing experiments were performed similar to using an inverted microscope (Axiovert 200M; Carl Zeiss MicroImaging, Inc.) equipped with a PALM laser system (Mikrolaser Technologie). The pulsed laser was focused to an ∼1-μm-thick spot in the focal plane. embryos were mounted on the inverted microscope. Because the spindle does not go to the center of the cell in embryos, the spindle was cut along the midzone when at a posterior position in both wild-type and animals, which corresponds to the onset of anaphase B. The wild-type centrosome speeds that were measured in these experiments are comparable with those measured previously (), indicating that severing the spindle when it is at a posterior position does not preclude the accurate measurement of centrosome velocity. For monitoring spindle severing, one DIC image was captured every second. Measurements of peak velocities were performed by manual tracking with ImageJ (tracking the center of the aster; National Institutes of Health).
Indirect immunofluorescence of embryos was performed as described previously (). We used rabbit anti–GPR-1/2 (1:80; ), mouse anti–α-tubulin (1:1,000; DM1A; Sigma-Aldrich), and rabbit anti-GFP (1:150; Abcam) as primary antibodies. Secondary antibodies were anti–rabbit AlexaFluor488, anti–rabbit AlexaFluor568, and anti–mouse AlexaFluor568 (1:500 each). DNA was visualized with DAPI.
To prepare embryonic extracts, wild-type embryos were obtained by hypochlorite treatment, and newly hatched synchronized L1 larvae were grown on bacteria expressing dsRNA until adulthood. Adult worms were collected and treated with hypochlorite to recover embryos. These embryos were resuspended in 1 vol of 4× Laemmli buffer and boiled for 10 min at 95°C before Western blot analysis. For immunoprecipitation experiments, the embryos were washed in immunoprecipitation lysis buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM MgCl, 1 mM EGTA, 1 mM DTT, 1% Triton X-100, and protease inhibitor cocktail; ), resuspended in an equal volume of buffer, and frozen at −80°C. They were then broken with glass beads (Lysing Matrix C; MP Biomedicals) in a bead beater (3 × 30 s with a 1-min interval at 4°C). The lysate was spun down at 14,000 rpm for 15 min at 4°C. The extract was incubated with 3 mg anti-GFP antibodies (Roche) or 2 mg anti–LIN-5 antibodies (). Antibody and protein A–Sepharose bead binding was performed as described previously (). For SDS-PAGE and Western blotting, standard procedures were used. The GPR-1/2 antibody used for blotting was described previously (). Although both or could be depleted by dsRNA injection in adult hermaphrodites expressing GFP–DYRB-1, the depletion of either or by feeding dsRNA to GFP–DYRB-1 L1 or L3/L4 animals (in liquid or on solid media) resulted in strong sterility, thereby preventing us from preparing embryonic extracts for immunoprecipitation experiments. We note that the amount of GPR-1/2 that was coimmunoprecipitated with anti–LIN-5 antibodies in appears proportionally high compared with the loading control. This could indicate that the anti–LIN-5 antibodies have a higher affinity for the pool of LIN-5 protein that is in a complex with GPR-1/2.
Fig. S1 shows the early development of ) mutant embryos. Fig. S2 shows that the hyperactive nuclear movement of mutant embryos is suppressed in embryos. Fig. S3 shows that DYRB-1–depleted embryos have normal polarity. Videos 1 and 2 show wild-type (Video 1) and (Video 2) embryos expressing GFP-tubulin and imaged at 22°C. Videos 3 and 4 show (Video 3) and ; (Video 4) embryos expressing GFP-tubulin and imaged at 22°C. Video 5 shows embryos expressing GFP-tubulin and imaged at 22°C. Video 6 shows a combined video with , , and embryos. Video 7 shows a combined video with spindle-cutting experiments in a wild-type embryo, a embryo, and a embryo. Table S1 shows that pronuclear migration is delayed in –depleted embryos. Online supplemental material is available at . |
Parenchymal cells require integrin-dependent attachment to solid structures to survive. As a consequence of such anchorage dependence, most tissue cells start to undergo apoptotic death within hours of being forced into suspension in a fluid environment, a process termed anoikis (). Physiological anoikis occurs during the involution of mammary and prostate glands, whereas pathological loss of anoikis is thought to accompany the malignant cell's acquisition of metastatic capabilities.
Presently, anoikis is understood largely in terms of withdrawal of integrin-related outside-in survival signals. Thus, constitutive activation of the survival proteins focal adhesion kinase (FAK), Src, Akt, and Ras all bypass detachment-induced death (; ; ). However, anoikis can be circumvented in the absence of Ras, phosphoinositide 3-kinase, or FAK activation (; ; ), suggesting that the activation of specific death pathways may be necessary in addition to the disabling of survival signals. Furthermore, integrin ligation without structural matrix rigidity is insufficient to prevent anoikis, indicating that integrins also control anoikis by sensing mechanical properties of the environment. Soluble matrix fragments or solid-phase RGD peptides bound to microbeads, e.g., ligate and cluster integrins, but neither prevent rounding nor increase the survival rate of floating endothelial cells (), which is consistent with observations that cell rounding by itself forces death upon attachment-dependent cells ().
The shortcomings of a unidirectional outside-in signaling model to explain attachment sensation is consistent with the concept that cells initiate processes designed to gauge substrate stiffness in an “inside-outside-in” feedback loop (). The molecular machinery responsible for such mechanosensation and its relationship to anchorage dependence have remained recondite; hematopoietic cells, for instance, lack such machinery, as they are both anchorage independent and insensitive to substrate stiffness (). Interestingly, hematopoietic cells also repress expression of p66, the long isoform of the integrin-associated adaptor Shc (). Whereas p52 is known to facilitate survival and proliferative signals, in large part through Ras recruitment and activation, p66 is best known as a proapoptotic protein. Here, we show that p66 permits activation of RhoA, which leads to tension-dependent death in floating cells. p66 may thus be important in the efferent limb of a mechanosensory loop reporting detachment.
p66 is poorly expressed in floating hematopoietic cell lines (), suggesting that either p66 may mediate adhesion or that its repression may be important for survival in suspension. We studied φNx-293 cells, as this epithelioid line forms adherent cultures yet is relatively resistant to death while floating. Indeed, these cells did not express p66 even when grown as adherent monolayers, and suspension over low attachment plates for 16–24 h did not increase cell death (). In preliminary studies, transient reexpression of p66 conferred sensitivity to anoikis in 293 cells. In contrast, overexpression of p52 had a minimal effect on cell death in attached or floating cells (). To avoid artifacts related to cell death from transfection itself, φNx-293 cells stably expressing Flag-p66 (293-p66 cells) were studied. In agreement with transient transfection studies, stable expression of p66 had no effect on cell death in adherent cells but caused death after detachment (). Cell death was accompanied by the activation of caspase 7, an executioner caspase activated during anoikis (; ), and the release of cytochrome (Fig. S1, A and B, available at ). More specifically, selective induction and mitochondrial localization of Bim has been shown to comprise a characteristic hallmark of anoikis (). Accordingly, detachment of p66-expressing cells caused a marked induction of mitochondria-associated Bim, without induction of Bax or Bid or suppression of Bcl-X ().
As the phosphorylation of Ser36 has been shown to be necessary to cause apoptosis in response to oxidative stress (; ), we introduced either S36A or S36E mutations into p66 and again created stable 293 lines. Surprisingly, the S36A mutant was able to confer anoikis, albeit not to the same magnitude as wild-type (wt) p66, whereas the phosphomimetic S36E mutation abrogated the anoikis-conferring property of p66 (), consistent instead with a possible protective effect of S36 phosphorylation against anoikis. Indeed, phosphorylation of Ser36 was present in adherent 293-p66 cells that did not undergo cell death and did not occur in 293-p66 (S36A) cells undergoing anoikis (). Thus, Ser36 phosphorylation was neither necessary nor sufficient for anoikis, suggesting a mechanism of cell death through p66 different from that previously described for other stimuli (; ). As further confirmation of the effects of p66 and Ser36 on anoikis, we found that whereas vector-transfected 293 control cells were able to form colonies in soft agar, 293-p66 and 293-p66 (S36A) cells formed only abortive colonies (Fig. S1 C). In contrast, 293-p66 (S36E) cells, which escape anoikis, displayed prominent anchorage-independent growth.
Because p66 contains all the domains of the nonapoptotic p52, it seemed unlikely that the former protein functioned simply as a dominant-negative version of p52, unless the unique collagen homologous (CH) 2 N-terminal domain of p66 masked other Shc domains through intra- or transmolecular interactions. We found, however, that neither Flag-p66 nor the isolated Flag-CH2 domain coprecipitated with endogenous p66, p52, or the smaller p46 (unpublished data). Indeed, transient expression of the CH2 domain decreased rather than increased anoikis (), which is consistent with a unique role for p66 in causing anoikis.
We also studied nonimmortalized primary cells, human umbilical vein endothelial cells (HUVEC), based on their epithelioid behavior and marked sensitivity to anoikis. As expected, HUVEC expressed substantial levels of p66 and prominent detachment-induced cell death () with the induction of Bim (Fig. S1 D). Knockdown of p66 significantly decreased but did not eliminate anoikis. Although this incomplete protection may be caused by the partial effect of siRNA, death pathways independent of p66 may exist. As with 293 cells, phosphorylation of Ser36 did not increase as HUVEC underwent anoikis (Fig. S1 E). Again supporting a specific role for p66, transient expression of the isolated CH2 domain decreased anoikis by ∼33%, consistent with the transfection efficiency of HUVEC ().
Finally, sensitivity to anoikis is known to vary with the level of confluence in monolayers of nontransformed lines such as MDCK epithelial cells (). To investigate whether the endogenous regulation of p66 might account for anoikis susceptibility within a given cell line, we correlated levels of p66 with anoikis at sparse and confluent conditions. p66 progressively increased with cell density, paralleling the increase in anoikis seen in confluent cells (). In contrast, the increase in p66 in confluent cells did not cause cell death in attached MDCK cells. Notably, 3T3 cells, nontransformed fibroblastic cells known to undergo contact inhibition, also displayed increased p66 protein expression and sensitivity to anoikis upon reaching confluence (). Thus, in endothelial, epithelial, and mesenchymal cells, both externally manipulated and endogenously regulated levels of p66 determine sensitivity to anoikis.
Although the Shc proteins are known to associate with integrins, specific integrin structures and binding mechanisms are not well understood. Focal adhesions are thought to function as predominant anchoring structures, so we investigated the targeting of GFP-fused p66 to native HUVEC focal adhesions. Using total internal reflection fluorescence (TIRF) microscopy to image the ventral cell surface, we found that p66-GFP targeted well-formed zyxin-positive focal adhesions in HUVEC (). In parallel, transient expression of p66-GFP in 293 cells, which lack endogenous p66, conferred anoikis (). Focal adhesion targeting was preserved in p52 and did not occur with the isolated CH2 domain (). Association of Shc proteins with β1 integrins is known to occur indirectly through caveolin and Fyn through the Fyn Src homology (SH) 3 domain, presumably through proline-rich motifs contained within the p66 CH1 domain (). Disruption of either of two polyproline motifs within the CH1 domain (P411S/P414S or P472S/P475S), however, had no effect on focal adhesion targeting (Fig. S2, A and B, available at ) and did not diminish p66-related anoikis (). Furthermore, deletion of the entire CH1 domain neither altered focal adhesion targeting nor diminished anoikis ( and S2 C). This latter deletion also argues against the possibility that p66 causes death by sequestering Grb2, which binds phosphotyrosine residues within the CH1 domain (). Similarly, deletion of the C-terminal SH2 domain, known to bind growth factor receptors, N-cadherin, and the β4 integrin subunit (; ), did not abrogate association with focal adhesions and did not decrease anoikis ( and S2 D).
Deletion of the phosphotyrosine-binding (PTB) domain, however, caused complete loss of both focal adhesion targeting and anoikis activity (). Structural studies of this domain have revealed independent binding surfaces for phosphotyrosine-containing peptides and acidic phospholipids. We found that p66 harboring a single R285Q mutation, which selectively disrupts the phosphotyrosine-anchoring pocket (), also delocalized p66 from focal adhesions and completely lacked anoikis activity (). However, replacing the p66 PTB domain with the focal adhesion targeting (FAT) domain of FAK, while restoring focal adhesion localization, did not restitute anoikis ( and S2 E). Thus the specific molecular partner of p66 within focal adhesions may determine its ability to relay detachment signals. The lack of effect of the PTB/FAT domain swap also suggests that p66 does not simply compete with FAK for focal contact binding sites to diminish its survival. Indeed, p66 had little effect on FAK (Y397) phosphorylation in 293 cells (Fig. S2 G). FAK (Y397) phosphorylation was low in 293-vector and 293-p66 cells in both adherent and suspended states, suggesting that p66 does not induce cell death by decreasing FAK autophosphorylation during detachment.
Finally, p66 has recently been shown to translocate into the mitochondrial inner membrane, bind to cytochrome directly, and undergo a redox cycle, causing oxidant-dependent permeability transition and apoptosis in adherent cells undergoing apoptosis (). Disruption of cytochrome binding through mutation of Glu132/Glu133 completely disables the apoptotic activity of p66 after oxidative stress or staurosporine (; ). We found, however, that p66 (E132Q and E133Q) remained localized to focal adhesions and retained its full ability to cause anoikis ( and S2 F). Together with the persistence of anoikis in the S36A mutant, these data suggest a unique mechanism for p66 in mediating anoikis as opposed to other forms of apoptosis.
The targeting of p66 and p52 to focal adhesions is consistent with observations that cells lacking functional Shc proteins display adhesion defects and alterations in focal contact and actin fiber organization (), and that integrins that nucleate focal adhesions also permit Shc-dependent proliferative signals (). We questioned whether p66 specifically influenced activation of RhoA, known to drive focal adhesion formation (). Using the rhotekin rho-binding domain (RBD) fused to GFP (RBD-GFP) to report sites of RhoA activation, we again examined the ventral attachment surface of cells using TIRF microscopy. In control 293-vector cells, focal contacts identified by DsRed-zyxin were small and peripheral, often taking the appearance of focal complexes, and cells were rounded with little evidence of RBD-GFP accumulation and only occasional actin stress fibers (). In contrast, 293-p66 cells were more spread with typical focal adhesions and more abundant stress fibers present. Numerous accumulations of RBD-GFP indicated focal activation of RhoA at ventral sites corresponding to zyxin- containing contacts (). The influence of p66 on focal contact organization was also evident in HUVEC, which normally appear polarized with abundant focal adhesions and stress fibers. Knockdown of endogenous p66 caused a general loss of cell polarity with a substantial decrease in stress fibers and focal adhesions ().
Upon detachment, 293-vector cells displayed no activation of RhoA, whereas detachment of 293-p66 cells caused robust RhoA activation within 30 min (). 293-p66 (S36E) cells, being defective in anoikis, also failed to activate RhoA upon detachment. Dominant-negative RhoA (N19) or knockdown of endogenous RhoA both decreased p66-dependent anoikis, confirming a role for RhoA in detachment-induced death (). Conversely, active RhoA (V14) caused cell rounding (not depicted) and death in 293 cells lacking p66 even while attached, and further augmented cell death while floating (), suggesting a default decision of death by active RhoA in the absence of p66. In support of this, RhoA (V14) did not increase cell death in attached 293-p66 cells, whereas it increased cell death in floating 293-p66 cells (). Thus p66 not only initiates RhoA activation after detachment but appears to restrain RhoA-induced death in attached cells, in essence reporting appropriate anchorage through RhoA context. Again examining HUVEC as primary culture cells sensitive to anoikis, we found similar robust RhoA activation within 30 min of detachment (). Antagonizing RhoA through either expression of RhoA (N19) or knockdown of endogenous RhoA suppressed anoikis, confirming a role for RhoA in this process in primary cells ().
Because RhoA is known to increase tension across attachment sites with resultant focal adhesion and stress-fiber formation, we surmised that in floating cells such tension would be applied against unanchored points, allowing a mechanical readout for detachment. Accordingly, antagonists of actinomyosin contraction (myosin light chain kinase inhibitor ML-7; actin–myosin interaction inhibitor 2,3 butanedione monoxime [BDM]; and myosin II inhibitor (−)blebbistatin) decreased p66-dependent anoikis (), indicating a role for the generation of cytoskeletal tension in detachment sensing. Surprisingly, the Rho kinase inhibitor Y-27632 had only minimal effects on anoikis, suggesting a larger contribution of non-Rho kinase effectors.
Shc is best known as a transforming protein, acting to facilitate growth factor–dependent Ras activation through Grb2/Sos recruitment. However, growth factor receptors can directly bind Grb2 without Shc, suggesting that Shc dictates an additional level of control for this and perhaps other pathways. One possibility consistent with its scaffolding function is that it may link distal signaling with other transmembrane complexes, thus assigning environmental context to GTPase signals. For example, Shc may modify VEGF-receptor signaling in response to cell–cell contact through its association with vascular endothelial–cadherin (). In addition, Shc permits matrix-dependent EGF signaling to Ras and MAPKs (; ) and, by association with specific integrins, allows Rac1-dependent cell cycle progression after attachment to extracellular matrix proteins ().
Our data suggest that p66 may follow this paradigm and report anchorage context through RhoA activation. Like Shc, RhoA is best known as a proliferative agent, at least in part working through Rho kinase–dependent increases in cytoskeletal tension (). Such RhoA-dependent tension is clearly coupled with the mechanical resistance offered by the cell's environment, as epithelial cells grown in increasingly stiff matrices acquire increasingly malignant phenotypes (). A similar RhoA-dependent test of matrix rigidity is used by mesenchymal stem cells to determine which differentiation program to initiate (; ). In the extreme case of detachment from a solid matrix, however, a liquid environment would be expected to provide no resistance to such tension, which would instead have to be borne entirely by internal struts such as microtubules. Notably, the anoikis-associated BH3-only protein Bim associates with microtubules and initiates apoptosis only upon its release from the cytoskeleton (). Although the mechanism by which unopposed tension causes death is as yet unclear, the situation may be similar to fibroblasts grown in a stressed collagen matrix, which undergo apoptosis when the matrix is mechanically unloaded despite continued attachment to collagen (). Another comparable situation may occur in MDA-MD-231 breast cancer cells, in which overexpression of the motor protein tropomyosin-1 increases susceptibility to anoikis in a manner partially reversible by Y27632 ().
Human p66 was PCR cloned from a previously constructed HUVEC library and subcloned into pCINF containing an N-terminal Flag tag to create pCINF-p66. S36A and S36E mutations were subsequently introduced into pCINF-p66 by PCR mutagenesis. The isolated CH2 domain and the remaining p52 were subcloned into pCINF, and expression was assessed by immunoblot for the Flag tag. PCR strategies were used to subclone full-length p66, p52 (residues 111–583 of p66), and the CH2 domain (residues 1–110) onto the N terminus of EGFP (pEGFP-N3; CLONTECH Laboratories, Inc.). The P411S/P414S, P472S/P475S, E132Q/E133Q, and R285Q point mutations were each introduced into p66-GFP in single mutagenic PCR reactions. Starting with p66-GFP, domain deletions were accomplished by opposing PCR primers tailed with AgeI sites, resulting in deletions of the PTB (residues 119–317), CH1 (residues 318–485), and SH2 (residues 486–584) domains. The p66-ΔPTB plasmid was reopened with AgeI, and the FAT domain of FAK was ligated to create p66-PTB–FAT. The RBD of rhotekin was subcloned into pEGFP-N3 to form RBD-GFP. DsRed-zyxin was a gift from A. Huttenlocher (University of Wiconsin, Madison, WI) and actin-GFP was obtained from CLONTECH Laboratories, Inc. RhoA (N19) and RhoA (V14) were obtained from the University of Missouri at Rolla cDNA Resource Center and subcloned into pCI-neo (Promega). Adenovirus-harboring RhoA (N19) was a gift of C. Chen (University of Pennsylvania, Philadelphia, PA). New constructs were confirmed by sequencing.
φNx-293 cells were obtained from American Type Culture Collection with permission from G. Nolan (Stanford University, Stanford, CA). MDCK and NIH 3T3 cells were also obtained from American Type Culture Collection. Stable transfectants of 293 cells with pCINF, pCINF-p66, pCINF-p66 (S36A), or pCINF-p66 (S36E) were selected with G418. Single cell clones were tested for comparable expression of p66 between mutants and relative to p52. Mixed clones, particularly of 293-p66, progressively lost expression of p66 and thus were not used. G418 was removed for at least two passages before anoikis studies. HUVEC (Clonetics) were used at passages four and five.
The following antibodies were used: Shc, Bim, Bid, Bak, BclX (BD Biosciences); p66 (pS36; Qbiogene); Bax, RhoA, cleaved caspase 7 (Cell Signaling Technology); cytochrome (MitoSciences); FAK (Santa Cruz Biotechnology, Inc.); FAK (pY397; Zymed Laboratories); and actin (Millipore). ML-7, BDM, and (−)blebbistatin were obtained from Sigma-Aldrich, and Y-27632 was obtained from Calbiochem.
Cells were plated on either cell culture–treated or low-attachment 24-well plates (Corning Inc.). DNA fragmentation was assessed using the cell death ELISA (Roche). Mitochondrial release of cytochrome was assessed by digitonin permeabilization. Preliminary titration studies were performed in 293 cells to determine optimal digitonin concentrations for selective plasma membrane permeabilization. Cells were washed and permeabilized in 10 ng/ml digitonin in sucrose buffer on ice for 10 min. Mitochondria were harvested by centrifugation at 1,000 for 5 min. Fractions were then immunoblotted for cytochrome . Activated caspase 7 was assessed with antisera specific for cleaved caspase. Induction of BH domain proteins was assessed by fractionating cells into mitochondrial and cytosolic fractions and immunoblotting for BclX, Bax, Bid, and Bim. Anchorage-independent growth was assessed by plating cells in 0.35% agar on a bed of 0.9% agar in DME with 10% fetal calf serum at a density of 2,000 cells/ml. Visible colonies were scored after 14 d.
HUVEC were electroporated with expression constructs 6–8 h after release from thymidine-induced G1 arrest as described previously (). 293 cells were electroporated unsynchronized. Control cells were transfected with empty pCI-neo or pCINF. siRNA against the CH2 domain of p66 (nt 42–60), RhoA (nt 355–375), and the negative control luciferase were obtained from Dharmacon. Transfection of siRNA was accomplished with TransIT-TKO (Mirus). Adenoviral infection of HUVEC was performed at an MOI of 1:100, titrated to protein expression. Control cells were infected with the same titer of Ad-lacZ.
After transfection, cells were plated on fibronectin-coated chambered coverslips and observed live in full media without fixation 24–48 h later. TIRF microscopy was performed using a TE2000-U system (Nikon). Sequential red and green channels were acquired at 37°C through a 60×/NA 1.45 oil-immersion objective (Nikon) with a digital camera (Coolsnap ES; Roper) using Metamorph software (Molecular Devices). Morphological features were scored by examining all cells in at least five high-power fields per chamber in at least four chambers.
RhoA GTP loading was assessed by a pulldown technique (). The RBD of mouse rhotekin (residues 7–89) was previously ligated into pGEX-2TK (), and recombinant proteins were purified with GSH-sepharose (GE Healthcare). Active GTP-loaded Rho proteins from the 10,000- supernatant of cell lysate were pulled down and immunoblotted with antisera for RhoA.
Fig. S1 shows that p66 increases caspase 7 cleavage and cytochrome release, decreases anchorage-independent growth in 293 cells, and increases Bim in suspended HUVEC. Fig. S2 shows that p66 (P411S and P414S), p66 (P472S and P475S), p66 (ΔCH1), p66 (ΔSH2), p66 (FAT-PTB), and p66 (E132Q and E133Q) display persistent focal adhesion targeting, and that p66 does not affect FAK (Y397) phosphorylation. Online supplemental material is available at . |
Damage to tissues and organs is frequent in the life of vertebrates: tissues can be ripped, squashed, or wounded by mechanical forces, mishaps, or predators. Freezing or burns, chemical insults (strong acids or bases or cytotoxic poisons produced by invading bacteria), radiation, or the withdrawal of oxygen and/or nutrients can also kill cells. Thus, the ability to repair damaged tissues is essential for evolutionary success. Very often the new cells that replace the dead ones migrate from specific niches within the tissue or from distant districts such as the bone marrow. Although the mechanism of cell migration has been intensely studied, the orchestration of the physiological responses that bring the relevant cells to the required sites is much less understood.
We and others have found that high mobility group box 1 (HMGB1), an abundant component of the cell nucleus, when present in the extracellular space, signals tissue damage (). HMGB1 is released by cells undergoing necrosis (accidental cell death) but not by cells undergoing apoptosis (). Extracellular HMGB1 then promotes the ingression of inflammatory cells (), but also the migration and proliferation of stem cells (; ). Thus, HMGB1 has the expected characteristics of a signal that can orchestrate tissue regeneration, although it is not expected to be the only one ().
In particular, we previously described that extracellular HMGB1 can attract mesoangioblasts, both in vitro and in vivo (). Mesoangioblasts are a specific population of mesodermal stem cells that are associated with the walls of fetal and postnatal vessels (). They grow extensively in culture and can differentiate into most mesodermal cell types. When injected into the general circulation of dystrophic mice and dogs, they migrate into muscles and contribute to their regeneration and functional recovery (, ).
Here, we have investigated the signaling pathways that activate cell migration toward extracellular HMGB1 and allow mesoangioblasts to navigate to damaged muscles. HMGB1 is known to activate MAPKs and nuclear factor κB (NF-κB); we show that NF-κB activation proceeds via extracellular signal-regulated kinase (ERK) phosphorylation. Surprisingly, mesoangioblasts and fibroblasts do not migrate toward HMGB1 if NF-κB activation is blocked. This same NF-κB dependency applies to stromal derived factor (SDF)–1/CXCL12, which also directs the migration of stem cells, but not to TNF-α, the archetypal NF-κB activating signal.
Fibroblast cell lines such as 3T3 and wild-type (wt) mouse embryonic fibroblasts (MEFs), either primary or immortalized with polyoma large T antigen (), respond chemotactically to HMGB1 in Boyden chambers (). The migration is directional, as shown by the tracking of living 3T3 fibroblasts in chemoattractant gradients formed between the inner well and the external ring chamber of a Dunn chemotaxis apparatus (). Most cells migrated toward HMGB1 or PDGF, with mean paths of ∼70 and 55 μm, respectively, but were immobile or moved randomly (mean path of 20 μm) in the absence of chemoattractants (). Movement occurred within ∼10, 15, and 25 min in the presence of HMGB1, PDGF, and serum-free medium, respectively (). Similar results were obtained with primary and immortalized MEFs (unpublished data).
Extracellular HMGB1 has been reported to engage multiple receptors, including the receptor for advanced glycation end products (RAGE; ) and Toll-like receptors 2 and 4 (). RAGE has been reported to activate MAPKs; both RAGE and Toll-like receptors activate NF-κB (; ).
We had previously shown that U0126, a specific inhibitor of MAPK/ERK kinase (MEK) 1/2, which phosphorylates ERKs, abrogates the migration of smooth muscle cells in response to HMGB1 (). Likewise, HMGB1 induced the rapid phosphorylation of ERK1/2 in 3T3 fibroblasts () and U0126 inhibited the HMGB1-induced migration ().
NF-κB is a family of transcription factors consisting of dimers of five different proteins—p65 (RelA), RelB, c-Rel, NF-κB1 (p105/p50), and NF-κB2 (p100/p52)—and is essential for most innate and adaptive immune responses (; ). Different types of inactive NF-κB cytoplasmic complexes are activated by a host of stress stimuli by two routes. The classical or canonical NF-κB pathway begins with the activation of NEMO/IκB kinase γ (IKK), IKKβ, and IKKα in a cytoplasmic IKK signalsome complex. IKKβ phosphorylates the NF-κB inhibitor IκBα at two amino-terminal serines, targeting it for polyubiquitination and proteasomal destruction. This leads to the nuclear translocation of NF-κB p50/p65 and p50/c-Rel heterodimers and transcription of their target genes. The alternative or noncanonical pathway is IκBα independent and depends solely on IKKα, which phosphorylates p100 to promote its proteasomal processing to mature p52, thereby causing nuclear translocation of RelB/p52 heterodimers and a subset of p50/p65 heterodimers residing in cytoplasmic p100 complexes ().
Indirect immunofluorescence showed that 3T3 fibroblasts accumulate p65 in their nuclei in response to extracellular HMGB1, but to a lesser extent than in response to TNF-α (). IκBα phosphorylation on Ser-32 and Ser-36 peaked between 15 and 30 min after exposure to extracellular HMGB1 (). These data indicate that HMGB1 activates the canonical NF-κB pathway.
We next tested whether the activation of ERK and NF-κB in response to HMGB1 are parallel or consecutive events. In the presence of HMGB1 and U0126, NF-κB is not translocated to the nucleus of 3T3 cells () and IκBα phosphorylation is impaired (). Thus, HMGB1 activates the canonical NF-κB pathway via ERK.
ERK participates in cytoskeleton remodeling, but NF-κB is not expected to be involved (). To formally rule out a role for canonical NF-κB in HMGB1-elicited cell migration, we tested immortalized MEFs, genetically deficient in both p50 and p65. To our surprise, p50/p65 knockout fibroblasts failed to migrate toward HMGB1, although they migrated as expected toward PDGF (). In further support of this result, wt MEFs stably expressing an IκBα super-repressor (IκBαSR), which cannot be phosphorylated and degraded (), did migrate as expected toward PDGF but not toward HMGB1 (). These experiments indicate that the classical NF-κB pathway controls HMGB1-elicited cell migration, but the possibility remains that NF-κB dimers might interact with the cytoskeleton instead of controlling transcription.
We then showed that SN50, a cell-permeable peptide that competes with the nuclear transport of p50 (), interfered in HMGB1-elicited cell migration, whereas the scrambled control peptide SN50M had no effect (). Significantly, neither SN50 nor SN50M affected PDGF-elicited cell migration (unpublished data). These results suggest that the nuclear function of NF-κB is required for cell migration toward HMGB1. Indeed, fibroblasts pretreated for 30 min with cycloheximide, an inhibitor of protein synthesis, or 5,6-dichloro-1-β-D- ribobenzimidazole (DRB), an inhibitor of transcription, failed to migrate toward HMGB1 but migrated toward 1% serum (). Similar results were obtained with endothelial and smooth muscle cells (unpublished data). These data show unequivocally that the transcriptional activity of NF-κB is necessary for cell migration toward HMGB1.
We next investigated whether NF-κB activation is also required for the HMGB1 migration response of mesoangioblasts. HMGB1 induces p65 accumulation in mesoangioblast nuclei, although to a lesser extent than TNF-α (). In addition, HMGB1 induced the transcriptional activation of the endogenous IκBα gene, a direct target of NF-κB ().
It was previously shown that mesoangioblasts respond chemotactically to several cytokines present in dystrophic muscle, including SDF-1/CXCL12 and TNF-α (). TNF-α is a well-known activator of the NF-κB pathway (); CXCL12 has been reported to activate a variety of pathways, including NF-κB, in pre-B cell lines (). Indeed, CXCL12 induces p65 nuclear translocation in mesoangioblasts ().
NF-κB activation and mesoangioblast migration are causally related, but only in response to a subset of chemoattractants. In fact, mesoangioblasts transiently expressing IκBαSR showed an impaired chemotactic response to HMGB1 and CXCL12, whereas their response to TNF-α was unaffected ().
We next tested whether NF-κB activity is required for mesoangioblast ingression into the diseased muscles of α-sarcoglycan (α-SG) null dystrophic mice. Mesoangioblasts were transfected with plasmids expressing GFP and IκBαSR, or GFP alone. 48 h after transfection ∼25% of cells in each population were fluorescent. We injected 450,000 cells into the femoral artery of dystrophic mice (two per group). 6 h later we recovered filter organs (liver, spleen, and lung) and muscles (gastrocnemius, quadriceps, and tibialis) from the side of injection and the contralateral leg. Mesoangioblasts had migrated within the tissues and had not simply positioned themselves within or just outside microvessels ().
To estimate the fraction of transfected mesoangioblasts arriving into the tissues, we quantified GFP mRNA in each organ and in the mesoangioblast populations before injection. In each of the muscles on the injected and contralateral sides, mesoangioblasts expressing IκBαSR and GFP were substantially fewer than those expressing GFP alone (13 ± 0.1 vs. 32.0 ± 0.2%, P < 0.001, two-tailed test; ). In contrast, filter organs contained more IκBαSR-expressing than control mesoangioblasts (20.0 ± 0.1 vs. 27.3 ± 1.9%, P < 0.05), which is consistent with the notion that mesoangioblasts not homing to muscle are mostly trapped in filter organs. This experiment was replicated three times, with similar results. Interestingly, IκBαSR reduces mesoangioblast ingression into dystrophic muscle but does not abrogate it completely, which is consistent with our in vitro results showing that not all chemotactic responses require NF-κB activation.
Collectively, our results indicate that the canonical NF-κB activation is required for migration of fibroblasts and mesoangioblasts toward specific chemoattractants associated with tissue damage. Although NF-κB is well known to direct the synthesis of cytokines and chemokines that induce the migration of immune effector cells, reports on its role in the migrating cells themselves are scarce. IκBαSR abrogates chemotaxis toward Fgf-7 in immortalized human pancreatic ductal epithelial cells and in fibroblasts forcedly expressing the Fgf-7 receptor FGF123R/IIIb (). In a clone of the human osteogenic sarcoma cell line forcedly expressing the leukotactin-1 receptor CCR1, leukotactin-1 activates NF-κB to transcribe LZIP, and LZIP protein enhances cell migration by binding to CCR1 ().
Our results also indicate that NF-κB activation is required for mesoangioblast ingression into damaged tissue, under conditions corresponding to the procedures being developed for cell-based therapies of muscular dystrophy (). An important consequence of the requirement for NF-κB activation for tissue regeneration is that pharmacological regimens that suppress inflammation by interfering with NF-κB (including corticosteroids, which are commonly prescribed to dystrophic patients) might also suppress tissue regeneration and interfere with stem-cell therapies.
NF-κB activation is not required for the mechanical actions involved in cell migration because it is not needed for migration toward PDGF or formyl-met-leu-phe (fMLP). Moreover, NF-κB activation is not required for migration toward TNF-α, the archetypal canonical NF-κB activating signal. HMGB1 and CXCL12 both contribute to tissue repair (; ; ) and both belong to a small group of chemoattractants that direct the navigation of stem cells (; ; ; ). We speculate that, mechanistically, the difference between HMGB1/CXCL12 and TNF-α may be caused by differences in the timing and intensity of NF-κB activation or the concomitant activation of other signaling pathways. Physiologically, the difference may reflect specific requirements for the initiation and maintenance of migration of differentiated and stem cells in response to tissue damage.
Full-length, LPS-free recombinant HMGB1 protein was provided by HMGBiotech, human recombinant PDGFBB and TNF-α by R&D Systems, CXCL12 by Preprotec, and fibronectin by Roche. Anti–β-actin mAbs, fMLP, DRB, and cycloheximide were obtained from Sigma-Aldrich. U0126 and antibodies against IκBα, pIκBα, ERK, and phospho-ERK were obtained from Cell Signaling Technology. Rabbit polyclonal antibodies against p65/RelA were obtained from Calbiochem and Santa Cruz Biotechnology, Inc. SN50 and SN50M were obtained from Calbiochem.
3T3 mouse fibroblasts were grown in DME supplemented with 10% FCS. Mesoangioblasts (D16 clone) were grown in DME, supplemented with 20% FCS. Immortalized MEFs, wt or deficient for p50 and p65 (p50/p65 −/−; provided by A. Beg, H. Lee Moffitt Cancer Center & Research Institute, Tampa, FL), were cultured in DME supplemented with 10% FCS, 10 mM Hepes, 50 μM β-mercaptoethanol, 10 mM nonessential amino acids, and 10 mM sodium pyruvate. Primary and immortalized MEFs were obtained as described previously (). Constitutive expression of a transdominant IκBαSR in wt MEFs was obtained by retroviral transduction as previously described (; ). A Moloney retroviral vector, which coexpresses an SV40 promoter-driven GFP gene and a long terminal repeat-driven IκBαSR Flag tagged at the amino terminus, was generated by subcloning the Flag-tagged IκBαSR ORF isolated from the pcDNA4-Flag IκBαSR vector (; provided by D.W. Ballard, Vanderbilt University Medical Center, Nashville, TN) into the unique BamHI site of the MP9 Moloney retroviral vector, a derivative of the PINCO vector (; provided by L. Lanfrancone, Istituto FIRC di Oncologia Molecolare, Milan, Italy) in which the cytomegalovirus enhancer was replaced with the SV40 promoter/enhancer. Cells were transiently transfected with each of the aforementioned vectors using FuGENE 6 reagent, according to the manufacturer's instructions (Roche).
Boyden chamber assays were performed as described previously (). In the Dunn chamber, chemoattractants added to the outer well of the device diffuse across the bridge to the inner blind well and form a gradient within ∼30 min (). This apparatus allows for determination of the direction of migration in relation to the direction of the gradient. In our experiments, the outer well of the Dunn chamber was filled with 10 ng/ml PDGFBB and 100 ng/ml HMGB1 or serum-free medium, and the concentric inner well contained only serum-free medium. Fibroblasts were seeded on coverslips coated with 50 ng/ml fibronectin. Coverslips were inverted onto the chamber, and cell migration through the annular bridge between the concentric inner and outer wells was recorded with a microscope (Axiovert S100TV; Carl Zeiss MicroImaging, Inc.), with a still frame every 3 min for 6 h.
Cells were serum starved for 16 h in DME and stimulated with 100 ng/ml HMGB1, 100 ng/ml CXCL12, or 20 ng/ml TNF-α for the indicated times. Western blotting was done as described previously (). Indirect immunofluorescence was done as described previously (), using rabbit polyclonal anti-p65 antibody at 4°C and AlexaFluor 594–conjugated goat anti–rabbit IgG (Invitrogen). Slides were mounted in 90% glycerol, 20 mM Tris, pH 8.8, and 0.5% -phenylenediamine. Images were taken at 37°C on a DeltaVision system consisting of a microscope (Olympus IX70), PlanApo 40×/1.35 and 60×/1.4 oil immersion objectives (Olympus), and a camera (HQ CoolSnap; Roper Scientific). Image acquisition and deconvolution (10 iterations) were done with Softworx 3.5.0 (Applied Precision).
Mesoangioblasts (100,000 per well) were stimulated with 100 ng/ml HMGB1 or 20 ng/ml TNF-α for the indicated times. cDNA was obtained with Illustra RNAspin mini RNA isolation kit (GE Healthcare) and amplified by real-time PCR on a LC480 instrument (Roche), using the relative quantification software, with LightCycler 480 SYBR Green I Master mix and primers for mouse β-actin (TGACGGGGTCACCCACACTGTGCCCATCTA and CTAGAAGCATTGCGGTGGACGATGGAGGG) and IκBα (CTTGGCTGTGATCACCAACCAG and CGAAACCAGGTCAGGATTCTGC).
D16 mesoangioblasts were transfected with FuGene 6 (according to the manufacturer; Roche) with expression vectors encoding IκBαSR and GFP (pMP9IκBαSR) or GFP alone (pMP9), and grown for 48 h. The experiment was then conducted as previously described (). Cells were resuspended in PBS, and the same amount of cells (450,000 cells in 25 μl) were subjected to RT-PCR for GFP expression, or injected into the exposed femoral artery of anesthetized α-SG −/− dystrophic female mice (two per group). After 6 h, mice were killed, and the amount of GFP expression was measured by RT-PCR in filter organs (liver, lung, and spleen) and muscles (tibialis, gastrocnemius, and quadriceps) on the injected and contralateral sides. Results are expressed as the percentage of GFP signal in the specific tissue, setting GFP expression before injection as 100%. Muscles and filter organs were also processed for immunostaining as described previously (), using rabbit anti-GFP polyclonal antibody (Chemicon) and AlexaFluor 488–conjugated donkey anti–rabbit secondary antibody (Invitrogen) and mouse anti-myosin heavy chain mAb MF20 () and AlexaFluor 594–conjugated donkey anti–mouse secondary antibody (Invitrogen). Slides were mounted in fluorescent mounting medium (DakoCytomation). Images were taken at room temperature with a microscope (CTR 6000; Leica) equipped with a 40×/0.60 objective and a camera (DFC 350 FX; Leica); the acquisition software was LAS AF application suite 1.6.2 (Leica).
Digital Images were elaborated using Photoshop 8.0 (Adobe); the luminosity of brightest and dimmest pixel in each channel were adjusted to obtain the best visual reproduction, taking care to maintain linearity in the brightness scale. Images were included in figures using Illustrator 11.0 (Adobe).
Pairwise comparisons between continuous data were done using unpaired two-tailed Student's test; statistical analysis involving more than two groups was done using an analysis of variance model. Prism 4.0b software was used (GraphPad Software, Inc.). |
In somatic cells, DNA damage or stalled DNA replication can activate the S-phase checkpoint, resulting in delayed cell cycle progression to allow the damage to be repaired (for reviews see ; ). S-phase checkpoint signaling is mediated by ataxia telangiectasia mutated and Rad3 related (ATR) and Chk1 protein kinases. Replication forks that stall at sites of DNA damage activate ATR, which then phosphorylates and activates Chk1. Finally, cell cycle progression is delayed by activated Chk1 through the modulation of core cell cycle regulators, such as the Cdc25 protein phosphatase.
In contrast to somatic cells, early embryonic cell cycles typically lack a checkpoint response to DNA damage (for review see ). In both and , this is because an insufficient number of nuclei are present in early embryos, and, thus, an insufficiently robust checkpoint signal is generated to thwart the mitosis-promoting activity of maternally supplied and abundant Cdk1–cyclin B complexes. In both flies and frogs, it is only later in embryogenesis that the checkpoint signal produced by replication stress is strong enough to neutralize Cdk1–cyclin B, and this is caused by the accumulation of nuclei (; , ; ; ; ; ). In , the situation is quite different. The ATR–Chk1 pathway is present and active from the first division onwards in worms, and it plays an important role in controlling the timing of cell division during the early cycles (). Checkpoint function is restricted to the P lineage, or future germ line, in embryos, and its activation by as of yet undetermined developmental cues results in the delayed division of P cells relative to their sisters. This asynchrony in cell division is critical for embryonic and germ line development, as reducing the delay through inactivation of the ATR–Chk1 pathway results in germ line developmental failure and sterility, whereas extending the delay through hyperactivation of the ATR–Chk1 pathway results in patterning defects and embryonic lethality (; ; ; ).
Although differs from and in that the ATR–Chk1 pathway controls the pace of the early embryonic cycles, what is common between them is that like frog and fly embryos, the checkpoint is nonresponsive to DNA damage in early nematode embryos. This is not the result of insufficient signal strength but rather of the presence of an active silencing mechanism that suppresses the checkpoint response to DNA damage but allows the checkpoint to respond to developmental cues (). This silencing mechanism has presumably evolved to prevent unscheduled checkpoint activation, which would cause extended delays in cell division and, ultimately, embryonic lethality. Our laboratory identified this checkpoint silencing mechanism, and, to date, we have isolated three genes that are required for silencing: the SUMO E3 ligase, the translesion synthesis DNA polymerase, and the mutationally defined but uncloned gene (). Previous work has shown that and silence the checkpoint through their ability to promote the rapid replication of damaged DNA (), whereas the role of in silencing was as of yet unknown.
The mutation was isolated 25 yr ago in a screen for mutations causing embryonic sensitivity to DNA-damaging agents (). Follow-up phenotypic analysis of showed that mutant animals were competent for excision repair and that the period of DNA damage sensitivity was restricted to early embryogenesis (; ; ). More recently, we have shown that is a component of the silencing pathway that suppresses activation by DNA damage in early embryos (). This conclusion was based largely on effects of the mutation on the timing of cell division in early embryos exposed to DNA-damaging agents. Wild-type embryos did not delay the cell cycle after exposure to either methyl methanesulphonate (MMS) or UV-C or UV light, whereas mutant embryos showed a substantial delay. Importantly, the damage-induced delay in embryos was reversed upon the RNAi-mediated depletion of . These genetic experiments indicated that antagonizes the pathway during the early embryonic DNA damage response and prompted us to further explore function in checkpoint silencing.
In this study, we report the cloning of and show that the phenotype is caused by mutations in the gene. is an evolutionally conserved regulatory subunit of protein phosphatase 4 (PP4; or in ) and has recently been shown to control lifespan in the worm (). We report that the roles of in checkpoint silencing and longevity are distinct, and we show that the function of SMK-1 in silencing is to recruit PPH-4.1 to replicating chromatin so that it may antagonize checkpoint signaling during a DNA damage response. These results link PP4 to negative regulation of the ATR–Chk1 checkpoint, provide a targeting function for the SMK-1 regulatory subunit, and illustrate how during development primordial inputs into the ATR–Chk1 pathway such as DNA damage may be bypassed so that the checkpoint can respond exclusively to developmentally programmed inputs.
To gain cytological and biochemical evidence that antagonizes during a DNA damage response, we examined the phosphorylation status of CHK-1 in wild-type and embryos exposed to MMS. To do this, we used an antibody that recognizes the Ser345-phosphorylated (CHK-1–S345-P) and activated form of the enzyme and examined early embryos by immunofluorescence microscopy (). Wild-type (N2) embryos displayed a punctate staining pattern with this antibody that was specific for the P lineage in both two-cell () and four-cell () embryos, and this signal was largely reduced in RNAi embryos (Fig. S1, available at ). Exposure of N2 embryos to MMS did not substantially alter the CHK-1–S345-P signal intensity (), which is consistent with the checkpoint being silenced in wild-type embryos (). In contrast to wild type, however, embryos showed a noticeable increase in CHK-1–S345-P signal intensity after exposure to MMS (). To confirm these cytological observations biochemically, we prepared whole embryo extracts for the purpose of detecting activated CHK-1 by immunoblotting. As shown in , activated CHK-1 was not readily detected in control or MMS-exposed N2 embryos. In contrast, slightly more activated CHK-1 was observed in embryos, and this was substantially increased upon MMS exposure. To ensure equal loading, we also probed the blots for total CHK-1 and PCN-1, the worm orthologue of proliferating cell nuclear antigen, and found that equivalent amounts of these factors were present in all extracts. Image densitometry of the blot in revealed that approximately threefold more activated CHK-1 was present in embryos relative to wild type after exposure to MMS (). Based on the data in , we conclude that DNA damage activates CHK-1 to a greater extent in embryos relative to wild type.
The data in show that activated CHK-1 localizes to punctate cytoplasmic structures in P cells that are reminiscent of P granules. To determine directly whether these structures are indeed P granules, we performed colabeling experiments using antibodies against activated CHK-1 and the P granule component PGL-1 (). As shown in , the activated CHK-1 and PGL-1 signals overlapped, and, from this, we conclude that activated CHK-1 resides in P granules. To determine whether P granule residency was controlled by , we also stained early embryos with these antibodies and found that activated CHK-1 still resides in P granules despite the loss of function (). We conclude that activated CHK-1 localizes to cytoplasmic P granules in a –independent manner. The mechanism by which activated CHK-1 accumulates in P granules and the importance of this for CHK-1's ability to control the cell cycle is not yet known and is currently under investigation.
Having found that negatively regulates during the DNA damage response in early embryos, we next asked whether function was restricted to early embryogenesis or whether it was required throughout the embryonic period. Earlier studies had shown that plating embryos on media containing MMS did not prevent hatching, whereas exposing adults to MMS prevented the hatching of their progeny (; ). This suggested that very early embryogenesis represented the period of DNA damage sensitivity in mutants; therefore, we sought a more direct test of this hypothesis. To do this, we collected early embryos from gravid adults by bleaching and plated these embryos. Next, we UV irradiated the embryos and determined survival as a function of both dose and time of administration of the UV light (). Early embryos (i.e., those irradiated immediately after plating) were more sensitive to UV light than early wild-type embryos at all doses of UV that were tested. Interestingly, there was little difference in the UV light sensitivities of relative to wild type if the UV light was administered ≥4 h after plating (). From this, we conclude that early but not late embryos require to survive DNA damage.
In , there are two sources of rapidly proliferating cells: the early embryo and the adult hermaphrodite gonad (for review see ). We have previously shown that the pathway responds to DNA damage in the gonad but is silenced in the early embryo (). Therefore, it was of interest to determine whether function was restricted to early embryos or whether it was also required in the germ line to survive DNA damage. To do this, we UV light irradiated hermaphrodites to damage the germ cells and mated them to untreated males harboring a GFP–ribonucleotide reductase (RNR) transgene (). We then asked whether viable cross progeny could be produced from the UV-irradiated germ cells. We performed the mating step because we required a source of undamaged sperm so that all effects on the survival of progeny would be through DNA damage inflicted specifically in the mitotic zone of the hermaphrodite gonad. As shown in , the cross progeny from this experiment were viable, but the self progeny were not. This result indicates that mitotically dividing germ cells in the hermaphrodite gonad do not require function to survive DNA damage. The fact that the self progeny in this experiment were sensitive to DNA damage likely reflects the inability of early embryos to survive the damaged DNA supplied by the UV-irradiated sperm.
To pursue these observations further, we next asked whether the mutation hyperactivates the ATR–Chk1 pathway in the gonad, as it does in early embryos. Previous work has shown that mitotically dividing germ cells in the distal tip of the gonad arrest in an –dependent manner after exposure to UV light (). This arrest is reflected by a reduction in the number of nuclei at the distal tip (or mitotic zone) and an increase in their size. Therefore, we compared cell cycle arrest in wild-type versus gonads after exposure to UV light (). If the loss of function hyperstimulates the ATR–Chk1 pathway in germ cells, we would expect a more pronounced reduction in the number of mitotic nuclei at the distal tip in relative to wild-type gonads. We observed that UV light caused a reduction of 17.4 mitotic zone nuclei on average in wild-type animals and a reduction of 10.1 nuclei in gonads. These data show that the loss of function in distal tip germ cells does not reduce the number of UV light–exposed mitotic zone nuclei beyond what is observed in wild type and, in fact, that gonads are modestly more refractory to –dependent cell cycle arrest than are wild-type gonads. We conclude that the stimulatory effect of the mutation on the ATR–Chk1 pathway is specific for the early embryonic cell cycle.
To pursue these observations further, we next sought to identify the gene encoding . Previous genetic analysis of had mapped the position of the gene to 1.09 ± 0.46 cM on chromosome V (). Using a combination of bulk segregation analysis, three-factor crosses, and single nucleotide polymorphism (SNP) mapping, we were able to refine this position to the interval between 1.38 and 1.88 cM. To identify , we performed an RNAi screen across this interval using the soaking method. We initially searched for genes that would render embryos sensitive to UV light after depletion by RNAi. This resulted in the identification of at position 1.49 cM as a candidate gene encoding . To pursue this further, we performed more detailed analysis of the RNAi phenotype. Two different regions of the gene, the central region and the 3′ end, were targeted for RNAi knockdown. RNAi against the central region (RNAi#1) resulted in a low level of embryonic lethality, and this was greatly increased when embryos were exposed to MMS (). Therefore, RNAi#1 phenocopies . RNAi against the 3′ end (RNAi#2) resulted in high embryonic lethality even in the absence of MMS. When either RNAi#1 or #2 were combined with the mutation, embryonic lethality was higher than that observed in any single case alone (). These results show that is an essential gene and that RNAi#1 represents a hypomorphic condition. These results are also consistent with the idea that represents a hypomorphic allele of the gene.
A hallmark of the phenotype is that these embryos show a checkpoint-dependent delay in cell cycle progression in response to DNA damage. This is in contrast to wild-type embryos, which silence their checkpoint responses during a DNA damage response. If represents a hypomorphic allele of , RNAi#1 should phenocopy for checkpoint silencing. To address this, we timed cell cycle progression in early embryos as described previously (). In both and RNA#1 embryos, the first cell cycle occurred normally in the absence of DNA damage but was substantially delayed after exposure to MMS (). Importantly, in both cases, this MMS-induced delay was reversed after RNAi. These results show that RNAi#1 phenocopies the checkpoint silencing defect of . To determine whether a wild-type copy of the gene could rescue the phenotype, we made an –GFP fusion transgene () driven by the promoter and introduced the gene into animals by particle bombardment to produce the (––GFP) strain. Transformants were selected by virtue of GFP signals and were tested for sensitivity to DNA-damaging agents. Introduction of wild-type coding sequences into animals increased resistance to both MMS and UV light (). Furthermore, when the timing of cell division was examined in early embryos, we observed that (––GFP) embryos did not delay the cell cycle to the same extent as mutants after exposure to MMS (). From this, we conclude that RNAi#1 phenocopies the DNA damage response phenotypes of and that introduction of an –GFP transgene into mutants partially suppresses these phenotypes.
As further evidence that represents an allele of , we sought to link to a previously identified phenotype of , longevity. The gene was first identified in as a regulator of lifespan (). RNAi against reduces both the lifespan of wild-type animals and the extended lifespan of mutant animals. Therefore, we performed longevity assays on animals and animals exposed to RNAi and compared these lifespans with N2 and N2 RNAi animals. As shown in , in both cases, the N2 animals lived longer than animals. Thus, like RNAi, the allele reduces the lifespan of otherwise wild-type animals, and it reduces the extended lifespan that results from the depletion of .
To determine the molecular basis of the mutation, we sequenced the gene in the strain. encodes an evolutionally conserved regulatory subunit of PP4 (). Homologues of include human PP4R3, yeast , and (; ; ). Three differences were found in the gene from relative to wild-type strains (E497G, D580G, and D703G; ). Of particular interest is the mutation occurring at position 703, as this aspartic acid residue is absolutely conserved from yeast to humans () and is found within a highly conserved subdomain of the SMK-1 protein, conserved region 3 (). Collectively, our data show that RNAi phenocopies the allele for both DNA damage response and lifespan phenotypes, that a transgene can partially suppress the phenotype, and that the gene from the strain contains mutations, including an amino acid substitution at an evolutionally conserved position. We conclude that the phenotype is caused by mutations in the gene.
Recent work has demonstrated that functions in lifespan regulation by controlling transcriptional activity of the forkhead box O (FOXO) transcription factor (). Thus, it was possible that the effects of on checkpoint silencing were through the regulation of . If so, we would expect that mutant embryos would be sensitive to DNA-damaging agents, but this was not the case (). These results show that although is an allele of , the role of in checkpoint silencing is distinct from its role in –mediated longevity.
In other organisms, orthologues form complexes with PP4 (). To see whether SMK-1 did the same, we performed coimmunoprecipitation experiments using proteins expressed by in vitro transcription/translation in rabbit reticulocyte lysate. Lysates expressing PPH-4.1, the homologue of PP4, were mixed with lysates expressing epitope-tagged SMK-1. The mixtures were then immunoprecipitated with an antibody that recognizes the tag on SMK-1, and, as shown in , PPH-4.1 was found in these immune complexes. PPH-4.1 was not found in the immune complexes when epitope-tagged SMK-1 was omitted from the reaction or when nonspecific antibody was used in the coimmunoprecipitation, demonstrating specificity. We conclude that SMK-1 interacts with PPH-4.1. We next asked whether the D703G mutation in the allele of , which lies in conserved region 3 of the protein, influenced interaction between SMK-1 and PPH-4.1. As shown in , PPH-4.1 did not efficiently coimmunoprecipitate with a mutant form of SMK-1 containing the D703G substitution (SMK-1 D703G). These data suggest that at least in part, the phenotype is caused by a compromised interaction between SMK-1 and PPH-4.1.
To pursue these observations further, we assessed DNA damage response phenotypes for embryos depleted of by RNAi. Unlike mutants, embryos depleted of were very sensitive to both UV light and MMS (). Furthermore, –depleted embryos displayed a DNA damage–dependent delay in progression through the first cell cycle in a manner similar to embryos (). Based on these data, we conclude that the phenotype is caused by an inability of SMK-1 to control PPH-4.1 function during the DNA damage response.
To learn more about how performs its checkpoint silencing function, we used the (––GFP) strain to localize SMK-1–GFP in early embryos (). The fusion protein was nuclear throughout all stages of the cell cycle. At prophase, SMK-1–GFP colocalized with condensed chromosomes, indicating that SMK-1 is a chromosomal protein (). To make certain that these localization patterns were not an artifact of the exogenous promoter used in our construct, we repeated this analysis with a strain driving SMK-1–GFP off the endogenous promoter and obtained identical results (unpublished data). To see whether chromosomal occupancy of SMK-1 was dependent on DNA replication, we treated (––GFP) animals with RNAi against the replication initiation factor . As shown in , the chromosomal localization of SMK-1–GFP was abolished in RNAi embryos. This was not the case for embryos expressing a histone H2B-GFP fusion protein, which localized to condensed chromatin regardless of the depletion of (). We also asked whether abrogation of the ATR pathway influenced the chromosomal localization of SMK-1 and found that SMK-1 localization was not perturbed by RNAi (). The effectiveness of the RNAi in this experiment was ascertained by the high level of embryonic lethality that resulted, which is a known consequence of RNAi (). From this experiment, we conclude that SMK-1 is recruited to chromatin in a replication-dependent and checkpoint-independent manner.
The results obtained thus far indicate that SMK-1 and PPH-4.1 form a complex, that both proteins confer DNA damage resistance to early embryos, and that SMK-1 is recruited to chromatin in early embryos in a manner dependent on DNA replication. To pursue the chromatin-binding properties of SMK-1 further, we developed a chromatin-binding assay for early embryos () based on previously published procedures (). Large quantities of early embryos were isolated from adults and sonicated to produce an embryo extract. The extract was centrifuged to produce two fractions: a supernatant (A) and the chromatin-containing pellet (B). The pellet fraction was then treated with micrococcal nuclease to degrade the DNA and to release the DNA-bound chromatin proteins. This reaction was then centrifuged again to produce a supernatant (C) and pellet (D) fractions. Proteins that were originally in the first pellet fraction (B) but were found in the second supernatant fraction (C) after micrococcal nuclease treatment were defined as chromatin proteins and identified by immunoblotting. As shown in , the known chromatin protein PCN-1 was found in fractions B and C but not in fraction D as expected. In contrast, the nonchromatin protein tubulin was found exclusively in fraction A, verifying that this procedure can identify chromatin proteins. We also examined the behavior of SMK-1–GFP and PPH-4.1 under these fractionation conditions. As expected, based on the localization data in , SMK-1–GFP was found in the chromatin protein–containing fraction C. PPH-4.1 was also found in fraction C, and some was observed in fraction D. It may be that a subset of PPH-4.1 associates with a nonchromosomal, easily sedimenting structure such as the centrosome ().
We next used this assay to monitor the chromatin association of SMK-1–GFP and PPH-4.1 under different conditions. As shown in , SMK-1–GFP was found in the chromatin protein–containing C fraction in both control and MMS- exposed embryos (lanes 2 and 3). RNAi-mediated depletion of , another checkpoint silencing gene, had no effect on the chromatin binding of SMK-1–GFP (, lanes 4 and 5), whereas RNAi against itself did prevent the recovery of SMK-1–GFP in the chromatin fraction (, lane 6) as expected. PCN-1 was used as a control for these experiments and was found in the chromatin fraction under all conditions. To pursue these observations further, we extended this analysis to PPH-4.1. The PPH-4.1 protein was found in the chromatin fraction of both control and MMS-exposed wild-type embryos (, lanes 1 and 2). Importantly, the amount of PPH-4.1 that associated with chromatin in embryos was noticeably reduced relative to wild-type embryos (, lanes 3 and 4). The overall level of PPH-4.1 in versus wild-type extracts was only modestly reduced. To confirm these data using an alternative method, we immunostained MMS-exposed wild-type and embryos with antiserum directed against PPH-4.1. As shown in Fig. S2 (available at ), the PPH-4.1 signal was nuclear in wild-type embryos but not in embryos. Based on these results, we conclude that SMK-1 functions to recruit the PPH-4.1 phosphatase to chromatin and that a failure to do so, such as in embryos, leads to hyperactivation of the response to DNA damage and subsequent embryonic lethality.
In this study, we have shown that mutations in the gene cause the phenotype. We have also shown that although the mutation has a strong effect on early embryonic DNA damage resistance, it does not affect damage resistance in proliferating cells of the germ line. Consistent with a role for in early embryos but not the germ line is published data showing that an SMK-1–GFP fusion protein expressed off the endogenous promoter is abundant in early embryos as well as other tissues of the worm but is not readily observed in the germ line (). Therefore, it may be that the embryonic specificity of the checkpoint silencing pathway is achieved through preferential expression of the SMK-1–PPH-4.1 complex in embryos relative to germ cells. The lack of a phenotype in germ cells must be interpreted with caution, however, given the hypomorphic nature of the allele.
The work presented here has uncovered a role for SMK-1 in silencing DNA damage–based CHK-1 activation in early embryos. In , SMK-1 also functions in the insulin-mediated control of longevity (). In longevity, SMK-1 modulates the activity of the DAF-16 transcription factor through an unknown mechanism to regulate the expression of DAF-16 target genes. We have shown here that DAF-16 is not required for checkpoint silencing, and, thus, it appears that the roles for SMK-1 in aging and checkpoint silencing are distinct. DAF-16 is a member of the FOXO superfamily of transcriptional regulators, and, therefore, it is possible that SMK-1 functions with a FOXO transcription factor that is distinct from DAF-16 in the checkpoint silencing pathway. We do not favor this hypothesis, however, as it is generally true that early embryonic cell cycle control is driven by maternally supplied regulators and not via zygotic transcription. Although the roles of in longevity and checkpoint silencing can be unlinked in the embryo, we note that , the target for silencing, has been shown to reduce lifespan in the worm by acting in postmitotic cells (). Therefore, it may be that antagonizes the effect on lifespan, and experiments are in progress to test this hypothesis.
SMK-1 is an evolutionally conserved regulatory subunit of the PP4 phosphatase. Links between the PP4 complex and DNA damage response have been uncovered before, although not in the context of regulation of the ATR–Chk1 pathway as has been reported here. In , loss of the SMK-1 orthologue causes sensitivity to the DNA-damaging agent cisplatin (). In yeast, the SMK-1 orthologue Psy2 and the PP4 orthologue Pph3 have been shown to control the phosphorylation status of the histone variant H2AX after DNA damage (). In this case, dephosphorylation of H2AX by Pph3 is required for attenuation of the checkpoint response to double-strand breaks. This is somewhat similar to the results reported here, in which SMK-1 and PPH-4.1 negatively regulate the ATR–Chk1 pathway after DNA damage; however, the mechanism in is clearly different, as worms do not have H2AX. Although these previous reports clearly linked SMK-1 orthologues to the DNA damage response (; ), they did not explain the role of SMK-1 in this process. We report here that SMK-1 is a chromosomal protein and that its recruitment to chromatin is dependent on ongoing DNA replication. Furthermore, we show that SMK-1 is required to recruit PPH-4.1 to chromatin, the site of CHK-1 activation during a DNA damage response. Collectively, these data supply a function for SMK-1 during the DNA damage response (the targeting of PPH-4.1 to chromatin) and suggest that the SMK-1–PPH-4.1 complex may be a general regulator of the ATR–Chk1 pathway in metazoan cells.
Although our data clearly identify SMK-1–PPH-4.1 as an important negative regulator of the checkpoint response to DNA damage in early nematode embryos, we do not at present know the critical target for this phosphatase complex in attenuating the checkpoint response. Chk1 is known to be regulated directly by protein phosphatase 2A (), by PPM1D, a type 2C phosphatase (), and by Dis2, a type I phosphatase in fission yeast (). Preliminary results from our laboratory have shown that PPH-4.1 and CHK-1 form a complex (unpublished data), and, thus, it may be that PP4-type phosphatases are also capable of the direct regulation of Chk1. Regulation of Chk1 is likely to be complex in any given cell type, and is likely to involve multiple phosphatases controlling Chk1 under different circumstances and in different subcellular locations. Our data show that a site for regulation of the ATR–Chk1 pathway by PP4 is chromatin, and this is consistent with the embryo's requirement that the Chk1 pathway be rapidly inactivated so as to prevent potentially lethal delays in cell cycle progression. To completely understand how the checkpoint is silenced in early embryos, it will be necessary to identify the SMK-1–PPH-4.1 target and to determine how this target is accessed by SMK-1–PPH-4.1 on replicating DNA.
The wild-type N2 Bristol strain was used in all control experiments (). SP488 (), CF1038 (), and AZ212 (+) strains were provided by T. Stiernagle ( Genetics Center, University of Minnesota, Minneapolis, MN). An RNR-GFP strain (103+) was provided by E. Kipreos (University of Georgia, Athens, GA; ).
Previous genetic mapping had determined the chromosome location of to be V:1.09 ± 0.461 cM (). To confirm and extend these initial mapping data, the TH37 strain containing and markers located at V:0.00 cM and V:1.88 cM, respectively, was used in three-factor cross mapping of . TH37 hermaphrodites (−+−) were mated with males (+−), and cross progeny were isolated. These worms were allowed to self-fertilize, and recombinants representing the −+ and +− genotype were identified. Once it has been determined that recombinant worms were homozygous for each marker, the status of the gene was determined for each recombinant. 27 −+ and 21 +− homozygous recombinants were screened for MMS sensitivity. The fact that −− and −− recombinants were isolated confirms that is to the right of and to the left of , or between 0.00 and 1.88 cM. In a total of 48 recombination events, 30 events occurred between and , and 18 events occurred between and . These numbers translated to map ratios of 0.625 and 0.375 for and , respectively. Therefore, the three-factor cross indicated that lies closer to at V:1.175 with a 95% confidence interval of ±0.273 cM.
To further narrow the region of the gene, we performed SNP mapping. − worms were mated to the Hawaiian CB4856 strain, and five recombinants that were and wild type for (based on MMS sensitivity) were isolated. SNPs located between 1.14 and 1.46 cM were PCR amplified from recombinant worm lysates, and the origin of DNA at each locus was determined by either snip-SNP analysis or sequencing of the SNP. This analysis revealed that among the recombinants, DNA could be found at positions 1.14, 1.27, and 1.38 cM but not at 1.46 cM. This analysis positioned the locus to the right of 1.38 cM and, in combination with the three-factor crosses, defined the interval between 1.38 and 1.88 cM as the location of the gene.
and RNAi by soaking method was performed as described previously (). , , , , and RNAi by feeding method was performed as described previously (), and RNAi was performed as described previously ().
To examine whether function was restricted to early embryogenesis or whether it was required throughout the embryonic period, UV sensitivity assay was performed using embryos prepared by bleaching N2 and gravid hermaphrodites on the basis of published protocols (). About 50 early embryos were plated on fresh plates and exposed to UV light at 0, 10, and 25 J/m at the indicated times in . 24 h after UV irradiation, the unhatched eggs were counted. Embryonic lethality was determined by dividing the number of eggs remaining after 24 h by the total number plated.
To examine whether function was restricted to early embryos or whether it was also required in the germ line to survive DNA damage, UV light–irradiated hermaphrodites were crossed with untreated male worms harboring an RNR-GFP transgene, and the UV light sensitivity of progeny was examined. To do this, 10 L4-stage N2 and hermaphrodites were exposed to 100 J/m UV light followed by plating eight males harboring an RNR-GFP transgene. 48 h after transferring males, all worms were removed from the plate, and GFP and non-GFP embryos were counted. After 24 h, the embryos were scored for survival to determine embryonic lethality.
Using a combination of bulk segregation analysis, three-factor crosses, and SNP mapping, the position of the gene was refined to the genetic interval between 1.38 and 1.88 cM. To clone the gene, UV-sensitive genes across this genetic interval were initially identified by UV sensitivity assay after depletion by soaking RNAi and were analyzed further by MMS sensitivity assay and timing of cell division in living embryos, which were performed as described previously (). For longevity assay of , lifespan and statistical analyses were performed as described previously ().
To construct an –GFP fusion transgene, a full-length cDNA of the gene was cloned into a –GFP germline expression vector (). The –GFP fusion transgene was introduced into the mutant by microparticle bombardment to generate the (GFP) strain. Using this transgenic strain, rescue of the mutant was assessed by restoring normal embryonic viability and timing of cell division in living embryos in response to UV light and MMS exposures. Additionally, we also monitored the behavior of SMK-1–GFP expressed under the control of an endogenous promoter in early embryos of a transgenic strain, , which were generated previously (). For genomic DNA sequencing, genomic DNA corresponding to the gene was isolated from the mutant and cloned into pCR2.1-TOPO vector (Invitrogen). Mutations in the genomic DNA were identified by DNA sequencing performed by Agencourt Bioscience Corp.
proliferating cell nuclear antigen orthologue PCN-1 antibody was generated by immunizing rabbits with the peptide DIDSEHLGIPDQDYAVVCE (Bethyl Laboratories). PP4 orthologue PPH-4.1 antibody was a gift from M. Yamamoto (University of Tokyo, Tokyo, Japan; ). Antibodies against phospho-Chk1 Ser345 (Cell Signaling Technology), Chk1 (Santa Cruz Biotechnology, Inc.), GFP (Abcam), OIC1D4 (Developmental Studies Hybridoma Bank), α-tubulin (Sigma-Aldrich), and c-myc (Santa Cruz Biotechnology, Inc.) were purchased. To prepare whole embryo extracts, embryos were obtained by bleaching gravid hermaphrodites and were suspended in twice the pellet volume of homogenization buffer (50 mM Hepes, pH 7.6, 200 mM KCl, 1 mM EDTA, 1 mM EGTA, 0.2% Triton X-100, 5% glycerol, and protease inhibitors). The embryo suspension was sonicated and clarified by centrifugation at 12,000 at 4°C for 30 min. The pellet was resuspended in half the pellet volume of homogenization buffer. Immunoblotting was performed by the standard procedures with HRP-conjugated mouse or rabbit secondary antibodies (GE Healthcare), and protein bands were detected by enhanced chemiluminescence (Pierce Chemical Co.).
To prepare embryo extract fractions containing chromatin proteins, large quantities of embryos were obtained by bleaching gravid hermaphrodites and were suspended in twice the pellet volume of homogenization buffer. The embryo suspension was sonicated briefly on ice until the mixture had lost its viscosity. The sonicated embryo mixture was clarified by centrifugation at 8,000 at 4°C for 5 min, and the pellet was resuspended in half the pellet volume of homogenization buffer. The chromatin proteins were extracted from the pellet by adding micrococcal nuclease (Roche Applied Science) followed by centrifugation at 15,000 for 20 min at 4°C. The supernatant fraction containing chromatin proteins was identified by immunoblotting with antibodies against α-tubulin and PCN-1, which are nonchromatin and chromatin proteins, respectively.
The full-length cDNAs of and genes were cloned into pCS2+MT vector containing myc epitope tags and pSP72 vector, respectively. The myc-tagged SMK-1 mutant displacing an aspartic acid residue at position 703 to a glycine (myc-SMK-1 [D703G]) was generated by the Quik-Change II Site-Directed Mutagenesis kit (Stratagene). The myc–SMK-1, myc–SMK-1 (D703G), and untagged PPH-4.1 were transcribed and translated (TT reaction) in the presence of [S]methionine according to the manufacturer's instructions (Promega). For coimmunoprecipitation, 10 μl TT reactions were mixed in 400 μl of binding buffer (20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM MgCl, 10% glycerol, and 0.1% NP-40) and incubated with 0.5 μg of anti–mouse myc antibody at 4°C for 4 h. A mouse IgG (Santa Cruz Biotechnology, Inc.) was used as a nonspecific antibody for demonstrating specificity. After an overnight incubation with 20 μl of protein A/G PLUS-Agarose (Santa Cruz Biotechnology, Inc.) at 4°C, immunoprecipitated beads were washed three times with binding buffer. The protein bound to the beads was eluted by boiling in 30 μl of 2× Laemmli sample buffer. The samples were run on SDS-polyacrylamide gels and detected by autoradiography.
Embryos and gonads were dissected form adult hermaphrodites and were fixed and stained by Hoechst 33258 as described previously (, ). The images of nuclei in the gonad () were captured on camera (2.1.1; Diagnostic Instruments) and processed using SPOT Advanced version 3.2.4 software (Diagnostic Instruments). UPlanAPO 40× NA 1.40 oil objective lenses were used. The nuclei in the mitotic zone of the gonad were then counted as described previously (). For immunostaining, the fixed embryos were incubated with antibodies against phospho-Chk1 (Ser345), OIC1D4, and PPH-4.1 overnight at 4°C followed by a 2-h incubation with FITC- or rhodamine-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories). All confocal images (; ; ; S1, A–F; and S2, A–I) were obtained by a confocal system (LSM510 META; Carl Zeiss MicroImaging, Inc.) attached to a laser-scanning microscope (Axiovert 100M; Carl Zeiss MicroImaging, Inc.). Plan-Neofluar 40× NA 1.30 oil objective lenses (Carl Zeiss MicroImaging, Inc.) were used. All microscopic experiments were performed at room temperature.
Fig. S1 shows that the activated CHK-1 signal is largely reduced in RNAi embryos. Fig. S2 shows that the nuclear localization of PPH-4.1 is abolished in MMS-exposed and RNAi embryos. Table S1 shows the embryonic lethality in A. Online supplemental material is available at . |
Bloom's syndrome (BS) is a rare autosomal recessive disease characterized by a predisposition to a wide variety of cancers (). The responsible gene product, BLM, is a member of the highly conserved RecQ family of DNA helicases (). Cells derived from BS patients exhibit elevated frequencies of sister chromatid exchanges (SCEs), chromosomal breaks, interchanges between homologous chromosomes, and sensitivity to several DNA-damaging agents. The single RecQ helicase homologues of (Sgs1; ) and (Rqh1; ) play critical roles in the maintenance of genomic stability. In human cells, five genes encoding RecQ helicase homologues have been identified, and defects in two of them, and , cause Werner's syndrome and Rothmund-Thomson syndrome, respectively, which are disorders associated with genomic instability (; ).
Biochemical analyses indicated that BLM demonstrates G4 DNA unwinding, branch migration, and canonical DNA helicase activity (). More importantly, BLM with DNA topoisomerase IIIα (TOP3α) was shown to resolve double Holliday junctions (HJs) to yield noncrossover products (). BLM interacts physically with several proteins involved in various aspects of DNA metabolism, such as RAD51 and WRN (; ). In addition, BLM is a component of the BRCA1-associated genome surveillance complex (), which contains BRCA1, ATM, MRE11, RAD50, NBS1, MSH2, MSH6, and RFC, and a complex containing the five Fanconi anemia complementation proteins (FANCA, FANCG, FANCC, FANCE, and FANCF), RPA, TOP3α, and BLAP75 (; ). BLAP75 was recently shown to stimulate the dissolution of double HJs by BLM and TOP3α (; ).
Despite the accumulation of biochemical data on the activities and binding partners of BLM, little is known about its biological functions. Recently, we have reported a possible involvement of BLM and TOP3α in the dissolution of sister chromatids during the late stage of DNA replication using corresponding gene-disrupted chicken DT40 cells (). However, in BS cells, the molecular basis of the elevated frequencies of SCE, interchanges between homologous chromosomes, and their sensitivity to several DNA-damaging agents is not well understood. To address the mechanism that confers these BS cell phenotypes and to understand the functions of BLM in the cell, we used chicken DT40 cells to establish and characterize double and triple mutants bearing mutations in other genes involved in various repair pathways, including homologous recombination (, , and ). Based on our data, we discuss BLM function and propose an error-free lesion bypass mechanism that involves XRCC3 and BLM.
To investigate the function of BLM under DNA damage–inducing conditions, we generated double and triple mutants of bearing mutations in genes of the epistasis group (, , and ) that are relevant for homologous recombination, and in genes involved in postreplication repair () and nonhomologous end-joining () using the chicken B cell line DT40 (; , ; ; ). We generated five mutants with gene–disrupted () cells containing human () and transgenes, which can be deleted by activating the Cre recombinase with 4-hydroxy tamoxifen (). Although growth of -complemented cells was slightly slower than that of wild-type cells, the sensitivities of the cells to methyl methanesulfonate (MMS) and UV were indistinguishable from those of wild-type cells (unpublished data). Thus, we considered -complemented cells to be equivalent to wild-type cells. A scheme for the systematic generation of the various double and triple mutants from (“wild-type”) cells is shown in . Gene disruption was confirmed by RT-PCR () and genomic PCR (not depicted). Generation of , , and cells is shown in and described in Materials and methods.
A characteristic feature of BS cells is a high incidence of SCE. A possible explanation for this property is that BLM with TOP3α dissolves double HJs in a manner that does not produce crossovers, and that the BLM defect results in crossovers that are detected as SCE. Double HJs are formed by the activities of proteins involved in homologous recombination such as RAD51 (; ). Thus, we investigated the frequency of SCE in relevant mutant cells. As shown in , the SCE frequency in cells was lower than in cells, and the elevated SCE frequency in (“”) cells was greatly reduced by deletion of (, bottom). We previously reported that disruption of considerably reduces the frequency of SCE in cells (). As expected, the (“”) cells generated in this study showed a lower SCE frequency than cells (). In contrast, disruption of did not affect the increased frequency of SCE in cells ().
We also examined the functional relationship between BLM and TOP3α in the suppression of SCE. As TOP3α-depleted cells exhibit lethality, we previously generated and cells carrying a mouse transgene placed under the control of the doxycyclin-repressible promoter (). The cells ceased to grow within 3 d after the addition of doxycyclin, and they showed an increase in SCE frequency 2 d after the treatment, as similarly observed for cells (). Moreover, the SCE frequency in cells 2 d after doxycyclin addition was almost the same as that of cells, indicating that TOP3α functions with BLM to suppress the formation of SCE. Notably, disruption of did not increase the SCE frequency ().
To identify the pathway in which BLM functions under DNA damage–inducing conditions, we performed colony survival assays of various mutant cells in the presence of MMS. Double mutant and cells were more sensitive to MMS than either single mutant (). The same tendency was also seen in cells.
Interestingly, the MMS sensitivity of cells was partially suppressed by disruption of (, a, and S1 A, available at ). All clones of cells derived from the same parental cells showed almost the same sensitivity to MMS (unpublished data), which excluded the possibility that the suppression of MMS sensitivity was caused by mutations occurring during mutant isolation. Thus, BLM appears to function downstream of XRCC3 under damage-inducing conditions.
It has been reported that cells show synergistic or additive increases in sensitivity to genotoxic agents including MMS, compared with either single mutant (), suggesting that BLM and WRN perform nonoverlapping functions. The MMS sensitivity of cells was higher than that of either single mutant (, b), suggesting that WRN functions independently of XRCC3 in response to MMS-induced damage. However, the MMS sensitivity of triple mutant cells was not higher than that of (“/”) or cells (, c), indicating that BLM and XRCC3 function in the same pathway, even in the background.
In contrast to what we found for , we previously observed that cells show higher sensitivity to genotoxic agents, including MMS (), compared with either single mutant. cells were similarly more MMS sensitive than either single mutant (, a). In , proteins belonging to the RAD52 epistasis group, such as RAD51, RAD54, RAD55, and RAD57, are involved in recombinational repair (). Thus, it is possible that XRCC3, a RAD51 paralogue in higher eukaryotic cells, functions in a recombinational repair pathway involving RAD54. Therefore, we examined the MMS sensitivity of and cells. The MMS sensitivity of the cells was higher than that of either single mutant (, b, and S1 B), and the sensitivity of cells was almost the same as that of cells (, c). Thus, the genetic data obtained here are compatible with the notion that BLM and XRCC3 function in the same DNA repair or damage tolerance pathway after MMS treatment, but probably not in the canonical recombinational repair pathway.
To investigate the functional relationship between BLM and XRCC3, we examined RAD51 focus formation after exposure to MMS (unpublished data). After MMS treatment, an increase in the number of cells exhibiting RAD51 foci was observed in both and cells. However, and cells showed little increase in the number of RAD51 foci after MMS exposure. In contrast, MMS-induced RAD51 focus formation was observed in and cells. These results indicated that RAD51 focus formation does not correlate with survival after exposure to MMS.
To understand the mechanism of the suppression of MMS sensitivity in cells after disruption of , we examined , , and related cell lines for a variety of chromosomal aberrations (the types of chromosome aberrations analyzed are presented in Fig. S2, A and B, available at ). As shown in , the number of chromosomal aberrations increased 12 h after exposure to MMS in all cells examined; cells had a higher number of chromosome aberrations than and cells, and the defect in cells was suppressed by deletion of .
In contrast, a slight increase in chromosomal aberrations was observed in cells compared with either single mutant after exposure to MMS (), indicating that disruption of but not specifically suppresses the defect in cells.
We next focused our attention on chromatid exchanges. Although we observed chromatid exchanges between nonhomologous chromosomes or different regions of homologous chromosomes (Fig. S3 A, available at ), the majority of chromatid exchanges observed after exposure to MMS involved homologous chromosomes, which is a typical feature of chicken DT40 cells (). This type of chromosomal aberration is called a mitotic chiasma because it resembles the chiasma structure seen in meiosis. A slight increase of mitotic chiasmata was observed in cells compared with cells (, a). This phenotype is reminiscent of the increased interchanges between homologous chromosomes in BS cells (). As shown in , MMS-induced mitotic chiasmata in and cells were almost completely suppressed by deletion of (, a), whereas disruption of had no effect on the formation of mitotic chiasmata (b). In chromosomes forming mitotic chiasmata, it is generally held that the events of homologous recombination, but not the separation of the recombinant chromosomes linked by sister chromatid cohesion, have been completed (Fig. S3 B; ; ). If this were the case, RAD54 would also be likely to be required for the formation of mitotic chiasmata. Indeed, RAD54 was required for mitotic chiasma formation in the presence or absence of BLM (unpublished data).
As described in the previous section, we analyzed cells and -related double mutant cells after exposure to MMS. However, MMS induces a variety of DNA lesions including base alkylation and generation of single- and double-strand breaks. Thus, we examined these cells after irradiation with UV light that generates a specific lesion, thymine dimers. We also examined the sensitivity of these cells to x rays because x rays and MMS are reported to induce double-strand breaks. As shown in (a), cells were as sensitive to x rays as cells. The sensitivity of cells to x rays was not higher than that of either single mutant. Note that the x-ray sensitivity of cells was very mild compared with that of cells, which are sensitive to ionizing radiation (). Similar results were previously reported with DT40 cells (; ). However, hamster irsSF cells carrying a mutation in are reportedly sensitive to ionizing radiation (; ). The relatively high rate of recombination in DT40 cells may account for this inconsistency.
In contrast to their ionizing radiation sensitivity, cells were mildly UV sensitive compared with cells, and this sensitivity was suppressed by deletion of to the level of cells (, b). Chromosomal aberrations increased gradually during incubation after UV irradiation (, a). Details of various types of chromosome aberration are shown in Fig. S2 C. Chiasmata began to appear 6 h after UV irradiation and their frequency increased thereafter (, b). This type of chromosomal aberration was increased in cells. Deletion of suppressed various types of chromosome aberrations in cells to the level seen in cells. This especially concerned chromosome-type aberrations that manifest gaps or breaks at the same positions on sister chromatids ( [a] and S2 D). The induction of mitotic chiasmata by UV irradiation in and cells was almost completely suppressed by deletion of (, b).
Cells derived from BS patients exhibit elevated levels of SCE, interchange between homologous chromosomes, and sensitivity to several DNA-damaging agents. In this paper, we performed a systematic genetic analysis of mutant chicken DT40 cells to explore the function of BLM and identify a putative mechanism underlying the phenotype of BS cells.
We demonstrated that and are required for the elevated levels of SCE and MMS- or UV-induced mitotic chiasmata observed in cells. The MMS sensitivity of cells was partially suppressed by disruption of , but not by disruption of , 54, , 18, or 70. The suppression of -associated phenotypes upon disruption of , particularly cell viability and increased chromosomal aberrations, was clearly evident in response to UV irradiation. Thus, the increased frequency of SCE, the sensitivities to MMS and UV, and the elevated frequency of mitotic chiasmata are caused by the function of XRCC3.
Cells were cultured in RPMI 1640 supplemented with 10% fetal bovine serum, 1% chicken serum (Sigma-Aldrich), and 100 μg kanamycin/ml at 39.5°C. For gene targeting, 10 DT40 cells were electroporated with 30 μg of linearized targeting constructs using a Gene Pulser apparatus (Bio-Rad Laboratories) at 550 V and 25 μF. Drug-resistant colonies were selected in 96-well plates. Genomic DNA was isolated from drug-resistant clones. Gene disruption was confirmed by RT-PCR.
For generating double or triple mutants of with mutations in , , and/or , these genes were disrupted in conditional cells as described in . After gene disruption, the expression cassette was excised by Cre recombinase activated by 4-hydroxytamoxifen (). The gene was also disrupted in , , and cells. The cell strains used in this study are summarized in .
Total RNA was isolated using TRIzol (Invitrogen) and converted to cDNA with SuperscriptIII (Invitrogen). A part of each gene was amplified with polymerase. Primers were used to amplify (sense, 5′-GATTTGGATCTACTGGACCTGAATCCCAG-3′; antisense, 5′-GAGCTGCGTCCGGCCAGCTCAGTGATG-3′), (sense, 5′-ACCAGCGTGTGTCTCTGCTG-3′; antisense, 5′-CTACAGATTTTGGAAGGGAAGC-3′), (sense, 5′-CAGTGGAAAGTGATACATTCTGTTTTAGAAGAC-3′; antisense, 5′-CACCTGCAATTATCACAGCACTCTTC-3′), (sense, 5′-CTGGCCAAGAGGAAGGCGGGCGGCGAGGA-3′; antisense, 5′-TTAGGGAATCCCTCGCTGCTCTTCATGGG-3′), (sense, 5′-CGGCTCATACCATGAAGATGTGGG-3′; antisense, 5′-CCTGTTACGAGTTGTCATCTGGTGACG-3′), (sense, 5′-CCAGCAAAATTATTAGTAGTGACAAGGATCTG-3′; antisense, 5′-CTGCATATGGTAGGAAAATGATGTGGAAACC-3′), (sense, 5′-CCCATAACTATTGTTCCCTTTGCATACGG-3′; antisense, 5′-GGGATTTAGAGAATCACACTGAGCATTATACACGTGC-3′), and (sense, 5′-ATGACAGCTGTGGAAGTGCTA-3′; antisense, 5′-TCAGTCAAGAACAACAGGTTGGTCATCTC-3′; as a control) by RT-PCR.
5 × 10 cells were cultured for two cycles in a medium containing 10 μM BrdU and pulsed with 0.1 μg/ml colcemid for 2 h. The cells were harvested and treated with 75 mM KCl for 12 min at room temperature and fixed with methanol-acetic acid (3:1) for 30 min. The cell suspension was dropped onto wet glass slides and air dried. The cells on the slides were incubated with 10 μg/ml Hoechst 33258 stain in phosphate buffer, pH 6.8, for 20 min and rinsed with MacIlvaine solution (164 mM NaHPO and 16 mM citric acid, pH 7.0). The cells were exposed to a black light (352 nm) at a distance of 1 cm for 30 min, incubated in 2× SSC (0.3 M NaCl and 0.03 M sodium citrate) at 58°C for 20 min, and stained with 3% Giemsa solution (Merck) for 25 min.
To determine MMS sensitivity, 4 × 10 cells were inoculated into 60-mm dishes containing various concentrations of MMS in a medium supplemented with 1.5% (wt/vol) methylcellulose, 15% fetal bovine serum, and 1.5% chicken serum. For UV sensitivity, cells were suspended in 1 ml phosphate-buffered saline, inoculated into 6-well plates, and irradiated with various doses of UV. For x-ray sensitivity, cells were suspended in 1 ml phosphate-buffered saline, inoculated into 1.5-ml tubes, and irradiated with various x-ray doses. UV- or x-irradiated cells were inoculated into 60-mm dishes containing a medium supplemented with 1.5% (wt/vol) methylcellulose, 15% fetal bovine serum, and 1.5% chicken serum. Colonies were counted after 7–14 d, and the percent survival was determined relative to the number of colonies of untreated cells. We performed the same survival experiments several times (Fig. S1). After confirming that all the data gave similar results, we presented representative data (Fig. S1, surrounded by red square). In each experiment, we tested each cell genotype in duplicate.
Fig. S1 shows survival curves of various mutant cells exposed to MMS. Fig. S2 shows the classification of chromosomal aberrations. Fig. S3 shows mitotic chiasma. Fig. S4 shows characterization of cells. Fig. S5 is a schematic model of SCE formation. Online supplemental material is available at . |
Control of translation initiation allows a rapid and dynamic cellular response to stress. We have previously shown the dramatic inhibitory impact of stress on translation initiation in yeast (, ; ). This global inhibition has an immediate effect on protein levels and, thus, conserves resources. Additionally, this regulation affords the cell a respite period, allowing a redirection of resources toward stress survival (; ).
Two established regulatory mechanisms target distinct steps in translation initiation. First, the selection of mRNAs can be regulated. eIF4E and Pab1p select mRNA via interaction with the 5′ cap and 3′ poly(A) tail, respectively. eIF4G interacts with both factors, promoting a closed loop messenger RNP (mRNP) complex (). The formation of the closed loop mRNP complex can be inhibited either by eIF4E-binding proteins (4E-BPs) or by eIF4E homologous proteins (4EHPs).
4E-BPs competitively inhibit the eIF4G–eIF4E interaction, thereby preventing translation initiation either in a global or mRNA-specific manner. In higher eukaryotes, inactivation of the mammalian target of rapamycin pathway activates 4E-BPs, leading to a general down-regulation of 5′ cap–dependent protein synthesis (). Alternatively, 4E-BPs such as Maskin in or Cup in are targeted to specific mRNAs by interaction with mRNA-binding proteins (). has two 4E-BPs, Caf20p and Eap1p, which translationally regulate some mRNAs yet are unlikely to act as global translational regulators (; ).
4EHPs have the capacity to interact with the mRNA cap structure but do not interact efficiently with eIF4G (). Therefore, the closed loop mRNP complex may be compromised by 4EHP competition with eIF4E for the 5′ cap structure. For instance, this mechanism explains the translational regulation of mRNA in , in which 4EHP is specifically targeted to the mRNA 3′ untranslated region (). A similar mechanism has recently been proposed to explain the impact of Argonaute proteins on translation initiation. Here, micro-RNAs target Argonaute protein to specific transcripts, where they interact with the 5′ cap structure to inhibit translation initiation ().
A second regulated step in the translation initiation pathway involves activation of the stress-responsive eIF2α kinases. The initiator methionyl tRNA (Met-tRNA) forms a ternary complex (TC) with eIF2-GTP and is recruited to the 40S ribosome. GTP hydrolysis generates eIF2-GDP as a by product of translation initiation, and this is recycled to eIF2-GTP by a guanine nucleotide exchange factor, eIF2B. Phosphorylation of eIF2 by the eIF2α kinases inhibits this recycling to reduce the level of TC, which ultimately limits translation initiation (). The yeast eIF2α kinase Gcn2p responds in this manner to stresses such as amino acid starvation ().
In a variety of experimental systems, pools of mRNA are specifically sequestered into cytoplasmic granules or bodies during periods of translational inactivity (). In germ cells, specific germ cell granules are sites where repressed mRNAs congregate during oogenesis (; ). After stress in mammalian cells, translationally repressed mRNAs assemble into particles called stress granules (SGs; ). In neuronal cells, mRNAs that are in transit to their dendritic site of translation are held in neuronal-specific RNA granules (). Finally, in yeast and mammalian systems, RNA processing bodies (P-bodies) have been defined as cytoplasmic bodies that harbor many of the enzymes involved in the 5′ to 3′ pathway of mRNA decay (; ). A diverse array of functions has been ascribed to these RNA granules, including roles in mRNA localization, degradation, and storage as well as in the micro-RNA pathway ().
P-bodies harbor mRNA along with various components of the mRNA metabolic machinery (). The earliest described P-body components in yeast were the mRNA degradation components Dcp1p, Dcp2p, Dhh1p, Pat1p, Lsm1p, and Xrn1p (). These components are highly conserved and can be thought of as core components. Less well-conserved components include micro-RNA repressor factors, nonsense-mediated decay enzymes (Upf1-3p), and viral factors (for review see ). In yeast, translation factors were described as absent from P-bodies, although eIF4E was found in human P-bodies (; ; ; ).
Translational shut-off induced either genetically or via environmental stress causes P-bodies to increase in both size and abundance. P-bodies return to normal levels after a restoration of translation (; ). It has been suggested that P-bodies represent accumulations of translationally silent mRNPs undergoing degradation or storage (). A decrease in global translation would increase free mRNP concentration and, thus, increase flux into P-bodies. Mutation of mRNA degradation and P-body components (such as Dcp1p and Xrn1p) also increases P-body size and abundance, in this case inhibiting degradative flux through the P-body (, ).
An inhibition of translation initiation by eIF2α phosphorylation induces SGs in higher eukaryotes but not in yeast (). SGs harbor ribosomal proteins and early translation initiation factors such as eIF2, eIF3, eIF4E, and eIF4G. Therefore, they are thought to contain abortive translation initiation complexes (). SGs are compositionally and functionally distinct from P-bodies, yet several factors populate both structures in vivo (; ). It has been suggested that P-bodies dock with SGs and degrade designated mRNAs, whereas the SGs remain a site of mRNA triage ().
In this study, we provide important insight into mRNP fate in translationally repressed cells. This work highlights which step in the yeast translation initiation pathway is inhibited by glucose depletion. Moreover, we show that this translational shutdown causes eIF4E, eIF4G, and Pab1p to redistribute away from ribosomal subunits to cytoplasmic granules. Quantitative analyses of dual-tagged strains indicate a partial colocalization with P-body components. However, a substantial population of the translation factor granules do not contain core P-body components. Critically, de novo formation of such granules demonstrates a kinetic and spatial distinction from P-bodies. We term these granules EGP-bodies and propose that they are sites where mRNAs are stored during periods of translational inactivity. This would suggest that EGP-bodies share at least functional analogy with higher eukaryotic SGs.
Glucose depletion causes a severe but reversible inhibition of translation initiation by an unidentified mechanism (). Formaldehyde cross-linking of cells before polysome analysis has been used to provide mechanistic insight into translational regulation (). Using this methodology, we compared glucose starvation with amino acid starvation, a stress that inhibits translation initiation by activating the eIF2α kinase Gcn2p to give lower TC levels (). Importantly, the abundance of the translation factors tested in this analysis is not appreciably altered by glucose starvation (Fig. S1, available at ). Both glucose and amino acid starvation cause an accumulation of 80S monosomes via polysome run-off (). This is characteristic of the inhibition of translation initiation () and is accompanied by a movement of the ribosomal proteins Rps3p and Rpl35p toward the 80S fractions ().
Both eIF3 and eIF2 interact with the 40S ribosomal subunit in a 43S complex before mRNA recruitment () and are therefore detected in 40S fractions (). Additional protein is detected in polysomal regions and likely represents the formation of initiation complexes on heavily translated polysome-bound mRNAs. After amino acid starvation, a reduced level of eIF2α was observed cosedimenting with the 40S subunit (). This is consistent with the translational consequence of amino acid starvation: reduced TC levels, which would ultimately produce 40S ribosomal subunits with decreased associated eIF2 (). In contrast, eIF3 is still associated with the 40S subunit after amino acid starvation (). eIF3 is also maintained on the 40S ribosome after eIF2 depletion (). Therefore, it seems that a partial 43S complex can still form when eIF2 function has been abrogated.
After glucose starvation, we observed maintained or even slightly increased levels of eIF3 and eIF2 with the 40S ribosomal subunit accompanied by decreases in polysomal fractions (). Therefore, in contrast to amino acid starvation, glucose starvation does not decrease the level of eIF2 with the 40S ribosomal subunit. Any increase in eIF2 and eIF3 with the 40S subunit after glucose starvation is likely to result from the stress-induced polysome run-off. This causes polysomal 43S complexes in unstressed cells to relocate to the 40S region.
The most striking observation from the polysome analysis is that after glucose starvation, there is a marked increase in the levels of eIF4G and Pab1p sedimenting in submonosomal fractions (, fractions 1–3; +/− glucose). Similarly, the proportion of eIF4E in these fractions becomes enriched, although submonosomal eIF4E in glucose replete extracts somewhat masks this effect (). This demonstrates a reduced association of these factors with ribosomes and contrasts noticeably with the sedimentation of these factors after amino acid starvation (). Therefore, the key translational components present in the closed loop mRNP appear to resediment away from ribosomal fractions after glucose starvation. Given that there is a low affinity interaction between eIF4A and eIF4G in yeast () and that eIF4A is one of the most abundant translation initiation factors (present at roughly 40 times the number of eIF4G molecules; ), it is unsurprising that the sedimentation pattern of eIF4A remains submonosomal regardless of the stress condition (). Even if a weak eIF4A–eIF4G interaction persisted after stress, the considerable level of free eIF4A would mask changes in ribosome-associated material.
Overall, these data suggest a reorganization of the mRNP closed loop translation complex after glucose starvation, whereby the cosedimentation of eIF4E, eIF4G, and Pab1p with ribosomal complexes is compromised. Amino acid starvation produces a contrasting effect, which is consistent with a fundamentally different mode of action. Indeed, the observed decrease in eIF2 cosedimentation with the 40S ribosomal subunit agrees with the established model for amino acid starvation.
As part of our analysis, we assessed the localization of translation initiation factors in response to both glucose and amino acid starvation. We show that glucose starvation induces genomic GFP fusions of eIF4E, eIF4G1, eIF4G2, and Pab1p to accumulate as cytoplasmic granules (). In contrast, the localization of eIF3b, eIF4AI, eIF2α, or eIF2Bγ is unaffected by glucose starvation. Previously, we have shown that eIF2α and eIF2Bγ localize to a large cytoplasmic focus (). Irrespective of the stress condition, we still observe this defined localization.
Strikingly, those factors forming stress-induced granules (eIF4E, eIF4G1/2, and Pab1p) are identical to those with altered gradient sedimentation properties (). The translation initiation factor granules occur with an approximate frequency of four to five granules per cell (see ). In correlation with the cosedimentation analysis, amino acid starvation failed to induce cytoplasmic granules for any of these factors after 30 min ().
It is intriguing that glucose starvation causes eIF4E, eIF4G, and Pab1p to relocalize to cytoplasmic bodies, whereas amino acid starvation does not elicit this response or induce the marked resedimentation of eIF4E, eIF4G, and Pab1p. We attribute this to either variation in the severity of the imposed stresses or fundamental differences in their modes of action. Interestingly, an exposure of cells to 10 min of glucose starvation induced a shift in eIF4E, eIF4G, and Pab1p of lower magnitude than that observed for cell extracts from cells starved for 30 min. In accordance with this observation, 10 min of glucose starvation also failed to induce cytoplasmic granules of eIF4E, eIF4G, and Pab1p (Fig. S2, available at ). Collectively, these data imply a requirement for the large-scale release of eIF4E, eIF4G, and Pab1p from polysomes to induce visible cytoplasmic aggregations. Short periods of glucose starvation are insufficient to induce such an effect as a result of an apparent lag in factor release. Amino acid starvation induces minimal eIF4E, eIF4G, and Pab1p resedimentation and aggregation into cytoplasmic granules after either 30 or 60 min (, , and S2). These data further corroborate the fundamental difference between these nutritional stresses and their mechanism of inhibition of translation initiation.
An obvious question is whether eIF4E, eIF4G, and Pab1p localize to the same cytoplasmic bodies after glucose starvation. To investigate this, we generated dual-tagged strains bearing specific genomic GFP and RFP fusions. As shown in , we compared eIF4G1-GFP with eIF4E-RFP, eIF4G1-GFP with Pab1p-RFP, and eIF4E-GFP with Pab1p-RFP. For all of these experiments, we observed >90% colocalization ( and see ).
As eIF4E, eIF4G, and Pab1p colocalize to the same cytoplasmic granules after glucose starvation, a key question is whether they relocate to these granules as mRNPs (i.e., bound to mRNA). To test the ability of eIF4E, eIF4G, and Pab1p to interact with each other as well as the cap and poly(A) tail structures after glucose starvation, we used cap and poly(A) affinity chromatography approaches (). The levels of eIF4E and copurifying eIF4G on the cap affinity column remain largely unchanged in extracts from glucose-starved or unstarved cells (, left). Similarly, glucose starvation does not alter the level of Pab1p, eIF4G, or eIF4E associated with the poly(A) column (, right). Therefore, the interaction of these factors with mRNA is likely to persist after glucose starvation. We also used a strategy that enables a specific mRNA to be followed in live cells (; ). Here, the mRNA has multiple U1A-binding sites inserted into the 3′ untranslated region and can be visualized using fluorescence microscopy when coexpressed with a plasmid bearing a U1A-GFP fusion. When this mRNA localization system is used during the translational repression induced by glucose starvation, transcripts formed distinct cytoplasmic GFP foci, which precisely colocalize with the eIF4E-RFP granules (, top). Controls using either the reporter plasmid or the U1A-GFP fusion plasmid in isolation generate no GFP foci.
These results suggest that the dramatic effect of glucose starvation on translation initiation does not result from a reduced capacity to form mRNA closed loop complexes. Furthermore, the combination of live cell imaging and affinity chromatography experiments suggests that mRNA accompanies eIF4E, eIF4G, and Pab1p to cytoplasmic granules. Thus, it appears that there is a relocalization of closed loop mRNP complexes to cytoplasmic granules after severe translational repression.
After stresses such as glucose starvation, mRNA can become localized to P-bodies (). Originally in yeast, translation initiation factors were not identified in P-bodies (; ). However, while this manuscript was in preparation, this question was reevaluated, and the authors now suggest that eIF4E, eIF4G, and Pab1p do enter P-bodies (). In our experiments, we also investigated the relationship between P-bodies and the granules containing eIF4E, eIF4G1, and Pab1p. We simultaneously visualized eIF4E-, eIF4G1-, or Pab1p-RFP versus Dcp1p-GFP as a marker for P-bodies in living cells. After a 30-min glucose depletion, eIF4E, eIF4G1, and Pab1p were found in most Dcp1p-containing bodies (). In addition, Dcp1p cosediments with these translation factors across sucrose gradients, and there is an increase in eIF4G, eIF4E, and Pab1p coimmunoprecipitation with Dcp1p in glucose-starved versus unstarved extracts (Fig. S3, available at ). Strikingly, however, Dcp1p was not always found in the translation initiation factor granules (). Furthermore, we identified a mean of four to five granules per cell for each of the translation initiation factors, which is approximately twice that observed for Dcp1p (). This value for Dcp1p correlates well with the two to three granules per cell previously published for another core P-body component, Dhh1p ().
Overall, these data suggest that after glucose depletion for 30 min, a mean of two to three P-bodies form per cell, which also contain eIF4E, eIF4G1, and Pab1p. Intriguingly, a mean of approximately two additional bodies form per cell that contain eIF4E, eIF4G1, and Pab1p but do not harbor Dcp1p. Similar results were obtained when a second marker for P-bodies, Dcp2p, was used ( and not depicted). This is consistent with observations that Dcp1p colocalization with mRNA is incomplete in that mRNA granules exist that do not contain Dcp1p (, bottom). Overall, these results suggest that previously unrecognized cytoplasmic bodies exist that contain the translation initiation factors eIF4E, eIF4G, and Pab1p but do not contain the key P-body markers Dcp1p and Dcp2p. We have termed these bodies EGP-bodies, as thus far their only known protein constituents are eIF4E, eIF4G, and Pab1p.
Distinct possibilities exist as to the mode of EGP-body formation, which have important functional implications. First, EGP-bodies may form via a P-body maturation process in which core P-body components such as Dcp1p and Dcp2p are lost. Alternatively, EGP-bodies could arise de novo having never contained core P-body components.
To investigate these possibilities, we undertook detailed time course experiments following individual cells after glucose starvation, simultaneously visualizing eIF4E-RFP as a marker for EGP-bodies and Dcp1p-GFP as a marker for P-bodies (). The complex experimental setup dictates that the earliest possible time point for image acquisition is 10 min after glucose starvation. At this time point, it can be seen that in accordance with previously published data, Dcp1p has accumulated in P-bodies (; ). In contrast, eIF4E enters P-bodies after 20–25 min of glucose starvation, and, ultimately, most P-bodies recruit eIF4E (). Perhaps the most striking observation from these time course experiments concerns the EGP-bodies. The EGP-bodies harboring just eIF4E and not Dcp1p always arise spontaneously in the cell and have never been observed to accumulate as a result of the loss of Dcp1p from eIF4E-containing P-bodies (). This defined population of de novo EGP-bodies is quantified in , in which there is a gradual increase in the formation of EGP-bodies starting after 20–25 min. Overall, therefore, we conclude that P-bodies recruit eIF4E after prolonged glucose starvation but do not mature into EGP-bodies by losing core P-body components. In fact, EGP-bodies arise spontaneously as independent entities within the cell.
Glucose starvation causes the most severe stress-induced inhibition of translation initiation in that has yet been characterized (). Paradoxically, the mechanism for such a catastrophic alteration in protein synthesis is unknown. In this study, we demonstrate that the mode of regulation is distinct from the well-studied impact of amino acid starvation on TC levels. We also show that the capacity of different translation initiation factors to recognize the mRNA is unaltered after glucose starvation. Indeed, the mRNA and a core of mRNA-interacting translation initiation factors relocalize to cytoplasmic granules as a consequence of this stress.
Glucose starvation also does not alter the level of eIF2 or eIF3 associated with the 40S ribosomal subunit. Accordingly, mutants that are incapable of phosphorylating eIF2α and so do not elicit an amino acid starvation response still respond to glucose starvation to give wholesale translational repression (). Equally, glucose starvation does not rapidly induce established responses to amino acid starvation such as eIF2α phosphorylation or derepression (Fig. S1 and not depicted). Therefore, our data are inconsistent with a model in which glucose regulates translation by altering TC levels or affecting the interaction of eIF2 or eIF3 with the 40S ribosomal subunit. Regulation at steps further downstream in the translation initiation pathway can be ruled out, as the closed loop mRNP relocalizes away from the ribosomal pool to cytoplasmic granules. This pinpoints recruitment of the 43S complex to mRNA as the regulated step in the translation pathway after glucose starvation. There are precedents for regulatory mechanisms targeting this step. For instance, positioning of the RNA cis-acting iron response element to a cap-proximal position sterically prevents 43S complex binding to mRNA (; ). More recently, phospo-L13a has been shown to inhibit 43S subunit recruitment to γ interferon inhibitor of translation element–containing mRNAs (). In these examples, the inhibition of translation initiation is exerted at the level of a specific mRNA, whereas glucose starvation elicits a global translational repression (), and, thus, the regulatory mechanism studied here represents a new means by which to dramatically attenuate protein synthesis.
In this study, we also show that eIF4E, eIF4G, and Pab1p localize to P-bodies after glucose starvation. Previously, mRNA has been shown to enter P-bodies during translational shutoff, where it can either be degraded by virtue of the mRNA decay factors present within the P-body or stored (; ). The mechanism by which mRNA could be stored in the presence of high local concentrations of mRNA decay factors has yet to established. The factors eIF4E, eIF4G, and Pab1p could be viewed as the minimal requirement to protect the mRNA 5′ and 3′ ends from degradation (; ). We observe a recruitment of eIF4E to P-bodies after ∼25 min of glucose starvation, concomitant with resedimentation of eIF4E, eIF4G, and Pab1p away from ribosomal fractions of a sucrose density gradient. This is consistent with a model in which closed loop mRNPs are recruited to P-bodies in a manner kinetically distinct from the translation inhibition and P-body induction that occurs in <10 min.
In contrast, after extended periods of translational inhibition induced by amino acid starvation, we observe little or no P-body induction and no increase in mRNP relocalization. Once again, this highlights fundamental differences in the impact of the glucose and amino acid starvation stresses on translation initiation. Interestingly, in the case of amino acid starvation, it is particularly important for cellular adaptation that translation initiation events still occur on mRNAs such as (). This may provide a physiological explanation as to why mRNA does not appear to be at liberty to transit through P-bodies. That is, translation initiation events can still occur on mRNA after this stress even though these may be partial or nonproductive in many cases.
After glucose starvation, we show that there is a kinetic separation of P-body formation and closed loop mRNP redistribution. This poses interesting questions as to the relationship between these two events. Cycloheximide prevents P-body accumulation by freezing mRNAs in polysomes (). This denotes a polysomal origin for mRNAs that enter P-bodies. The resedimentation and relocalization kinetics of eIF4E, eIF4G, and Pab1p suggest that after glucose starvation, several mRNAs are retained in polysomes as closed loop mRNPs, whereas others are relocalized to P-bodies without associated eIF4E, eIF4G, and Pab1p. The mechanism by which the cell distinguishes between these populations or initiates the secondary redistribution of closed loop mRNPs requires further study.
As well as P-body–localized eIF4E, eIF4G, and Pab1p, we observe cytoplasmic bodies bearing these factors that do not contain the key decapping enzymes (and P-body markers) Dcp1p and Dcp2p. These EGP-bodies may be sites for the long-term storage of mRNA and translation initiation factors under specific stress conditions (). Furthermore, as translation initiation factors remain associated with mRNAs under conditions of translational repression in EGP-bodies, the rapid resumption of translation initiation on these mRNAs would be favored should conditions improve (). Thus, although lacking in 43S complex components, these EGP-bodies could be viewed as at least functionally analogous to the SGs that are found in higher cells (). A particularly intriguing possibility is that gene regulation might occur when mRNA fate is decided such that transcripts that are still valuable to the cell can be stored in EGP-bodies, whereas those that are not required would be targeted to P-bodies. The closed loop complex is perceived to be associated with high levels of translation initiation and, therefore, is a target for translational regulation (). However, the aforementioned results are particularly striking, as they suggest that the closed loop complex exists on mRNAs that are translationally repressed. Therefore, as well as identifying a new strategy for translational regulation, these results challenge the notion that the formation of a closed loop complex predisposes an mRNA to translation initiation.
Strain genotypes are listed in the Table S1 (available at ). Proteins were C-terminally tagged and verified by both PCR and immunoblotting using a previously described method (), and plasmid reagents were provided by G. Pereira (German Cancer Research Centre, Heidelberg, Germany). Anti-Rps3p and Rpl35p were gifts from M. Pool (University of Manchester, Manchester, UK), and anti-eIF2α and eIF2Bδ were gifts from G. Pavitt (University of Manchester, Manchester, UK). Anti-eIF3a was generated against an in vitro–synthesized peptide (Genosphere). The 2μ RNA localization construct plasmids pPS2037 () and pRP1187 () were gifts from R. Parker (Howard Hughes Medical Institute, University of Arizona, Tucson, AZ).
Cells were grown to OD 0.6 in rich media containing yeast extract, peptone, and glucose (YPD; ). Cells were harvested, and, for starvation experiments, cell pellets were resuspended in 50 ml of either YPD or YPD lacking glucose (YP) and incubated for the indicated times at 30°C. Extracts were prepared by centrifugation and immediate freezing of cell pellets in liquid N before hand grinding to a powder. The powder was resuspended in buffer A (2 mM MgOAc and 30 mM Hepes, pH 7.5, 100 mM KOAc, 1 mM PMSF, and 1× Complete Mini-EDTA free protease inhibitors [Roche]) and centrifuged for 10 min at 800 . Supernatant was isolated, and protein content was assayed by Bradford reagent assay (Bio-Rad Laboratories). Western blotting was conducted as described previously ().
For cap and poly(A) affinity chromatography assays, 3 mg of protein was preincubated with 4B Sepharose (GE Healthcare) for 30 min before incubation with 25 μl 7-methyl-GTP Sepharose (GE Healthcare) or poly(A) Sepharose (GE Healthcare) for 2 h. Beads were washed three times in buffer A before elution in 50 μl Laemmli buffer at 95°C for 2 min. In the case of the cap-binding assay, an additional 1-h wash step in the presence of 0.1 mM GTP was conducted. For Western analysis, 10 μl of the input was loaded for comparison with 10 μl of pellet fractions.
Cells were grown as described in Strain growth and extract preparation and were fixed by the addition of 1% (vol/vol) formaldehyde to media. They were incubated on ice for 1 h before the addition of 0.1 M glycine and lysed with glass beads. 7.5 OD U of extract was loaded onto 15–50% sucrose gradients and centrifuged at 40,000 for 2.5 h. 500-μl gradient fractions were collected while the Abs was continuously measured. Individual fractions were precipitated with 10% TCA, washed with 300 μl acetone, and resuspended in 50 μl Laemmli buffer. Proteins were analyzed by SDS-PAGE and immunoblotting.
Cells were grown to an OD of 0.6 in synthetic complete glucose media (SCD), washed twice, and resuspended in media lacking either glucose (SC) or amino acids (SCD-AA; ). After 30 min of incubation, cells were observed. Confocal images were taken at room temperature by a confocal microscope (SP5; Leica) using a 63× 0.6–1.40 NA plan Apo oil objective (Leica). Images were acquired using Application suite 1.6.3 (Leica). For densitometric analysis, a merged 12-image z-series was taken; for clarity, the images shown are single planes. Dual-label microscopy was conducted at room temperature using a microscope (Eclipse E600; Nikon) with a 100× 0.5–1.3 NA planFluor oil immersion objective (Nikon) and camera (Axiocam MRm; Carl Zeiss MicroImaging, Inc.). Images were acquired using Axiovision 4.5 software (Carl Zeiss MicroImaging, Inc.). Representative cells are shown from experiments repeated at least three times. Images were quantified using ImageJ (National Institutes of Health). >10 cells were analyzed for densitometric analysis of factor abundance in the granules. For counts of granules/cell, >75 cells were analyzed. For the colocalization scoring, three biological replicates, each with 50 cells, were used.
Table S1 details yeast strains used in this study and their genotypes. Fig. S1 shows that the abundance of translation factors and eIF2α phosphorylation are unaffected by glucose starvation for periods of at least 45 min. Fig. S2 shows by sucrose density gradient sedimentation analysis and confocal microscopy that after either 10 min of glucose starvation or 1 h of amino acid starvation, Pab1p, eIF4E, and eIF4G are not induced to resediment or form cytoplasmic bodies. Fig. S3 shows sucrose density gradient sedimentation analysis of Dcp1p, which is concentrated in submonosomal fractions (A). It also shows an induction of eIF4G1, eIF4E, and Pab1p association with Dcp1p-TAP (tandem affinity purification epitope) after glucose starvation by Western blotting against tandem affinity purification outputs (B). Online supplemental material is available at . |
In eukaryotic cells, most secretory and membrane proteins are folded and assembled in the ER. Impairment of this process is collectively called ER stress. Because accumulation of unfolded proteins is harmful to cells, defensive mechanisms against ER stress exist. The type I transmembrane protein Ire1 is conserved in fungi, animals, and plants and plays a central role in the unfolded protein response (UPR). Ire1 has protein kinase and RNase domains in the cytosolic region (; ; ). ER stress causes autophosphorylation of Ire1, which is followed by its activation as an RNase (; ). In budding yeast , Ire1 contributes to unconventional RNA splicing that converts the precursor form of mRNA (HAC1) to the mature form (HAC1; ). HAC1 is effectively translated into the transcription factor protein Hac1, which regulates a wide variety of genes to alleviate ER stress (; ). In contrast to yeast cells, which have only one known ER stress-sensing protein (Ire1), mammalian cells carry multiple ER stress sensors. There are two Ire1 paralogues, IREα and IREβ, as well as a second type I transmembrane protein, pancreatic ER kinase (PERK), which attenuates protein translation by phosphorylation of eukaryotic translation initiation factor 2α (). A third system is the membrane-anchored transcription factor ATF6. Upon ER stress, ATF6 is solubilized by proteolysis and up-regulates ER chaperone and other genes ().
Because the luminal domains of Ire1 and PERK show moderate amino acid sequence similarity, we believe that they sense ER stress by the same mechanism. One breakthrough finding toward understanding the ER stress-sensing mechanism was the observation that BiP binds to Ire1 and dissociates in response to ER stress (; ). It is highly likely that BiP binding negatively regulates Ire1 because activity of Ire1 is considerably attenuated in yeast BiP mutants in which dissociation of the mutated BiP proteins from Ire1 is impaired (, ). However, as described in the next paragraph, dissociation of BiP from Ire1 is not sufficient for activation of Ire1.
We previously predicted that the luminal domain of yeast Ire1 is composed of five subregions (I–V; ). This is because our systematic mutagenesis analysis demonstrated that 10 aa deletions in subregions II and IV, but not in subregions I, III, or V, inactivate Ire1 (). Our speculation that subregions I and V are loosely folded is supported by the finding that these subregions are highly sensitive to proteolysis in the context of a recombinant Ire1 luminal domain protein (). According to the crystal structure presented by , subregions II–IV form one tightly folded domain, which we termed the core stress-sensing region (CSSR). Both this crystal structure analysis and our systematic mutational analysis suggest that subregion III exists as a flexible stretch sticking out from the CSSR. The BiP binding site is located in subregion V (). Importantly, an Ire1 mutant that contains a deletion of almost all of subregion V (hereafter called ΔV; see for the position) is not constitutively active, but is regulated by ER stress as well as wild-type Ire1, even though BiP does not bind to this mutant ().
The crystal structure also indicates that a CSSR dimer forms a major histocompatibility complex (MHC)–like groove (). Analagous to the MHC, peptide fragments and, more speculatively, unfolded proteins may be captured by this groove. An idea emerging from these observations is that in addition to regulation by BiP, the CSSR, as its name denotes, directly senses unfolded proteins and regulates Ire1. However, based on the crystal structure of the mammalian IRE1α luminal domain, argued that this groove is not suited to capture unfolded proteins. Moreover, no biochemical evidence has been provided for direct binding of the CSSR to unfolded proteins.
The oligomerization status of Ire1 is also enigmatic (). Epitope-tagged Ire1 coimmunoprecipitates with differently epitope-tagged Ire1 from lysate of ER-stressed cells, but only to a small extent from that of nonstressed cells (). We reported recently, from coimmunoprecipitation analysis, that Ire1 seems to be fully self-associated, even in the absence of ER stress, when it carries both the ΔV mutation and another deletion of the N-terminal three-quarters of subregion I (hereafter called ΔI; see for the position; ). This finding suggests that the self-association is regulated both by subregion V (probably by binding and dissociation of BiP) and subregion I. Notably, activation of this ΔIΔV mutant Ire1 was still dependent on ER stress (although ΔIΔV Ire1, whose luminal domain consists of almost only the CSSR, was named “core mutant” in our previous paper, we do not use this name in the present paper because it can be confused with “a mutation in the CSSR”; ). According to density gradient fractionation of cell lysates, it seems that the oligomerization status of Ire1, upon activation by ER stress or when carrying the ΔIΔV mutation, is dimeric (; ). Furthermore, recombinant Ire1 luminal domain or CSSR exists in dimeric form in solution (; ). However, the crystal structure suggests that the CSSR forms higher order oligomers because one CSSR molecule is in contact with two CSSR molecules via different interfaces (). The authors argue that this higher order oligomerization is crucial for activation of Ire1 because point mutations that are deduced to deform either interface considerably weaken Ire1 activity. Nevertheless, formation of such higher order oligomers in vivo has not been demonstrated.
One scenario proposed by and modified by us () explains these confusing observations. In this explanation, a combination of BiP dissociation and release from a negative regulation by subregion I lead Ire1 to dimerize. Unfolded proteins bind and may tether dimerized CSSRs, which results in highly oligomerized and active Ire1. Because the higher order oligomer is unstable, it cannot be detected in cell lysates or in solutions of recombinant Ire1 fragments. In the present study, we provide some lines of evidence that in vivo Ire1 actually forms higher order oligomers, here called clusters, and interacts with unfolded proteins. Furthermore, we describe a new scenario in which these two events are positioned differently from the aforementioned scenario.
We recently reported Ire1 to be constitutively activated by a combination of ΔIΔV deletion and Ser103 to Pro (S103P) point mutations (see for the positions and its legend for the amino acid numbering; ). Here, we checked cellular localization of Ire1 and its mutants by immunofluorescent staining of C-terminally HA-tagged molecules. As shown in , HA- tagged wild-type Ire1 (Ire1-HA) demonstrated an ER-like staining pattern in nonstressed cells, whereas the S103PΔIΔV mutation changed localization of Ire1-HA to a clumped distribution. A dotlike distribution, hereafter called clusters, was also observed in wild-type Ire1-HA cells exposed to ER stress by treatment with tunicamycin (Tun) or DTT, although there was also residual ER-like distribution (). To our knowledge, this paper is the first demonstration of a localization change of Ire1.
In this immunostaining analysis and in the immunoprecipitation analyses shown in , , , and , we mainly used cells expressing Ire1-HA from multicopy plasmids. It should be noted that, as shown in Fig. S1 (available at ), this multicopy expression of wild-type Ire1-HA alone did not cause activation of this molecule. As expression level of Ire1 is reported to be low, we could not detect a meaningful HA signal from nonstressed cells containing the centromeric plasmid-borne wild-type Ire1-HA gene (, compare wild type [WT] to the vector control). However, probably because of assembly of several HA epitopes, cluster formation of S103PΔIΔV Ire1-HA in nonstressed cells and of wild-type Ire1-HA in ER-stressed cells was observed even with expression from centromeric plasmids (, S103PΔIΔV and WT [Tun]), although this fluorescent signal was faint. This finding indicates that the cluster formation is not an artifact caused by high expression from multicopy plasmids. Furthermore, we constructed an Ire1-HA gene knockin strain, here named YKY2005. Western blot detection of Ire1-HA showed that the expression level of Ire1-HA in YKY2005 cells is slightly higher than that in Δ cells in which Ire1-HA is expressed from a centromeric plasmid (Fig. S2). As expected, clusters of Ire1-HA were observed in YKY2005 cells when they were exposed to ER stress ().
Clustered Ire1-HA in DTT-treated cells quickly reverted to the normal ER-like distribution when the cells were incubated under nonstressed conditions (). This observation suggests that the cluster is not a simple aggregate of the protein.
Cluster formation of S103PΔIΔV Ire1-HA was observed even in cells carrying the Δ mutation () or when a kinase-inactive mutation, K702A, was introduced (). Thus, activation of either Ire1 or the UPR signaling pathway (Ire1- pathway) is not a prerequisite for cluster formation. For cluster formation, the combination of the ΔI and ΔV mutations was necessary and sufficient, whereas the S103P mutation did not contribute (). Total cellular amount of Ire1-HA did not change considerably upon introduction of any of these mutations (), and thus expression level is not an important determinant of the localization.
As presented in electron micrographs (), cells with multicopy expression of S103PΔIΔV Ire1-HA often showed abnormally folded cisternae that were probably derived from the ER. Importantly, immunogold labeling of anti-HA antibody demonstrated that the epitope was located on these membranous structures. These observations strongly suggest that Ire1 clusters on the membrane of the ER, which is deformed by the cluster formation. When the cluster is smaller, as in the case of single copy expression of Ire1-HA, the immunogold signals of the Ire1-HA clusters and the abnormal membranous structures are likely to be more difficult to find. Thus it is unclear whether cluster formation of endogenous Ire1 under ER-stress conditions leads to such morphological changes of the ER.
Many of the Ire1-HA clusters did not overlap with the DAPI signal (), indicating that the clustering is most likely occurring in regions that do not abut the nucleus. In , cells expressing both S103PΔIΔV Ire1-HA and ER marker protein GFP-Sec12 () were doubly stained with anti-HA and anti-GFP antibodies and observed using an ApoTome-based optical sectioning system. Most of the clusters (>90%) localized on or touched the anti-GFP–stained area, suggesting again that the clusters are formed in the ER. It is likely that not all of the clusters exactly localize on the anti-GFP–stained area because they may exclude other membrane proteins including GFP-Sec12. As shown in , the clusters did not colocalize with Golgi marker protein Mnn9.
According to the crystal structure (), two CSSR molecules contact via two different interfaces (I and II; see ), which suggests a possibility for formation of higher order oligomers. The crystal structure also predicts that the point mutations of Ire1, F247A and W426A (see for the positions), respectively deform interfaces I and II. As shown in , both the F247A and the W426A mutations abolished cluster formation, although F247A Ire1-HA clustered in ∼20% of total cells treated with Tun. This finding strongly suggests that the cluster formation is caused by the high order oligomerization of the CSSR.
Notably, our present observations provide the first evidence for in vivo high order oligomerization of Ire1, which, as described in the Introduction, is not observed in analyses of cell lysates. One explanation for this discrepancy is that the homomeric association via interface I or II is too weak to maintain the cluster structure after lysis of cells. When ΔIΔV Ire1-HA is coexpressed with a C-terminally FLAG-tagged version (Ire1-FLAG) of the same mutant, the FLAG-tagged version is efficiently coimmunoprecipitated with the ΔIΔV Ire1-HA from the cell lysate even in the absence of ER stress (; ). The F247A mutation, but not the W426A mutation, considerably reduced the level of coimmunoprecipitation (). This finding suggests low affinity of the association via interface II.
In Fig. S3 (available at ), cells were treated with the noncleavable cross-linking reagent disuccinimidyl suberate, and the lysate was analyzed by anti-HA Western blotting. The amount of ΔIΔV Ire1-HA migrating to the original position was significantly decreased by treatment of cells with disuccinimidyl suberate, whereas wild-type Ire1-HA was less affected. The smear signal produced by high molecular mass protein is more intense in lane 4 than in lane 2 (Fig. S3), although highly cross-linked proteins may be harder to detect for several reasons, including extremely low mobility in the gel. Altogether, this finding supports highly efficient cross-linking and thus high order oligomerization of ΔIΔV Ire1-HA.
We then monitored activity of the Ire1 mutants by using a UPR element (UPRE)-reporter construct, from which expression of β-galactosidase is driven by a UPRE (; ). The activity of ΔI Ire1 and ΔV Ire1 was as tightly regulated as that of wild-type Ire1 (, –4), whereas ΔIΔV Ire1 was slightly activated even in nonstressed cells (, ; ). Because extrinsic ER stress was required, even for full activation of ΔIΔV Ire1, which was constitutively clustered, we think that cluster formation is not sufficient for full activation of Ire1.
When either of the clustering-impaired mutations, F247A or W426A, was introduced into ΔIΔV Ire1 (, and ) or wild-type Ire1 (), the activity was considerably compromised. This finding strongly suggests that cluster formation is a prerequisite of Ire1 activation. Although the W426A mutation has a stronger negative effect on activity and cluster formation than the F247A mutation ( and ), coimmunoprecipitation of Ire1-FLAG with Ire1-HA was considerably impaired only by the F247A mutation (), suggesting again that a kind of homomeric association of Ire1 is not detected by the coimmunoprecipitation analysis.
A deletion of subregion III (hereafter called ΔIII; see for position) significantly reduced activity of wild-type and ΔIΔV Ire1 (, and ). This finding was further confirmed by directly checking splicing of mRNA (, lane 10). It should be noted that the ΔIII mutation did not impair cluster formation either by extrinsic ER stress or by the ΔIΔV mutation (), nor did it reduce cellular expression level of Ire1 (). These observations strongly suggest that in addition to cluster formation (see Step 1 in ), the CSSR is responsible for another step, hereafter called Step 2 (see ), which is also necessary for Ire1 activation.
To address the involvement of the transmembrane domain in the activity of Ire1, we performed mutation scanning of this domain. ΔIΔV Ire1 was mutagenized such that it had 4 aa serial deletions, from aa 527–530 to 567–570 (see for the positions), and was subjected to the UPRE-lacZ reporter assay, which showed that none of these mutations inactivate Ire1 (). On the contrary, ΔIΔV Ire1 was activated by the deletion of 4 aa residues, LLSK, located at the cytosolic end of the transmembrane domain even in the absence of ER stress (, 12; and see for the position). This mutation, hereafter called ΔLLSK, did not significantly alter activity of wild-type Ire1 ( and F[lanes 6 and 13]). However, the impaired- activation phenotype of the ΔIII mutation was suppressed by the ΔLLSK mutation ( and F [compare lane 14 to 10]). Shift of mobility on SDS-PAGE by Endo H digestion indicates that ΔLLSK Ire1, as well as wild-type Ire1, is N-glycosylated (). We also checked whether wild-type and ΔLLSK Ire1 could be extracted from the yeast microsome fraction by various reagents (). Although both of the Ire1 variants were resistant to extraction with sodium chloride and sodium carbonate, they were partially extracted by Triton X-100. Together with ER localization (, bottom left) and activity of ΔLLSK Ire1, these observations show that the ΔLLSK mutation does not prevent Ire1 from being an ER-located type I transmembrane protein, although the LLSK residues may be a part of the transmembrane domain. Cluster formation of Ire1 was not affected by the ΔLLSK mutation (). Collectively, our results demonstrate that the ΔLLSK mutation abolishes the requirement of Step 2 for activation of Ire1, but not the requirement of the cluster formation.
S103P, as well as ΔLLSK, is likely to be a mutation that does not facilitate cluster formation () but which activates clustered Ire1. The S103P mutation confers a constitutive-activation phenotype on ΔIΔV Ire1 (, ; ) but not on wild-type Ire1 (, ). Also, the S103P mutation suppressed the impaired-activation phenotype of the ΔIII mutation ( D and F [compare lane 12 to 10]).
It is notable that the constitutive-activation phenotypes of both S103PΔIΔV Ire1 and ΔIΔVΔLLSK Ire1 were only moderately compromised by the ΔIII mutation, whereas ΔIΔIIIΔV Ire1 carrying neither the S103P nor the ΔLLSK mutation was almost completely inactive (, compare 3 to 2). In contrast, the clustering-impaired mutations F247A and W426A almost completely attenuated basal activity of ΔIΔV Ire1, even when carrying the S103P or the ΔLLSK mutation (, and ).
shows the results of a coimmunoprecipitation experiment that probed the physical interaction between Ire1 variants and BiP. As described in our previous papers (; , ; ), Ire1-HA or mutants were expressed from 2-μm plasmids, and the cells were lysed and subjected to anti-HA immunoprecipitation. Double bands of Ire1-HA in the top and third panels (), which were also shown in 2, and S3, are caused by partial degradation in subregion I (). Anti-BiP Western blotting of the anti-HA immunoprecipitates (, bottom) indicates BiP binding to Ire1-HA and its dissociation upon ER stress, neither of which were affected by either the ΔIII or the ΔLLSK mutation. S103P Ire1-HA also shows BiP binding and dissociation at similar levels to wild-type Ire1-HA ().
Finally, the possibility that the CSSR binds directly to unfolded proteins was explored by monitoring its ability to inhibit aggregation of denatured proteins in vitro. For this experiment, the CSSR was bacterially expressed as an N-terminally maltose binding protein (MBP)–fused and C-terminally His-tagged recombinant protein, hereafter called MBP-CSSR. Integrity and purity of wild-type and mutant MBP-CSSRs and an unfused MBP control were verified by SDS-PAGE analysis (). Firefly luciferase and porcine citrate synthase were denatured by guanidine HCl, and then the aggregation induced by dilution of the denaturing mixtures was monitored by measuring the increase of turbidity. shows that aggregation of either denatured protein was not attenuated by the unfused MBP control, whereas MBP-CSSR showed significant effects in inhibiting the aggregation. This finding indicates that the CSSR can interact with two models of aggregation-prone substrates, which suggests that this domain binds to unfolded regions on polypeptides.
To demonstrate that activation of Ire1 is related to its direct interaction with unfolded proteins, we used the ΔIII mutation as an activation-impaired mutation. This is because, as shown in , the phenotype of the ΔIII mutation was suppressed either by the S103P or the ΔLLSK mutation, which indicates that the global structure of Ire1 is not perturbed by the ΔIII mutation. As shown in , the ΔIII mutant version of MBP-CSSR did not attenuate aggregation of the denatured proteins.
Does the groove-like structure described in the Introduction contribute to interaction of the CSSR with unfolded proteins? We modified MBP-CSSR to carry combined substitutions of 3 aa residues (M229A/F285A/Y301A) facing into the groove, which abolish activation of Ire1 (). This mutant MBP-CSSR did not inhibit aggregation of unfolded proteins (), supporting the idea that unfolded proteins are captured by the groove. Nevertheless, it is likely that the M229A/F285A/Y301A mutation confers more extensive damage to the CSSR than the ΔIII mutation, which may include perturbation of global protein structure, because the impaired-activation phenotype of M229A/F285A/Y301A Ire1 was not rescued either by the S103P or the ΔLLSK mutation (, ).
Here, we demonstrate that Ire1 clusters when activated. Impairment of the cluster formation, either by the F247A or the W426A mutation, strongly suggests that the molecular basis of the cluster formation is high order oligomerization of the CSSR (). Importantly and conversely, our finding provides evidence for the high order oligomerization of Ire1, which, as detailed in the Introduction, has been unsupported by many of the previous biochemical analyses of cell lysates or solutions of recombinant Ire1 fragments. We think that the homomeric interaction via interface II is so weak that the association is undetectable in samples used in the biochemical analyses, where the concentration of Ire1 is probably much lower than its local concentration on the ER membrane. It is likely that cluster formation is not a result but a prerequisite of activation of Ire1. This is because Ire1 clustered even with the kinase mutation K702A or in Δ cells and because activity of Ire1 was impaired by either the F247A or the W426A mutation. Clusters of wild-type Ire1 quickly dissociated upon removal of ER stress. This finding again implies that the cluster formation is biologically relevant. The molecular mechanism by which this cluster dissociation is promoted is currently unclear.
The cytosolic domain of yeast Ire1 carries a highly basic sequence, which, according to , acts as an NLS when fused with other proteins. However, our immunofluorescent analysis indicated that both unclustered and clustered Ire1 variants are distributed not only at the nuclear rim but also at other parts of the ER. We thus think that this highly basic sequence does not function as an NLS in the authentic Ire1 molecule.
Another feature of the CSSR that is predicted by the crystal structure but for which there is no supporting biochemical evidence is its direct binding to unfolded proteins (). Here, we also demonstrate that the CSSR actually interacts with unfolded proteins by monitoring its ability to inhibit aggregation of denatured proteins in vitro. Because this property of the CSSR was abolished by the ΔIII mutation, we think that this interaction is biologically meaningful and required for activation of Ire1. The result from the M229A/F285A/Y301A mutation supports the idea that unfolded proteins are captured by the groove-like structure of the CSSR. We failed to demonstrate in vivo interaction between unfolded model proteins and Ire1 by coimmunoprecipitation, even from cells treated with a chemical cross-linker, dithiobis succinimidyl propionate. The interaction may be weak and transient, and in addition, we speculate that because of a structural reason, the cross-linking between the proteins is inefficient.
Unlike a previous scenario detailed in the last paragraph of the Introduction, we now believe that there exists a regulatory step, called here Step 2, other than the cluster formation. This is because ΔIΔV Ire1 was constitutively clustered, but extrinsic ER stress was still required for full activation of this mutant. Furthermore, the S103P and the ΔLLSK mutations are likely to abolish the requirement of Step 2 for activation of Ire1; in contrast, the ΔIII mutation impairs progression of Step 2 but not of the cluster formation. Importantly, such phenotypes of these mutations support our proposal about requirement of Step 2 for activation of Ire1.
What triggers cluster formation or Step 2? Unlike ΔIΔV Ire1, either ΔI or ΔV single mutant showed normal ER-like localization. This finding indicates that the cluster formation requires both dissociation of BiP or the BiP-nonbinding mutation ΔV, and release from repression by subregion I. The mechanism by which subregion I negatively regulates Ire1 remains unclear. Considering BiP's ability to recognize a wide variety of unfolded proteins, it is an attractive idea that BiP acts as a sensor for unfolded proteins in the cluster-formation step, although Ire1 may positively contribute to its own dissociation from BiP (). In contrast, BiP is not involved in Step 2 because the BiP-nonbinding mutant ΔIΔV Ire1 undergoes regulation in Step 2 and because none of the Step 2 mutants (ΔIII, S103P, or ΔLLSK) affect BiP binding and its dissociation from Ire1. Impairment of the interaction between denatured proteins and the CSSR protein carrying the ΔIII mutation strongly suggests that Step 2 is regulated by direct interaction of unfolded proteins with Ire1. Considering the constitutive cluster formation of ΔIΔV Ire1, we think that direct interaction of unfolded proteins is not required in Step 1.
Another important question is what change is produced in Step 2. Because the ΔLLSK mutation, which is located at the cytosolic end of the transmembrane domain, abolishes the requirement of Step 2 for full activation of Ire1, it is likely that orientation of the cytosolic domain is tightly related to Step 2. We propose that, as illustrated in , physical interaction of unfolded proteins with the CSSR causes conformational change of the luminal domain, which leads to reorientation of the cytosolic domain, without changing oligomerization status. This proposal is similar to the case for some cytokine receptors, which, upon binding of ligands, undergo not only self-association but also conformational change, causing alteration of cytosolic-domain orientation.
As a result of this work, we propose a model that is illustrated in . Importantly, ER stress provokes multiple events that separately contribute to activation of Ire1 at different steps. Cluster formation probably leads to considerably higher local concentration of the cytosolic effector domain of Ire1, which may be required for efficient cleavage of the HAC1 mRNA. Regulation by dual steps in different manners is likely to be important for precision of response by ensuring that Ire1 is only up-regulated by ER stress. Indeed, either ethanol or high temperature inappropriately activates ΔV Ire1 ().
Nevertheless, it is likely that not all events noted in are required to obtain partial activation of Ire1. Indeed, the constitutively clustering mutant ΔIΔV Ire1 is slightly but significantly activated even without extrinsic ER stress. Furthermore, reported substantial activation of chimeric Ire1 mutants in which the luminal domain was replaced by dimer-forming fragments of transcription factor proteins. More recently, reported that transient activation of mammalian IRE1α in pancreatic β cells exposed to high levels of glucose does not accompany BiP dissociation.
Is the mechanism presented here applicable to Ire1 orthologues? Because mammalian and plant orthologues of Ire1 do not carry regions corresponding to subregion I, their self-association may be regulated solely by BiP. It should be noted that reported a crystal structure of the luminal domain of mammalian IRE1α, which suggests that neither high order oligomerization nor direct binding of unfolded protein is likely. Unlike cells of the unicellular organism yeast, which suffer direct environmental stress and carry only one known ER-stress sensor Ire1, mammalian cells live in sophisticatedly regulated conditions and have more complicated pathways to respond to ER stress. Thus, it is not unreasonable to postulate that the regulatory mechanisms of the mammalian ER-stress sensors are different from those of yeast Ire1. Nevertheless, we do not think that mammalian IRE1α is regulated solely by BiP because a mutant of this protein with a deletion of the entire region corresponding to subregion V was still regulated by ER stress (unpublished data).
Whereas PERK's activity is apparently regulated by binding and dissociation of BiP (; ), a recombinant fragment of its luminal domain has been shown to inhibit aggregation of unfolded proteins in vitro (Yohda, M. et al. 2006. Proceedings of the 20th International Union of Biochemistry and Molecular Biology Congress and the 11th Federation of Asian and Oceanian Biochemists and Molecular Biologists Congress). Therefore, it is likely that in a similar manner to yeast Ire1, PERK is regulated both by BiP and by direct binding of unfolded proteins. Interestingly, ATF6 may also be regulated dually in its activation upon ER stress, although it has no structural similarity to Ire1 or PERK. Activation (i.e., transport to Golgi apparatus) of ATF6 is negatively regulated by binding of BiP and by intra- and intermolecular disulfide bridge formation, both of which are lowered upon ER stress (; ). Finally, such multiplicity of regulatory mechanisms implies complexity of conditions in which these ER-stress sensors are individually activated, as suggested by .
Yeast congenic haploid strains KMY1015, KMY1516, and KMY1520 (), all of which carry an Δ null mutation, were used according to the difference of their mating and auxotrophic phenotypes. To generate the Δ Δ strain YKY1004, a EUROSCARF strain Y15650 (α ; provided by J.W. Goethe, University Frankfurt, Frankfurt, Germany) was modified to carry a complete deletion of the ORF by replacement with the gene. To obtain a YKY2005 strain, the Ire1-HA gene was knocked in to replace the endogeneous gene of wild-type strain KMY1005 (α ), which is congenic to KMY1015, KMY1516, and KMY1520. Cells were exponentially cultured in SD medium (2% glucose and 0.66% yeast nitrogen base without amino acids; Difco), supplemented with appropriate nutrients, at 30°C.
In our previous studies (; , ), plasmids pRS313-IRE1, pRS315-IRE1-HA, pRS423-IRE1-HA, and pRS426-IRE1-FLAG were generated by insertion of the gene (or its C-terminally epitope-tagged version) into yeast centromeric vectors pRS313 and pRS315 () and 2-μm vectors pRS423 and pRS426, respectively (). As described in , we introduced mutations into these plasmids by in vivo homologous recombination (gap repair) between the plasmids cleaved by restriction enzymes and mutant fragments created by overlap PCR, primers for which are listed in Table S1 (available at ). A UPRE-lacZ reporter plasmid pCZY1 ( 2 μm) was provided by K. Mori (Kyoto University, Kyoto, Japan). A GFP-Sec12 expression plasmid pSKY54-GFP-SEC12 was provided by A. Nakano (Institute of Physical and Chemical Research, Saitama, Japan). A pRS315-based derivative of pSKY54-GFP-SEC12 was generated by in vivo homologous recombination between PvuII-digested pSKY54-GFP-SEC12 and XhoI–NotI–digested pRS315.
For fluorescent microscopic examination, cells were fixed in 0.1 M potassium acetate buffer, pH 6.8, containing 3.3% formaldehyde for 2 h and processed according to . Antibodies used are listed in Table S2 (available at ), including anti-Mnn9 antiserum, which was provided by Y. Noda (The University of Tokyo, Tokyo, Japan). The mounting medium was 90% glycerol containing 0.1% -phenylenediamine. For conventional fluorescent microscopy, Axiophoto (Carl Zeiss MicroImaging, Inc.) was used with an oil immersion lens (Plan-Neofluor 100/1.30), and images were captured by a digital charge-coupled device (CCD) camera system (DP70; Olympus) carrying built-in software for image acquisition. For deconvolution microscopy, an Axiovert 200M (100/1.40 oil immersion Plan Apochromat objective; Carl Zeiss MicroImaging, Inc.) with the Apotome system was used. Unless noted, FITC was used as fluorochrome. Photoshop software (Adobe) was used for conversion to grayscale images (, , and ) and image overlapping ().
Cells were fixed basically as described previously in . Cells were frozen in a high pressure freezer (HPM010; Bal-Tec Inc.) and transferred to anhydrous acetone containing 2% OsO in an automatic freeze-substitution apparatus (EM AFS; Leica) in which the temperature was gradually sifted from −80 to 23°C. After washing three times with anhydrous acetone, the samples were infiltrated with increasing concentrations of Spurr's resin in anhydrous acetone, and finally with 100% Spurr's resin. After polymerization in capsules at 60°C, ultrathin sections were cut on a microtome (Ultracut UCT; Leica). The sections were immunostained with anti-HA antibody and 10 nm of gold immunogold conjugate EM goat anti–mouse IgG (Table S2) and stained with 3% uranyl acetate for 2 h. The sections were then examined with an electron microscope (H-7600; Hitachi) at 100 kV.
RNA extraction, Northern blotting, cell lysis for protein analyses, immunoprecipitation, and Western blotting were performed as described previously (, ). Detergent-free cell lysates used to obtain microsome fractions were prepared as described in . Radioactive signal from Northern blots was detected using a phosphor imager (BAS-2500; Fuji). For SDS-PAGE, lysates from 10 cells and IPs from 3 × 10 cells were run on precast gels (Multigel II Mini; Daiichi Pure Chemicals; 7.5% acrylamide), unless otherwise noted. The chemiluminescent signal from Western blots was captured by a cooled CCD camera system (LAS-1000plus; Fuji) and quantified using imaging software (ImageGauge; Fuji). When obtaining the Ire1-FLAG/Ire1-HA values in , we confirmed the linear relation between the quantified chemiluminescent signal and the actual amount of the band protein. Antibodies used, including that against yeast BiP (), are listed in Table S2.
To avoid artifactual disulfide bond formation, all versions of MBP-CSSR had all Cys residues changed to Ser because this amino acid replacement does not affect activity of Ire1 (). An gene partial fragment corresponding to the CSSR was PCR amplified from pRS315-IRE1(CS)-HA () using primer set P-5 and P-6, digested with BamHI and HindIII, and inserted into similarly digested pMAL-c2x (New England BioLabs, Inc.). To generate mutant variants, mutations were introduced by using the overlap PCR mutagenesis technique as described in . The PCR primers are listed in Table S1.
An strain BL21 codon plus (DE3)–RIL (Strategene) was transformed with one of the resulting plasmids and cultured at 37°C in 400 ml of 2× YT medium. Expression of MBP-CSSR or its mutant variants was induced by addition of IPTG (0.3 mM final concentration) into the culture, followed by further incubation at 30°C for 1 h. After harvest, cells were suspended in 15 ml of lysis buffer (50 mM Hepes, pH 8.0, 300 mM KCl, 5 mM MgCl, 10 mM imidazole, 1% Triton X-100, 2 mM phenylmethylsulfonyl fluoride, 0.4 mg/ml benzamidine, 0.4 mg/ml pepstatin A, 0.4 mg/ml leupeptin, 0.3 mg/ml lysozyme, and 14 U/ml DNase I; Takara) and disrupted by ultrasonication. The lysate was clarified by centrifugation (SRX-201; Tomy; 8,200 rpm for 10 min) and incubated with 0.5 ml of HisLink protein purification resin beads (Promega) for 12 h. The beads were packed into an 8-mm-diam column and sequentially washed with 6 ml of 50 mM Hepes, pH 8.0, 1 M KCl, 5 mM MgCl, and 0.1% Triton-X 100; 6 ml of 50 mM Hepes, 300 mM KCl, 5 mM MgCl, 0.1% Triton X-100, and 20 mM imidazole; 6 ml of 50 mM Hepes, 300 mM KCl, 5 mM MgCl, 0.1% Triton X-100, and 40 mM imidazole; 3 ml of 50 mM Hepes, 300 mM KCl, 5 mM MgCl, 0.1% Triton X-100, and 60 mM imidazole; 5 ml of 50 mM Hepes, 300 mM KCl, 5 mM MgCl, and 10 mM ATP; and 3 ml of 20 mM Hepes, 100 mM KCl, 5 mM MgCl, and 50% (vol/vol) glycerol. Bead-bound proteins were eluted with 50 mM Hepes, 100 mM KCl, 5 mM MgCl, 200 mM imidazole, and 50% (vol/vol) glycerol, and elution fractions (1 ml each fraction) were analyzed by SDS-PAGE.
Citrate synthase (Roche) was dialyzed against 20 mM Hepes, pH 7.0, 150 mM KCl, 2 mM MgCl, and 10% (vol/vol) glycerol, and stored at −80°C. Luciferase (25 μM final concentration) or citrate synthase (50 μM final concentration) were denatured by incubation in guanidine HCl– denaturing solution (guanidine HCl [6 M for luciferase or 4 M for citrate synthase], 20 mM Hepes, 50 mM KCl, and 2 mM MgCl) for 30 min at room temperature. The denaturing mixture was then diluted with assay buffer (20 mM Hepes, pH 7.2, 50 mM KCl, and 2 mM MgCl) in the presence or absence of MBP-CSSR (wild type or mutant) or MBP, and aggregate formation was monitored by absorbance at 320 nm with a spectrophotometer (DU640; Beckman Coulter) at room temperature.
Table S1 lists Ire1 mutations analyzed in this study. PCR primers used for generation of these mutations are also listed. Table S2 lists antibodies used in this study. Fig. S1 shows that even when expressed from a multicopy plasmid, Ire1-HA was activated in an ER stress-dependent manner. In Fig. S2, Ire1-HA was expressed from a single copy plasmid, a multicopy plasmid, and the knocked in gene, and the expression levels were compared. Fig. S3 shows highly efficient intermolecular homo-cross-linking of ΔIΔV Ire1-HA, which supports high order oligomerization of this molecule. Online supplemental material is available at . |
Protein targeting and transport across lipid bilayers is a fundamental energy-requiring process in all organisms. Up to approximately half of the proteins in an organism's proteome are inserted into or transported across membranes by protein translocation systems, or translocons (; ). Most bacterial proteins are transported using the conserved Sec translocation pathway (). However, a distinct set of proteins are transported in fully folded and assembled form by the twin-arginine translocation (Tat) pathway (; ). Tat substrates are characterized by a twin-arginine–containing consensus motif (SRRxFLK) present in the N-terminal signal peptide of precursor proteins. In , approximately two thirds of Tat substrates contain prosthetic groups, which are inserted into the proteins in the cytoplasm (). A poorly understood proofreading mechanism prevents transport of substrates until they are properly folded and assembled (; ).
The Tat translocation system contains four identified protein components: TatA, TatB, TatC, and TatE. TatA, TatB, and TatE each contain a single N-terminal transmembrane domain and a C-terminal cytoplasmic domain; the transmembrane domain is followed by an amphipathic helix that could preferentially interact with the lipid–water interface (; ). TatC, which contains part of the signal sequence binding site (; ), has six transmembrane domains with both N and C termini facing the cytoplasm (; ). Mutational analyses have shown that a functional Tat system minimally requires TatB, TatC, and either TatA or TatE (, ; ). Thus, TatA and TatE are structural and functional homologues.
Three main oligomeric Tat complexes have been found in the periplasmic membrane. TatA forms oligomers from <100 kD to >500 kD that have been characterized as ring-like structures by electron microscopy (; , ; ). TatBC oligomers have an average molecular mass of ∼500 kD () wherein the TatB/TatC ratio is ∼1:1 (). The average molecular mass of TatABC complexes as estimated by gel-filtration (; ) and blue-native gel electrophoresis () is ∼600 kD and ∼370 kD, respectively. TatA is found in large molar excess (as much as ∼20-fold) over TatB and TatC (), suggesting that the TatA complexes outnumber the TatBC complexes. It has been hypothesized that a pore composed of TatA oligomers allows the mature domain of the precursor protein to cross the membrane (). In such a model, the mature domain of a precursor protein bound to a TatBC complex through its signal sequence would have to be transferred through the TatA pore, perhaps as a result of oligomerization of a TatBC complex and a TatA complex.
The Tat system was first identified in plant thylakoids as a translocation system that requires the proton motive force (PMF), and not ATP, for transport. The energy stored in the PMF has two components, the electric field gradient (Δψ) and the pH gradient (ΔpH). From early experiments on thylakoids, it was concluded that the Tat system is energetically driven by the ΔpH alone (; ). This basic conceptual finding was recently challenged (), and more recent work indicates that the Δψ can also contribute to driving Tat transport in thylakoids (). Energetic studies of the bacterial Tat machinery have been hampered by the lack of an efficient in vitro assay. The first reported in vitro assay yielded a transport efficiency of <1% (). Subsequently, it was found that precursors can be transported with up to ∼20% transport efficiency if they are synthesized via in vitro translation in the presence of inverted membrane vesicles (IMVs) (). Here, we report the development of an efficient in vitro assay for the Tat machinery using purified overexpressed precursors. We show that two distinct Δψ-dependent steps are required for Tat transport. We did not detect a role for the ΔpH in influencing transport efficiency, despite the presence of substantial pH gradients. Our data are consistent with a model in which a relatively large Δψ of brief duration is required for an initial step in the transport process, and in which a small Δψ of long duration is required for a later step.
To avoid the assembly and folding complications inherent in using a Tat substrate that contains a cofactor, we have used pre-SufI, spSufI-GFP, and spTorA-GFP (). Although SufI is homologous to proteins of the multi-copper oxidase family, SufI does not appear to bind Cu () or possess a cofactor binding site (). Pre-SufI was isolated under native conditions. The artificial GFP substrates were isolated under denaturing conditions and folded in vitro. Proper folding of the GFP substrates was assumed based on their green fluorescence emission upon UV excitation ().
A typical in vitro transport assay consisted of the addition of TatABC-enriched IMVs to prewarmed (37°C) tubes containing 50 nM pre-SufI. NADH was used to generate a PMF. Reactions were incubated at 37°C for 30 min and then treated with protease (proteinase K) to digest any remaining untransported pre-SufI. The proteins in each reaction were resolved by SDS-PAGE and immunoblotted using SufI antibodies (see Materials and methods). Overexpression of TatA, TatB, and TatC was essential for detecting transport, and only these three Tat proteins were required (). In general, IMVs from JM109 were easier to work with and yielded better transport efficiencies than those from MC4100 (). Membrane orientation in vesicle preparations was estimated by the accessibility of the TatB cytoplasmic domain to protease digestion. Typical membrane preparations consisted of ∼90% IMVs and ∼10% right-side-out vesicles (). Addition of 0.05% Triton X-100 during protease treatment resulted in complete digestion of mSufI (), indicating that pre-SufI that had translocated into the vesicle lumen could be digested by protease after membrane permeabilization. Control experiments lacking added pre-SufI demonstrate that the high molecular weight bands observed on anti-SufI immunoblots arose from endogeneous proteins within the IMV preparations, and not from pre-SufI aggregates (, lanes 4 and 7). Detection with 6xHis antibodies confirms this result (). Mature- and precursor-length SufI were not always resolvable due to their small difference in molecular weights (e.g., compare ).
Having developed an efficient in vitro Tat transport assay, we then tested whether a common model substrate, spTorA-GFP (; ; ), could be efficiently transported under the conditions that yielded efficient pre-SufI transport. The spTorA-GFP protein was transported at a much lower efficiency (up to ∼15%; ) than pre-SufI (72 ± 7%). We considered the possibility that proteins with the TorA signal peptide could not be efficiently transported in our standard assay for unknown reasons. Therefore, we changed the signal peptide on spTorA-GFP to that of pre-SufI yielding spSufI-GFP (). No transport of spSufI-GFP could be detected ().
We expected that protein chimeras containing Tat signal peptides should at least interact with the signal peptide binding site of the Tat translocation machinery. If so, such chimeras should competitively inhibit transport of pre-SufI. Surprisingly, even a 50-fold molar excess of spSufI-GFP had no effect on transport of pre-SufI (). These data indicate that spSufI-GFP cannot bind (or only very weakly binds) to the Tat pathway receptor complex. In contrast, pre-SufI transport was almost completely inhibited by a 10-fold molar excess of spTorA-GFP (). These data likely indicate that spTorA-GFP was recognized by the Tat translocation machinery through the TorA signal peptide, and therefore, that spTorA-GFP competitively inhibited pre-SufI transport.
Having established optimum conditions for pre-SufI transport, we then examined the transport kinetics. Transport reactions were quenched at various times by plunging reaction tubes into an ice bath. Two notable features were immediately apparent: transport was a relatively lengthy process, occurring on the timescale of many minutes; and the transport kinetics were complex (i.e., not a single exponential), exhibiting a lag period before the appearance of transported protein (). A lag period was observed earlier for the thylakoid Tat system ().
When NADH is added to IMVs, the electron transport chain generates a PMF using dioxygen (O) as the final electron acceptor. Thus, the duration of the resultant pH gradient and electrical potential is limited by how rapidly the dissolved oxygen is consumed. Steady-state fluorescence spectroscopy was used to monitor the presence of a ΔpH and a Δψ (see Materials and methods). Because the detectable gradients (as reported by the dyes) decayed in seconds to tens of seconds after oxygen consumption, the duration that detectable gradients existed (ΔpH and Δψ) was somewhat longer than the time to anaerobiosis (ΔpH and Δψ), although the latter were more easily measurable due to discrete inflection points. ΔpH and Δψ were identical (within error) for a given set of conditions; ΔpH and Δψ were not identical. At least part of these differences can be attributed to the inherent slow response of the dye distributions to the gradients. As expected, both ΔpH and Δψ were inversely affected by an increase in IMV concentration, consistent with a faster enzymatic consumption of dissolved oxygen at higher IMV concentrations (). At the concentration of TatABC-enriched IMVs used in our standard transport assay (
= 5), the solutions became anaerobic in 10 ± 4 s (), and detectable ΔpH and Δψ gradients were observed for ∼15 s and ∼10 s, respectively (). However, mature SufI continued to accumulate for at least ∼12 min after the reaction became anaerobic (), indicating that Tat pathway–mediated transport of precursor proteins can be completed in the absence of a detectable ΔpH and Δψ.
We next tested whether the duration of the detectable ΔpH and Δψ influenced pre-SufI transport efficiency by varying the concentration of the IMVs.
= 5 to
= 0.5, the Δψ and ΔpH increased from ∼10 s to ∼42 s (). At a high precursor concentration (presumably saturating the Tat translocons over all the IMV concentrations tested), the amount of precursor transported per IMV was invariable (). These data indicate that the approximately fourfold increase in the duration of the detectable Δψ and ΔpH did not affect pre-SufI transport efficiency under these conditions.
We next examined the effect of the magnitude of the detectable Δψ and ΔpH on pre-SufI transport efficiency. Because ΔpH generation takes many seconds due to the large number of ions that must be translocated, the point at which the maximum ΔpH (ΔpH) was generated was easily estimated (). In contrast, Δψ generation is very fast due to the much lower number of ions that must be translocated, and it was not always clear if observed signals were due to the injection needle, mixing artifacts, or Δψ generation. Instead, we estimated the average Δψ (Δψ; ). We compared the pre-SufI transport efficiencies obtained with various uncoupler concentrations. When the ΔpH was selectively reduced with various concentrations of nigericin (an electroneutral K/H exchanger), pre-SufI transport was unaffected or slightly increased. Increased transport efficiencies correlated with an increased Δψ (). When the Δψ was selectively reduced with valinomycin (a K ionophore), the ΔpH remained high and transport efficiency again correlated with Δψ (). The Δψ could not be completely dissipated with valinomycin alone (see Fig. S3 for an explanation, available at ). However, in the presence of both valinomycin and nigericin, the detectable ΔpH and the Δψ were both completely collapsed and pre-SufI transport was completely inhibited ().
= 5), a low concentration of nigericin was necessary (). When the detectable ΔpH was first completely dissipated by a low concentration of nigericin and the Δψ was progressively dissipated by increased concentrations of valinomycin, the pre-SufI transport efficiency was highly correlated with Δψ (). When the Δψ was dramatically reduced by 25 mM NaSCN without decreasing the ΔpH gradient (), pre-SufI transport was almost completely inhibited (). Transport reactions with spTorA-GFP confirmed the pre-SufI results that Tat transport requires a Δψ (). The increased spTorA-GFP transport observed when nigericin was present () is likely explained by the increased Δψ observed under these conditions (), which likely results from compensation for the PMF decrease that results from loss of the ΔpH. The NaSCN data support the hypothesis that inhibition of pre-SufI transport by valinomycin was through collapse of the Δψ, rather than by a direct effect of valinomycin on the Tat translocation machinery. In total, these data indicate that pre-SufI transport was largely, if not completely, independent of the ΔpH and strongly dependent on the Δψ.
The data discussed in the previous two sections indicate that pre-SufI transport required a Δψ, yet transport occurred on a much longer timescale (many minutes) than the time a detectable Δψ (∼10 s) was maintained across IMV membranes. One possible explanation for these results is that the Δψ is required only for an early step in the transport process and that later steps of transport do not require a Δψ. To investigate this possibility, we added ionophores to dissipate any nonmeasurable gradients at various times after reaction initiation. As expected based on earlier results ( and ), pre-SufI transport efficiency was not affected by dissipation of the ΔpH by 10 μM nigericin either early or late in the transport process (). In contrast, addition of 10 μM nigericin and 10 μM valinomycin early in the transport process (e.g., at ∼1–5 min after reaction initiation) almost completely inhibited pre-SufI transport (), despite the fact that the detectable gradients had collapsed when the ionophores were added (). Our interpretation of these data is that undetectable ΔpH and Δψ gradients existed after the reaction solutions became anaerobic. One possibility is that the solutions were not completely anaerobic due to gas exchange at the aqueous–air interface, thereby allowing low level respiratory activity. Alternatively, a small Δψ was maintained by an anaerobic pathway. Regardless, our interpretation of these data is that a small, undetectable Δψ was maintained after collapse of the short duration but substantial Δψ spike (), and that this undetectable Δψ was essential for driving pre-SufI transport. The threshold sensitivity of our Δψ measurements is unknown. Other investigators have estimated a threshold sensitivity of ∼50 mV, but this value is strongly dependent on the lipid to dye ratio ().
To address whether a precursor protein must be present during the Δψ spike period, pre-SufI was added to IMVs at various times after addition of NADH. Efficient Tat transport (> ∼90% of control) occurred even if pre-SufI was added up to ∼5 min after membrane energization, long after collapse of the short, initial Δψ spike (). Transport efficiency decreased by ∼90% with an ∼30 min delay between precursor addition and membrane energization (), possibly due to consumption of NADH.
Because efficient pre-SufI transport resulted when the precursor was added after collapse of the Δψ spike, we considered the possibility that the undetectable Δψ gradient was sufficient to drive the entire Tat translocation cycle. To test this hypothesis, we measured the translocation yield after producing a second Δψ spike by addition of O-saturated buffer. We observed that the pre-SufI translocation yield increased by ∼60% when a second Δψ spike was generated ∼12.5–20 min after the first. A second Δψ spike had little to no effect if it was generated < ∼5 min after reaction initiation ().
The fact that a second Δψ spike generated within 5 min of the first had little to no influence on transport yield indicates that the translocons were predominantly at a stage in the precursor translocation cycle where the energy of the second Δψ spike could not be used. This suggested to us that perhaps the kinetics shown in reflected a single translocon turnover cycle, and that a Δψ spike might be required to initiate a second turnover cycle. To test this possibility, we reasoned that the observed kinetics should be essentially invariable with IMV concentration if they reported a single translocation cycle. In contrast, if the kinetics in reflected numerous enzyme turnovers, we reasoned that faster transport would be observed at higher IMV concentrations. We observed that the transport kinetics were largely independent of IMV concentration (). According to the predictions just discussed, these data are consistent with the picture that the observed transport kinetics reflect a single turnover cycle.
Considering the conclusion from the previous section that the transport kinetics are consistent with a single turnover cycle, we hypothesized that if spTorA-GFP was added after pre-SufI transport was initiated, it might no longer be competitive for transport. We found that pre-SufI transport was sensitive to the addition of spTorA-GFP competitor more than 5 min after reaction initiation. In fact, the transport inhibition versus the time delay of competitor addition curve () looks surprisingly similar to the transport kinetics of the reaction as determined by ionophore quenching ().
To maintain a high magnitude steady-state Δψ, ATP was used to generate a PMF through catalytic reversal of the FF ATPase. With this approach, the Δψ is independent of the dissolved O concentration and depends only on the ATP energy charge. Because ATP can be solubilized at higher concentrations than O and regenerated in situ, the Δψ gradient can be maintained for a longer period. When ATP (plus an ATP regenerating system) was used to energize IMVs, both ΔpH and Δψ remained detectable for at least 9 min (), and pre-SufI translocation was approximately threefold faster () without the detectable lag phase observed when IMVs were energized by NADH (). Despite the ability of ATP to establish a long-lived high-magnitude ΔpH, pre-SufI transport was, once again, largely, if not completely, independent of the ΔpH and strongly dependent on the Δψ (). These data support the hypothesis that translocation speed is dependent on the magnitude of the steady-state Δψ. When NADH was used to generate a PMF, the length of the lag period was observed to vary for different IMV preparations (not depicted). In light of the above ATP results, we surmise that the length of the lag period reflects the differential ability of different IMV preparations to support a Δψ.
The development of an efficient in vitro transport assay for the Tat machinery has allowed us to decipher a number of important fundamental features of the transport mechanism: the energetic driving force for transport comes largely, if not entirely, from the Δψ alone—we have no evidence that the ΔpH assists with promoting transport (, , and ); at least two distinct transport steps require a Δψ—one Δψ requiring step occurs early in the transport process and requires a Δψ of relatively high magnitude that may be short-lived (), and a second Δψ requiring step minimally requires a long duration, but relatively low magnitude Δψ (); transport speed is increased if the steady-state Δψ is increased (); transport efficiency is decreased if the average Δψ is decreased (); the first Δψ requiring step can occur in the absence of precursor protein (); and transport can be competitively inhibited long after transport initiation (). These findings provide important constraints for the transport mechanism. Some of the implications are discussed in the following paragraphs.
A natural question to arise from these studies is whether the Tat transport system is fundamentally unique or different than the Tat transport system found in photosynthetic organisms. Early in vitro studies in higher plant thylakoid systems support a picture wherein it is the ΔpH alone that provides the driving force for Tat precursor transport (; ). More recent data suggest a minimum substrate-specific ΔpH (). In contrast, in vivo studies in and barley leaves support a picture where it is the Δψ alone that provides the driving force for Tat precursor transport (). Precursor maturation in tobacco protoplasts does not require either a ΔpH or a Δψ, but it is not clear whether the mature domain can be transported under these conditions (). Possible explanations for the apparent discrepancies are that something is missing from in vitro assays and/or that both the Δψ and the ΔpH can support Tat transport under certain conditions (). The hypothesis that both the ΔpH and Δψ components of the PMF can energetically contribute to driving precursor transport is supported by recent in vitro studies in thylakoids (). The studies reported here, however, raise some additional possibilities: under some conditions the Δψ is not completely collapsed by valinomycin alone; and only a brief, yet relatively large, Δψ is necessary to initiate a transport cycle. In thylakoids, a brief but substantial Δψ does exist upon photoillumination, which may be sufficient to drive the first step of Tat transport, and a relatively low Δψ does exist during steady state under in vivo conditions and under some in vitro conditions (), which may be sufficient to drive the completion of Tat transport.
According to current models, an early step in Tat transport is precursor binding to the TatBC complex. In some models, recruitment of TatA molecules leads to formation of a highly selective pore that allows the precursor protein to cross the membrane (; ). In thylakoids, transport can occur without migration of the signal peptide from the TatBC binding site (). We show here, however, that transport was blocked by an excess of a precursor that was added many minutes after transport initiation. If transport was inhibited by competitive binding, which seems likely, one interpretation is that the signal sequence binding interaction must be readily reversible. This conclusion is somewhat puzzling because it contradicts the thylakoid data, and because it would seem to lead to slow and inefficient transport. However, if different types of translocons existed (e.g., arising from different oligomeric states of TatA), a reversible signal sequence binding interaction on TatBC would be an effective way for the precursor to sample many different structures. One way that this could happen without losing the precursor to the bulk solution is if the signal peptide binds to the lipid alone, as has been recently demonstrated (). This picture could explain how spTorA-GFP can compete with pre-SufI for transport long after reaction initiation (). On the other hand, if multiple TatBC complexes can bind to the same TatA oligomer, multiple precursors could compete for translocation through the same gated pore. Hence, spTorA-GFP added after transport of pre-SufI has begun could bind to TatBC oligomers that are a part of TatABC–pre-SufI complexes, and thereby reduce the efficiency of pre-SufI transport. According to this picture, the signal sequence binding interaction could be relatively strong and not readily reversible, consistent with the thylakoid data that indicate that transport can occur without migration of the signal peptide from the TatBC binding site ().
There are at least two interpretations of the transport kinetics shown in . We discussed the single turnover cycle model in the Results section. According to this interpretation, a second turnover cycle cannot initiate without a second high-magnitude Δψ pulse, and this pulse must occur at, or near, the end of the first turnover cycle (compare with ). An alternative interpretation is that each high-magnitude Δψ pulse is sufficient to drive numerous turnover cycles. The initial lag in the observed kinetics would then be consistent with the hypothesis that the first turnover cycle is relatively slow, and subsequent turnover cycles are faster. The existence of “slow” and “fast” enzymatic forms is not a novel concept (). In contrast, as discussed earlier, if the translocons function with only one translocation speed under a given set of conditions, then it is expected that higher translocon concentrations would yield faster overall translocation kinetics, which was not observed (). In conclusion, both the single and multiple turnover cycle models (with the slow and fast caveat for the latter) are consistent with our data.
The nature of the coupling between the Δψ and protein transport by the Tat machinery remains unclear. We have not ruled out the possibility that a concentration gradient other than the ΔpH is coupled to Tat transport—e.g., a sodium or other ion gradient. The coupling of ion flow to protein transport would imply that a portion of the energy stored in a potential gradient is “consumed” to drive transport because ion flow would actually reduce the gradient. However, energy “consumption” is not strictly required. Tat protein transport occurs down a concentration gradient because precursor proteins are only found on the cytoplasmic side of the cytoplasmic membrane (the concentration of mature protein in the periplasm can be much higher, but this does not affect the thermodynamics of transport because these proteins cannot be transported after signal peptide cleavage). Thus, instead of being consumed to energetically drive transport, it is possible that the Δψ is instead coupled to protein conformational changes (“gating reactions”) that are required for transport. In the case of voltage-gated ion channels, the membrane Δψ is coupled to the opening and closing of pores required for ion transport. Current models postulate that a charged region of voltage-gated ion channels moves in response to a Δψ, thereby causing the channel to open (). In the case of the Tat pathway, it is unlikely that the Δψ would simply open large pores allowing folded proteins to diffuse across the membrane because this would result in the immediate collapse of the PMF and other ion gradients. However, a reasonable scenario is that movement of certain charged regions within Tat proteins could be induced by a Δψ. For example, the movement of Δψ sensing domains could lead to the oligomerization or rearrangements of subunits, leading the translocon to be primed and ready for catalyzing transport. We emphasize that this picture still requires additional gating reactions (e.g., signal peptide recognition) for controlling access to the translocation channel.
In summary, we have demonstrated an efficient in vitro transport assay for the Tat machinery. When NADH was used to produce a PMF, transport was relatively slow and occurred with a half-time of ∼10 min. Because faster translocation could be achieved with a higher magnitude and longer lasting PMF, we predict that in vivo transport rates are faster, and that our current transport rates are limited by the leakiness of IMVs. Nonetheless, under our conditions, precursor transport is ΔpH independent and there exist at least two Δψ-dependent transport steps.
strains MC4100, MC4100ΔTatABCDE, JM109, and BL21(λDE3) have been described earlier (; ; ; ). Overexpression cultures were grown in Luria-Bertani (LB) medium at 37°C supplemented with appropriate antibiotics (), unless otherwise noted.
The spTorA-GFP protein was expressed from plasmid pTorA-GFP. Plasmid pTorA-GFP was constructed by QuikChange site-directed mutagenesis (Stratagene) from pJDT1 () using primers prTorAHisC-F and prTorAHisC-R as forward and reverse primers, respectively (Table S1, available at ). This process added a C-terminal HHHHHHC tag to the encoded protein.
The spSufI-GFP protein was expressed from plasmid pSufI-GFP, which was constructed in two steps by replacing the TorA signal peptide in pTorA-GFP with the SufI signal peptide. First, a unique SacII restriction endonuclease site was generated 130 bp downstream of the translation start site encoding spTorA-GFP in pTorA-GFP by the QuikChange protocol using prTorA-ScII-F and prTorA-ScII-R as forward and reverse primers, respectively (Table S1), generating plasmid pTorA-GFP2. Then, the PCR product generated from amplification of pET-SufI () with primers prSufI-sp-F and prSufI-sp-R (Table S1) was digested with EcoRI and SacII and inserted into pTorA-GFP2 digested with the same two restriction enzymes, generating plasmid pSufI-GFP. Coding regions were confirmed by DNA sequencing.
TatA, TatB, and TatC were detected by Western blotting using rabbit polyclonal TatA, TatB, and TatC antibodies (1:5,000 dilutions) () and GFP was detected by using rabbit polyclonal GFP antibodies (1:10,000; Santa Cruz Biotechnology, Inc.) in 1× PBS (137 mM NaCl, 2.7 mM KCl, 10 mM NaHPO, and 2 mM KHPO, pH 7.4) with 2% nonfat dry milk, 0.1% Triton X-100, and 0.1% Tween. SufI was detected as described above by using rabbit 6xHis antibodies (1:1,000; Santa Cruz Biotechnology, Inc.), or by using SufI antibodies (1:15,000) () in 1× PBS as described above except with higher detergent concentrations (0.5% Triton X-100 and 0.5% Tween 20). Goat polyclonal anti–rabbit IgG-HRP conjugate (1:15,000; Santa Cruz Biotechnology, Inc.) was used as the secondary antibody, and bands were visualized by chemiluminescence (). Band intensities were quantified with a PhosphorImager (model FX; Bio-Rad Laboratories).
Cells for IMV isolation were grown in low-salt LB medium (1% bactotryptone, 0.5% yeast extract, and 0.25% NaCl) supplemented with 5% glycerol.
reached ∼1. TatABC expression from pTatABC () was induced with 0.7% arabinose and growth was continued for another 4 h. Cultures were plunged into an ice bath and cells were harvested by centrifugation at 4,000 for 8 min at 4°C. The cell pellet (10–12 g) was suspended in 50 ml ice-cold Buffer A (1 mM MgSO, 0.5% polyvinylpyrrolidine [MW 360,000], 450 mM mannitol, 2 mM DTT, 50 μg/ml DNase I, 10 μg/ml RNAase, 1 mM KCl, and 100 mM Tricine, pH 7.5) with 0.4 mg/ml lysozyme, 0.5 mM EDTA, and protease inhibitors (10 mM PMSF, 100 μg/ml trypsin inhibitor, 20 μg/ml leupeptin, and 100 μg/ml pepstatin), and incubated on ice for 20 min to produce spheroplasts. The spheroplasts were sedimented by centrifugation at 4,000 for 10 min at 4°C, resuspended in Buffer A, and passed through a French pressure cell once at ∼6,000 psi to produce IMVs. The IMV solution was centrifuged at 4,000 for 10 min at 4°C to remove debris. The supernatant (8-ml portions) was layered over 6-ml sucrose cushions (Buffer B [1 mM KCl, 1 mM MgSO, 2 mM DTT, and 10 mM Hepes, pH 7.0] with 2.2 M sucrose) and centrifuged for 75 min at 108,000 . The band from the top of the sucrose cushion in each tube was collected. IMVs were pooled, diluted 1:4 with Buffer B, and centrifuged for 45 min at 108,000 . The pellet was resuspended in 4–6 ml of Buffer B with 50% glycerol, and frozen immediately at −80°C.
in 2% SDS.
≈ 50–60.
Pre-SufI was overexpressed from pET-SufI () in BL21(λDE3) and purified under native conditions by Ni-NTA chromatography.
reached ∼3. The pH of the cultures was raised with 25 ml of 0.5 M CAPS buffer (pH 9.0), and induced with 0.5 mM IPTG for 2.5 h. Cultures were chilled in an ice bath and centrifuged at 5,000 for 8–12 min at 4°C. Pellets were rapidly resuspended on ice in 50 ml Buffer C (100 mM Tris and 25 mM CAPS, pH 9.0) containing 250 mM NaCl, 20 mM imidazole, 0.2% Triton X-100, and protease inhibitors. Cells were then passed through a French pressure cell once at 16,000 psi. The cell lysate was cleared of cellular debris by centrifugation at 50,000 for 10 min at 4°C, and then stirred with 2 ml Ni-NTA Superflow resin (QIAGEN) that had been pre-equilibrated with Buffer C for 10 min on ice. The resin was loaded onto a 10 × 1 cm column, and sequentially washed with: 100 ml of Buffer D (10 mM Tris-HCl, 1 M NaCl, and 20 mM imidazole, pH 8.0) with 0.1% Triton X-100; 20 ml of Buffer D; 20 ml of 100 mM NaCl and 10 mM imidazole, pH 8.0; and 10 ml of 100 mM NaCl, 10 mM imidazole, and 50% glycerol, pH 8.0. Pre-SufI was eluted with 100 mM NaCl, 250 mM imidazole, and 50% glycerol, pH 8.0 as 1-ml fractions and stored at −80°C. Typical yield was 8–10 mg protein/L of culture.
The spSufI-GFP and spTorA-GFP proteins were overexpressed from plasmids pSufI-GFP and pTorA-GFP, respectively, in MC4100ΔTatABCDE. These GFP proteins were purified under denaturing conditions and refolded by dilution/dialysis from 9 M urea.
reached ∼5 (∼4 h). Cells were harvested by centrifugation, resuspended in fresh LB medium with 1.5% arabinose to induce protein expression, and incubated at 37°C for 2.5 h. Cells (∼7 g) were harvested by centrifugation, resuspended in 50 ml Buffer E (20 mM CAPS, 2 mM DTT, and 0.2% Triton X-100, pH 9.0) with protease inhibitors, and stirred on ice for 30 min. The cell suspension was passed through a French pressure cell once at 16,000 psi. The cell lysate was diluted to 400 ml with Buffer E, and centrifuged at 15,000 for 30 min. Pellets containing inclusion bodies were suspended in Buffer F (10 mM CAPS, 5 mM DTT, and 9 M urea, pH 9.0), and stirred at room temperature (RT) for 30 min. The resuspension was centrifuged at 50,000 for 30 min. The supernatant was stirred with 4 ml Ni-NTA Superflow resin equilibrated with Buffer F for 30 min at RT, and then loaded onto a 10 × 1 cm column. The resin was washed sequentially with: 20 ml Buffer F; 20 ml Buffer F with 0.1% Triton X-100; and 10 ml Buffer G (50 mM MOPS, 9 M urea, 100 mM KCl, 10% glycerol, 5 mM Mg acetate, 50 mM glycine-glycine, and 5 mM DTT, pH 8.0) with 10 mM imidazole. Proteins were eluted with Buffer G with 250 mM imidazole. Protein concentration was quantified using ɛ = 2.06 × 10 M cm (), and solutions were diluted to 50 ng/ml with Buffer G. Proteins were folded by dialyzing (10 kD molecular weight cut-off) out the urea using a four-step gradient (9 M → 7 M → 5 M → 3 M; ∼3 h per step) in the dark (typical yield was 95–99% folded protein). The dialyzate was centrifuged at 15,000 for 30 min. Folded protein was recovered from the supernatant using 5 ml Ni-NTA Superflow resin equilibrated with 50 mM MOPS, 100 mM KCl, 5 mM Mg acetate, and 5 mM DTT, pH 8.0. The Ni-NTA–adsorbed GFP proteins were washed and eluted identically as described above for the purification of pre-SufI. Typical yield was ∼2 mg protein/L of culture.
Standard in vitro translocation assays used a 35-μl reaction volume containing 50 nM pre-SufI and 4 mM NADH in Translocation Buffer (TB; 5 mM MgCl, 50 mM KCl, 200 mM sucrose, 57 μg/ml BSA, 25 mM MOPS, and 25 mM MES, pH 7.0).
= 5). After a 30 min incubation at 37°C, reactions were quenched in an ice bath for 2 min. Samples were digested with 0.73 mg/ml proteinase K for 40 min at RT. Digestions were quenched with 68 mM PMSF, diluted twofold with 2× Gel Buffer (4% SDS, 10% glycerol, 0.04% bromophenol blue, 0.4% β-mercaptoethanol, 10 M urea, and 200 mM Tris, pH 6.8), and incubated in a boiling water bath for 10 min. Samples were centrifuged briefly at 16,000 , and then were resolved by 8% SDS-PAGE with known standards. Gels were electroblotted onto PVDF membranes and immunoblotted with SufI antibodies. For spSufI-GFP and spTorA-GFP translocation assays, both protease-treated and untreated IMVs were sedimented at 100,000 at 4°C for 15 min (or 16,100 at 4°C for 45 min). The IMVs (pellets) were washed with 500 μl TB with 20 μg/ml BSA, resuspended in 35 μl TB containing 68 mM PMSF, resolved by 10% SDS-PAGE with known standards, and immunoblotted using GFP antibodies.
The TatB protein has a single transmembrane domain near the N terminus and a C-terminal cytoplasmic domain (). TatB is unlikely to undergo topology inversion (), as has been reported for TatA (; ). The percentage of IMVs with an inside-out orientation was determined based on the protease accessibility of the TatB C-terminal domain using TatB antibodies raised against peptides within this domain (). IMVs in TB without BSA were incubated with 2 mg/ml proteinase K at RT for 30 min. Digestions were quenched with 20 mM PMSF, diluted twofold with 2× Gel Buffer, and incubated in a boiling water bath for 10 min. Samples were centrifuged briefly at 16,000 , and then were resolved by 18% SDS-PAGE with known standards. Gels were electroblotted onto PVDF membranes and immunoblotted with TatB antibodies.
The presence of ΔpH and Δψ gradients across IMV membranes was determined by fluorescence spectroscopy using 2.5 μM quinacrine (EX = 420 nm, EM = 510 nm) and 100 nM oxonol VI (EX = 610 nm, EM = 645 nm), respectively (). IMVs were preincubated in TB at 37°C for ∼5 min before the addition of 4 mM NADH or 4 mM ATP. Conditions were identical to those used for gel-based transport assays. Control experiments indicated that the effects observed when ionophores were added were not due to dilution or solvent.
All errors are standard deviations.
Table S1 summarizes the DNA primers used for plasmid construction. Fig. S1 shows the effect of IMV formation method and transport buffer on transport efficiency. Fig. S2 shows the transport efficiency dependence on energy source. Fig. S3 shows the effect of high valinomycin concentration on ΔpH and Δψ. Online supplemental material is available at . |
Sphingolipids comprise a relatively small but vital fraction of the mammalian membrane lipids (). Sphingomyelin (SM) carries a phosphocholine headgroup on a ceramide backbone and occurs in every mammalian cell, just like glucosylceramide (GlcCer). GlcCer serves as the basis for a highly polymorphic set of complex glycosphingolipids (GSLs). The unique physicochemical properties of sphingolipids allow different modes of interaction with their environment. Sphingolipids are concentrated at the cell surface and endocytotic membranes, where their bulk presence provides the membranes with chemical and mechanical stability. In addition, sphingolipids have the tendency to cluster with cholesterol in an environment of glycerolipids. The roles of sphingolipids in protein sorting, signaling, and membrane deformation may therefore be explained by their ability to form lateral domains that specifically include or exclude membrane proteins. In addition, the broad diversity in glycosidic structure allows individual GSLs to interact specifically with proteins, including viral and bacterial pathogens like Shiga and cholera toxin. The mode of action of sphingolipids in cellular processes depends on their concentration in the various subcellular organelles and the trans-bilayer and lateral distribution in those membranes. How cells sense and control the sphingolipid concentration of their membranes is largely unknown, but the spatial organization of metabolism, action of translocators, and selectivity of transport are three important determinants that are intrinsically linked. Many but not all enzymes of sphingolipid metabolism have been identified, some of them only very recently: the SM synthase (SMS) family (; ) and two nonlysosomal glucocerebrosidases (; ; ). One contemporary challenge is to unravel how sphingolipid metabolism is organized and controlled in the cellular context.
Ceramides are synthesized in the ER by various ceramide synthases (). Some cell types express the galactosylceramide (GalCer) synthase (GalCS), which acts on the lumenal side of the ER (, ). However, all other synthetic enzymes of ceramide-containing lipids are located in the Golgi, with the exception of SMS2 at the plasma membrane (). Newly synthesized ceramides are transported from the ER to the Golgi, where they are converted by the GlcCer synthase (GCS) and SMS1. An important question is how cells regulate the ceramide supply to these enzymes. A first clue has come from the finding that the synthesis of SM but not GlcCer depended on ceramide transport to the trans-Golgi by ceramide transport protein (CERT; ), a pathway that is regulated via phosphoinositides, sterols, and CERT phosphorylation (). However, both SMS1 and GCS have been localized biochemically to the cis-medial Golgi, whereas GCS has also been assigned to pre-Golgi and trans-Golgi membranes (; , ; ; ). The current agreement is that GlcCer is synthesized on a cytosolic surface and translocates across the Golgi membrane for higher GSL synthesis in the late Golgi (, ).
However, the consensus picture is based on evidence obtained before the recent identification of early enzymes in sphingolipid metabolism and transport. In addition, the bulk of the lipid data has been collected using short-chain sphingolipid analogues by the lack of tools to study natural sphingolipids. Such short-chain analogues have much higher off rates from membranes and spontaneously exchange between membranes (). In addition, they are less well ordered in the membrane (). Therefore, we have addressed the following questions: Where are the various enzymes situated along the Golgi stack? Is natural GlcCer translocated across the Golgi membrane by multidrug transporters ()? Is there a function for the glycolipid-binding proteins glycolipid transfer protein (GLTP) in the cytosol () and FAPP2 on the trans-Golgi () in GSL metabolism and transport (; ; , )? Using a novel assay for the transmembrane translocation of natural GSLs, we report that GlcCer can reach the plasma membrane via nonvesicular transport and translocates to the cell surface. In contrast to the short-chain lipid analogues, natural GSLs are not translocated by the multidrug transporters but by a novel mechanism that, in turn, does not recognize short-chain lipids. In addition, we have uncovered a pathway used by most GlcCer to reach the Golgi lumen. It involves transport from the Golgi to the ER by the glycolipid-binding protein FAPP2.
Because the intra-Golgi localization of SMS1 and GCS is a matter of controversy (see Introduction), we first addressed their intracellular localization by immunofluorescence and immuno-EM (IEM). Attempts to localize the endogenous enzymes of mammalian cells with specific antibodies have failed so far, most likely because of their low expression levels. As an alternative, we ectopically expressed epitope-tagged constructs in HeLa cells and determined their cellular distribution using fluorescence microscopy ( and Fig. S1 A, available at ) and EM (). In the confocal fluorescence microscope, both GCS and SMS1 displayed a virtually identical staining pattern with endogenous GM130, a cis-Golgi marker (), in transiently transfected HeLa cells (). Neither SMS1 nor GCS was found in the nuclear envelope and ER. To determine the intra-Golgi localization of GCS and SMS1, HeLa cell lines stably expressing these enzymes were analyzed by IEM. Golgi stacks containing five cisternae were used for statistical analysis, and endogenous GM130 was used as a cis-Golgi reference for every image. The intra-Golgi distribution of a protein was determined by counting all gold particles over a Golgi stack ( = 20 per protein), after which the number of gold particles found over a specific cisterna was expressed as a percentage of total gold (). Indeed, GM130 was restricted mostly to the cis-most cisterna (81%), with only a little label in the next cisterna (). In contrast to GM130, labeling for GCS was found in all five cisternae but with higher concentrations in cisternae 3–5 (medial–trans). SMS1 was mostly present at the trans-side of the Golgi and clearly peaked in the fourth cisterna. Quantitation showed that GCS and SMS1 had a substantially differential distribution: almost 50% of GCS localized to the first three cisternae, whereas only 33% of SMS1 was found there. Notably, SMS1 and GM130, which showed almost complete overlap by fluorescence microscopy, displayed different intra-Golgi distributions by IEM: GM130 was concentrated at the cis-side, and SMS1 was concentrated at the trans-side of the Golgi.
Lactosylceramide (LacCer) synthesis not only depends on the presence of GlcCer but also on localization of the LacCer synthase (LCS) and the UDP-Gal transporter (UGT). LCS was found in all cisternae but peaked in cisternae 4 and 5. The location of the UGT was more restricted to the medial–trans-Golgi and peaked in cisternae 3 and 4. LacCer is further converted to GM3 (sialyllactosylceramide) by the transfer of sialic acid from CMP–sialic acid. The GM3 synthase (GM3S) and the CMP–sialic acid transporter (CST) localized exclusively to the Golgi stack, predominantly to the trans side (). Although a 100-fold overexpression of a Golgi-resident protein had no effect on the distribution of this protein within the HeLa Golgi (), overexpression of a tagged version of a protein could lead to mislocalization. As one internal control, we therefore determined the localization of the UDP-GlcNAc transporter (NGT), which provides the medial-Golgi GlcNAc transferase with its substrate. It localized to the three medial cisternae with only little labeling in the first and last Golgi cisternae (). Collectively, the data indicate that the enzymes synthesizing LacCer and GM3 show a preferential medial–trans-localization, like SMS1, whereas the GCS is more evenly distributed over the Golgi stack.
Because GlcCer is synthesized at the cytosolic aspect of the Golgi (see Introduction), we next addressed the question of whether GlcCer directly flips to the lumenal side, where it can be used for complex GSL synthesis. Exogenously added -6-NBD-aminohexanoyl (C-NBD)–GlcCer was readily converted to C-NBD–LacCer in a postnuclear supernatant (PNS) of HeLa or CHO cells (; ; ; ). The translocation was 70% reduced by PSC833, an inhibitor of the ABC transporter ABCB1, a translocator for C-NBD–GlcCer (), and vanadate, which blocks the nucleotide-binding domain of ABC transporters. To test whether natural GlcCer translocates across the Golgi membrane, we monitored the conversion of exogenous [H]GlcCer to LacCer after insertion into the cytosolic leaflet of membranes in the HeLa PNS by the small cytosolic GLTP (Fig. S2, A and B; available at ). However, only a little [H]LacCer synthesis was observed unless the Golgi had been fused to the ER by brefeldin A (BFA) before the experiment, in which case the synthesis of [H]LacCer but not C-NBD–LacCer increased fivefold (). The translocation of [H]GlcCer was not inhibited by PSC833 or indomethacin, inhibitors of the ABC transporters ABCB1 and ABCC1, nor by concanamycin A, a vacuolar proton pump inhibitor. ABCB1 and -C1 translocate short-chain GlcCer (; ), but the translocation of natural GlcCer across the ER–Golgi membrane was independent of these multidrug transporters.
To see whether GlcCer could translocate toward the outer leaflet of the plasma membrane, from where it can reach the site of LacCer synthesis (; ), we labeled the newly synthesized lipids with [C]palmitate and, after various times, extracted the GSLs from the cell surface by GLTP. Based on previous experience, the assay was optimized starting from a 1,000-fold higher GLTP concentration than that used for the delivery of GlcCer to PNS membranes in , and the standard assay conditions were defined as 45 min at 37°C with 1.5 mg/ml GLTP (Fig. S2, C and D). The addition of acceptor liposomes to the medium inhibited rather than stimulated GlcCer extraction by GLTP (Fig. S2 E), most likely because GLTP interacted with the liposomes even if they contained no GSLs, thereby lowering the free GLTP concentration. Under these conditions, GLTP extracted 30–50% of the radiolabeled GlcCer and 20–70% of the GM3 from various cell types in 45 min but <0.5% of the SM and glycerophospholipids (). To check whether the phospholipid distribution across the membrane was affected by the extraction of the GSLs, exposure on the cell surface of inner leaflet phosphatidylserine was measured using FITC-labeled annexin V. No substantial annexin binding was observed in the GLTP-treated cells as compared with a positive control (). BFA induces fusion between the Golgi and ER and blocks vesicular traffic from the merged ER–Golgi to the plasma membrane. In mouse fibroblast (MF) cells, BFA did not affect GM3 synthesis (Fig. S4, available at ) but fully blocked its transport (), which is in line with the relocation of GM3S to the ER and the absence of a nonvesicular pathway for GM3 transport (). The same was observed for GalCer sulfate (SGalCer) in the oligodendrocytic D6P2T cells (). In contrast, GlcCer transport persisted in the presence of BFA. GalCer (and galactosyldiacylglycerol), which is synthesized in the ER lumen of CHO cells transfected with GalCS and in the ER lumen of D6P2T cells, behaved identically to GlcCer.
The fact that GlcCer and GM3 were synthesized in the same compartment in the presence of BFA () and that GlcCer but not GM3 reached the cell surface under these conditions implied that GlcCer had reached the plasma membrane in the absence of vesicular traffic. Because natural GlcCer, in contrast to short-chain C-NBD–GlcCer, does not readily exchange through the cytosol, we tested whether transport in the presence of BFA was mediated by a transfer protein. FAPP2, a protein that shares the glycolipid-binding domain with GLTP, has been found to be ubiquitously expressed (). However, the knockdown of FAPP2 in MEB4 cells had no effect on the transport of newly synthesized GlcCer to the cell surface in the presence of BFA (). Likewise, a knockdown of GLTP in D6P2T cells did not reduce the transport of GlcCer (or GalCer; not depicted) to the plasma membrane in the presence of BFA (), indicating that transport from the merged ER–Golgi to the plasma membrane in the presence of BFA was not mediated by these proteins or that they are redundant. On the other hand, GLTP overexpression stimulated the transport of GlcCer (and GalCer) in the presence of BFA twofold, showing that GLTP in vivo is able to fulfill this function (). Unexpectedly, FAPP2 knockdown in MEB4 cells reduced transport to the plasma membrane by 30% in the absence of BFA, and the knockdown of GLTP in D6P2T cells reduced the transport of GlcCer (and also of GalCer; unpublished data) by 30–50%, implying that >30–50% of the GlcCer that reached the cell under normal conditions had been transported by FAPP2 or GLTP followed by translocation across a post-Golgi membrane. We did not succeed in making a double knockdown for FAPP2 and GLTP, possibly because this is a lethal combination. Transport of GM3 in MEB4 cells was not reduced by FAPP2 knockdown. As before, it was effectively blocked by BFA (unpublished data).
We have previously reported that a variety of short-chain GlcCer analogues, including C-NBD–GlcCer, was transported to the cell surface by the multidrug transporters ABCB1 and -C1 (). Therefore, we addressed whether the transport of natural GlcCer to the cell surface in the presence of BFA () was mediated by these ABC transporters by using an MF cell line derived from the Abcb1a/Abcb1b/Abcc1 triple knockout (TKO) mouse () vs. TKO cells stably transfected with human ABCB1 (MDR1). The transport of C-NBD–GlcCer was strongly reduced in TKO cells (down to zero with BFA) and was partially restored by transfection with ABCB1 (), demonstrating that the translocation of C-NBD–GlcCer across the plasma membrane of MF cells was fully the result of Abcb1 and Abcc1. In line with this, the C-NBD–GlcCer translocation in MF cells was reduced by specific inhibitors (PSC833 for Abcb1a/b) or general inhibitors (glibenclamide and vanadate) of ABC transporters (). In strong contrast, knockout of the multidrug transporters had no effect on the transport of natural GlcCer either in the absence or presence (unpublished data) of BFA. Transfection of the TKO cells with ABCB1 did not increase the translocation of natural GlcCer, nor was there a substantial effect of the various inhibitors, with one exception: concanamycin A, a specific inhibitor of the vacuolar proton ATPase (), essentially abolished the translocation of natural GlcCer without effect on C-NBD–GlcCer (). Also, in the absence of BFA, GlcCer transport to the cell surface was virtually abolished by concanamycin A (), which cannot be explained by an inhibition of vesicular traffic as indicated by the mild reduction in GM3 transport. Concanamycin had no effect on the synthesis of GM3 from newly synthesized GlcCer (unpublished data).
Because concanamycin did not affect the rate of GM3 synthesis, GlcCer destined for GM3 synthesis must have followed a different transport pathway independent of the post-Golgi translocation that was inhibited by concanamycin. Therefore, we tested the possibility that this GlcCer reached the site of LacCer synthesis in the lumen of the Golgi via the ER. To test whether GlcCer is present in the ER lumen, we introduced an enzyme into the ER lumen, the SGalCer synthase, that transfers a sulfate from 3′-phosphoadenosine 5′-phosphosulfate (PAPS) to GalCer and GlcCer () but that normally acts in the trans-Golgi in conjunction with the PAPS transporter (). A chimera (SGCS [chimeric protein consisting of an HA-tagged PAPS transporter and SGalCer synthase]) was constructed of the SGalCer synthase to an HA-tagged PAPS transporter. In MEB4 cells, SGCS showed a diffuse reticular staining and clear staining of the nuclear envelope. The patterns of SGCS and the cis-Golgi marker GM130 were mutually exclusive (). In accordance with its location in the ER, SGCS remained sensitive to EndoH digestion (), implying the presence of high mannose -glycans. The molecular mechanism retaining the SGCS in the ER has remained unclear. It may contain a previously hidden ER retention signal, or maybe the complex is partially unfolded. When SGCS-transfected MEB4 cells were labeled with [S]HSO, one major radioactive band appeared on TLC plates that ran faster than SGalCer (). When GlcCer and [S]PAPS were added to a PNS of MEB4 cells transfected with the SGCS construct, a lipid product was made that ran above SGalCer. Synthesis depended on GlcCer, on PAPS, and on the transfection. A lipid with the same mobility on TLC was found in D6P2T cells. The band disappeared when the cells had been incubated with the GCS inhibitor -butyldeoxygalactonojirimycin, identifying the unknown lipid as SGlcCer ( and Fig. S3 A, available at ). In addition, it was absent after a GCS knockdown but not an LCS knockdown (Fig. S3 B), which efficiently inhibited the synthesis of LacCer (not depicted). Finally, similar amounts of the [S]lipid were synthesized in SGCS-transfected CHO and mutant CHO-lec8 cells (Fig. S3 C), which have threefold reduced LacCer levels (). Thus, GlcCer reached the lumen of the ER, where it was converted to SGlcCer by SGCS.
Selective GlcCer transport from the cytosolic surface of the Golgi back to the ER would be most easily explained by the activity of a cytosolic transfer protein followed by a translocation across the ER membrane. Therefore, we tested the involvement of the glycolipid-binding protein FAPP2 in this pathway. When lamin and FAPP2 knockdown cells were transfected with SGCS and labeled with [S]HSO, SGlcCer synthesis in the FAPP2 knockdown cells was only 30% of that of the control (). In MEB4 cells, FAPP2 knockdown resulted in a twofold decrease in the synthesis of GM3 with a concomitant accumulation of GlcCer (). Similar results () were obtained in MDCKII cells, in which FAPP2 expression was reduced by retrovirus-mediated RNAi (). GalCer levels in these cells were normal, indicating that ceramide levels in the ER were unaffected. FAPP2 specifically affected the conversion of GlcCer to GM3 via LacCer. These experiments show that FAPP2, which binds to the Golgi (Fig. S1 A) via PI4P (), plays an important role in the transport of GlcCer to the sulfation site in the ER and to the site of LacCer and GM3 synthesis in the lumen of the Golgi. Concanamycin A did not affect GlcCer translocation across the ER membrane (), nor did it affect GM3 synthesis from newly synthesized GlcCer (not depicted), showing that ER translocation occurred by a different mechanism and that the pathways of newly synthesized GlcCer to the Golgi lumen and to the plasma membrane are independent.
Until recently, sphingolipid assembly in the Golgi was thought to be organized according to the same simple linear model as glycoprotein processing (). Because it is not water soluble, ceramide synthesized in the ER would be transported by the secretory pathway to the cis-Golgi, where it would be converted to SM by SMS (; ) and to GlcCer by GCS (; ). Subsequently, GlcCer would encounter the enzymes of complex GSL synthesis in an ordered array along the Golgi stack (). The finding that GlcCer is synthesized on the cytosolic surface but converted to LacCer in the lumen forced the questions whether and how GlcCer crosses the Golgi membrane. This seemed solved by the finding that short-chain GlcCer analogues were able to cross the Golgi membrane (, ; ), the identification of the multidrug transporters ABCB1 and -C1 as floppases for these molecules (), and a correlation between ABCB1 activity and complex GSL synthesis in living cells (). However, the mechanism of SM synthesis is much more intricate in that SMS receives its substrate via the cytosolic transfer protein CERT, which binds to ER and trans-Golgi and is regulated at various levels via phosphorylation (; ; ).
Also, the synthesis and processing of GlcCer are more sophisticated than suspected (): The GCS is concentrated in the trans-Golgi, not in the cis-Golgi, and its activity was inhibited more than twofold upon CERT knockdown. In contrast to short-chain GlcCer, natural GlcCer did not flop efficiently across the Golgi membrane and was not a substrate for the multidrug transporter ABCB1. Instead, newly synthesized GlcCer reached the outside of the plasma membrane by a nonvesicular transport pathway and was translocated via a mechanism that was inhibited by concanamycin A, an inhibitor of the vacuolar ATPase, suggesting the involvement of a proton gradient. Finally, most GlcCer reached the LCS in the Golgi lumen via the ER, with a role for the trans-Golgi glycolipid-binding protein FAPP2 in shuttling GlcCer to the ER. This suggests that GlcCer may play a role in transport and sorting events at the ER. In addition, FAPP2 appears to regulate complex GSL synthesis. The finding that multiple pathways remove GlcCer from the cytosolic surface of the trans-Golgi suggests the possibility that GlcCer exerts a physiological function at that location.
BSA, fraction V, and other chemicals were purchased from Sigma-Aldrich unless indicated otherwise and were used in the highest purity available. Silica TLC plates were obtained from Merck, organic solvents were purchased from Riedel de Haën, and cell culture media, reagents, and FCS were obtained from PAA. Cell culture plastics were purchased from Costar. Tran[S]label (>37 TBq/mmol), [S]HSO (74 MBq/mmol), D-[1-C]galactose (1.8 GBq/mmol), [U-C]palmitic acid (18 GBq/mmol), [9,10(n)-H]palmitic acid (1.5 TBq/mmol), and L-[3-C]serine (1.9 GBq/mmol) were obtained from GE Healthcare and MP Biomedicals. Lipids and lipid standards were obtained from Avanti Polar Lipids, Inc. C-NBD–fatty acid was purchased from Invitrogen. C-NBD– and [H]palmitoyl-GlcCer were synthesized as described in supplemental Materials and methods (available at ). EST clones were purchased from RZPD. Rabbit antisera against the V5 and HA epitope (Y11) were obtained from Sigma-Aldrich and Santa Cruz Biotechnology, Inc., respectively. The rabbit peptide antibody against mouse GLTP was a gift from K. Aikawa-Kojima (Ochanomizu University, Tokyo, Japan). The rabbit serum against GM130 was a gift from E. Sztul (University of Alabama, Birmingham, AL). Mouse anti-V5 antibody was purchased from Invitrogen, mouse mAb 16B12 anti-HA was obtained from BabCO, mouse anti-GM130 antibodies were obtained from BD Biosciences, and mouse anti–glyceraldehyde-3 phosphate dehydrogenase was purchased from Applied Biosystems. Fluorescent secondary goat antibodies were obtained from Jackson ImmunoResearch Laboratories, Inc. HRP-conjugated secondary goat anti–rabbit IgG was purchased from DakoCytomation. Recombinant GLTP was produced as described in supplemental Materials and methods.
Plasmids are described in . ORFs were amplified by PCR using cDNA clones as a template. PCR products were gel purified and cloned into mammalian expression vectors containing a sequence encoding for a triple HA tag (3*YPYDVPDYA) at the 5′ or 3′ end or a single HA or V5 tag (GKPIPNPLLGLDST) at the 3′ end. SGCS is a chimeric construct containing a triple HA tag, the rat PAPS transporter, a double myc tag as spacer (2*EQKLISEEDL), and the mouse SGalCer synthase. Primers containing RNAi sequences of FAPP2 and lamin were inserted between the BglII and HindIII sides of the RNAi plasmid pKoen (). All synthetic constructs were verified by restriction analysis and dye termination sequencing of both strands.
HeLa, CHO, and CHO-lec8 cells were purchased from American Type Culture Collection. D6P2T cells were a gift from S. Pfeiffer (University of Connecticut Medical School, Farmington, CT; ). CHO-GalCS cells have been described previously (as CHO-CGalT; ). MEB4 cells were purchased from the Institute of Physical and Chemical Research Cell Bank. The MF and TKO lines (; ) were obtained from P. Borst (Netherlands Cancer Institute, Amsterdam, Netherlands). MDCK strain II cells infected with recombinant retroviruses carrying short hairpin RNA against FAPP2 have been described previously (). All cells were grown in DME, stable glutamine, 4.5 g/liter glucose, and 10% FCS at 37°C with 5% CO. Stable transfectants were grown in the presence of 200 U/ml hygromycin B or 0.6 mg/ml geneticin (G418). Cells were transfected using ∼1 μl LipofectAMINE 2000 and ∼0.3 μg DNA per squared centimeter of cells. For transient protein expression, the cells were used 1 d after transfection. For the generation of stable transfectants, the cells were trypsinized 24 h after transfection and divided over four 15-cm dishes in culture medium containing hygromycin B or G418. Medium was refreshed weekly, and, after ∼3 wk, individual colonies were selected for green fluorescence in the nucleus () and trypsinized using a metal cylinder.
Stably transfected HeLa cells were fixed with 2% PFA + 0.2% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4, and incubated for 4 h at room temperature followed by an overnight incubation at 4°C. After fixation, cells were rinsed with PBS and with PBS containing 0.02 M glycine, scraped, and pelleted in 12% gelatin. Small blocks of the embedded cell pellets were infiltrated overnight with 2.3 M sucrose, mounted on aluminum pins, and frozen in liquid nitrogen. Ultrathin cryosections were cut at −120°C, picked up with 1% methylcellulose and 1.2 M sucrose, thawed, and collected on grids. After washing with PBS containing 0.02 M glycine, sections were double labeled as described previously () with antibodies against the tags (mouse) and the cis-Golgi marker GM130 (rabbit). For sub-Golgi localization of the tagged enzymes, immunogold particles were quantified per cisterna. For each enzyme, 20 Golgi stacks were selected on double-labeled sections based on the following criteria: clear visibility of five cisternae per Golgi, which we denoted as cisterna 1–5 (), the presence of at least five gold particles for the specific tag, and the presence of GM130 label as a cis-Golgi reference. The cisterna containing the GM130 marker was located at one end of each Golgi stack and was denoted as cisterna 1. EM pictures of the selected Golgi stacks were made at a magnification of 30,000× or 40,000×. The number of gold particles for the specific tag was counted per cisterna and expressed as a percentage of the total gold particles within that Golgi stack. The results for each enzyme are expressed as the mean percentage ± SEM ( = 20) of immunogold label for each cisterna. The reliability of the sample size was determined by accumulating data (for a given enzyme) until mean percentages remained stable within 10% of the end value for the last added five Golgi stacks. 200–500 gold particles were localized for each protein. The significance of the peak value of the gold particle percentage in a given cisterna was determined by performing paired tests between cisternae with the lowest and highest percentage of gold label (P ≤ 0.05). Confocal laser-scanning microscopy was performed as described in supplemental Materials and methods.
24-well dishes with ∼2 × 10 MF, CHO-GalCS, or D6P2T cells were preincubated with various drugs (1 μg/ml BFA, 5 μM PSC833, 50 μM glibenclamide, and 1 mM vanadate) for 0.5 h and metabolically labeled in the presence of inhibitors with radioactive precursors or C-NBD–ceramide at 37°C. For natural lipids and proteins, cells were labeled with 24 kBq/ml [1-C]palmitate, 15 kBq/ml [3-C]serine, 3,600 kBq/ml [S]HSO, or 72 kBq/ml [C]galactose in culture medium for 1.5 h at 37°C and 5% CO. Serine labeling was performed in MEMα because of its lower level of serine. The medium was replaced by HBSS with 20 mM Hepes, pH 7.4 (HBSS′), containing purified GLTP to extract GSLs from the cell surface at 37°C. After varying times and GLTP concentrations (see Results; typically 45 min with 1.5 mg/ml GLTP ± BFA), the medium was removed, and the cells were washed twice with 0.5 ml HBSS′. Wash buffers and medium were pooled and centrifuged for 5 min at 800 to remove detached cells. The lipids were extracted from medium and cells as described in Lipid analysis. Transport is calculated as [C]- or [S]lipid in the medium as the percentage of the total amount of that lipid in medium plus cells. Effects of drugs or RNAi treatments on synthesis were determined by measuring the sphingolipid of interest against an internal standard like [C]phosphatidylserine in triplicate. For C-NBD–GlcCer labeling, cells were incubated for 1 h with 1 μM C-NBD–ceramide in HBSS′ + 1% BSA (wt/vol) to back-exchange newly synthesized C-NBD–lipids appearing at the surface at 37°C. The buffer was replaced by HBSS′ + 1% BSA and incubated for 0.5 h on ice. The lipids were extracted from the combined label and washing buffers and from the cells as described in the Lipid analysis section. Transport is expressed as C-NBD–lipid in the medium as the percentage of total C-NBD–lipid. Measurements were performed in triplicate. For protein transport analysis, cells were incubated with 800 kBq/ml [S]amino acids in culture medium for 135 min. Cells and media were analyzed as described in the SDS-PAGE and Western blotting section. Transport was calculated as the percentage of control.
15-cm dishes with ∼1.5 × 10 HeLa cells were preincubated in the presence or absence of 1 μg/ml BFA in culture medium for 0.5 h at 37°C in 5% CO. Cells were then washed twice with PBS and scraped in homogenizing buffer (120 mM K-glutamate, 15 mM KCl, 5 mM NaCl, 0.8 mM CaCl, 5 mM MnCl, 2 mM MgCl, 1.6 mM EGTA, 20 mM Hepes/KOH, pH 7.2, 1 μg/ml apoprotein, 1 μg/ml leupeptin, 1 μg/ml pepstatin, 5 μg/ml antipain, and 1 mM benzamidine). Cells were centrifuged at 300 for 5 min at 4°C, taken up in 800 μl of homogenizing buffer, and broken up by passing 10 times through a homogenizer (0.016 ball size; European Molecular Biology Laboratory). Nuclei and unbroken cells were removed from PNS by spinning for 10 min at 300 at 4°C. Soluble GLTP–[H]GlcCer complexes were prepared as follows: 0.4 MBq [H]GlcCer in 10 μl ethanol were injected into 1,200 μl of homogenizing buffer containing 12 μg GLTP. The mix was incubated for 20 min at 37°C and spun down for 5 min at 20,000 to remove aggregates. 100 μl PNS (0.25 mg/ml) was incubated for 1 h at 37°C in the presence or absence of BFA, with 100 μl GLTP–[H]GlcCer complexes, 2 mM UDP-Gal, and 2 mM ATP in a volume of 500 μl. Lipids were extracted as described in the next section, and LacCer synthesis was measured as the percentage of control.
Lipids were extracted and applied to TLC plates, which, when used to separate GalCer from GlcCer, had been dipped in 2.5% wt/vol boric acid in MeOH and dried (all as described previously; ). Lipids were generally separated by 2D TLC using in the first dimension either CHCl/MeOH/25% vol/vol NHOH/water (65:35:4:4 vol/vol) or CHCl/MeOH/0.2% aqueous CaCl (55:45:10 vol/vol) with the acidic solvent CHCl/MeOH/acetone/HAc/water (50:20:10:10:5 vol/vol) for the second dimension. C-NBD–lipids were analyzed by 1D TLC in the acidic running solvent. 1D TLC plates of [S]HSO-labeled cells and of PNS were developed in the CaCl mixture. Radiolabeled spots were detected by exposure of phosphorimaging screens and read-out on a Personal FX phosphorimager. TLC plates with fluorescent lipids were directly developed using a phosphorimager (STORM 860; Molecular Dynamics). Spots were identified by comparison with standards and quantified using Quantity One software (Bio-Rad Laboratories).
Cells were washed three times with PBS and were resuspended in protein sample buffer (200 mM Tris-HCl, pH 6.8, 3% wt/vol SDS, 12% vol/vol glycerol, 1 mM EDTA, 0.003% wt/vol bromophenol blue, and 50 mM DTT). Media were centrifuged at 3,000 to remove cell debris, and one-third volume of 4× protein sample buffer was added. Samples containing SGCS were incubated for 10 min at room temperature and for 0.5 h at 50°C. All other samples were heated for 5 min at 95°C and resolved by SDS-PAGE on 7.5% minigels. Radioactive gels were dried and analyzed by fluorography or a Personal FX phosphorimager (Bio-Rad Laboratories) using Quantity One software (Bio-Rad Laboratories). For Western blotting, nitrocellulose transfers were blocked for 1.5 h in PBS, 5% Protifar (Nutricia), and 0.2% Tween 20 (Blotto). Primary antibody incubations were performed for 1 h in Blotto. Detection was performed with HRP-conjugated secondary antibodies using enhanced chemiluminescence (GE Healthcare).
Fig. S1 shows that FAPP2 is localized in the perinuclear region of HeLa cells, where it partially colocalizes with the medial Golgi marker mannosidase II, whereas GLTP distributes all over the cytosol except for the nucleus. Knocking down CERT resulted in an 80% reduction in the synthesis of SM and a 70% reduced synthesis of GlcCer and GM3, which is in line with a trans-Golgi location of a large fraction of GCS. Fig. S2 shows the purified recombinant GLTP as a single band in a Coomassie-stained gel. The protein extracted C10-pyrene-GlcCer but not SM from membranes. Extraction of radiolabeled GlcCer from fibroblasts was concentration and time dependent, whereas exogenous liposomes inhibited the extraction. Fig. S3 shows that the product of the ER sulfotransferase construct SGCS was not SLacCer. Product synthesis was inhibited by the GCS inhibitor -butyldeoxygalactonojirimycin and by GCS knockdown but not by knocking down LCS. In addition, there was no reduced synthesis of the compound in Lec8 cells, which synthesize far less LacCer because they lack the galactose importer in their ER and Golgi. Fig. S4 shows that BFA does not inhibit GM3 synthesis but does inhibit the synthesis of Gb3 in HeLa cells. Fig. S5 shows that BFA fully inhibited protein secretion in melanocytes but that FAPP2 knockdown had no effect on protein secretion. Supplemental Materials and methods provides information about the synthesis of sphingolipids containing fluorescent or radiolabeled fatty acids, the purification of GLTP, and confocal laser-scanning microscopy. Online supplemental material is available at . |
Developing oocytes use both actin and microtubule cytoskeletal systems to construct and maintain internal landmarks that define the dorsal-ventral and anterior-posterior axes (; ; ; ). The and genes encode actin filament nucleation factors (), and mutation of either gene disrupts localization of the earliest known polarity determinants (; ). and were originally identified in the same genetic screen (), and loss of either results in the premature onset of microtubule-dependent fast cytoplasmic streaming during oogenesis, loss of oocyte polarity, and female sterility (; ). demonstrated a genetic interaction between and by showing that premature cytoplasmic streaming occurs in flies heterozygous for mutations in both genes. Mutation of profilin (Chickadee) or addition of the actin-depolymerizing toxin cytochalasin D (; ) also cause premature fast cytoplasmic streaming. Together, these data suggest that actin polymerization driven by Spire (Spir), Cappuccino (Capu), and Chickadee suppresses fast cytoplasmic streaming until the appropriate point in oogenesis ().
Consistent with genetic data, recently showed that Spir and Capu proteins interact directly. These authors found that the N-terminal region of Spir, which contains the kinase noncatalytic C-lobe domain (KIND) and a cluster of actin-binding WH2 domains (Wiskott-Aldrich syndrome protein homology domain 2), binds to the formin homology 2 (FH2) domain of Capu. Their data suggest that interaction is mediated by direct binding of the WH2 cluster to the FH2 domain. These authors report that the Spir–Capu interaction has no effect on actin nucleation by either protein but that interaction with Spir inhibits FH2-dependent cross-linking of actin filaments and microtubules. Based on these data, propose a model in which Spir and Capu inhibit premature cytoplasmic streaming by cross-linking microtubules to actin filaments in the oocyte cortex.
Interactions between Spir and Capu have been studied only in , but there is evidence linking the two proteins in other organisms. In sequenced metazoan genomes, Capu family formins appear only in organisms that also contain Spir family genes (). Mammals have two copies of each gene. Arthropods, including , contain at least one and one gene, whereas nematodes, such as , contain neither. Because nematodes diverged from arthropods long after Deuterostomes diverged from Protostomes, it appears that nematodes lost both genes at some point in their evolution. found that the patterns of and () expression are nearly identical in developing and adult mice.
We also note that Spir and Capu homologues are found in a variety of polarized cells, including and oocytes (Eg6 or Spir-2; ), mammalian eggs (Fmn2; ), neurons (; ), and polarized epithelial cells (formin-1; ). In oocytes, the mRNA of Spir-2 (Eg6) localizes to the vegetal cytoplasm and marks the posterior end of the developing embryo (). Knockout of Fmn2 in the mouse produces a maternal effect phenotype in which females are sterile as a result of mispositioning of the meiotic spindle ().
In this study, we investigate the molecular basis of the interaction between Spir and Capu and how the interaction influences actin nucleation. We find that Spir and Capu interact in vivo as well as in vitro. Similar to , we find that the Spir-WH2 cluster interacts with the Capu-FH2 domain. However, we also find that the Spir-KIND domain binds the Capu-FH2 domain with several orders of magnitude higher affinity than the WH2 cluster. This interaction has three functional consequences: the KIND domain potently inhibits actin nucleation by Capu; interaction between the KIND domain and Capu leads to enhanced actin nucleation by Spir; the KIND domain competes with actin filaments and microtubules for binding to the FH2 domain of Capu. The KIND–FH2 interaction is evolutionally conserved, as we observe the same results using both and mammalian Spir and Capu family proteins. The direct interaction of Spir and Capu, the fact that the expression patterns of Spir-1 and Fmn2 exactly overlap in the developing nervous system (), and the fact that their evolutionary conservation appears to be linked lead us to speculate that Spir and Capu function as part of a complex whose job is to assemble cytoskeletal landmarks for polarity in many systems.
We find that full-length Spir is sufficient to rescue the mutant phenotype. The FlyBase Genome lists four gene products, which are all derived from a single gene: Spir-PA, -PB, -PC, and -PD (GenBank/EMBL/DDBJ accession nos. , , , and , respectively). Spir-PA and -PB are ∼1,000 amino acids and differ by a 29–amino acid insert. Spir-PD is equivalent to the first 584 amino acids of Spir-PA, whereas Spir-PC is approximately the C-terminal half of Spir-PA. detected two distinct bands in Northern blots of RNA from fly oocytes, which they named Spire long form and short form (GenBank/EMBL/DDBJ accession nos. and ). These correspond to Spir-PA/PB and Spir-PD, respectively. There is no published evidence for the expression of Spir-PC. We made transgenic flies that express monomeric RFP (mRFP)–tagged full-length Spir (we refer to PA/PB as full length) in the germline. The localization of Spir fusions was enriched in the oocyte cortex and diffuse in the oocyte cytoplasm (Fig. S1 A, available at ). expressed GFP fusions of two putative spliceoforms of Spir (GFP-SpirC and GFP-SpirD) in egg chambers. Consistent with our observation, they found both proteins associated with the oocyte cortex. They also observed GFP-SpirC in punctae and GFP-SpirD diffuse throughout the oocyte. flies are putative nulls with the stereotypical phenotypes, including female sterility. Both mRFP-Spir and Spir-mRFP rescue female sterility in flies, demonstrating that the full-length transcript is sufficient during oogenesis and that the shorter spliceoforms are not essential.
We next determined the localization of endogenous Spir in wild-type egg chambers by immunofluorescence microscopy (). To distinguish specific from nonspecific staining, we compared wild-type egg chambers with those of homozygous mutants (Fig. S1, B and C). From early oogenesis through stage 9, Spir localizes specifically to the actin-rich cortex of the oocyte (). We cannot confirm the diffuse cytoplasmic localization observed in mRFP-Spir flies with immunofluorescence because we also observe it in flies (Fig. S1 B). At stage 10, near the onset of cytoplasmic streaming, Spir staining disappears from the cortex (). Because the loss of Spir produces precocious cytoplasmic streaming, this result suggests that cytoplasmic streaming is normally triggered by the destruction or displacement of Spir from the oocyte cortex.
Spir and Capu have been shown to interact in vitro (). To determine whether these proteins interact in vivo, we immunoprecipitated Capu from wild-type ovary lysates and probed the precipitated material with anti-Spir antibodies. Spir coimmunoprecipitates with Capu but not with beads alone or beads with nonspecific IgG, indicating that Spir and Capu are part of a protein complex in vivo ( and S1 E).
To further examine the in vivo interaction between Spir and Capu, we studied their subcellular localizations when expressed individually or together in NIH 3T3 fibroblasts. We compared and mammalian Spir and Capu family proteins and used truncation mutants to map domains required for interaction. As we reported previously, full-length Spir localizes to punctae () that correspond to the trans-Golgi network, post-Golgi vesicles, and recycling endosomes (). Full-length Capu (myc tagged) is distributed uniformly throughout the cytoplasm (). Coexpression of Spir together with Capu induces a striking change in Capu localization. Capu shifts from a diffuse distribution to discrete punctae that coincide with the localization of Spir (). Using truncation mutants, we found that the N-terminal portion of Spir and the C-terminal portion of Capu are necessary for colocalization (Fig. S2 B, available at ). We then coimmunoprecipitated EGFP-Capu-FH2 with myc-Spir-NT from cells expressing both constructs, demonstrating that colocalization corresponds with interaction ().
The N-terminal half of the Spir proteins, which is necessary for the colocalization of Spir and Capu, contains two different structural motifs: one KIND domain and a cluster of four WH2 domains (). mapped the interaction between Spir and Capu to the Capu-FH2 domain and the Spir-WH2 cluster. They also reported a weak interaction with the ∼150–amino acid region adjacent to the WH2 cluster containing the C-terminal half of the KIND domain. However, they did not test for an interaction with the intact KIND domain. We found that the KIND domain is sufficient for colocalization with an EGFP-tagged Capu-FH2 (). We targeted the KIND domain to membranes using a C-terminal Ha-Ras-CAAX motif (). When expressed in NIH 3T3 fibroblasts, KIND-CAAX localizes to the plasma membrane and to cytoplasmic spots (, red). Coexpression of an EGFP-Capu-FH2 led to colocalization with the membrane-targeted KIND (). We could not test the WH2 domain in this context because the CAAX motif did not effectively drive WH2 localization to the plasma membrane or distinct punctae (unpublished data).
We also observed the colocalization of mammalian Spir and Capu family proteins (Spir-1 and Fmn2; Fig. S2 C). The Spir-1–KIND and Fmn2-FH2 domains were sufficient to mediate this interaction ( and S2 C). The interaction is specific because a KIND domain from the protein very-KIND (VKIND-KIND-CAAX; ) does not colocalize with or pull down Fmn2-FH2, nor does the FH2 domain of the formin mDia1 (mDia1-FH2) colocalize with or pull down Spir-1–KIND (Fig. S2, D and E). These data suggest that the interaction between Spir and Capu family proteins is specific and conserved.
To further examine the interaction between Spir and Capu, we determined the affinity of purified KIND for purified Capu-FH1FH2 using fluorescence polarization anisotropy. Capu-FH2 and Capu-FH1FH2 behave similarly, but the longer construct is more stable, so for the majority of experiments, we used Capu-FH1FH2. We labeled an endogenous cysteine in KIND with AlexaFluor488 and measured changes in polarization anisotropy induced by Capu-FH1FH2. We determined the affinity by fitting the data with a quadratic binding curve (K = 1 ± 2 nM; ). To determine whether the label affected binding, we also determined the affinity of unlabeled KIND by using it to compete with the labeled protein (K = 5 ± 3 nM; , inset). The agreement indicates that attachment of the fluorophore has little effect on the interaction. The affinity measured using Capu-FH2 rather than FH1FH2 was nearly indistinguishable (K = 9 ± 6 nM; Fig. S2 F).
We found that the WH2 cluster binds weakly to Capu-FH1FH2. The addition of Capu-FH1FH2 to AlexaFlour488-labeled WH2 produced a saturable change in fluorescence intensity, so we used fluorescence intensity as a metric for binding. We determined an affinity of 2.4 ± 0.9 μM (), which is roughly three orders of magnitude weaker than the affinity of KIND for Capu-FH2. We could not measure the affinity of unlabeled WH2 by competition because higher concentrations of the WH2 domain produced dose-dependent light scattering. This probably reflects aggregation caused by the highly charged WH2 cluster.
We found that the stoichiometry of the KIND–FH2 complex is 2:2 (two KIND monomers/one FH2 dimer). Using velocity sedimentation and equilibrium centrifugation, we first determined that the KIND domains from human Spir-1 and Spir are both monomeric and highly asymmetric (). To measure the stoichiometry of the complex, we combined AlexaFluor488-labeled Capu-FH1FH2 with KIND at three different ratios and spun the mixtures to equilibrium at multiple speeds. We determined the equilibrium distribution of Capu-FH1FH2 by measuring the absorbance of the attached fluorophore. These data were best fit by a single-species model with a molecular mass close to that predicted for two KIND domains plus one Capu-FH1FH2 dimer (predicted, 223.6 kD vs. measured, 225 kD; ). The fact that the data fit a single-species model is consistent with a high affinity interaction between KIND and Capu-FH1FH2. We detected no evidence of the Capu-FH1FH2 dimer either free or bound to a single KIND domain.
Spir family proteins inhibit actin nucleation by Capu family formins ( and Fig. S3, available at ). Because Spir binds the nucleation domain of Capu, we used pyrene-actin fluorescence assays to determine the effect on Capu activity. Both Capu-FH2 and Capu-FH1FH2 promote rapid actin filament assembly. The addition of KIND caused a dose-dependent decrease in nucleation activity ( and S3 A). Because of local concentration effects, the addition of a weak interaction to a stronger one can have an effect on overall affinity. Thus, we also tested a mutant form of NTSpir (NTSpir[A*B*C*D*] from ), which includes both the KIND domain and the WH2 cluster but nucleates only very weakly (Fig. S3 B). By plotting the rate of nucleation versus the concentration of KIND ( and S3 D) and fitting the data with a quadratic binding curve, we determined inhibition constants (K). In all cases, Spir inhibited FH2-dependent nucleation by >90% and with comparable apparent affinities (5–10 nM; ). These Ks agree well with the K measured by polarization anisotropy, but, from our data, we cannot determine whether the binding of one or two KIND domains is required for inhibition. We observe the same effect with the mammalian proteins (Fig. S3 C). The K and K of the Spir-1–KIND–Fmn2-FH2 interaction are higher than observed with isoforms but, in general, agree with each other (300 ± 60 and 190 ± 40 nM; Figs. S2 G and S3 E).
Neither nor mammalian KIND affects spontaneous actin assembly (Fig. S4, A and B; available at ), demonstrating that the effect is specific to the activity of the Capu-FH2 domain. Also, Spir-1–KIND had no effect on actin nucleation by FH2 domains from Diaphanous family formins mDia1 and mDia2 (Fig. S4 C), indicating that the inhibitory effect of Spir-KIND domains is specific to Capu family formins.
Capu does not inhibit actin nucleation by Spir (). To assess the effect of Capu on Spir-dependent nucleation, we mutated Capu-FH1FH2. Mutating Ile-706 to Ala (analogous to Ile-1431-Ala in Bni1; ) almost completely abolishes nucleation activity. The addition of Capu-FH1FH2(I706A) to NTSpir enhanced nucleation activity (, green vs. blue traces). The effect increased with increasing concentrations of Capu-FH1FH2(I706A) until approximately equimolar concentrations of proteins were present. Further increases in Capu-FH1FH2(I706A) concentration decreased activity. This dose response is consistent with an enhancement mechanism dependent on the dimerization of Spir via the KIND–FH2 interaction. We observed the same effect with mammalian isoforms (unpublished data). Although we detect binding between the two domains, the Capu-FH1FH2 domain has no effect on actin nucleation by the WH2 cluster alone (, inset), confirming that enhancement depends on interaction between the KIND and FH2 domains.
When active Capu-FH1FH2 and NTSpir are combined, the measured nucleation rate reflects a combination of inhibition and enhancement activity (). For example, the rate of polymerization in the presence of 100 nM Capu-FH1FH2 and 250 nM NTSpir falls between the rates of either nucleator alone. Nucleation by 100 nM Capu-FH1FH2 has virtually no lag, whereas nucleation by 250 nM NTSpir has a marked lag (∼30 s). When the two are combined, a long lag is observed that is consistent with the potent inhibition of Capu-FH1FH2 nucleation (, inset). did not observe such synthetic activity when they combined Capu-FH2 and SpirD (equivalent to NTSpir). One possible explanation is that the KIND domain was not folded correctly in these experiments. When we treat NTSpir with denaturant (e.g., GnHCl) or the KIND domain is absent (WH2 alone), Spir retains nucleation activity, but Spir and Capu do not interact in the polymerization assay (, inset; and not depicted).
Spir-KIND competes with microtubules for binding to Capu-FH2 (; ). The FH2 domain of formins is known to bind microtubules in vitro and in vivo (). reported that Capu-FH2 cross-links actin and microtubules and that this activity is modulated by Spir. We also assessed the ability of Capu family formins to bind microtubules and tested the effect of the KIND domain on this interaction. We found that both Capu and Fmn2 cosediment with microtubules (Fig. S5 A, available at ), whereas KIND domains do not detectably bind microtubules (Fig. S5 B). We confirmed that Capu-FH1FH2 cross-links microtubules by examining solutions of taxol-stabilized microtubules mixed with Capu-FH1FH2 by fluorescence microscopy and by performing polymerization assays under conditions that require a factor that stabilizes or cross-links tubulin nuclei (Fig. S5, E and F; ). The addition of KIND to Capu and microtubules decreased microtubule binding by Capu in a dose-dependent manner (). About 2.5 μM KIND is necessary to compete half of the Capu-FH1FH2 away from 2 μM of polymerized tubulin, indicating that microtubules and KIND bind Capu-FH1FH2 with similar affinity. By fitting the data to a competition binding curve, we measured an affinity of Capu for microtubules of <1 nM (). We found no difference in competition with NTSpir versus KIND alone, indicating that the WH2 cluster does not contribute measurably to this inhibitory interaction (Fig. S5 C).
Spir-KIND also regulates actin bundling by Capu (). In addition to binding barbed ends, some formins also bind the sides of actin filaments and bundle them (; ). To test for actin bundling, we mixed 0.5 μM Capu-FH1FH2 with 2 μM phalloidin-stabilized actin. We observed bundles directly by fluorescence microscopy (Fig. S5 G) and indirectly with a low speed pelleting assay, in which only cross-linked networks or bundles of actin sediment (; ). The majority of the actin was bundled in the low speed pelleting assay. As a control, we used 0.5 μM α-actinin, a known actin cross-linker. Actin is in the supernatant when alone and in the pellet when α-actinin is added. Examination of the actin showed tight bundling in the presence of Capu-FH1FH2 similar to other formins and distinct from the loose networks created by α-actinin (Fig. S5 G; ; ). Capu-FH1FH2 bundles more effectively than α-actinin, most likely reflecting a difference in off rates. We measured the effect of Spir on actin bundling by mixing 0.5 μM Capu-FH1FH2 with 2 μM actin and a range of concentrations of KIND or NTSpir. We then performed high speed cosedimentation assays and low speed cross-linking assays. We quantified Capu-FH1FH2 in the supernatants and pellets as a function of KIND concentration. By fitting the data to a competition binding curve, we found that the K of Capu for the sides of actin filaments is 7 ± 1 nM (high speed) or 6 ± 2 nM (low speed; , B and C; and not depicted).
The and genes have been linked since their discovery in a genetic screen 17 yr ago. We find that the KIND domain of Spir binds with high affinity to the Capu-FH2 domain at a stoichiometry of 2:2 (two KIND monomers to one FH2 dimer). We also find that the WH2 cluster of Spir interacts with Capu-FH2 but that this interaction is three orders of magnitude weaker than that between the Capu-FH2 and the KIND domain. Although we detect binding between the two domains, the Capu-FH2 domain has no direct effect on actin nucleation by the Spir-WH2 cluster. However, if the KIND domain is present and correctly folded, binding of the FH2 dimer increased nucleation activity of the WH2 cluster. On the other hand, the KIND domain potently inhibits actin nucleation by the Capu-FH2 domain. Constructs containing both the KIND and WH2 cluster do not enhance the inhibition of Capu-FH2–mediated actin nucleation or microtubule bundling over that observed for the KIND domain alone. For these reasons, we propose that the KIND–FH2 interaction is more physiologically relevant than the WH2–FH2 interaction. Additional structural and functional studies of the KIND domain are required to determine how many KIND domains are required to inhibit actin nucleation and to compete for actin and microtubule binding.
We originally identified the KIND module as a conserved region in the N-terminal half of Spir proteins () and named the region based on its sequence similarity to the C-lobe of the protein kinase fold (). The KIND domain is found only in metazoa, and its consensus sequence lacks catalytic residues required for kinase activity. Because the substrates of protein kinases interact with α-helical regions in the C-lobe (; ) we hypothesized that the KIND domain evolved from a functional kinase into a protein–protein interaction domain. The discovery that the Spir KIND domains bind specifically to Capu family FH2 domains supports this hypothesis.
What role do Spir and Capu play in oogenesis? Spir disappears from the oocyte cortex at stage 10, when rapid streaming normally begins and its absence in mutant flies leads to premature streaming. This strongly suggests that Spir plays an inhibitory role in rapid streaming. We do not yet know whether endogenous Capu has the same restricted temporal pattern observed for Spir. This information will be essential to understanding the nature of the Spir–Capu complex and its role during oogenesis. We find that Spir and Capu interact in the oocyte, and found that GFP fusions of these protein both exist at the oocyte cortex, placing them in an ideal location to coordinate actin and possibly anchor microtubules. Rapid streaming is, in part, characterized by bundling and movement of microtubules. Capu bundles microtubules, which is an activity regulated by Spir. If Spir is removed at stage 10 but Capu remains, Capu could play a role in reorganizing the microtubule cytoskeleton and possibly coordinating it with the actin cytoskeleton. A complete understanding of how Spir and Capu achieve this coordination depends on knowing when and how the Spir–Capu complex is regulated.
Capu and other members of the formin family nucleate de novo actin filament assembly and remain associated with elongating barbed ends of newly formed filaments (; ). The activity of most formin family proteins is regulated by an autoinhibitory interaction between an N-terminal sequence (the Diaphanous inhibitory domain [DID]) and a C-terminal sequence (the Diaphanous autoinhibitory domain [DAD]). Small G proteins of the Rho family stimulate nucleation activity by binding to the DID domain and disrupting its interaction with DAD. However, Capu family formins lack both DID and DAD domains (). In fact, did not observe autoinhibition when combining the N terminus of Capu with the FH2 domain, as has been observed for mDia1 (). Our results argue strongly that Capu activity is regulated in trans by interaction with Spir.
The mechanism of actin nucleation by Spir is very different from that of formins like Capu. Spir binds four actin monomers using four closely apposed binding sites and then assembles them into a filament nucleus. After nucleation, Spir proteins remain associated with the slow-growing pointed end of the new filament. If Spir and Capu always function together as a single filament-forming complex, we suggest that their activities might synergize. One intriguing possibility is that Spir nucleates filaments whose free barbed ends are then handed off to Capu. Such a mechanism would enable the independent control of filament nucleation and barbed end binding. The tight binding that we measure suggests that Spir and Capu may not dissociate upon nucleation but that actin and microtubules do bind competitively. This idea begs two important questions: Does the activation of Capu require the complete dissociation of Spir, or can the two proteins function together as a single filament–forming unit? How is the Spir–Capu interaction modulated by upstream signaling systems? Recent data implicate the GTPase Rho as a regulator of Spir–Capu interaction in (). The Spir–Capu interaction is evolutionally conserved, but whether or not this mode of regulation is conserved remains to be tested.
flies were used as wild type (provided by R. Bainton, University of California, San Fransisco, San Fransisco, CA). and ,,,
/CyO flies were obtained from T. Schupbach (Bloomington Stock Center, Indiana University, Bloomington, IN). Transgenes were cloned into a pUASp vector and expressed in the germline with VP16nos:Gal4 or pCog:Gal4; NGT40:Gal4; VP16nos:Gal4 triple maternal driver lines (provided by L. Cooley, Yale, New Haven, CT). For immunofluorescence, they were fixed and stained according to methods described by . For actin visualization, ovaries were incubated in 1–2 U rhodamine-conjugated phalloidin (Invitrogen). For Spir immunolocalization, ovaries were incubated with ∼1 μg/ml antibody, and AlexaFluor488-conjugated goat anti–rabbit secondary antibody (Invitrogen) was used at a 1:1,000 dilution. Samples were mounted in fluorescence mounting medium (DakoCytomation). For live imaging, ovaries were dissected and teased apart under Halocarbon 700 oil (Sigma-Aldrich) at room temperature. In both cases, images were collected with a plan-Neofluor 25× 0.8 NA objective lens on a confocal microscope (LSM 510 META; Carl Zeiss MicroImaging, Inc.) with its proprietary software. Final rotations, cropping, and conversion to TIF format were performed in ImageJ (National Institutes of Health).
Standard PCR and cloning methods were used to make DNA constructs.
NIH 3T3 mouse fibroblasts were cultured in DME supplemented with 10% FCS, glutamate, penicillin, and streptomycin at 37°C in a CO (10%) incubator. The cells were transiently transfected with eukaryotic expression vectors with LipofectAMINE (Invitrogen). Immunofluorescence was performed as described previously (). Cells were assayed 36 h after transfection. 5 μg/ml myc 9E10 mouse monoclonal antibodies (Santa Cruz Biotechnology, Inc.) and anti-TRITC–conjugated donkey anti–mouse (1:200; Dianova) were used. Fixed samples were mounted in a solution of 15 g Moviol, 60 ml PBS, 30 ml glycerol, and 2.25 g -propyl-gallate. Images were collected with a 100× NA 1.3 oil U-V-I objective lens (Leica) on a fluorescence microscope (DMIRBE; Leica). Recordings were made with a camera (C4742-95; Hamamatsu) using Openlab 4.0.4 software (Improvision). Images were contrast enhanced and saved as TIF files using Openlab. All work was performed at room temperature.
His-tagged proteins pQE-80L-m-Fmn2-FH2 (Fmn2-FH2), pQE-80L-hu-Spir1-KIND (Spir-1–KIND), pET-20b-p150-Spir-KIND (KIND), and pET-20b-p150-NTSpir (NTSpir) were expressed in BL-21(DE3)pLysS cells. Cells were grown at 37°C to an optical density (A) of 0.6–0.8, induced with 0.25 mM IPTG, and harvested 3 h later. pET-20b-p150-Spir-WH2 (WH2) and pET-20b-p150-Spir-KCK-WH2(C459S) (KCK-WH2) were harvested after only 1.5 h. pET-20b-Capu-FH2 (Capu-FH2) and pET-20b-Capu-FH1FH2 (Capu-FH1FH2) were expressed in bacteria (Novagen). Cells were grown at 37°C to an optical density (A) of 0.4–0.6, cooled to 21°C, induced with 0.1 mM IPTG, and grown overnight. His-tagged proteins were purified with BD Talon resin (CLONTECH Laboratories, Inc.), and most were further purified by anion exchange (monoQ) chromatography. See for a summary of constructs and their extinction coefficients.
GST-tagged protein bacteria was transformed with GST fusion protein expression vectors (GST-Spir1-KIND and GST-VKIND-KIND). Cells were grown at 37°C to an optical density (A) of 0.6–0.8, induced with 0.1 mM IPTG, and incubated at 21°C overnight. After centrifugation, the bacteria were suspended in TBS-Tween buffer (150 mM NaCl, 10 mM Tris-HCl, pH 7.4, and 0.1% Tween) and sonicated. The soluble extract was incubated with glutathione–Sepharose 4B beads (GE Healthcare) for 2 h at 4°C. The beads were washed twice with TBS-Tween buffer and resuspended in the same buffer for pull-down assays.
Immunoprecipitation from fly ovary was performed according to the methods of ; ∼100 flies were used for each condition). Immunoprecipitation from tissue culture cells was performed as follows: 36 h after transfection, NIH 3T3 cells were lysed with immunoprecipitation lysis buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM EDTA, 2 mM EGTA, 10% glycerol, 0.1% NP-40, 2 μg/ml leupeptin, 2 μg/ml aprotinin, 1 mM PMSF, 10 mM NaF, and 0.2 mM NaVO). Anti–myc 9E10 monoclonal antibodies (Santa Cruz Biotechnology, Inc.) were added to the cleared cell lysate to a final concentration of 5 μg/ml and incubated for 1 h on ice. Protein G–agarose (Roche) was added, and the sample was rotated for 150 min at 4°C. The beads were washed twice with immunoprecipitation lysis buffer, and bound proteins were analyzed by Western blotting.
actin was labeled with pyrene iodoacetamide as described previously (). Purified KIND (human and ) was incubated with a 1–2.5-fold molar excess of AlexaFluor488-maleimide (Invitrogen) in labeling buffer (50 mM KCl, 50 μM Tris(2-carboxyethyl) phosphine [TCEP], and 10 mM Hepes, pH 7) for 30 min at 24°C. The reaction was quenched by the addition of 10 mM DTT. Free dye was removed by gel filtration and verified by SDS-PAGE. Protein concentration was determined by absorbance at 280 nm using molar extinction coefficients (human, 22,620 cm; , 17,452 cm; ) and correcting for dye absorbance (0.11*A). The concentration of incorporated dye was determined by absorbance at 496 nm using an extinction coefficient of 71,000 cm. The labeling efficiency of Spir-1–KIND varied from 85 to 94% at all dye/protein ratios above 1:1, suggesting that Spir-1–KIND contains a single accessible and reactive cysteine. The labeling efficiency of KIND was >100% at dye/protein ratios above 1:1. Therefore, we underlabeled KIND (15–45%) to increase the probability of only labeling one cysteine per protein.
Solution molecular weights were determined by spinning samples to equilibrium in an analytical ultracentrifuge (XL-I; Beckman Coulter) and measuring protein concentration as a function of radius. For individual protein, we used three different concentrations (∼0.5, 0.25, and 0.125 mg/ml). For KIND and Capu-FH1FH2 together, we used three molar ratios of AlexaFluor488-labeled Capu-FH1FH2 and KIND (1 [Capu dimer]:1 [KIND], 1:2, and 1:4). We spun samples to equilibrium at three speeds and measured protein concentration as a function of radius by absorbance at 280 or 500 nm to track the fluorophore. We then globally fit all of the sedimentation curves obtained at different concentrations and speeds (nine datasets) to different models () using Winnonln (J. Lary and D. Yphantis, National Analytical Ultracentrifugation Facility, Storrs, CT) or Ultrascan (B. Demeler, University of Texas Health Science Center, San Antonio, TX).
We performed pyrene actin assembly assays as described previously (). In brief, we used 4 μM actin doped with 5% pyrene-labeled actin. Ca-actin was converted to Mg-actin before each reaction by a 2-min incubation with 50 μM MgCl and 200 μM EGTA. All components except actin were combined before the initiation of polymerization. All polymerization reactions contained 50 mM KCl, 1 mM MgSO, 1 mM EGTA, and 10 mM imidazole, pH 7.0. Pyrene fluorescence was measured by a multifrequency fluorometer (K2; ISS), and data were analyzed using KaleidaGraph (Synergy Software) and in-house software.
Low concentrations (10–20 nM) of the protein labeled with AlexaFluor488 (Spir-1–KIND, KIND, or WH2) was mixed with unlabeled target protein (Fmn2-FH2, Capu-FH1FH2, or Capu-FH2) at the indicated concentrations in KMEH (50 mM KCl, 20 mM Hepes, pH 7.0, 1 mM MgCl, 1 mM EGTA, and 1 mM TCEP), and polarization anisotropy was measured at 24°C using a multifrequency fluorometer (K2; ISS) and analyzed using KaleidaGraph (Synergy Software). Under the conditions used in our study, anisotropy is a measure of the rotational mobility of the labeled protein. We excited the fluorophore with plane polarized light at 488 nm (with a 488-nm bandpass filter) and measured emission at 520 nm (KV500 and KV520 filters) at polarizations both parallel (I) and perpendicular (I) to the excitation light. We simultaneously monitored total intensity to ensure that the quantum efficiency of the fluorophore was independent of the protein complex. In the case of labeled WH2, the intensity changed in a concentration-dependent saturable manner, so we used this value instead of anisotropy. We determined equilibrium dissociation constants using a quadratic binding model as previously described ().
In competition binding experiments, we determined the dissociation constant of the unlabeled protein by fitting anisotropy data to the function () where K is the dissociation constant of the nonfluorescent competitor, [C] is the total concentration of the competitor, and [R] is the concentration of free Capu-FH1FH2 when [C] = 0. For this analysis, [R] and K are determined from the anisotropy in the absence of competitor, and K is determined from fitting the above equation to experimental data. This function is an approximation and is only valid when unlabeled KIND and formin are in excess over the labeled protein. These conditions were met in our competition binding experiments.
Purified porcine brain tubulin was provided by A. Carter (laboratory of R. Vale, University of California, San Fransisco, San Fransisco, CA). Microtubules were stabilized with taxol, and all experiments were performed in KMEH. Capu-FH1FH1, Fmn2-FH2, KIND, and Spir-1–KIND were cleared by centrifugation at 100,000 for 20 min at 24° C before each assay. Microtubules were added at four- to eightfold molar excess. Samples were incubated for 15 min and centrifuged at 100,000 for 10 min at 24°C. Supernatants were removed, and pellets were washed before resuspending. Both supernatants and pellets were analyzed by SDS-PAGE.
In competition experiments, KIND or NTSpir was added in a twofold concentration series ranging from 16 to 0.0625 μM (i.e., 16, 8, 4, …0.0625 μM), and 20 μg/ml BSA was added as a loading standard. These gels were stained with Sypro-Red (Invitrogen) and quantified using a multiformat imager (Typhoon 9400; GE Healthcare) and ImageQuant software (GE Healthcare). We determined the dissociation constant of the unlabeled protein by fitting binding data to the equation above. In this case, K is the dissociation constant of the competitor (KIND or NTSpir) for Capu-FH1FH2, [C] is the concentration of the competitor, and [R] is the concentration of free Capu-FH1FH2 when [C] = 0. For this analysis, K was determined by fluorescence anisotropy (5 nM), and K, the dissociation constant of Capu-FH1FH2 for microtubules, is determined by fitting the above equation to experimental data. This function is an approximation and is only valid when the competitor and microtubules are in excess over Capu-FH1FH2. Only the data points that met these conditions were fit.
Two actin-binding assays were used. An actin cross-linking assay was performed according to with minor modifications. 50-μl solutions containing 2 μM phalloidin-stabilized actin plus 0.5 μM Capu-FH1FH2 or α-actinin or each component separately were mixed in KMEH, allowed to stand at RT for 10 min, and centrifuged at 16,000 for 5 min. 40 μl of the supernatant was removed, and the pellet was washed once and resuspended in 50 μl. Equal amounts of supernatants and pellets were analyzed by SDS-PAGE. For microscopy experiments, we substituted AlexaFluor488-phalloidin for unlabeled phalloidin. After standing for at least 10 min, solutions were diluted 1:100 in KMEH and added to poly--lysine–coated flow chambers at room temperature. Images were collected with a plan Apo 60× 1.2 NA objective lens and camera (C4742-98; Hamamatsu) on a microscope (TE300; Nikon) with Simple PCI software (Compix, Inc.).
Two microtubule cross-linking assays were used. The first assay was performed according to the methods of . 50-μl solutions containing 10 μM tubulin (doped with 10% rhodamine-tubulin [Cytoskeleton, Inc.]) plus 0.5 μM Capu-FH1FH2 or GST-Spastin(E542A) (a mutant in the Walker B site that does not sever []) were mixed on ice in 80 mM Pipes, pH 6.9, 1 mM EGTA, 1 mM MgCl, 1 mM GTP, and 25% glycerol. They were incubated at 37°C for 25 min and fixed with 1% gluteraldehyde. Samples were diluted 1:100, introduced into a flow chamber, and examined by fluorescence microscopy. For the second assay, the microtubules were prepolymerized at high concentration (18 μM), taxol stablized, and diluted and combined with Capu-FH1FH2 or GST-Spastin(E542A) (10:0.5 μM). These solutions were allowed to stand at room temperature for at least 15 min before dilution (1:100) and visualization in a flow chamber, with the same equipment used for actin cross-linking assays.
Fig. S1 shows a stage 9 egg chamber expressing Spir-mRFP (A), immunofluorescence and Western blots showing the specificity of anit-Spir antibody (B–D), complete Western blots from E (E), and a Western blot of an oocyte coimmunoprecipitated with and without latrunculin (F). Fig. S2 shows additional images of Spir and Capu expression in NIH 3T3 cells (A–D), GST pull-down with mammalian isoforms of Spir and Capu (E), an anisotropy experiment with Capu-FH2 (F), and an anisotropy experiment with mammalian isoforms of Spir and Capu (G). Fig. S3 shows the inhibition of Capu-FH2–mediated actin nucleation by KIND (A–C), sample analysis of polymerization assays (D), the inhibition curve for mammalian isoforms of Spir and Capu (E), and alternate inhibition analysis (F). Fig. S4 shows that KIND does not influence spontaneous actin polymerization (A and B) and that Spir-1–KIND does not interact with mDia1 or mDia2 (C). Fig. S5 shows gels of FH2 domains and KIND mixed with tubulin (A and B), a cosedimentation assay with Capu-FH1FH2, tubulin, and NTSpir (instead of KIND; C and D), and images of Capu-FH1FH2–mediated tubulin polymerization and bundling of microtubules and actin (E–G). Online supplemental material is available at . |
In vertebrates, all skeletal muscles of trunk and limbs originate from somites, which are formed sequentially in a rostral-caudal direction through segmentation of the paraxial mesoderm during embryogenesis (). In response to signals from the neural tube, notochord, and ectoderm, somites further differentiate into ventral-medially positioned sclerotome and dorsally located dermatome with the muscle-forming myotome sandwiched in between (). In myotome, the muscle precursor cells establish their myogenic fate to form proliferating myoblasts by selectively expressing one or a few myogenic regulatory factors (MRFs). Under appropriate conditions, the myoblasts withdraw from the cell cycle to differentiate into mononucleated myocytes, which, in turn, align with each other and fuse to form multinucleated myotubes or myofibers.
Muscle stem cells, which are also called muscle satellite cells (MSCs), start to form at the late stage of vertebrate embryo development (; ; ; ). In the adult, most of the MSCs are quiescent and uniquely located between basal lamina and the plasma membrane of the myofibers. Several molecular markers, including Pax7, c-Met, M-cadherin, and CD34, are expressed in quiescent MSCs. In contrast, MyoD is not expressed in quiescent MSCs. In response to muscle injury or exercise, these quiescent MSCs become activated, as indicated by the expression of MyoD, reenter the cell cycle, and actively proliferate to form myoblasts. Eventually, these proliferating myoblasts irreversibly withdraw from cell cycles, differentiate, and fuse with existing myofibers. Accumulating evidence indicates that myoblasts are the primary cell types responsible for muscle regeneration in vivo (; ; ).
Elucidation of the molecular mechanisms underlying myogenic differentiation has been greatly facilitated by the availability of several immortalized myogenic cell lines, including C2C12 cells, which are derived from mouse MSCs (). Thus, C2C12 cells represent an excellent cell culture model to study the proliferation and differentiation of MSC-derived myoblasts. Two families of transcription factors play critical roles in myogenesis. MRFs consist of Myf5, MyoD, MRF4, and myogenin. MRFs normally heterodimerize with gene E2A products (i.e., E12/E47) and bind to the consensus sequence of CANNTG (E box) in the promoters of many muscle-specific genes (; ; ). Id, a negative regulator for myogenesis, represses myogenic differentiation by binding to and sequestering either MRFs or E proteins, thus preventing MRFs from binding to the E box (). In addition to MRFs, myocyte enhancer–binding factor 2 (MEF2), which consists of MEF2A, 2B, 2C, and 2D, is also essential for myogenesis (; ). MRFs and MEF2 physically interact with each other to synergistically activate many muscle-specific genes ().
Many intracellular signaling molecules/pathways are known to modulate myogenic differentiation by regulating myogenin gene expression. Among them, the p38 MAPK, insulin-like growth factors/phosphatidylinositol 3-kinase/Akt, calcium/calmodulin-activated protein kinase, and calcineurin positively regulate myogenic differentiation (; ; ; ; ; ; ; ), whereas extracellular signal-regulated kinase (ERK) has dual roles: it inhibits differentiation at the early stage of differentiation but promotes myocyte fusion at the late stage of differentiation (; ; ). The Janus kinase (JAK)–signal transducer and activator of transcription (STAT) pathway represents one of the best-characterized cellular signaling pathways (). Four s (i.e., , , , and ) and seven s (i.e., , , , , , , and ) have been identified in the mouse and human genomes. Although it is well established that the JAK–STAT pathway plays essential roles in hematopoiesis and antimicrobial immune response (), it remains unclear whether the JAK–STAT pathway plays any essential role in myogenesis. Several lines of evidence suggest that the JAK–STAT pathway may have a role in myogenic differentiation. In regenerating rat muscles, proliferating myoblasts were found to contain higher levels of phosphorylated (i.e., active) STAT3 (). In response to leukemia inhibitory factor (LIF), proliferating primary myoblasts grown in culture were also found to contain higher levels of phosphorylated STAT3 (; ). Recently, MyoD was found to interact with STAT3 in an overexpression study (). Nevertheless, the aforementioned evidence was mainly correlative in nature, and none of these studies has addressed the question of whether and how the JAK–STAT pathway is involved in myogenic differentiation.
In this study, we report that the JAK1–STAT1–STAT3 pathway plays dual roles in proliferating myoblasts: it is required for myoblast proliferation and also serves as a key checkpoint to prevent myoblasts from premature differentiation. Specific knockdown of either or (to a lesser extent) reduces cell proliferation and induces precocious myogenic differentiation. LIF engages the JAK1–STAT1–STAT3 pathway to promote proliferation and to prevent the premature differentiation of myoblasts.
To explore the role of JAK1 in myogenic differentiation, we first examined its expression profiles in both immortalized C2C12 cells and primary myoblasts by Western blotting. As shown in , JAK1 was expressed in both C2C12 cells and primary myoblasts both before and after differentiation. To explore its functional role in myogenic differentiation, we first knocked down the endogenous in C2C12 cells with siRNA. Surprisingly, compared with the control cells transfected with an siRNA against the gene encoding the jellyfish GFP, a JAK1-specific siRNA dramatically accelerated myogenic differentiation, as indicated by an increased number of myogenin-positive and myosin heavy chain (MHC)–positive cells and precocious formation of large multinucleated myotubes (). Western blotting confirmed that JAK1-siRNA was specific and effective and potently induced the expression of myogenin and MHC (). In contrast, a JAK2-specific siRNA inhibited C2C12 differentiation, as indicated by decreased myogenin and MHC expression levels compared with the GFP-siRNA control (). Consistently, JAK1-siRNA greatly activated two myogenic luciferase reporter genes driven by a fragment of the native myogenin promoter and muscle creatine kinase (MCK) promoter (i.e., G133-luc and MCK-luc, respectively; ). To find out whether the knockdown had a similar effect on primary myoblasts, we infected the primary myoblasts (i.e., MyoD) with adenoviruses expressing either the JAK1-specific short hairpin RNA (shRNA) or human lamin-specific shRNA (control). Compared with cells infected with lamin-shRNA, those infected with JAK1-shRNA underwent potent and accelerated differentiation, as manifested by the considerably increased number and size of MHC-positive myotubes (). In addition, we determined the kinase activity of the endogenous JAK1 in C2C12 cells before and after differentiation. We found that the kinase activity of JAK1 decreased upon differentiation ().
Because the knockdown of stimulates myogenic differentiation (), we next asked what happens if we overexpress in myogenic cells. We generated stable C2C12 cells expressing either the wild-type JAK1 or an empty vector and subjected these cells to differentiation. As shown in , upon differentiation, cells overexpressing JAK1 had much lower myogenin levels and undetectable MHC compared with those with an empty vector. Consistently, the overexpression of inhibited the activity of MCK-luc (). The inhibitory role of JAK1 at the early phase of myogenic differentiation was reminiscent of that of the ERK pathway (; ; ). To study the potential cross talk between the two pathways, we first examined the effect of JAK1-siRNA on the levels of the active ERKs (i.e., phospho-ERK) in C2C12 cells. As a control, we also examined the status of the active p38 MAPK and Akt, both of which are indispensable for myogenic differentiation (; ; ). We found that the knockdown did not affect the levels of the active ERKs, p38 MAPKs, and Akt in C2C12 cells (unpublished data), suggesting that the potent prodifferentiation effect of JAK1-siRNA was not mediated by these pathways. We then tested whether JAK1 cooperates with the ERK pathway to repress myogenic differentiation. Although cells overexpressing either MEK1ΔN4, which constitutively activates ERK (), or JAK1 already led to a reduced expression of myogenin, cells overexpressing both further repressed myogenin expression (). Conversely, cells treated with either JAK1-siRNA or U0126, a specific inhibitor for the ERK pathway (), differentiated faster as indicated by the accelerated induction of myogenin (, compare lanes 1 and 3 with lane 2). Importantly, cells treated with both JAK1-siRNA and U0126 together had a more potent myogenin induction than either treatment alone (, lane 4). Our data suggest that JAK1 and the ERK pathway cooperate to repress early myogenic differentiation.
To uncover the molecular mechanisms underlying the prodifferentiation effect of JAK1-siRNA, we examined several key players involved in early myogenic differentiation. As shown in , the protein levels of both MyoD and MEF2 were clearly induced by JAK1-siRNA at the early stage of differentiation. This induction mainly occurred at the transcription level, as the mRNA levels of , , , and were also induced by JAK1-siRNA as judged by semiquantitative RT-PCR (unpublished data). In contrast, Id1, a known negative regulator for myogenic differentiation (), was down-regulated at a faster rate in the presence of JAK1-siRNA (). In addition, we also examined the impact of JAK1-siRNA on both MyoD and MEF2-dependent gene transcription. Although JAK1-siRNA had less effect on the activity of 4xRE-luc, a MyoD-dependent reporter, it substantially elevated the activity of 3xMEF2-luc, an MEF2-dependent reporter (). Consistently, the overexpression of JAK1 greatly inhibited the MEF2-dependent gene transcription but had less effect on MyoD-dependent gene transcription ().
While we were studying the impact of JAK1-siRNA on myogenic differentiation, we also noticed that JAK1-siRNA considerably affected cell proliferation. In a time course experiment, we found that C2C12 cells transfected with JAK1-siRNA displayed a reduced proliferation rate compared with the control cells transfected with GFP-siRNA (). Similarly, a lower percentage of C2C12 cells treated with JAK1-siRNA incorporated BrdU compared with cells treated with GFP-siRNA (). To closely examine the effect of JAK1-siRNA on the cell cycle, we subjected the siRNA-treated C2C12 cells to flow cytometry analysis. We found that a higher proportion of cells transfected with JAK1-siRNA were arrested in G1 phase compared with the GFP-siRNA–transfected control cells no matter whether nocodazole was used or not (to reduce the in terference from either untransfected cells or cells already in S or early G2 phase; ). To investigate the underlying molecular mechanisms, we examined several key cell cycle regulators. Although JAK1-siRNA had no obvious effect on the levels of CDK4 and total Rb, it induced p21Cip1 and p27Kip1, two prominent CDK2 inhibitors, in both proliferating cells as well as in cells undergoing early differentiation (). Consistently, the levels of Rb phosphorylation at Ser795, a site mainly phosphorylated by CDK2 and CDK4 (), also decreased in cells with elevated p21Cip1 and p27Kip1 ().
Although our aforementioned data indicated that JAK1 negatively regulates myogenic differentiation, it remained unclear whether its kinase activity is required in this process. To address this issue, we generated two siRNA-resistant JAK1 constructs: one encoding the wild-type protein and the other encoding a kinase-dead mutant. We then transfected C2C12 cells with JAK1-siRNA together with expression vectors encoding either the wild-type JAK1 (i.e., siRNA sensitive) or the siRNA-resistant JAK1. As shown in , although both forms of the siRNA-resistant were expressed at high levels in the presence of JAK1-siRNA, only the one retaining the kinase activity effectively reversed the prodifferentiation effect of JAK1-siRNA, as indicated by reduced myogenin expression and a complete lack of MHC expression (lane 2). This suggested that the kinase activity of JAK1 is essential for its repressive effect during myogenic differentiation. To identify downstream mediators of JAK1 in myogenic differentiation, we focused on STATs that are well-established downstream targets and mediators of JAKs (). We first designed siRNAs to individually knock down STATs that are present in C2C12 cells (i.e., STAT1, 2, 3, 5A, and 5B as judged by RT-PCR; unpublished data). Among them, only STAT1-siRNA led to a precocious induction of myogenin (), an effect similar to that of JAK1-siRNA (). In contrast, the siRNAs against and other inhibited myogenin expression ( and unpublished data). Transfection of C2C12 cells with a second set of siRNAs targeting a different region of and generated similar results (unpublished data). To extend the study to primary myoblasts, we first confirmed that was expressed in primary myoblasts both before and after differentiation (). Consistently, primary myoblasts transfected with STAT1-siRNA differentiated faster than those with GFP-siRNA, as indicated by a substantially increased number of MHC-positive cells (). Furthermore, we found that the repression of myogenin and MHC mediated by the overexpression of could be rescued by STAT1-siRNA but not by STAT2-siRNA ().
LIF is a known mitogen for both primary myoblasts and C2C12 cells (; ; ). Because JAK1 is also required for myoblast proliferation (), naturally, we wanted to test whether LIF promotes myoblast proliferation through JAK1 and STAT1. When C2C12 cells were exposed to LIF, cell proliferation was clearly accelerated as measured by WST-1 assays (). To reveal the composition of the STAT complex induced by LIF, we performed electrophoretic mobility shift assays (EMSAs) using an oligonucleotide containing a consensus STAT-binding site as a probe (). As shown in , we did detect specific STAT complexes. Supershift assays with several STAT-specific antibodies revealed that the LIF-induced STAT complexes mainly consist of STAT1 and STAT3 but not STAT2 (, lanes 6 and 7). Furthermore, JAK1, STAT1, and STAT3 were all activated as early as 10 min after LIF treatment, which was evident by an increase in their tyrosine-phosphorylated (i.e., active) forms even though their total protein levels did not change much ( and unpublished data). Moreover, the LIF-induced tyrosine phosphorylation of STAT1 and STAT3 could be reduced by JAK1-siRNA but not by JAK2-siRNA, suggesting that JAK1 mediates the LIF-induced phosphorylation of STAT1 and STAT3 in C2C12 cells (). In addition to its role in myoblast proliferation, LIF was also known to inhibit myogenin expression and myogenic differentiation (). To uncover the underlying mechanisms, we examined the expression status of MyoD and MEF2. We found that LIF repressed the expression of MEF2 in both proliferating myoblasts and cells undergoing differentiation (). In contrast, LIF had less effect on the expression levels of MyoD (unpublished data). Importantly, LIF-mediated down-regulation of both MEF2 and myogenin was efficiently rescued by JAK1-siRNA and was partially rescued by STAT1-siRNA (). Furthermore, LIF-induced C2C12 proliferation was also greatly inhibited by JAK1-siRNA and, to a lesser extent, the siRNAs against either or (). Thus, our aforementioned data suggested that JAK1–STAT1–STAT3 act downstream of LIF and mediate its effect on myoblast proliferation and differentiation.
Because MSC-derived primary myoblasts are mainly responsible for injury-induced muscle regeneration, as an initial attempt, we examined the status of JAK1, STAT1, and STAT3 in regenerating tibialis anterior (TA) muscles in response to cardiotoxin-induced muscle injury (). In this injury-induced muscle regeneration model, the satellite cell– derived myoblasts actively proliferated in the first 2–3 d (). A majority of myoblasts started to differentiate by day 3 as indicated by a peak expression of myogenin (). The damaged area was largely repaired by day 15 (). Consistently, we found that the total levels of both JAK1 and STAT1 increased upon injury, peaked at day 3 after injury, and gradually decreased afterward (). Importantly, both the kinase activity of JAK1 and levels of the active STAT1 increased as early as 1 d after injury (). As to STAT3, although its total levels did not change much in response to the injury, the levels of active STAT3 substantially increased 1 d after injury and peaked at day 3 ().
In this study, we provide evidence showing that the JAK1– STAT1–STAT3 pathway has two distinct roles in myogenic differentiation: on one hand, it is required for myoblast proliferation as a result of its involvement in regulating the expression of p21Cip1, p27Kip1, and Id1. On the other hand, it prevents myoblasts from premature differentiation by actively repressing genes essential for differentiation (e.g., , , and ). In this sense, the status of the JAK1–STAT1–STAT3 pathway can be viewed as a key checkpoint for differentiation, as shutdown of this pathway is a prerequisite for myoblasts to initiate the differentiation program. The diverse roles of the JAK1–STAT1–STAT3 pathway are especially important during injury-induced muscle regeneration. MSCs normally undergo three distinct phases during injury-induced regeneration: activation, proliferation, and differentiation (). In the activation phase, the quiescent MSCs are activated by an ill-defined mechanism in response to injury. Certain cytokines/chemokines released by inflammatory cells could potentially serve as the trigger (; ). In the proliferation phase, the activated MSCs actively proliferate to generate a sufficient number of myoblasts. Premature differentiation of myoblasts is undesired at this stage. In the final differentiation phase, myoblasts differentiate and fuse with existing myofibers to repair the damaged muscles. Based on the properties of the JAK1–STAT1–STAT3 pathway in myoblast cultures, we expect that the pathway mainly operates in the proliferation phase during muscle regeneration. Consistently, our preliminary study shows that JAK1, STAT1, and STAT3 are all activated in regenerating muscles at a time when myoblasts actively proliferate (). Because the invading inflammatory cells, including neutrophils, macrophages, and T cells, are also present in regenerating muscles, it remains to be further clarified whether the changes in JAK1–STAT1–STAT3 specifically occur in proliferating myoblasts.
Although STAT3 is commonly associated with cell proliferation in many different cell types, STAT1 is rarely associated with proliferation (). In fact, the activation of STAT1 often reduces cell proliferation (; ). Therefore, it is quite unique that STAT1 is required for myoblast proliferation. Although it remains unclear what dictates these different outcomes, presumably, it is a cell context–dependent phenomenon, which implies that STAT1 has to cooperate with other molecules/pathways to bring about different effects on cell proliferation. The involvement of STAT3 in myoblast proliferation is supported by our following findings: STAT3 and STAT1 form complexes in response to LIF stimulation and the knockdown of STAT3 reduces LIF-induced myoblast proliferation. Consistently, an earlier study showed that LIF activates STAT3 in proliferating myoblasts (). Unexpectedly, the knockdown of does not accelerate differentiation the same way as the knockdown of does (). A possible explanation is that STAT3 may be required for both the proliferation and differentiation of myoblasts, which is supported by our findings that the levels of active STAT3 gradually increase during differentiation in C2C12 cells (unpublished data) and that the knockdown of inhibits differentiation (). It is likely that STAT3 may function at different phases of differentiation by associating with different partners. Further investigation is needed to clarify the role of STAT3 in myogenic differentiation.
So far, several cytokines and growth factors, including LIF, hepatocyte growth factor, and basic FGF, have been shown to stimulate the proliferation of myoblasts in vitro (; ). The only in vivo data came from the LIF knockout mice. It was shown that injury-induced muscle regeneration is delayed in LIF mice and that the defect can be rescued by the injection of exogenous LIF (). Thus, LIF is essential for the proliferation of myoblasts both in vivo and in vitro. As a member of the IL-6 family of cytokines, LIF is known to exert its diverse functions mainly through the JAK–STAT pathway (; ). Consistent with this notion, we demonstrate here that LIF specifically utilizes the JAK1–STAT1–STAT3 pathway to regulate the proliferation and differentiation of myoblasts. It remains to be seen whether other members of the IL-6 family are capable of regulating myogenic differentiation through similar mechanisms.
Although several different JAK family members are usually associated with the same receptor complex (e.g., JAK1 and JAK2 associate with the α and β chains of interferon-γ receptor, respectively) in cytokine signaling (), JAK1 seems to form homodimers during myogenic differentiation, as the knockdown of either or , the remaining two members of the JAK family present in both primary and immortalized myoblasts, has the opposite effect as compared with that of ( and unpublished data). This suggests that multiple JAK–STAT pathways function to control the proliferation and differentiation of myoblasts with opposing effects. In the future, it is essential for us to understand how these different JAK–STAT pathways coordinate with each other to control myogenic differentiation and how they cross talk with other signaling pathways. With such knowledge, it is possible to accelerate injury-induced muscle regeneration by differentially modulating different JAK–STAT pathways with small molecules. We believe that our work has provided a new direction in studying the biology of muscle stem cells.
C2C12 cells were maintained in DME with 20% FBS, 100 U/ml penicillin, and 100 μg/ml streptomycin (also called growth medium [GM]) in a 37°C incubator with 5% CO. To induce differentiation, cells were grown in DME containing 2% horse serum (also called differentiation medium [DM]). Flag-JAK1 (human) and two STAT1 constructs (i.e., the wild-type and Y701F mutant) were gifts from Z. Wen (Hong Kong University of Science and Technology, Hong Kong, China). G133-luc, MCK-luc, 4xRE-luc, and 3xMEF2-luc were described previously (; ). Silent mutation was introduced into Flag-JAK1(wild type) and the kinase-dead Flag-JAK1(K896R) constructs to generate siRNA-resistant R-Flag-JAK1(wild type) and R-Flag-JAK1(K896R), respectively, using the oligonucleotide 5′-ACCCGAAAGCGGCGGCAATCACATAGCTGATCTGAAAAAG-3′ (top strand). U0126 and LIF were purchased from Calbiochem and Chemicon, respectively.
The sources of the antibodies used in this study are listed as follows: JAK1, JAK2, and STAT1 were purchased from Upstate Biotechnology; JAK1, Rb, CDK4, myogenin, MyoD, Id1, p21Cip1, p27Kip1, β-actin, and MEF2 were obtained from Santa Cruz Biotechnology, Inc.; anti-Flag was purchased from Sigma-Aldrich; phosphor-Rb (Ser795), phosphor-JAK1 (Tyr1022/1023), phopho-STAT1 (Tyr705), and phosphor-STAT3 (Tyr705) were obtained from Cell Signaling; and MHC was purchased from Developmental Studies Hybridoma Bank. Polyclonal STAT3 antibody was a gift from Z. Wen. Immunoblotting was performed according to standard procedures (). For immunostaining, FITC- and rhodamine-conjugated secondary antibodies and DAPI were used to label selected molecules and to counterstain the nuclei, respectively. The images were acquired at room temperature by a CCD camera (Spot RT; Diagnostic Instruments) mounted on a fluorescent microscope (IX70; Olympus) using SPOT software (version 4.0.9; Diagnostic Instruments). UPlanFL 10× NA 0.3 and LCPlanFL 20× NA 0.4 objectives (Olympus) were used. The brightness and contrast of the images were adjusted by Photoshop 6.0 (Adobe).
After removal of the culture medium, cells were washed once with PBS before addition of the lysis buffer (50 mM Hepes, pH 7.6, 1% vol/vol Triton X-100, 150 mM NaCl, 1 mM EGTA, 1.5 mM MgCl, 100 mM NaF, 20 mM -nitrophenylphosphate, 20 mM β-glycerophosphate, 50 μM sodium vanadate, 2 mM DTT, 0.5 mM PMSF, 2 μg/ml aprotinin, 0.5 μg/ml leupeptin, and 0.7 μg/ml pepstatin). Cells were lysed for 10 min at 4°C followed by centrifugation to remove the insoluble cell debris. The concentration of the whole cell extracts (WCEs) was determined by a protein assay reagent (Bio-Rad Laboratories).
Isolation of primary MSCs was performed as described previously with minor modifications (). In brief, skeletal muscles of 2-mo-old C57BL/6J mice were isolated, minced, and digested in 1.25 mg/ml protease type XVII (Sigma-Aldrich) for 1.5 h in a 37°C water bath. Satellite cells were purified by discontinuous Percoll gradient centrifugation and collected from the interface of 20 and 60% Percoll. Purified satellite cells were then cultured in GM (DME containing 20% FBS and 2% chicken embryo extracts) in culture dishes coated with 4 mg/ml Matrigel (BD Biosciences) to generate primary myoblasts. Myoblasts were induced to differentiate in DM (DME with 5% horse serum).
TA muscles of 6–8-wk-old C57BL/6 mice were injected with 25 μl of 10 μM cardiotoxin (Sigma-Aldrich). At different time points after injury, mice were killed by cervical dislocation, and the TA muscles were surgically isolated and homogenized in the lysis buffer followed by centrifugation to remove insoluble debris. Uninjured TA muscles were used as a control.
To deliver oligonucleotide-based siRNA, 50–70% confluent C2C12 cells were transfected with 100 nM siRNA using LipofectAMINE 2000 (Invitrogen). The following siRNAs were synthesized at Dharmacon, Inc.: JAK1 (5′-GCCUGAGAGUGGAGGUAAC-3′), JAK2 (5′-GCAAACCAGGAAUGCUCAA-3′), STAT1 (5′-GCGUAAUCUCCAGGAUAAC-3′), STAT3 (5′-CTGGATAACTTCATTAGCA-3′), and enhanced GFP (5′-GCUGACCCUGAAGUUCAUC-3′). To generate shRNA from an adenoviral vector, we used the Block-iT RNAi expression kit (Invitrogen) according to the manufacturer's instructions. 20 μl of viruses (∼10 plaque-forming U/ml) was used to infect 60% confluent primary myoblasts in 35-mm plates.
JAK1 was first immunoprecipitated from 1 mg WCEs, washed extensively, and reconstituted in the kinase buffer (10 mM Hepes, pH 7.4, 50 mM NaCl, 5 mM MgCl, 5 mM MnCl, 100 μM NaVO and 0.25 mCi/ml γ-[P]ATP). The reaction was incubated at room temperature for 30 min and terminated by adding sample loading dye. The reaction was separated on 8% SDS-PAGE, and the gel was dried and subjected to autoradiography.
C2C12 cells were first seeded into 96-well plates in triplicate at a density of 10 cells/well followed by transfection with various siRNAs the next day. WST-1 reagent (Boehringer) was then added at fixed time points according to the manufacturer's instructions. The absorbance at 480 nm was measured using a microtiter plate reader.
EMSA was performed as described previously (). The top strand sequence of the probe is 5′-GTCGACATTTCCCGTAAATC-3′.
An in situ cell proliferation kit (Roche) was used to measure BrdU incorporation. Cells were labeled with 10 μM BrdU for 1 h. Nuclei incorporating BrdU were visualized by fluorescein-conjugated monoclonal anti-BrdU antibody. For FACS analysis, cells were trypsinized, washed with PBS, fixed in 70% ethanol for 30 min on ice, harvested by centrifugation at 300 for 5 min, and resuspended in 400 μl PBS. To stain the DNA, 5 μl propidium iodide (5 mg/ml) and 5 μl RNase A (10 mg/ml) were added to the solution. Cells were incubated at 37°C for 30 min before being subjected to FACS analysis. |
Subsets and Splits