text
stringlengths 7
107k
|
|---|
All peaks within 0-35 min
retention time and 100 Da to 5000 Da MS1 precursor mass were aggregated into compound
24
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . groups using a 15 ppm. mass tolerance and 1.0 min retention time tolerance. Peaks were
excluded if peak intensity was less than 1 x 105, peak width was greater than 1.0 min, signal-
to-noise ratio was less than 1.5, or intensity was < 3-fold greater than blank. For each feature,
compound formula was predicted based on the molecular weight with 15 ppm. mass tolerance. Annotation of the compound formula was assigned by searching database Chemspider with
molecular weight or formula. Compound identification was carried out by searching the MS2
spectra against online database mzClound (https://www.mzcloud.org/) as well as a local
customized MS2 spectrum library. The customized library is composed of 60 in-house
acquired standards, 598 spectra of polar compounds generated from Bamba lab, and whole
collection of LCMS MS2 spectra (120K entries) from North American Mass Bank
(https://mona.fiehnlab.ucdavis.edu/downloads). Identifications were filtered with a match
threshold of 60 for mzCloud hits, and threshold of 30 for customized library hits. We normalized all the measurements of each sample to correct for bias brought by different
amounts of starting material. First, we calculated the sum intensity value of all the metabolite
features that were detected in each sample. Original intensity values were divided by the sum
intensity of each sample. The resultant intensity values were multiplied by the mean of the
sum intensities of all the samples. Normalized intensity values were Log2 transformed before
downstream analysis. Gene set enrichment analysis (GSEA)
We used the clusterProfiler R package to conduct GSEA analysis on gene ontology biological
process terms (Yu et al. 2012). We used a minimum gene set size of 10, a maximum gene set
size of 500, and performed 10,000 permutations. For the transcriptome data, we used a gene
list that is ordered by log2(Fold Changes) from the differential expression analysis. In the
proteomic analysis, the gene name for each protein were retrieved and the gene list was then
ordered based on fold change values of the protein quantity obtained from the differential
analysis of the proteomic level. Pairwise comparisons between layers were performed by
overlapping the genes that are up or down with an FDR adjusted p-value of 0.1. Transcriptomic/proteomic GO Enrichment analysis
All significantly changing genes / proteins (adjusted p-value < 0.05 and an absolute fold
change > 1) were split into 8 groups based on the combination of direction of the fold change,
includes genes up & down regulated significantly in a single analysis (i.e.
|
transcriptome OR
proteomics). An overrepresentation analysis using GO gene sets was performed on each of
the eight groups to determine the main – if any – gene sets changing. Variant analysis
25
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Variants for the CB4856 strain were retrieved from the “Caenorhabditis elegans Natural
Diversity Resource” (https://elegansvariation.org/). Soft filtered variant file was retrieved from
http://storage.googleapis.com/elegansvariation.org/releases/20200815/variation/WI.2020081
5.soft-filter.vcf.gz. Hard filtered variant file was retrieved from
http://storage.googleapis.com/elegansvariation.org/releases/20200815/variation/WI.2020081
5.hard-filter.vcf.gz. These variants were then filtered for the CB4856 strain keeping only 1/1
variants with high impact consequence. Acknowledgements
We thank the Caenorhabditis Genetics Center for providing the C. elegans strains. We thank
all team members of the J. Auwerx laboratory for helpful discussions. This work was supported
by grants from the École Polytechnique Fédérale de Lausanne, European Research Council
(ERC-AdG-787702) and Swiss National Science Foundation (31003A_179435) and a Global
Research Laboratory grant
from
the National Research Foundation of Korea
(2017K1A1A2013124). A.W.G. was supported by the Accelerator prize given by the United
Mitochondrial Disease Foundation (PF-19-0232). T.Y.L. was supported by the Human Frontier
Science Program (LT000731/2018-L). Author contributions
AWG, GEA and JA conceived and designed the project. AWG, AL, TYL, and KH performed
the experiments. MM and RHH performed lipidomics analysis, YZ, KO, ES, and JJC performed
proteomics and metabolomics analysis, AWG, GEA, AL, TYL, KH, MM, and MBS performed
the data analysis. JA supervised the work. AWG, GEA, AL and JA wrote the manuscript with
comments from all authors. Competing fanatical interests
The authors declare no competing interests. Data availability
The RNAseq data has been deposited in the National Center for Biotechnology Information
Gene Expression Omnibus database (accession number: GSE159228). 26
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . References:
Andreux PA, Houtkooper RH, Auwerx J. 2013. Pharmacological approaches to restore
mitochondrial function. Nature reviews Drug discovery 12: 465-483. Ashrafi K, Chang FY, Watts JL, Fraser AG, Kamath RS, Ahringer J, Ruvkun G. 2003. Genome-wide RNAi analysis of Caenorhabditis elegans fat regulatory genes.
|
Nature 421: 268-272. Banse SA, Lucanic M, Sedore CA, Coleman-Hulbert AL, Plummer WT, Chen E, Kish JL, Hall D, Onken B, Presley MP et al. 2019. Automated lifespan determination across Caenorhabditis strains and species reveals assay-specific effects of chemical interventions. Geroscience 41: 945-960. Baruah A, Chang H, Hall M, Yuan J, Gordon S, Johnson E, Shtessel LL, Yee C, Hekimi S, Derry WB et al. 2014. CEP-1, the Caenorhabditis elegans p53 homolog, mediates opposing longevity outcomes in mitochondrial electron transport chain mutants. PLoS Genet 10: e1004097. Benedetti C, Haynes CM, Yang Y, Harding HP, Ron D. 2006. Ubiquitin-like protein 5
positively regulates chaperone gene expression in the mitochondrial unfolded protein response. Genetics 174: 229-239. Chinetti G, Fruchart JC, Staels B. 2000. Peroxisome proliferator-activated receptors (PPARs): nuclear receptors at the crossroads between lipid metabolism and inflammation. Inflamm Res 49: 497-505. Copeland JM, Cho J, Lo T, Jr., Hur JH, Bahadorani S, Arabyan T, Rabie J, Soh J, Walker
DW. 2009. Extension of Drosophila life span by RNAi of the mitochondrial respiratory chain. Curr Biol 19: 1591-1598. Cox J, Hein MY, Luber CA, Paron I, Nagaraj N, Mann M. 2014. Accurate proteome-wide
label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ. Mol Cell Proteomics 13: 2513-2526. Cutler RG, Thompson KW, Camandola S, Mack KT, Mattson MP. 2014. Sphingolipid
metabolism regulates development and lifespan in Caenorhabditis elegans. Mech Ageing Dev 143-144: 9-18. D'Amico D, Mottis A, Potenza F, Sorrentino V, Li H, Romani M, Lemos V, Schoonjans K, Zamboni N, Knott G et al. 2019. The RNA-Binding Protein PUM2 Impairs Mitochondrial Dynamics and Mitophagy During Aging. Mol Cell 73: 775-787 e710. D'Amico D, Sorrentino V, Auwerx J. 2017. Cytosolic Proteostasis Networks of the Mitochondrial Stress Response. Trends Biochem Sci 42: 712-725.
de Sena Brandine G, Smith AD. 2019. Falco: high-speed FastQC emulation for quality
control of sequencing data. F1000Res 8: 1874. Dillin A, Hsu AL, Arantes-Oliveira N, Lehrer-Graiwer J, Hsin H, Fraser AG, Kamath RS,
Ahringer J, Kenyon C. 2002. Rates of behavior and aging specified by mitochondrial function during development. Science 298: 2398-2401. Dingley SD, Polyak E, Ostrovsky J, Srinivasan S, Lee I, Rosenfeld AB, Tsukikawa M, Xiao
R, Selak MA, Coon JJ et al. 2014. Mitochondrial DNA variant in COX1 subunit
27
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . significantly alters energy metabolism of geographically divergent wild isolates in Caenorhabditis elegans. Journal of molecular biology 426: 2199-2216. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M,
Gingeras TR.
|
2013. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29: 15-21. Doroszuk A, Snoek LB, Fradin E, Riksen J, Kammenga J. 2009. A genome-wide library of
CB4856/N2 introgression lines of Caenorhabditis elegans. Nucleic Acids Res 37: e110. Durieux J, Wolff S, Dillin A. 2011. The cell-non-autonomous nature of electron transport
chain-mediated longevity. Cell 144: 79-91. Evans DR, Guy HI. 2004. Mammalian pyrimidine biosynthesis: fresh insights into an ancient
pathway. The Journal of biological chemistry 279: 33035-33038. Goudeau J, Bellemin S, Toselli-Mollereau E, Shamalnasab M, Chen Y, Aguilaniu H. 2011. Fatty acid desaturation links germ cell loss to longevity through NHR-80/HNF4 in C. elegans. PLoS Biol 9: e1000599. Green CL, Mitchell SE, Derous D, Wang Y, Chen L, Han JJ, Promislow DEL, Lusseau D, Douglas A, Speakman JR. 2017. The effects of graded levels of calorie restriction: IX. Global metabolomic screen reveals modulation of carnitines, sphingolipids and bile acids in the liver of C57BL/6 mice. Aging cell 16: 529-540. Harlow PH, Perry SJ, Stevens AJ, Flemming AJ. 2018. Comparative metabolism of
xenobiotic chemicals by cytochrome P450s in the nematode Caenorhabditis elegans. Scientific reports 8: 13333. Haynes CM, Yang Y, Blais SP, Neubert TA, Ron D. 2010. The matrix peptide exporter HAF-
1 signals a mitochondrial UPR by activating the transcription factor ZC376.7 in C. elegans. Mol Cell 37: 529-540. Herholz M, Cepeda E, Baumann L, Kukat A, Hermeling J, Maciej S, Szczepanowska K,
Pavlenko V, Frommolt P, Trifunovic A. 2019. KLF-1 orchestrates a xenobiotic detoxification program essential for longevity of mitochondrial mutants. Nat Commun 10: 3323. Herzog K, Pras-Raves ML, Vervaart MA, Luyf AC, van Kampen AH, Wanders RJ, Waterham HR, Vaz FM. 2016. Lipidomic analysis of fibroblasts from Zellweger spectrum disorder patients identifies disease-specific phospholipid ratios. J Lipid Res 57: 1447- 1454. Hoshino A, Wang WJ, Wada S, McDermott-Roe C, Evans CS, Gosis B, Morley MP, Rathi
KS, Li J, Li K et al. 2019. The ADP/ATP translocase drives mitophagy independent of nucleotide exchange. Nature 575: 375-379. Houtkooper RH, Mouchiroud L, Ryu D, Moullan N, Katsyuba E, Knott G, Williams RW,
Auwerx J. 2013. Mitonuclear protein imbalance as a conserved longevity mechanism. Nature 497: 451-457. Huang X, Withers BR, Dickson RC. 2014. Sphingolipids and lifespan regulation. Biochimica
et biophysica acta 1841: 657-664. 28
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Jang JY, Blum A, Liu J, Finkel T. 2018. The role of mitochondria in aging. J Clin Invest 128:
3662-3670. Jovaisaite V, Mouchiroud L, Auwerx J. 2014. The mitochondrial unfolded protein response, a
conserved stress response pathway with implications in health and disease.
|
J Exp Biol 217: 137-143. Kamkina P, Snoek LB, Grossmann J, Volkers RJ, Sterken MG, Daube M, Roschitzki B,
Fortes C, Schlapbach R, Roth A et al. 2016. Natural Genetic Variation Differentially Affects the Proteome and Transcriptome in Caenorhabditis elegans. Mol Cell Proteomics 15: 1670-1680. Kaur S, Aballay A. 2020. G-Protein-Coupled Receptor SRBC-48 Protects against Dendrite Degeneration and Reduced Longevity Due to Infection. Cell Rep 31: 107662. Kim C, Kim J, Kim S, Cook DE, Evans KS, Andersen EC, Lee J. 2019. Long-read
sequencing reveals intra-species tolerance of substantial structural variations and new subtelomere formation in C. elegans. Genome Res 29: 1023-1035. Kim HE, Grant AR, Simic MS, Kohnz RA, Nomura DK, Durieux J, Riera CE, Sanchez M,
Kapernick E, Wolff S et al. 2016. Lipid Biosynthesis Coordinates a Mitochondrial-to- Cytosolic Stress Response. Cell 166: 1539-1552 e1516. Lane AN, Fan TW. 2015. Regulation of mammalian nucleotide metabolism and biosynthesis. Nucleic Acids Res 43: 2466-2485. Lê S, Josse J, Husson F. 2008. FactoMineR: An R Package for Multivariate Analysis. Journal of Statistical Software; Vol 1, Issue 1 (2008). Li TY, Sleiman MB, Li H, Gao AW, Mottis A, Bachmann AM, El Alam G, Li X, Goeminne LJE,
Schoonjans K et al. 2021. The transcriptional coactivator CBP/p300 is an evolutionarily conserved node that promotes longevity in response to mitochondrial stress. Nat Aging 1: 165-178. Liu Y, Samuel BS, Breen PC, Ruvkun G. 2014. Caenorhabditis elegans pathways that
surveil and defend mitochondria. Nature 508: 406-410. Liu YJ, McIntyre RL, Janssens GE, Williams EG, Lan J, van Weeghel M, Schomakers B, van der Veen H, van der Wel NN, Yao P et al. 2020. Mitochondrial translation and dynamics synergistically extend lifespan in C. elegans through HLH-30. The Journal of cell biology 219. Loffler M, Fairbanks LD, Zameitat E, Marinaki AM, Simmonds HA. 2005. Pyrimidine
pathways in health and disease. Trends Mol Med 11: 430-437. Melber A, Haynes CM. 2018. UPR(mt) regulation and output: a stress response mediated by
mitochondrial-nuclear communication. Cell Res 28: 281-295. Merkwirth C, Jovaisaite V, Durieux J, Matilainen O, Jordan SD, Quiros PM, Steffen KK,
Williams EG, Mouchiroud L, Tronnes SU et al. 2016. Two Conserved Histone Demethylases Regulate Mitochondrial Stress-Induced Longevity. Cell 165: 1209- 1223. 29
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Molenaars M, Janssens GE, Williams EG, Jongejan A, Lan J, Rabot S, Joly F, Moerland PD,
Schomakers BV, Lezzerini M et al. 2020. A Conserved Mito-Cytosolic Translational Balance Links Two Longevity Pathways. Cell metabolism 31: 549-563 e547. Molenaars M, Schomakers BV, Elfrink HL, Gao AW, Vervaart MAT, Pras-Raves ML, Luyf
AC, Smith RL, Sterken MG, Kammenga JE et al.
|
2021. Metabolomics and lipidomics in Caenorhabditis elegans using a single-sample preparation. Dis Model Mech 14. Mouchiroud L, Molin L, Kasturi P, Triba MN, Dumas ME, Wilson MC, Halestrap AP, Roussel
D, Masse I, Dalliere N et al. 2011. Pyruvate imbalance mediates metabolic reprogramming and mimics lifespan extension by dietary restriction in Caenorhabditis elegans. Aging cell 10: 39-54. Moullan N, Mouchiroud L, Wang X, Ryu D, Williams EG, Mottis A, Jovaisaite V, Frochaux
MV, Quiros PM, Deplancke B et al. 2015. Tetracyclines Disturb Mitochondrial Function across Eukaryotic Models: A Call for Caution in Biomedical Research. Cell Rep 10: 1681-1691. Nagarathnam B, Kalaimathy S, Balakrishnan V, Sowdhamini R. 2012. Cross-Genome
Clustering of Human and C. elegans G-Protein Coupled Receptors. Evol Bioinform Online 8: 229-259. Naresh NU, Haynes CM. 2019. Signaling and Regulation of the Mitochondrial Unfolded
Protein Response. Cold Spring Harb Perspect Biol 11. Nargund AM, Fiorese CJ, Pellegrino MW, Deng P, Haynes CM. 2015. Mitochondrial and
nuclear accumulation of the transcription factor ATFS-1 promotes OXPHOS recovery during the UPR(mt). Mol Cell 58: 123-133. Nargund AM, Pellegrino MW, Fiorese CJ, Baker BM, Haynes CM. 2012. Mitochondrial
import efficiency of ATFS-1 regulates mitochondrial UPR activation. Science 337: 587-590. Nunnari J, Suomalainen A. 2012. Mitochondria: in sickness and in health. Cell 148: 1145-
1159. Olson MV. 1999. When less is more: gene loss as an engine of evolutionary change. American journal of human genetics 64: 18-23. Pees B, Yang W, Zarate-Potes A, Schulenburg H, Dierking K. 2016. High Innate Immune
Specificity through Diversified C-Type Lectin-Like Domain Proteins in Invertebrates. Journal of innate immunity 8: 129-142. Pellegrino MW, Nargund AM, Kirienko NV, Gillis R, Fiorese CJ, Haynes CM. 2014. Mitochondrial UPR-regulated innate immunity provides resistance to pathogen infection. Nature 516: 414-417. Quiros PM, Mottis A, Auwerx J. 2016. Mitonuclear communication in homeostasis and
stress. Nature reviews Molecular cell biology 17: 213-226. Risso D, Ngai J, Speed TP, Dudoit S. 2014. Normalization of RNA-seq data using factor
analysis of control genes or samples. Nat Biotechnol 32: 896-902. 30
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, Smyth GK. 2015. limma powers
differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43: e47. Robinson MD, McCarthy DJ, Smyth GK. 2010. edgeR: a Bioconductor package for
differential expression analysis of digital gene expression data. Bioinformatics 26: 139-140. Shishkova E, Hebert AS, Westphall MS, Coon JJ.
|
2018. Ultra-High Pressure (>30,000 psi)
Packing of Capillary Columns Enhancing Depth of Shotgun Proteomic Analyses. Anal Chem 90: 11503-11508. Shpilka T, Haynes CM. 2018. The mitochondrial UPR: mechanisms, physiological functions and implications in ageing. Nature reviews Molecular cell biology 19: 109-120. Sorrentino V, Romani M, Mouchiroud L, Beck JS, Zhang H, D'Amico D, Moullan N, Potenza F, Schmid AW, Rietsch S et al. 2017. Enhancing mitochondrial proteostasis reduces amyloid-beta proteotoxicity. Nature 552: 187-193. Srinivasan S. 2015. Regulation of body fat in Caenorhabditis elegans. Annu Rev Physiol 77:
161-178. Sun N, Youle RJ, Finkel T. 2016. The Mitochondrial Basis of Aging. Mol Cell 61: 654-666. Taubert S, Van Gilst MR, Hansen M, Yamamoto KR. 2006. A Mediator subunit, MDT-15, integrates regulation of fatty acid metabolism by NHR-49-dependent and - independent pathways in C. elegans. Genes & development 20: 1137-1149. Tharyan RG, Annibal A, Schiffer I, Laboy R, Atanassov I, Weber AL, Gerisch B, Antebi A. 2020. NFYB-1 regulates mitochondrial function and longevity via lysosomal prosaposin. Nat Metab 2: 387-396. Thomas JH. 2006. Adaptive evolution in two large families of ubiquitin-ligase adapters in
nematodes and plants. Genome Research 16: 1017-1030. Thompson OA, Snoek LB, Nijveen H, Sterken MG, Volkers RJ, Brenchley R, Van't Hof A, Bevers RP, Cossins AR, Yanai I et al. 2015. Remarkably Divergent Regions Punctuate the Genome Assembly of the Caenorhabditis elegans Hawaiian Strain CB4856. Genetics 200: 975-989. Vafai SB, Mootha VK. 2012. Mitochondrial disorders as windows into an ancient organelle. Nature 491: 374-383. Van Gilst MR, Hadjivassiliou H, Yamamoto KR. 2005. A Caenorhabditis elegans nutrient
response system partially dependent on nuclear receptor NHR-49. Proceedings of the National Academy of Sciences of the United States of America 102: 13496- 13501. West AP, Shadel GS. 2017. Mitochondrial DNA in innate immune responses and
inflammatory pathology. Nat Rev Immunol 17: 363-375. Wu Z, Senchuk MM, Dues DJ, Johnson BK, Cooper JF, Lew L, Machiela E, Schaar CE,
DeJonge H, Blackwell TK et al. 2018. Mitochondrial unfolded protein response transcription factor ATFS-1 promotes longevity in a long-lived mitochondrial mutant through activation of stress response pathways. BMC Biol 16: 147. 31
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Yoneda T, Benedetti C, Urano F, Clark SG, Harding HP, Ron D. 2004. Compartment-
specific perturbation of protein handling activates genes encoding mitochondrial chaperones. J Cell Sci 117: 4055-4066. Yu G, Wang LG, Han Y, He QY. 2012. clusterProfiler: an R package for comparing biological
themes among gene clusters.
|
OMICS 16: 284-287. 32
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Supplemental Figures
Supplemental Figure S1. Gene expression changes at mRNA and protein level in both N2 and
CB4856 worms upon Dox. Related to Figure 1. (A) Venn diagrams depicting the distribution of
transcripts/mRNAs, proteins, lipids, and metabolites in N2 and CB4856 worms upon Dox compared to
those at control condition. Shared elements (numbers in the middle), N2-specific (numbers on the left
side), and CB4856-specific (numbers on the right side) elements in response to Dox treatment are
indicated within each pairwise comparison. Upregulation (red), downregulation (blue), and reciprocal
regulation (black font) of significantly impacted elements (mRNAs, proteins, lipids and metabolites) are
depicted. (B) Venn diagrams demonstrate genes significantly altered upon Dox in both N2 and CB4856
worms at mRNA and protein level. Genes that are changed at both mRNA and protein levels (numbers
in the middle), only at mRNA level (numbers on the left side), and only at protein level (numbers on the
right side) upon Dox treatment are indicated in each pairwise comparison, respectively. Upregulation
(red), downregulation (blue), and reciprocal regulation (black) of significantly impacted genes. 33
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Supplemental Figure S2. GSEA maps reveal strain-specific differences at transcript and protein
level at basal condition. Related to Figure 2. The color of the dots represents the adjusted p-value;
The size of the dots indicates the number of genes within each gene set. The edges represent the
presence of mutual genes between the gene sets it connects. 34
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Supplemental Figure S3. Strain-specific changes at transcriptomic level upon Dox. Related to
Figure 3. GO term enrichment analysis was performed on the significantly up-regulated genes
exclusively in N2 (A), and CB4856 (B), respectively. GO term enrichment (biological process) of these
genes using David and ReviGO showed upregulated GO terms. The size of the dots indicated the
frequency of the GO term in the underlying Gene Ontology Annotation database; the plots are color-
coded according to significance (Log10-transformed); level of significance increases from red to blue.
|
GO terms belonging to the same cluster were grouped and circled in dark grey dashed lines. 35
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Supplemental Figure S4. Limits in mitochondrial protein detection revealed less alterations in
Dox-treated worms. Related to Figure 4. Changes of mitochondrial proteins are of limited detection. Outer ring: N2; inner ring: CB4856. Orange bars: upregulated genes upon Dox; blue bars:
downregulated genes upon Dox. Grey lines: mitochondrial proteins that were not able to detect. 36
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
doi:
https://doi.org/10.1101/2021.07.20.453059
;
this version posted July 20, 2021.
made available under a
CC-BY-NC-ND 4.0 International license . Supplemental Figure S5. Lipidomics data profiles were significantly altered in Dox-treated
worms. Related to Figure 5. (A) Overview of measured lipids and their belonging classes. (B-C)
Percentage distributions of significantly altered lipids species in N2 (B) and CB4856 (C) worms upon
Dox. (D) Dox effect in both worm strains on the single chain lipids. Np+: normal phase with positive
scan; Np-: normal phase with negative scan; Rp+: reverse phase with positive scan; Rp-: reverse phase
with negative scan. (E) The majority of fatty acids remained unchanged in Dox-treated worms. The
dotted-line indicates the significance threshold (p-value <0.05). 37
|
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Sex-dependent parameters of social behavior show marked variations between distinct laboratory mouse strains and their mixed offspring
Natalia Kopachev1, Shai Netser1 and Shlomo Wagner1*
1Sagol Department of Neurobiology, Faculty of Natural Sciences, University of Haifa, Haifa,
Israel
Correspondence: [email protected], +97248288773
1
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Abstract
Background: The survival of individuals of gregarious species depends on their ability to
properly form social interactions. In humans, atypical social behavior is a hallmark of several
psychopathological conditions, such as depression and autism spectrum disorder, many of which
have sex-specific manifestations. Various strains of laboratory mice are used to reveal the
mechanisms mediating typical and atypical social behavior in mammals. Methods: Here we used three social discrimination tests (social preference, social novelty
preference, and sex preference) to characterize social behavior in males and females of three
widely used laboratory mouse strains (C57BL/6J, BALB/c, and ICR). Results: We found marked sex- and strain-specific differences in the preference exhibited by
subjects in a test-dependent manner. Interestingly, we found some characteristics that were
strain-dependent, while others were sex-dependent. Moreover, even in the social preference test,
where both sexes of all strains prefer social over object stimuli, we revealed sex- and strain-
specific differences in the behavioral dynamics. We then cross-bred C57BL/6J and BALB/c
mice and demonstrated that the offspring of such cross-breeding exhibit a profile of social
behavior which is different from both parental strains and depends on the specific combination of
parental strains. Conclusions: We conclude that social behavior of laboratory mice is highly sex- and strain-
specific and strongly depends on genetic factors. Keywords: social behavior; mouse strains; social preference; social novelty preference; sex preference; sex-specific behavior. 2
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license .
|
1. Introduction
The survival and success of individuals of gregarious mammalian species depend on their
ability to form social interactions properly (1, 2). In humans, atypical social behavior is a
hallmark of several psychopathological conditions and neurodevelopmental diseases (NDDs) (3),
such as social anxiety disorder (4), autism spectrum disorder (5), and schizophrenia (6). Notably,
many of these conditions have gender-specific manifestations and exhibit a robust diagnostic
gender-bias (7, 8). However, unraveling the biological basis of gender-specific manifestations in
Human NDDs is extremely difficult due to the strong influence of culture, education, and living
style, all of which are heavily gender-biased (9). One way to overcome this difficulty is by using
animal models, which are not affected by cultural factors (8, 10). Animal models are an
important tool for exploring the biological basis of social behavior in general and particularly are
used to unravel impaired mechanisms which underlie atypical social behavior in NDDs (11-14). Specifically, genetically modified mouse models carrying mutations in NDD-associated genes
are widely used for such research (15). Notably, different laboratories use distinct laboratory
mouse strains, some of which are inbred strains (16), for their research, and these strains also
serve as a genetic background for the various mouse models of NDDs (17, 18). While multiple
previous studies have explored inter-strain differences in murine social behavior (19-21), the
effect of sex on aspects of social behaviors which are not associated with aggression, parenting,
or sexual behavior has been poorly studied. Moreover, nothing is known about the consequences
of mixing distinct strains by cross-breeding, regarding the social behavior of the offspring. Several behavioral paradigms have been developed for assessing social behavior in mouse
models. Of these, the most famous is the three-chamber test (22), which employs two types of
social discrimination tasks: social preference (SP) and social novelty preference (SNP). We have
3
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . previously presented a novel experimental system for automatic assessment of murine social
discrimination behavior (23) and used it to analyze the behavior of C57BL/6J mice in the SP and
SNP paradigms (21, 24). Here we added a third paradigm of social discrimination, the sex-
preference (SxP) paradigm (25), in order to systematically examine sex- and strain-dependent
behavioral parameters using three distinct strains of laboratory mouse strains (C57BL/6J,
BALB/c, and ICR). Moreover, we characterized the behavior of offspring generated by cross-
breeding of C57BL/6J and BALB/c mice.
|
2. Results
2.1 Sex- and strain-dependent differences in social discrimination behavior
To assess sex- and strain-specific differences in social behavior, we employed our
computerized behavioral system (23, 24) to analyze the level and dynamics of investigation
behavior exhibited by mice in three distinct social discrimination tests: SP, SNP, and SxP. We
used all three tests to systematically characterize the behavior of male and female subjects of
three distinct laboratory strains: C57BL/6J, BALB/c, and ICR (CD-1). Of these strains, the
former two are inbred while the latter is an outbred strain. In all tests, we automatically measured
the time dedicated by the subject to investigating each of two stimuli, simultaneously presented
in distinct chambers, which are located at the opposite corners of the experimental arena. A
statistically significant difference in investigation time between the two stimuli was considered
to reflect a preference for the stimulus investigated by the subject for a longer duration. Notably,
all experiments were repeated at least twice using distinct cohorts of animals, and the minimum
number of mice subjected to each test was nineteen. 4
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . As apparent in Fig. 1, we found sex- and strain-specific differences in the performance of
the subjects in a test-dependent manner. While a preference for the social stimulus compared to
the object stimulus was observed for all strains and both sexes (Fig. 1a. C57: Males: t57=7.754,
p<0.0001; Females: t26=4.044, p<0.0001. BALB/c: Males: t20=10.210, p<0.0001; Females:
t18=3.080, p=0.006. ICR: Males: t22=9.496, p<0.0001; Females: t34=6.110, p<0.0001, paired t-
test), the results of the other two tests were different. A preference for the novel social stimulus
over the familiar one in the SNP test was observed for females of all strains. However, when
examining males, only C57BL/6J mice showed such a preference (Fig. 1b. C57: Males:
t57=5.780, p<0.0001; Females: t24=3.088, p<0.005. BALB/c: Males: t20=1.423, n.s. ; Females:
t19=4.287, p<0.0001. ICR: Males: t23=1.199, n.s. ; Females: t34=3.774, p<0.001, paired t-test). The most variable pattern was found in the SxP test (Fig. 1c), where C57BL/6J and ICR males
showed a clear preference for the opposite-sex stimulus, while BALB/c males did not show any
preference between the stimuli (C57: t44=4.281, p<0.0001; BALB/c: t21=1.812, n.s. ; ICR:
t37=9.508, p<0.0001, paired t-test). In contrast, in none of the strains did females show a
preference for the opposite-sex stimulus. However, ICR females showed a clear preference for
the same-sex stimulus (C57: t49=-0.982, n.s. ; BALB/c: t19=0.739, n.s. ; ICR: t34=5.833,
p<0.0001, paired t-test).
|
5
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Figure 1. Sex- and strain-specific differences in social behavior during three distinct social
discrimination tests: SP, SNP, and SxP. 6
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Mean investigation time (±SEM) measured separately for each stimulus across the SP (a), SNP
(b), and SxP (c) tests, performed by females (left) and males (right) of all three mouse strains
(denoted above). Sample size is denoted below the bars (**p<0.01, ***p<0.001, ns – not
significant, paired t-test). For direct statistical comparison between animal groups, we calculated the difference in
investigation time between the two stimuli (henceforth termed
∆
IT) and compared it across sexes
and strains separately for each test. As apparent in Fig. 2a, for the SP test, we found a significant
main effect of the sex (F1,177=19.241, p<0.0001, two-way ANOVA). Post hoc analysis revealed a
significantly higher
∆
IT for males, as compared to females, for all three strains (C57: t83=-2.085,
p=0.040; BALB/c: t38=-2.113, p=0.041; ICR: t56=-3.787, p<0.0001, independent t-test). In the
SNP test (Fig. 2b), there was a significant interaction between sex and strain (F2,177=4.864,
p<0.009, two-way ANOVA). Post hoc analysis revealed significantly higher
∆
IT for female than
male ICR mice (t57=2.009, p<0.049), while no significant difference was observed between male
and female C57BL/6J and BALB/c mice (C57: t81=-1.782, n.s. ; BALB/c: t39=1.901, n.s.,
∆
independent t-test). Between strains, C57BL/6J male mice showed significantly higher
IT than
ICR male mice (p<0.040, Tukey’s HSD post hoc test), while no significant difference was
observed between females of the three strains. As for the SxP test (Fig. 2c), we observed again a
significant interaction between sex and strain (F2,204=12.705, p<0.0001, two-way ANOVA). Post
hoc analysis revealed significantly higher
∆
IT for C57BL/6J and ICR males than their respective
females (C57: t93=-3.428, p<0.001; ICR: t71=-10.690, p<0.0001, independent t-test). Across
strains, significantly lower
∆
IT values were found for ICR females compared to both C57BL/6J
and BALB/c females (C57: p<0.001; BALB/c: p<0.019, Tukey’s HSD post hoc test), while ICR
7
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
|
It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . ∆
male mice showed significantly higher
IT than BALB/c male mice (p<0.007, Tukey’s HSD
post hoc test). Overall, these data suggest sex- and strain-specific differences in the preference
exhibited by subjects during the SP, SNP, and SxP tests, in a test-dependent manner. Figure 2. Sex- and strain-dependent difference in investigation time between the two
stimuli in the SP, SNP, and SxP tests. Mean (±SEM) difference in investigation time (
∆
IT) between the two stimuli, measured
separately for each animal group during the SP (a), SNP (b), and SxP (c) tests. *p<0.05,
**p<0.01, ***p<0.001, ns – not significant, post hoc unpaired t-test or Tukey’s HSD test,
following main effect in two-way ANOVA test. 2.2 Strain-, but not sex-dependent differences in the dynamics of transitions between
stimuli
In a previous study, we found that C57BL/6J mice and Sprague-Dawley (SD) rats exhibit
distinct dynamics of transitions between the two stimuli during the SP test and linked these
differences to their distinct dynamics of social motivation (21). Therefore, in the current study
we investigated the dynamics of transitions during all three tests for all animal groups. To avoid
bias that may affect the behavioral dynamics towards the end of the experiment, we calculated
8
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . the transitions that took place during the first four minutes of each trial. Interestingly, unlike the
differences in investigation time (Fig. 1), we found no difference between males and females of
any of the strains in the dynamics of transitions between stimuli (Fig. 3). We observed, however,
marked differences among the various strains in a test-dependent manner. This difference was
most notable between C57BL/6J mice and BALB/c mice. In accordance with our previous
report, both male and female C57BL/6J mice exhibited high levels of transitions at the beginning
of the test, which gradually declined during later stages. In contrast, BALB/c mice exhibited a
rather constant pattern of transitions, except for the SxP test, were a lower level at the beginning
of the test was observed among females. For ICR mice, the pattern of transitions was rather
constant across the various tests with one exception: in the SxP test both male and female ICR
mice exhibited a pattern of transitions which was similar to the one displayed by C57BL/6J mice. 9
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
|
It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Figure 3. The dynamics of transitions between stimuli across the three behavioral tests
a. Mean number of transitions made by male (left) and female (right) subjects of the distinct
mouse strains (denoted above) during the SP test (1-min bins). b. As in a, for the SNP test. c.
As in a, for the SxP test. p<0.05, **p<0.1, ***p<0.01, one-way repeated ANOVA. For statistical comparison between the groups, we calculated the difference in transition
number between the first and fourth minute of the test (henceforth termed
∆
Transition), for each
animal group. These time bins were selected since they displayed the greatest difference in
transition number in most cases. For the SP test (Fig. 4a), we found a significant main effect of
the strain but not sex (Strain: F2,177=14.330, p<0.0001; Sex: F1,177=2.872, n.s., two-way
ANOVA). Post hoc analysis revealed a significant difference between C57BL/6J males and males
of both other strains (C57 vs. BALB/c: p<0.0001; C57 vs. ICR: p<0.0001, Tukey’s HSD post hoc
test), while C57BL/6J female mice showed a significant difference from BALB/c females only
(C57 vs. BALB/c: p<0.026, Tukey’s HSD post hoc test). A slightly different pattern was found
for the SNP test (Fig. 4b), with post hoc analysis following main effect of strain (Strain:
F2,177=19.962, p<0.0001, two-way ANOVA) revealing significant differences between C57BL/6J
female mice and females of both other strains (C57 vs. BALB/c: p<0.0001; C57 vs. ICR:
p<0.0001, Tukey’s HSD post hoc test), while male C5BL/6J mice showed significant difference
from ICR males only (C57 vs. ICR: p<0.008, Tukey’s HSD post hoc test). For the SxP test (Fig. 4c), post hoc analysis following main effect of strain (Strain: F2,204=11.639, p<0.0001, two-way
10
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . ANOVA) revealed that C57BL/6J and ICR males differed from BALB/c males, while ICR
females differed from BALB/c females (Males: C57 vs. BALB/c: p<0.001; BALB/c vs. ICR:
p<0.002; Females: BALB/c vs. ICR: p<0.011, Tukey’s HSD post hoc test). Overall, these results
suggest that the transition dynamics are a strain-, but not sex-specific feature, which is kept
rather unchanged across the various tests, at least for the inbred C57BL/6j and BALB/c mice. Figure 4. Sex- but not strain-dependent difference in the dynamics of transitions between
the two stimuli. Mean (±SEM) difference in the number of transitions between the two stimuli between the first
and fourth minute of each test (
∆
Transition), measured separately for each animal group during
the SP (a), SNP (b), and SxP (c) tests.
|
*p<0.05, **p<0.01, ***p<0.001, post hoc unpaired t-test
or Tukey’s HSD test, following main effect in two-way ANOVA test. 2.3 Sex- and strain-dependent differences in the dynamics of social preference behavior
We next analyzed the dynamics of investigation behavior across the various strains and
sexes. For this analysis, we focused on the SP test for two reasons. First, since all groups showed
a significant preference of the social stimulus over the object, we reasoned that differences in
behavioral dynamics during this test could not be attributed to preference variations. Second, as
11
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . this test was always the first to be performed, it could not be affected by other tests in a strain- or
sex-dependent manner. We observed an apparent difference between males and females in the
behavioral dynamics of the SP test (Fig. 5a). While males of all strains showed social preference
that was very strong at the beginning of the test and declined over time, females showed a
relatively stable preference throughout the test. To statistically analyze these apparent
∆
differences, we compared
IT between the first and last two minutes of the test, across sex and
strains (Fig. 5b). We found an interaction between time and sex (F1,177=16.468, p=0.0001,
mixed-model ANOVA). Post hoc analysis revealed a significant difference between the first and
last two minutes only for males, but not for females of all strains (C57: Males: t57=7.017,
p=0.0001; Females: t26=1.182, n.s., BALB/c: Males: t20=2.242, p=0.036; Females: t18=-1.170,
n.s., ICR: Males: t22=3.079, p<0.005; Females: t34=-0.022, n.s., paired t-test). Thus, the
behavioral dynamics during the SP test were sex-dependent. 12
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Figure 5. Sex-dependent dynamics of investigation behavior during the SP test. 13
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . (a) Mean investigation time (±SEM), measured separately for each stimulus (20-s bins) along the
time course of the SP test session (20-s bins) for females (upper panels) and males (lower panels)
of the three strains.
|
Note the transiently strong preference of males as compared to the stable but
weaker preference of females. (b) Mean
∆
IT (±SEM) measured separately for the first and last two minutes of SP test. *p<0.05,
**p<0.01, ***p<0.001, ns – not significant, post hoc paired t-test following the main effect in
mixed-model ANOVA test. 2.4. Sex-dependent differences in bout duration during the SP test
We have previously shown that in C57BL/6J mice the difference in investigation time
between the social stimulus and the object during the SP test is reflected only in long (>6 s)
≤
investigation bouts, while shorter bouts (
6 s) do not differ between the two stimuli (24). To
examine this point across sexes and strains, we plotted the fraction of
∆
IT as a function of
investigation bout duration for all strains and compared it between males and females. As for the
transitions, we calculated only the bouts that took place during the first four minutes of each trial,
in order to avoid bias affecting the behavioral dynamics towards the end of the experiment. As
apparent in Fig. 6a, in all strains there seems to be a clear difference between males and females. Males of all strains showed longer bouts than females. In addition, BALB/c and ICR females
∆
showed more short bouts than males. We then statistically compared
IT separately for short and
long bouts across sexes and strains. For short bouts (Fig. 6b), we found a significant interaction
between sex and strain (F2,177=6.834, p<0.001, two-way ANOVA), with a post hoc analysis
showing a higher
∆
IT for female BALB/c and ICR, but not C57BL/6J mice, as compared to
males (BALB/c: t38=4.286, p<0.0001; ICR: t56=2.211, p=0.039, independent t-test). Also,
14
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . ∆
BALB/c showed a significantly higher
IT as compared to C57BL/6J females, while no
difference was found between the males (Females: C57 vs. BALB/c: p<0.005; Tukey’s HSD post
hoc test). In contrast, long bouts (Fig. 6c) showed the main effect in
∆
IT only for the sex
∆
(F1,177=36.022, p<0.0001), with a post hoc analysis revealing a significantly higher
IT for
males as compared to females in all strains (C57: t83=-2.130, p=0.036; BALB/c: t38=-3.916,
p<0.0001; ICR: t56=-4.616, p<0.0001, independent t-test). Thus, it seems as if social preference
of males is generally expressed by longer bouts, while females express their social preference
using shorter investigation bouts. 15
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license .
|
Figure 6. Sex-dependent differences in bout duration during the SP test. (a) Cumulative
∆
IT plotted against bout duration during the SP test, separately for males (blue)
and females (brown) of the three strains. Inset – same data on a larger scale for duration<10 s.
∆
≤
(b) Mean
IT (±SEM), calculated separately for short (
6 s) investigation bouts generated
during the SP test by each animal group. (c) As in B, for long (>6 s) bouts. 16
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . p<0.05, **p<0.01, ***p<0.001, ns – not significant, post hoc unpaired t-test or Tukey’s HSD
test, following main effect in two-way ANOVA test. 2.5. Offspring of a cross-breeding between C57BL/6J and BALB/c mice exhibit distinct
social behavioral characteristics which depend on their parental combination
Finally, as we found significant differences in social behavior between the distinct strains,
we sought to explore the effect of cross-breeding two distinct strains. For this purpose, we
focused on male offspring generated by crossing C57BL/6J and BALB/c mice, both of which are
inbred strains of a similar size. We separately analyzed litters of C57BL/6J mothers and BALB/c
fathers, termed by us CB mice, and litters of BALB/c mothers and C57BL/6J fathers, termed BC
mice. Interestingly, we found that the BC and CB offspring showed different behavioral profiles,
neither of which recapitulated the profiles of any of their parental strains. In the SP test (Fig. 7a),
we found that only CB mice showed a significant social preference (BC: t35=-0.298, n.s. ; CB:
t46=2.903, p=0.006, paired t-test). Nonetheless, the two cross-bred strains were similar to each
other in the dynamics of their behavior, with an initial preference at the beginning of the test and
a loss of preference at a later stage (Fig. 7b), a pattern not observed in any of the parental strains
(Fig. 5a). As for the transitions, both groups showed a pattern resembling their mothers, with BC
mice showing a pattern similar to BALB/c mice and CB mice to C57BL/6J mice (Fig. 7c). In the
SNP test (Fig. 7d), both BC and CB mice did not show any preference, similar to BALB/c males
but in contrast to C57BL/6J mice (BC: t18=0.531, n.s. ; CB: t23=1.137, n.s., paired t-test). The
behavioral similarity of both groups to BALB/c mice was also apparent in the dynamics of the
investigation behavior (Fig. 7e) and transitions (Fig. 7f). In the case of the SxP test, while both
CB and BC mice exhibited a significant preference for the female over the male stimulus (Fig. 17
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
|
It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 7g), this preference looked stronger in CB mice, similar to C57BL/6J males, and much weaker in
BC mice, similar to BALB/c males, which did not show any sex preference (BC: t32=3.430,
p<0.002; CB: t44=12.926, p<0.0001, paired t-test). These differences were also apparent from
the dynamics of investigation behavior (Fig. 7h) and transitions (Fig. 7i). Thus, in the SxP test,
both types of cross-bred offspring showed a behavioral pattern which resembles males of their
maternal strain. 18
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Figure 7. Offspring of a cross-breeding between C57BL/6J and BALB/c mice exhibit
distinct social behavioral characteristic
19
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . (a) Mean investigation time for each stimulus during the SP test for BC (green) and CB (blue)
mice. **p<0.01, paired t-test. (b) Mean investigation time measured separately for each stimulus (20-s bins) along the time
course of the SP test session (20-s bins) for BC (upper panel) and CB (lower panel) mice. (c) Mean number of transitions between the two stimuli made during the SP test for BC (upper
panel) and CB (lower panel) subjects. ***p<0.001, one-way repeated ANOVA. (d-f) As in a-c, for the SNP test. (g) As in a, for the SxP test. **p<0.01, ***p<0.001, paired t-test. (h) As in b, for the SxP test. (i) As in c, for the SxP test. **p<0.01, one-way repeated ANOVA
∆
For statistical comparison between the groups, we calculated the
IT (Fig. 8a) and
∆
Transitions (Fig. 8b) and compared between BC and CB male mice separately for each test. For
both parameters, we found that CB mice exhibited significantly higher values than BC mice in
both the SP (
∆
∆
IT: t81=-2.049, p<0.044; ∆
∆
Transitions: t81=-2.672, p<0.009, independent t-test)
and SxP (
IT: t76=-5.308, p<0.0001;
Transitions: t76=-2.944, p<0.004, independent t-test)
tests, while no differences were found in the SNP test (
∆
IT: t41=-0.466, n.s. ;
∆
Transitions: t41=-
∆
1.041, n.s., independent t-test). Notably, although the difference in
Transitions was not
significant in the SNP test, the trend was similar to the other two tests. Thus, regarding the
dynamics of transitions, it seems as if in all cases the offspring behavior resembled their
mothers’ behaviors.
|
We conclude that offspring of BALB/c and C57BL/6J mice exhibit a profile
of social behavior which depends on the parental combination and is different from both parental
strains, with a tendency to mimic the behavior of the maternal strain. 20
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Figure 8. Comparing offspring of C57BL/6J mothers and BALB/c fathers with offspring of
BALB/c mothers and C57BL/6J fathers
∆
(a-c) Mean
IT, for the SP (a), SNP (b), and SxP (c) results of BC and CB mice, shown in Fig. 7a, d, g, respectively. ∆
(d-f) Mean
Transitions, for the SP (d), SNP (e), and SxP (f) results of BC and CB mice, shown
in Fig. 7c, f, i, respectively. p<0.01, **p<0.01, ***p<0.001, ns – not significant, unpaired t-test. 21
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 3. Discussion
Sex-dependent differences in murine social behavior have been comprehensively explored
in previous studies, but this was done mainly regarding hormonal-driven aspects of social
behavior (26, 27), such as sexual (28), anxiety/aggressive (10), and parental behaviors (29). In
contrast, evidence for differences in other aspects, such as social preference and social novelty
preference have been reported only anecdotally, and in some cases with contradicting results (see
for example 22, 30). Here, we systematically explored such differences across three types of
social discrimination tasks and three distinct laboratory mouse strains. We have used a relatively
large number of animals and at least two independent cohorts for each animal group, to verify
replicability. We found a marked difference in social investigation behavior between males and
female for all three strains, especially in the SNP and SxP tests. While the SxP test may be
related to sexual behavior (31), the SNP test is not, and is well known to reflect social novelty
seeking. Thus, our results demonstrate that even social behaviors which are not directly related to
the sexual, aggressive, or parental aspects may be sex-dependent. Our computerized experimental system enables analyzing the dynamics of the social
discrimination behavior as well as categorizing each investigation bout according to its duration
(23, 24). Using these features, we found that even in the SP test, where all strains and sexes
exhibit a significant preference for a social stimulus over an object, the dynamics of the behavior
may markedly differ between strains and sexes.
|
This was most profoundly demonstrated by the
difference between ICR and BALB/c males, which exhibited a significant change over time in
their social preference behavior, and all other groups. We have previously shown that this
reduction over time in social preference reflects dynamic changes in the motivation for social
interactions. Thus, ICR and BALB/c males seem to exhibit a very strong drive for social
22
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . interactions at the beginning of the encounter, which gradually decreases over time, while the
other groups show a lower level of social motivation, which is kept rather constant over time. Another dynamic aspect of social preference found to be sex-specific is the duration of
investigation bouts. We have previously shown for male C57BL/6J mice that there is no
difference in investigation using short bouts between the stimuli. In contrast, the preference for
the social stimulus is specifically expressed by more investigation using long bouts. We thus
suggested that long bouts reflect the interaction of the subject with the social stimulus while
short bouts reflect curiosity per se. Here, we show for the first time that in females, specifically
in ICR and BALB/c females, the picture is quite different. In contrast to males, females express
their social preference by both long and short bouts, with no apparent difference between them. Moreover, in all strains, males exhibited longer investigation bouts towards the social stimulus
than females. This sex-specific difference in the duration of investigation bouts may reflect a
weaker motivation for interaction with a novel same-sex social stimulus exhibited by females,
compared to males. Another aspect of the behavioral dynamics we explored is the transitions between the two
stimuli. Interestingly, while most differences in investigation behavior were between males and
females, we found no sex-dependent differences in the pattern of transitions. Instead, this
characteristic seems highly strain-dependent. For example, the C57BL/6J and BALB/c strains
exhibited a rather uniform and specific pattern across both sexes and all tests, while ICR mice
showed a pattern similar to BALB/c mice in the SP and SNP tests and a pattern similar to
C57BL/6J mice in the SxP test. In a previous study (21) we suggested that similarly different
patterns of transitions between SD rats and C57BL/6J mice reflect their distinct dynamics of
social motivation. Accordingly, a recent study found a specifically high level of activity in
23
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.
|
It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . dopaminergic neurons of the ventral tegmental area (VTA) during transition from object to social
stimuli (30). Thus, the strain-dependent distinct patterns of transitions found by us may reflect
distinct dynamics of social motivation between the strains. The genetic basis of social behavior is well established throughout the animal kingdom. Significant differences between various mouse strains have been previously found using the
three-chamber test (19, 20, 22). Yet, to our knowledge, our study is the first to examine the
consequences of cross-breeding between distinct mouse strains which exhibit different social
behavior. Interestingly, the behavior of F1 male offspring of the cross-breeding scheme was
different from both parental strains in a test-dependent manner. This was most strikingly
demonstrated by the dynamics of the social preference behavior, where both BC and CB mice
showed such preference only at the beginning of the test. Such a phenomenon, of F1 behavior
which is weaker than both parental mouse strains, was previously shown for morphine analgesia
(32). Even more surprising is the observation that F1 offspring that were born to male C57BL/6J
mothers and BALB/c fathers, differ in their behavior compared to offspring of BALB/c mothers
and C57BL/6J fathers. Interestingly, in several aspects of their behavior, the F1 offspring showed
resemblance to the strain of their mothers. The question whether this difference between BC and
CB mice is caused by genetic, epigenetic, or environmental (i.e., the strain of the female taking
care of the newborn animals) factors should be addressed by future studies. Overall, we conclude that social behavior of laboratory mice, even if not related to sexual,
aggressive, or parental aspects, is highly sex-and strain dependent. Moreover, we show that the
behavioral outcomes of a cross-breeding between mouse strains is unpredictable and may differ
markedly between tests and breeding schemes. These conclusions should be taken into account
in future studies exploring modified social behavior in genetic mouse models. 24
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 4. Materials and Methods
Animals
The research subjects were naïve male and female adult (8-12 weeks old) mice of three distinct
laboratory strains: C57BL/6J, BALB/c, and ICR (CD-1) and adult male mice generated by
crossing C57BL/6J and BALB/c mice. The social stimuli were C57BL/6J, BALB/c, and ICR
(CD-1) juvenile (21-30 day-old) naïve male and female mice (SP, SNP), and C57BL/6J,
BALB/c, and ICR (CD-1) adult (8-12 weeks old) naïve male and female mice (SxP).
|
Mice were
commercially obtained (Envigo, Israel) and housed in Plexiglas cages in groups of 2-5 animals
per cage. They were kept at 22±2°C under a 12-h light/12-h dark cycle, with lights being turned
on at 7 p.m. each night. All animals had ad libitum access to food (standard chow diet; Envigo
RMS, Israel) and water. Behavioral experiments were performed during the dark phase, under
dim red light. All experiments were performed according to the National Institutes of Health
guide for the care and use of laboratory animals and approved by the Institutional Animal Care
and Use Committee (IACUC) of the University of Haifa. Behavioral assays
All social discrimination tasks were conducted using our published automated experimental
system (23). SP and SNP tests were conducted on the same day, as previously described (23). The SxP test was conducted as previously described (25). Briefly, it consisted of 15 min
habituation to the arena with empty chambers, followed by exposing the subject for 5 min to
both novel adult male and female social stimuli located in individual chambers at opposite
25
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . corners of the arena. All stimuli used in all three tests (SP, SNP, and SxP) were C57BL/6J,
BALB/c, and ICR (CD-1) mice. Data analysis
Video data analysis was conducted by our published custom-made TrackRodent software, as
previously described in detail (23). Statistical analysis
All statistical tests were performed using SPSS 23 (IBM) statistics software. Shapiro-Wilk test
was used for examining the normal distribution of the dependent variables. A 2-tailed paired t-
test was used to compare between parameters within a group, and a 2-tailed independent t-test
was used to compare a single parameter between distinct groups. For examining the influence of
one categorical independent variable on one continuous dependent variable, a one-way ANOVA
model was applied to the data. This model assesses the main effect of the independent variable
on the dependent variable. For examining the influence of two different categorical independent
variables on one continuous dependent variable, a two-way ANOVA model was applied to the
data. This model assesses the main effect of each independent variable and the interaction
between them. For comparison between multiple groups and parameters, a mixed-model analysis
of variance (ANOVA) model was applied to the data. This model contains one random effect
(ID), one within effect, one between effect, and the interaction between them. All ANOVA tests
were followed, if the main effect or interaction were significant, by post hoc Student’s t-test with
Bonferroni’s correction. Significance was set at p-value *<0.05.
|
Supplementary Materials: None
26
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . Author Contributions: Conceptualization, S.W. and S.N. ; methodology, S.N. ; software, S.N. ;
validation, S.W. and S.N. ; formal analysis, S.N. ; investigation, N.K. ; resources, S.W. ; data
curation, S.N. ; writing—original draft preparation, S.W; visualization, S.N and N.K..;
supervision, S.W. ; project administration, S. N.; funding acquisition, S.W. All authors have read
and agreed to the published version of the manuscript. Funding: This study was supported by ISF-NSFC joint research program (grant No. 3459/20 to SW), the Israel Science Foundation (ISF grants No. 1350/12, 1361/17 to SW), the
Ministry of Science, Technology and Space of Israel (Grant No. 3-12068 to SW) and the United
States-Israel Binational Science Foundation (BSF grant No. 2019186 to SW). Institutional Review Board Statement: All experiments were performed according to the
National Institutes of Health guide for the care and use of laboratory animals and approved by
the Institutional Animal Care and Use Committee (IACUC) of the University of Haifa. Code availability: All codes and datasets used for the current study are available on request from
the corresponding author within a reasonable time. The code used for the video analysis is
publicly available at the following link: [https://github.com/shainetser/TrackRodent]. Conflicts of Interest: The authors declare no conflict of interest. References
1. Robinson GE, Fernald RD, Clayton DF (2008): Genes and social behavior. Science. 322:896-900. 27
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 2. Stanley DA, Adolphs R (2013): Toward a neural basis for social behavior. Neuron. 80:816-
826. 3. Porcelli S, Van Der Wee N, van der Werff S, Aghajani M, Glennon JC, van Heukelum S,
et al. (2019): Social brain, social dysfunction and social withdrawal. Neurosci Biobehav
Rev. 97:10-33. 4. Leichsenring F, Leweke F (2017): Social Anxiety Disorder. N Engl J Med. 376:2255-2264. 5.
de la Torre-Ubieta L, Won HJ, Stein JL, Geschwind DH (2016): Advancing the
understanding of autism disease mechanisms through genetics. Nat Med. 22:345-361. 6. Mier D, Kirsch P (2017): Social-Cognitive Deficits in Schizophrenia. Curr Top Behav
Neurosci. 30:397-409. 7. Gobinath AR, Choleris E, Galea LAM (2017): Sex, Hormones, and Genotype Interact To
Influence Psychiatric Disease, Treatment, and Behavioral Research.
|
J Neurosci Res. 95:50-
64. 8. Palanza P, Parmigiani S (2017): How does sex matter? Behavior, stress and animal models
of neurobehavioral disorders. Neurosci Biobehav R. 76:134-143. 9. Grabowska A (2017): Sex on the brain: Are gender-dependent structural and functional
differences associated with behavior? J Neurosci Res. 95:200-212. 10. Bredewold R, Veenema AH (2018): Sex differences in the regulation of social and anxiety-
related behaviors: insights from vasopressin and oxytocin brain systems. Curr Opin
Neurobiol. 49:132-140. 11. Crawley JN (2012): Translational animal models of autism and neurodevelopmental
disorders. Dialogues Clin Neurosci. 14:293-305. 28
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 12. Fernando AB, Robbins TW (2011): Animal models of neuropsychiatric disorders. Annu
Rev Clin Psychol. 7:39-61. 13. Kaiser T, Zhou Y, Feng G (2017): Animal models for neuropsychiatric disorders:
prospects for circuit intervention. Curr Opin Neurobiol. 45:59-65. 14. Nestler EJ, Hyman SE (2010): Animal models of neuropsychiatric disorders. Nat Neurosci. 13:1161-1169. 15. McGraw CM, Ward CS, Samaco RC (2017): Genetic rodent models of brain disorders:
Perspectives on experimental approaches and therapeutic strategies. Am J Med Genet C
Semin Med Genet. 175:368-379. 16. Casellas J (2011): Inbred mouse strains and genetic stability: a review. Animal. 5:1-7. 17. Tam WY, Cheung KK (2020): Phenotypic characteristics of commonly used inbred mouse
strains. J Mol Med (Berl). 98:1215-1234. 18. Wade CM, Daly MJ (2005): Genetic variation in laboratory mice. Nat Genet. 37:1175-
1180. 19. Moy SS, Nadler JJ, Young NB, Nonneman RJ, Segall SK, Andrade GM, et al. (2008):
Social approach and repetitive behavior in eleven inbred mouse strains. Behav Brain Res. 191:118-129. 20. Moy SS, Nadler JJ, Young NB, Perez A, Holloway LP, Barbaro RP, et al. (2007): Mouse
behavioral tasks relevant to autism: phenotypes of 10 inbred strains. Behav Brain Res. 176:4-20. 21. Netser S, Meyer A, Magalnik H, Zylbertal A, de la Zerda SH, Briller M, et al. (2020):
Distinct dynamics of social motivation drive differential social behavior in laboratory rat
and mouse strains. Nat Commun. 11:5908. 29
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 22. Moy SS, Nadler JJ, Perez A, Barbaro RP, Johns JM, Magnuson TR, et al. (2004):
Sociability and preference for social novelty in five inbred strains: an approach to assess
autistic-like behavior in mice.
|
Genes Brain Behav. 3:287-302. 23. Netser S, Haskal S, Magalnik H, Bizer A, Wagner S (2019): A System for Tracking the
Dynamics of Social Preference Behavior in Small Rodents. J Vis Exp. 24. Netser S, Haskal S, Magalnik H, Wagner S (2017): A novel system for tracking social
preference dynamics in mice reveals sex- and strain-specific characteristics. Mol Autism. 8:53. 25. Jabarin R, Levy N, Abergel Y, Berman JH, Zag A, Netser S, et al. (2021): Pharmacological
modulation of AMPA receptors rescues specific impairments in social behavior associated
with the A350V Iqsec2 mutation. Transl Psychiatry. 11:234. 26. Dulac C, Kimchi T (2007): Neural mechanisms underlying sex-specific behaviors in
vertebrates. Curr Opin Neurobiol. 17:675-683. 27. Yang CF, Shah NM (2014): Representing sex in the brain, one module at a time. Neuron. 82:261-278. 28. Shelley DN, Choleris E, Kavaliers M, Pfaff DW (2006): Mechanisms underlying sexual
and affiliative behaviors of mice: relation to generalized CNS arousal. Soc Cogn Affect
Neurosci. 1:260-270. 29. Zilkha N, Scott N, Kimchi T (2017): Sexual Dimorphism of Parental Care: From Genes to
Behavior. Annu Rev Neurosci. 40:273-305. 30. Contestabile A, Casarotto G, Girard B, Tzanoulinou S, Bellone C (2021): Deconstructing
the contribution of sensory cues in social approach. Eur J Neurosci. 53:3199-3211. 30
bioRxiv preprint The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
doi:
https://doi.org/10.1101/2021.09.12.459932
;
this version posted September 15, 2021.
available under a
CC-BY-ND 4.0 International license . 31. Kondo Y, Hayashi H (2021): Neural and Hormonal Basis of Opposite-Sex Preference by
Chemosensory Signals. Int J Mol Sci. 22. 32. Miner LL, Elmer GI, Pieper JO, Marley RJ (1993): Aggression modulates genetic
influences on morphine analgesia as assessed using a classical mendelian cross analysis. Psychopharmacology (Berl). 111:17-22. 31
|
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Single cell based high-throughput Ig and TCR repertoire sequencing analysis in rhesus macaques
Running Title:
Single cell sequencing of rhesus Ig and TCR repertoire
Authors:
Evan S. Walsh*,†, Tammy S. Tollison*, Hayden N. Brochu*,†, Brian I. Shaw‡‡, Kayleigh R. Diveley*,§, Hsuan Chou*, Lynn Law¶, Allan D. Kirk‡‡, Michael Gale, Jr. ¶,||,#, and Xinxia Peng*,†,**
Author Affiliations:
Department of Molecular Biomedical Sciences, North Carolina State University College of Veterinary Medicine, Raleigh, NC 27607; † Bioinformatics Graduate Program, North Carolina State University, Raleigh, NC 27695; ‡ Department of Surgery, Duke University, Durham, NC 27710; § Genetics Graduate Program, North Carolina State University, Raleigh, NC 27695; ¶ Department of Immunology, University of Washington, Seattle, WA 98109; || Center for Innate Immunity and Immune Diseases, University of Washington, Seattle, WA 98109; # Washington National Primate Research Center, University of Washington, Seattle, WA 98121; and ** Bioinformatics Research Center, North Carolina State University, Raleigh, NC 27695
Corresponding author
Xinxia Peng, [email protected], 919-515-4481
Equal contributors:
Evan S. Walsh and Tammy S. Tollison contributed equally to this work. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Abstract
Recent advancements in microfluidics and high-throughput sequencing technologies have enabled recovery
of paired heavy- and light- chains of immunoglobulins (Ig) and VDJ- and VJ- chains of T cell receptors
(TCR) from thousands of single cells simultaneously in humans and mice. Despite rhesus macaques being
one of the most well-studied model organisms for the human adaptive immune response, high-throughput
single cell immune repertoire sequencing assays are not yet available due to the complexity of these
polyclonal receptors. Here we employed custom primers that capture all known rhesus macaque Ig and
TCR isotypes and chains that are fully compatible with a commercial solution for single cell immune
repertoire profiling. Using these rhesus specific assays, we sequenced Ig and TCR repertoires in over 60,000
cells from cryopreserved rhesus PBMC, splenocytes, and FACS-sorted B and T cells. We were able to
recover every Ig isotype and TCR chain, measure clonal expansion in proliferating T cells, and pair Ig and
TCR repertoires with gene expression profiles of the same single cells. Our results establish the ability to
perform high-throughput immune repertoire analysis in rhesus macaques at the single cell level.
|
Introduction
Immunoglobulin (Ig) and T-cell receptor (TCR) repertoire analysis plays a key role in understanding the
development of host immunity. These receptor molecules are responsible for recognizing a myriad of
foreign antigens from infectious agents. The ability of T and B lymphocytes to give rise to such a diversity
of receptor molecules with affinity to these potential antigens is, in part, due to their generation and
structure. Igs are tetrameric proteins typically composed of two identical light chains (IgL or IgK) and two
identical heavy chains (IgH).1 TCRs are heterodimeric proteins composed of paired beta (TCRβ) and alpha
(TCRα) or gamma (TCRγ) and delta (TCRδ) chains, respectively.2 The IgH chain, as well as TCRβ and
TCRδ chains, consist of Variable (V), Diversity (D), Joining (J), and Constant (C) region gene segments. Ig light chains, as well as TCRα and TCRγ chains, do not possess D gene segments. Germline V(D)J gene
segments exist at large loci within the genome and are somatically rearranged to produce functional and
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. diversified mRNA transcripts and proteins. It is estimated that, within a single individual, the number of
possible TCR and Ig V region domains is on the order of 1013 and 1017, respectfully.3
Increasingly, high-throughput single cell-based sequencing techniques are being employed to profile Ig and
TCR repertoires. Single cell immune repertoire sequencing (scIRS) is a promising, new sequencing
approach that allows for paired V(D)J repertoire analysis of thousands of cells simultaneously. For example,
it has been used to identify multiple neutralizing antibodies against SARS-Cov-2 infection in humans.4,5
However, scIRS assays are species specific and the development of scIRS assays relies on the complete Ig
and TCR reference sequences for the species of interest. Ig and TCR loci are characterized by high levels
of repetitive sequences and allelic variation, making targeted sequencing and assembly difficult technical
challenges.6 To our knowledge, current commercial scIRS assays are only available for human and mouse. Due to the close phylogenetic relationship and highly similar physiology to humans, rhesus macaques
(Macaca mulatta) have been one of the most popular and well-studied nonhuman primates (NHPs) for
modeling immune responses in humans.7,8 For example, rhesus macaques have been used to model the
adaptive immune response and progression of infectious diseases from such agents varicella zoster9, HIV10-
12, and SARS-Cov-213, as well as many other immune-related studies and diseases such as allograft
rejection14 and graft versus host disease (GvHD)15. Recently, using long read transcriptome sequencing, we
generated the first complete reference set of constant regions of all known isotypes and chain types of rhesus
Ig and TCR repertoires.16 We also designed in silico rhesus-specific scIRS assays that remove the need for
primers conventionally targeting variable regions.
|
In this study, we sought to experimentally validate and optimize rhesus specific scIRS assays that are fully
compatible with commercial solutions for single cell immune repertoire profiling. Based on the complete
rhesus macaque constant region reference set of Ig and TCR isotypes and chains16, we designed and
validated primers that target these constant regions in mRNA transcripts. We further adopted these rhesus-
specific primers into the human single cell immune profiling workflow provided by 10x Genomics. These
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. rhesus-specific scIRS assays were validated using cryopreserved PBMC and splenocytes as well as FACS-
sorted B and T cells from various rhesus animals. We were able to recover every known Ig and TCR isotype
and pair Ig/TCR repertoire analysis with transcriptome profiles from the same single cells. We also
observed clonal expansion in proliferating versus non-proliferating rhesus T cells. These results establish
the ability to perform high throughput scIRS analysis in rhesus macaques with comparable performances
to commercially available platforms. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Methods
Allogeneic mixed lymphocyte reaction (MLR) and FACS gating strategies
PBMC were obtained from peripheral blood of rhesus macaques using CPT tubes with Sodium Citrate (BD
Biosciences, San Jose, CA). Cells from spleens of rhesus macaques were obtained via manual maceration
of the spleen, and lysis of red blood cells using high yield lysing solution (Life Technologies, Carlsbad,
CA). Cells were cryopreserved in 10% dimethyl-sulfoxide (Sigma-Aldrich, Raleigh, NC) and 90% fetal
bovine serum (Corning, Corning, NY) using temperature controlled freezing containers. To prepare the
mixed lymphocyte reactions (MLRs), cryopreserved cells were obtained, thawed, and counted. CD14+
monocytes were isolated from splenocytes using magnetically labeled antibodies (Miltinyi, Auburn, CA). These cells were cultured in R10 Media [Consisting of RPMI 1640 supplemented with Pen-Strep at
100U/mL, L-Glutamine at 2mM (all Gibco, Gaithersburg, MD) and 10% FCS] with IL-4 (Miltinyi Biotech,
Auburn, CA) and GM-CSF (Miltinyi Biotech, Auburn, CA) for 7 total days in with the addition of TNF-α
(Novus Biologicals, Centennial, CO) on the 6th day to induce activated dendritic cells (DC). PBMC were
thawed and labeled using Violet Proliferation Dye-450 (BD Biosciences, San Jose, CA) and combined with
DCs at a 1:4 ratio. These cells were co-cultured in R10 for 5 days. At this time, the culture was stained with
CD3-APC Cy7 (BD Biosciences, San Jose, CA, 557757) and sent for cell sorting on a BD DIVA (BD
Biosciences, San Jose, CA).
|
Our gating strategy is shown in Supplemental Figure 1. Rhesus samples, single cell processing, and cDNA generation
Multiple sample types from rhesus macaques were used to optimize and validate rhesus specific scIRS
assays: cryopreserved PBMCs and splenocytes, FACS-sorted stimulated B-cells, and FACS-sorted
stimulated non-proliferating and proliferating T-cells. Cells were washed in resuspension buffer (RPMI,
10% FBS), collected by centrifugation at 700xg, and resuspended at an appropriate concentration, between
700 and 1200 cells/ul. All cells were counted by hemocytometer and assayed for viability using a Countess
II automated cell counter (Thermo Fisher Scientific, Waltham, MA). From each sample a maximum volume
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. containing 17,000 cells, having approximately 90% of viable cells, were loaded onto the 10x Chromium
(10x Genomics, Pleasanton, CA) controller for a targeted cell recovery of 10,000 cells (10x Genomics,
Pleasanton, CA, Chromium Next GEM Single Cell V(D)J Reagents Kits v1.1_RevE). Two replicates of the
same PBMC sample were established (PMBC1 and PMBC2); whereas, only one sample was prepped for
each of the remaining cell types. cDNA amplification was carried out at 13 cycles of amplification for all
samples and assayed for quality and concentration using a Bioanalyzer 2100 and High Sensitivity DNA Kit
(Agilent, Santa Clara, CA). Rhesus V(D)J and 5’GEX library construction and sequencing
Construction of V(D)J libraries required the implementation of rhesus-specific primers16 for two subsequent
enrichments of Ig and TCR transcripts as is typical for the 10x V(D)J workflow (Supplemental Table I). qPCR results of various primer pools, which varied primer concentrations to enrich for lower abundant
transcripts, showed no impact on final sequencing data. Therefore, an equimolar ratio of gene-specific
primers was used to construct each pool, whereby Ig assays utilized a final concentration of 0.5uM of each
gene specific primer paired with 1uM 10x forward primer, and TCR assays utilized a final concentration of
1um of each gene specific primer paired with 2uM of the 10x forward primer. Due to initial sequencing
results, later Ig and TCR primer pools were split into VDJ and t chain pools to improve transcript capture
and chain pairing efficiency. Primer pools were constructed in volumes of 10uL, requiring the nuclease-
free water volume added to the cDNA at the beginning of target enrichment to be reduced to 28ul from
33uL. Other than the needed changes to accommodate novel primers to the 10x VDJ workflow, VDJ and
5’GEX libraries were prepared according to the manufacturer’s instructions (10x Genomics, Pleasanton,
CA, Chromium Next GEM Single Cell VDJ Reagents Kits v1.1). Final V(D)J and 5’ GEX libraries were run on a Bioanalyzer 2100 with a High Sensitivity DNA kit (Agilent,
Santa Clara, CA) to assess library size and concentration.
|
Further analysis of library concentration was
performed using a Qubit 3 fluorometer (Thermo Fisher Scientific, Waltham, MA) coupled with the sizing
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. data from the Bioanalyzer to determine appropriate loading concentration for each library. Sequencing was
performed using a NextSeq 500 sequencer (Illumina, San Diego, CA) utilizing 150 cycle kits in a paired
end fashion. Runs were programmed to generate a 26bp read 1 sequence consisting of the 16bp 10x barcode
and 10bp 10x UMI, an 8bp index read, and a 133bp read 2 sequence of the cDNA insert to exhaust the
number of cycles available in the kit. Sequencing depth was targeted at a minimum of 5,000 reads per cell
for V(D)J libraries and 20,000 reads per cell for 5’ GEX libraries as recommended by the 10x workflow
(10x Genomics, Pleasanton, CA, Chromium Next GEM Single Cell VDJ Reagents Kits v1.1). Single cell V(D)J sequencing data analysis
Raw sequencing data were demultiplexed and converted to fastq files also using the Cellranger (v4.0.0)
command mkfastq. Next, raw reads were de novo assembled into contigs using the Cellranger vdj
command. A de novo, rather than reference-based, contig assembly was performed due incomplete
annotation of the rhesus germline V(D)J sequences. To annotate the constant regions of the assembled
V(D)J transcripts, assembled contigs were searched against inner-enrichment primers and a reference of
constant region amplification sequences using the ublast (e-value ≤ 1e-4) and usearch_global (sequence ID
threshold = 0.9) commands respectively from the USEARCH (v10.0.240)17 suite of sequence analysis tools. Subsequently, to annotate the V regions of the assembled V(D)J transcripts, the assembled contigs were
aligned to a custom IgBLAST (v1.8.0) 18 database of human and rhesus germline V(D)J sequences. An e-
value cutoff of 0.01 and default parameters were used in IgBLAST queries. After annotation, we
systematically filtered assembled contigs from our analysis as follows. Contigs representing misassembled
chimeras (e.g. IgL primer alignment and IgM constant region alignment) or off-target transcripts (e.g. constant region primer match at the 3’ end but without a corresponding constant region alignment) were
removed. Further, assembled contigs without a productive and complete V(D)J sequence or that had a
premature stop codon were removed. Only contigs that passed all of these quality control criteria were
considered in downstream analyses, including the assessment of pairing efficiency. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
|
We assessed read coverage of filtered contigs using the all_contig.bam file outputted by cellranger vdj,
where sequencing reads were aligned to contigs after assembly. Reads were only aligned to contigs of their
respective cell barcode. For each dataset, the samtools depth command was run on the bam file to generate
the read coverage per contig base position. To facilitate the comparison of read coverages across the contigs
of different lengths, the read coverages were normalized as follows. For each assembled contig, we first
labeled each base position as part of the 5’UTR, V region, or C region, based on the start positions of the
V gene segments and the end positions of the J gene segments from IgBLAST results. Base positions within
each labeled region were placed into 100 bins of equal length and the mean relative coverage per bin was
calculated. VDJ versus VJ chain pairing efficiency was assessed using B and T cells defined by the 5’ GEX data
collected from the same cell samples described below. After removing low-quality cells and identifying the
B and T cell clusters, the remaining cell barcodes were used as a ground truth to assess chain pairing
efficiency in V(D)J sequence data since the same cell barcodes were already independently validated as a
productive B or T lymphocyte based on the filtered V(D)J transcript contigs. The V(D)J sequences of each
cell were integrated into a Seurat object as metadata for gene expression and clonotype analysis. Expanded clones were defined by more than one cell sharing the same V germline gene segments and with
at least 85% identical CDR3 nucleotides sequences, for both VDJ and VJ contigs. Cells that had unpaired
contigs were not considered in the clonotyping analysis. CDR3 sequence clustering was performed using
the CD-HIT (v4.6.6)19 sequence analysis tool. Single cell 5’ gene expression profile analysis
Raw sequencing data were demultiplexed, aligned to the RheMac10 genome annotation, and UMI-
collapsed using the 10x Genomics cellranger (v4.0.0) commands mkfastq and count, respectively. Raw
gene expression matrices were normalized and scaled using the SCTransform20 method with the Seurat R
package (v3.1.5). Quality control was performed on each dataset to remove poor quality cells. For each
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. sample, cells that had greater than 5% of their UMIs map to mitochondrial genes were removed from the
analysis. Principal component analysis was performed using the normalized and scaled expression levels of the 3,000
most variable genes in each dataset. Based on the Seurat’s recommendations, the first 30 principal
components were used as input to UMAP dimension reduction and K-nearest neighbor cell clustering. Canonical marker gene expression and differential expression testing were used to determine the cell types
present in the tissue culture samples.
|
Differential expression analysis was performed using a Wilcoxon rank
sum test within the FindMarkers Seurat function. The potential for CD3+ B cells to be technical B+T cell doublets was assessed by simulating B+T cell
doublets by combining UMI counts from randomly selected pairs of B and T cells. The counts for gene j in
doublet i with parent cells a and b was defined as yi,j = xa,j + xb,j. The new simulated doublets were added
to the unnormalized gene expression matrix while the original pair of singlets was removed. Next, the
standard Seurat analysis workflow was performed as described as above. Once cells were embedded with
individual UMAP coordinates the Euclidian distance between CD3+ B cells and simulated doublets was
measured. This analysis was performed using between 10 and 100 simulated doublets by an increase of 10
cells. The simulation and distance calculations were repeated 10 times for each number of simulated
doublets. Wilcoxon rank sum tests were performed to test for a difference in distance between CD3+ B
cells to each other and to simulated B+T cell doublets. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Results
High throughput single cell Ig and TCR sequencing in rhesus macaque
As shown in Fig. 1, we adopted rhesus specific V(D)J primers into the human single cell immune profiling
assays commercially available from 10x Genomics and sequenced in total over 80,000 single cell barcodes
from cryopreserved rhesus PBMC, splenocytes, FACS-sorted memory B cells (CD20+ CD27+) and
stimulated T cells (CD3+; sorted into proliferating and non-proliferating fractions by dye dilution). De novo
assembly of raw sequencing reads resulted in over 150,000 unfiltered, unannotated rhesus V(D)J transcript
contigs. We devised a custom computational pipeline for contig isotype and germline annotation, as well
as quality control filtering standards (see Methods). Assembled VDJ and VJ transcript sequences were
searched against a custom IgBLAST database of rhesus and human germline segments and selected for
sequences encoding complete, productive variable domains. Filtering using these criteria yielded the final
set of 63,623 (78.8%) cell barcodes with a total of 97,932 (61.8%) V(D)J sequences from six rhesus samples
(two PBMCs, splenocytes, sorted memory B cells, proliferating and non-proliferating T cells)
(Supplemental Table II). Despite the 5’ capture method used in the 10x immune profiling assays; we did not observe significant bias
in read coverage toward the 5’ end of assembled V(D)J transcript contigs. Instead, we observed that the
read coverages were normally distributed across the 5’ untranslated regions (UTRs) and variable regions
of the obtained V(D)J contigs and were uniformly distributed across the constant region (Supplemental Fig.
|
2a ). We reasoned that coverage was distributed differently in the constant region due to the coincidence
that the read 2 length (133 bp) was similar to the lengths of the 5’ portion of constant regions captured by
our assays. The median length of constant region sequences amplified from our enrichment protocol was
between 73-108 bp (Supplemental Figure 2b). Therefore, those sequencing reads corresponding to the
unfragmented cDNA molecules would cover the entire amplified constant region. The average numbers of
reads supporting the final filtered contigs in Ig enrichment libraries were between 117 and 2,434 and the
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. numbers of unique molecular identifiers (UMI) were between 1 and 9. The median numbers of reads and
UMIs supporting the final filtered contigs in TCR enrichment libraries were comparable, ranging from 763
to 2,282 for reads and from 1 to 4 for UMIs (Supplemental Table II). To improve the recovery of paired VDJ and VJ sequences from the same cells (e.g. TCRβ and TCRα or
IgG and IgL chains) we devised a sequencing library construction strategy in which VDJ and VJ transcripts
were enriched in both separate and pooled sequencing libraries (Fig. 1a-b). Pairing efficiency, defined as
the percentage of cells with at least one final filtered VDJ and one final filtered VJ sequence, may be
affected by differences in VDJ- and VJ-chain transcript abundances as well as individual primer
amplification efficiencies. To facilitate the comparisons of pairing efficiency between VDJ and VJ split and
pooled libraries, we additionally sequenced and generated gene expression profiles of the same single cells
from five rhesus samples. B and T cell barcodes that passed gene expression quality control filtering served
as an independent “ground-truth” of viable B and T cells for assessing pairing efficiency. When VDJ and
VJ primers were used for separate library preparation, an increase in the fraction of cells with paired VDJ-
VJ repertoires was observed across all five samples and both Ig and TCR enrichments (Fig. 1c). Between
60 and 74% of filtered B cell barcodes from gene expression libraries had paired heavy and light chains
(Supplemental Table III). We observed between 30 and 68% of T cell barcodes that had paired VDJ and VJ
sequences (Supplemental Table III). Both Ig and TCR pairing efficiency ranges were comparable to other
scIRS assays in the split library configurations.21,22 We noted a relatively lower pairing efficiency rate
observed in T cells from the one splenocyte sample. Overall, we recovered a total of 7,539 B cells and 9,906
T cells with paired VDJ and VJ contigs as well as a filtered gene expression profile (Supplemental Tables
II & III). B cell receptor repertoire profiling in rhesus PBMC, splenocyte, and sorted B cell samples
We analyzed BCR repertoires from two separate PBMC samples (the total number of B cells sequenced
n=3,418 and n=3,035), one splenocyte (n=15,682), and one sorted B cell sample (n=9,746) (Supplemental
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021.
|
The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Table II). The number of cell barcodes with at least one filtered V(D)J sequence will be higher in the
absence of paired GEX data due to the inability to filter out doublets, non-viable cells, or ambient transcripts
assigned to cell barcodes. Overall, we detected every known Ig isotype (Fig. 2a) and were also able to
recover every known IgA allotype and IgG subclass identified in our previous Iso-Seq analysis, observing
preferential usage of IgA*01 and IgA*02 as well as IgG1, respectively.16 No IgE contigs were detected in
the PBMC2 and splenocyte samples, however IgE is known to be detected at the lowest levels among any
IgH isotype.22 Ig light chains (IgK and IgL) were more frequently detected than IgH contigs across all four
samples. We also investigated the ratio of cells with IgA, IgE, and IgG contigs to those with IgM and IgD. IgM and IgD isotypes are known to be expressed in mature naive B cells while IgA, IgE, and IgG are
expressed by activated (memory) B cells after undergoing an IgH class-switch recombination process.23
Overall, this IgH ratio ranged between 0.39 to 0.60 for our four Ig datasets, suggesting that we might have
captured about twice as many naive relative to activated B cells, on average (Supplemental Fig. 2d). As
expected, the largest ratio of activated to naïve B cells appeared in the FACS B cell dataset, which were
sorted using cell surface markers of B cell memory (CD20 & CD27). Despite the 5’ capture technology used to build the cDNA libraries, we observed a high rate of complete
assembly of the constant regions at the 3’ end of the obtained Ig V(D)J contigs. Only 1.6 to 2.5% of filtered
Ig V(D)J contigs did not map to an inner enrichment primer or expected amplified constant region sequence
(Fig. 2a). Ig V(D)J contigs that failed to map to the inner enrichment primer or amplified constant region
sequences also had a lower sequencing depth in terms of both reads and UMIs (Supplementary Fig. 1c). To
ensure that the 3’ constant regions of Ig V(D)J sequences were correctly assembled we examined the
distances of our primer and constant region alignments from the 3’ end of Ig V(D)J contigs. We found that
97.4-98.4% map directly to the 3’ end of Ig V(D)J contigs (Fig. 2c). The distribution of isotype contig
lengths was also largely consistent across the different datasets, providing confidence that our assembled
contigs represent true B cell receptor mRNA molecules (Fig 2b). bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. T cell receptor repertoire profiling in rhesus PBMC, splenocyte, and sorted T cell samples
Similarly, we analyzed T cell receptor repertoires from two separate PBMC samples (the total number of
T cells sequenced n=3,988, n=3,543), one splenocyte (n=3,162), and two samples derived from a MLR
experiment representing proliferating (P, n=12,964) and non-proliferating (NP, n=8,085) FACS-sorted T
cells (Supplemental Table II).
|
Like our B cell repertoire analysis, TCR V(D)J contigs were searched against
the inner-enrichment primers used to build TCR libraries in addition to a reference of expected amplified
constant region sequences. Importantly, in addition to the more commonly targeted α- and β- TCRs, we
also included primers capturing δ- and γ- TCRs. Notably, primers that target δ- and γ- chains are not
currently available in the human and mouse single cell immune profiling assays from 10x Genomics,
resulting in repertoire analyses solely covering αβ T cells. 25,26 γδ T cells have been reported to make up
only 4% of T cells, on average.27 As expected, we detected δ- and γ- chain types at a lower frequency than
the predominantly expressed α- and β- chain types (Fig. 3a). However, we observed TCRα+ cells to be only
as much as 1.59 times more abundant than TCRγ+ cells and TCRβ+ cells were about 15.8 times more
abundant than TCRδ+ cells in mixed cell datasets (i.e PBMC and splenocytes), suggesting that leaving out
γ- and δ- TCRs could potentially miss relevant biological insights. Interestingly, overall TCR VDJ contigs
were 1.38x more abundant than VJ contigs in αβ T cells, but 4.52x less abundant in γδ T cells (Fig 3a). Similar to our Ig repertoire analysis, we also observed a high rate of complete constant region assembly in
our TCR enrichment datasets. Only 1.5 to 2.2% of TCR V(D)J contigs did not map to an inner enrichment
primer or amplified constant region sequence (Fig. 3a). Furthermore, TCR V(D)J contigs that failed to map
to the inner enrichment primer sequences or constant region sequences also had a lower sequencing depth
in terms of reads and UMIs than those that did map to our constant region and primer references
(Supplemental Fig. 2c). Correct constant region assembly was confirmed by observing constant region
amplification sequences and inner-enrichment primers mapped directly to the 3’ end of the TCR V(D)J
contigs, between 97.8-98.5% (Fig. 3c). The distributions of TCR V(D)J contig lengths were also largely
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. consistent across different samples, supporting that our assembled TCR contigs represented the full-length
TCR V(D)J transcripts (Fig. 3b). Clonal expansion in proliferating T cells
To assess the potential of our rhesus scIRS assays to detect clonally expanded lineages, we compared the
clonal expansion rates in our proliferating (P) and non-proliferating (NP) T cell datasets. These cells were
isolated from a MLR, in which responding PBMCs were labeled with a proliferation dye and incubated
with MHC mismatched activated DCs for five days using a previously described protocol (Supplemental
Fig. 1).28 We then sorted these cells based on CD3 positivity and whether they had undergone proliferation
as indicated by a loss of proliferation dye.
|
Proliferating cells were enriched for those cells responding to
alloantigens from the MHC mismatched DCs. We defined T-cell lineages by grouping cells with identical
VDJ and VJ germline variable gene segments and required at least 85% nucleotide identity in both CDR3
regions, although the number of clones did not change significantly when this sequence identity threshold
was increased (Fig. 4a). As expected, we observed a substantial clonal expansion in the proliferating T cell dataset with 42.5% of
cells belonging to clonally expanded lineages compared to < 2% in the non-proliferating T cell dataset (Fig
4a-c). Despite the high clonal expansion rate in P T cells, we did not observe any lineages that dominated
the clonally expanded cells. Out of the 2,166 cells of expanded lineages in the P T cell dataset, 530 (24.5%)
belonged to a lineage of size two. The largest lineage size consisted of only 24 cells. Hierarchical clustering
of TCR VDJ and VJ variable germline gene segments in clonally expanded P T cells did not reveal any
modules of highly overlapping VJ and VDJ combinations (Supplemental Fig. 3a). MLRs typically result in
a generalized activation of some T cells predisposed for proliferation. There are a multitude of potential
antigens that can activate T cells, there is a propensity of stimulatory cytokines to induce by-stander
proliferation in previously primed cells (e.g. those with a memory phenotype), and heterologous cross-
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. reactive stimulation can be expected.29 Nevertheless, the products of an MLR are substantially enriched for
allospecific cells. Interestingly, among P T cells the most frequent VJ germline gene segment detected in expanded lineages
was TRGV10, which was co-detected in many T cells with TCRβ molecules (Supplemental Fig. 3a&c and
Supplemental Fig. 4). Additionally, TCRGV10 showed the largest increase in usage between expanded P
T cells and NP T cells among any VJ or VDJ germline segment (Fig 4d). This observation was intriguing,
given the high correlation of germline usage rate between expanded P and NP T cells (Supplemental Fig. 3b). In some of the clonotypes, no additional TCRβ contig was detected, but the frequency of αβ T cells
that also expressed a γ chain showed the largest increase from NP to expanded P T cells (Supplemental Fig. 4). Similar to what others have reported, we observed a diverse number of TCRα and TCRβ chains and a
limited number of TCRγ chains in our γ-αβ T cells in both NP and P + expanded T cell datasets.30 In both
NP and P + expanded γ-αβ T cells the TCRGV10 variable gene was used in over 80% of lineages (singleton
or expanded). We did not observe one TCRα and TCRβ variable gene used in over 12% of lineages. Some studies have reported T cells that co-express αβ and γδ receptors.
|
30,31 Others have characterized δ/αβ
T cells which possess chimeric contigs that encompass a δ variable gene but α joining and constant region
domains.32 This chimeric scenario is unlikely to be the case in this data because chimeric contigs as the one
described in δ/αβ T cells would be removed from our analyses (see Methods). The frequency of γ-αβ T
cells was between 9.32%-12.46% of the T cells in PBMC and splenotye samples, while γ-β T cells was
between 11.77%-25.39%. It shows that the observed co-expression was not unique to these FACS-sorted
T cell datasets. Since typical single cell based TCR repertoire sequencing analyses have focused on αβ T
cells, it is unclear if these αβ T cells co-expressing TCRγ or TCRδ tend to be undetected or are atypically
deleted at some point post-production. Additionally, the presence of TCRγ transcripts in αβ T cells does
not imply γδ TCRs are on the surface of the same T cells. Obviously future analyses of more samples are
needed. However, this complete chain pairing coverage highlights the unbiased potential of our rhesus
scIRS assays which could also enable future exploratory analysis of αβ and γδ co-expressions. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Integrative analysis of gene expression profiles and V(D)J repertoires of the same single cells
Parallel 5’ gene expression profiling analysis allowed us to cluster the same single cells into groups of cell
types, and to overlay individual cells with their Ig and TCR repertoire data (Fig. 5a-b). For example, in the
PBMC2 sample (Fig. 5a), we identified 2,100 B cells across 3 cell clusters. We recovered 1,344 (64.1%) B
cells with paired heavy and light chains, 636 (30.3%) with only a light chain, and 62 (3%) with only a heavy
chain. Among the 3,071 T cells in the same PBMC2 sample, we recovered 1,495 (48.7%) cells with paired
TCR VDJ and VJ chains, 651 (21.2%) cells with only a VDJ and 582 (19%) cells with only a VJ chain. The
rates of paired TCR chains were similar across the different T cell subtypes. Among the 3,531 B cells
identified in the splenocyte sample (Fig. 5b), we recovered 2,118 (60%) B cells with paired heavy and light
chains, 817 (23%) with only a light chain, 328 (9.3%) with only a heavy chain. In the same splenocyte
sample we captured 1,886 T cells. Of those, we recovered 575 (30.5%) cells with paired TCR VDJ and VJ
chains, 472 (25%) cells with only a VDJ, and 347 (18.4%) cells with only a VJ chain. In addition, we
observed 492 (26.1%) T cells without any TCR V(D)J contigs (Supplemental Table III). Intriguingly, among the cells sequenced in the PBMC2 dataset we observed a small cluster of 23 B cells
clustered separately from the others (Fig 5a). Upon investigation, we obtained both Ig and TCR V(D)J
contigs from a high fraction of these cells.
|
Of these 23 cells, Ig and TCR dual expression was detected in
22 (> 95%) cells. Dual expression of Ig and TCR molecules was also observed in 229 additional cells, but
these cells appeared to be randomly distributed in UMAP space. Therefore, we focused our exploratory
Ig/TCR dual expression analysis on this small cluster of 23 cells. Dual expression of Ig and TCRs or other B and T cell markers have been reported by others, but the role
that these hybrid lymphocytes play is largely unknown.33-35 The possibility of doublets, where two cells are
given the same cellular barcode, does arise when using droplet-based single cell technologies such as the
10x Genomics used here. To assess the potential for these cells to be B+T cell doublets we used differential
expression analysis to identify transcriptomic signatures that were absent in B and T cells. This small B cell
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. clusters not only expressed both Ig and TCR molecules, but the genes involved in B and T cell receptor
signal transduction. CD79A, CD79B, CD3E, CD3G, and CD3D were all upregulated in this small cluster
of B cells. As such we labeled this cluster as CD3+ B cells. The top differentially expressed genes appeared
to be constitutively expressed across these CD3+ B cells while absent in regular B and T cells (Fig. 5c). Furthermore, CD3+ B cells did not have a greater sequencing depth than other B cells in terms of number
of UMIs or unique features expressed (Fig 5d). To further demonstrate the transcriptional differences of
CD3+ B cells from technical B+T cell doublets we simulated B+T cell doublets and added their expression
profiles to the gene expression matrix to rerun dimension reduction and clustering (see Methods). To assess
the transcriptional similarity of CD3+ B cells and simulated B+T cell doublets we measured the Euclidean
distance between CD3+ B cells to each other and to simulated doublets. Consistently, CD3+ B cells were
closer to each other (3.06 ± 0.397) (i.e more transcriptionally similar) than to simulated doublets (5.14 ±
0.570). These results were additionally found to be statistically significant by Wilcoxon rank sum test
(adjusted p-value < 0.05). Taken together, these data suggest that these CD3+ B cells were less likely to be
B and T cell doublets, but more likely to be expressing both Ig and TCR transcripts. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Discussion
In this study, using extensive experimental optimization and validation, we established rhesus specific
single cell high-throughput assays for Ig and TCR repertoire sequencing analysis that recover every known
rhesus Ig and TCR isotype and chain type.
|
While rhesus macaque is one of the most widely used NHP
models, it is our understnading that this is the first such assay developed for rhesus macaques that are also
fully compatible with a commercially available platform. We also devised a custom computational pipeline
to process and analyze the generated rhesus single cell V(D)J sequencing data. The establishment of these single cell-based Ig/TCR repertoire sequencing assays will improve our
understanding of immune responses in rhesus macaques, which will in turn allow for applications to further
understanding the immune response in humans. For example, we were able to detect clonally expanded
lineages of T cells in a proliferating T cell sample, using the full-length productive VDJ and VJ sequences
paired within single cells. The overlay of cellular gene expression profiles allowed the identification of the
underlying cell types and subsets. We also observed rare cells expressing both Ig and TCR V(D)J transcripts
that could be investigated in the future. Compared to the human and mouse assays commercially available
from 10x Genomics, our assays also capture TCRγ and TCRδ chains. While αβ T cells are more frequent
in general, our results showed that the frequencies of γδ T cells could be extremely variable, and sometimes
their frequency could be as high as 26% of V(D)J contigs detected in a PBMC sample. This suggests that
the additional coverage of γ- and δ- TCRs may enable the discovery of relevant biological insights that are
overlooked by current assays. Unexpectedly, we detected γ-αβ T cells in all four types of rhesus samples we analyzed in this study: PBMC,
splenocyte, FACS sorted non-proliferating and clonally-expanded, proliferating T cell samples. Interestingly, we also observed a large increase in the frequencies of γ-αβ T cells in the clonally-expanded,
proliferating T cell sample compared to the non-proliferating T cell sample (Supplemental Figure 4). Generally, a mutually exclusive pattern of γδ- and αβ- expression is expected.36 However, reports have
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. demostrated that there are exceptions. Using FACS in combination with qPCR and sc-RNAseq reseraches
have observed and validated the presence of dual expressing γδ-αβ T cells at the mRNA and cell-surface
receptor level in multiple tissues and developmental stages of human and mice. 30,31 Furthermore, transgenic
mice constructed with a pair of productively rearranged γ and δ genes have been shown to produce a normal
number of αβ T cells37, in addition to other reported unconventional expression patterns of TCR
molecules.38,39 Considering potential techincal confounding factors such as ambinet transcripts and
techincal doublets, future independent analyses will be needed to experimentally investigate the rhesus γ-
αβ T cells observed here.
|
Analysis of the adaptive immune response has co-evolved with the advancement of high-throughput IRS
strategies. For RNA-based analysis, 5’ RACE has been used to reliably generate unbiased amplification of
Ig and TCR cDNA molecules and can be done using a simple primer set.40-42 The major limitations of 5’
RACE and similar strategies has been extracting the pairing of variable region information required for the
determination of antigen/epitope specificity. Therefore, to recover paired variable region sequences, single
cell high-throughput sequencing strategies started to emerge. Emulsion PCR, where TCRβ and TCRα RNA
molecules from single cells are fused together43,44 and pairSeq where single cells are isolated and mRNA
reverse transcription reactions occur individually within a 96-well plate became commonplace.45 Despite
the advancement in chain pairing of these IRS strategies, these methods are generally low-throughput in
terms of the number of cells sequenced in a single experiment and are not applicable to many Ig and TCR
chain types. The recent advent of droplet-based single cell sequencing technologies has greatly improved
the throughput of IRS analysis. The human and mouse immune profiling assays from 10x Genomics allow
for the recovery of paired, full-length V(D)J sequences as well as gene expression profiles of thousands of
cells simultaneously. However, the development of similar scIRS assays for other species is challenging. For a species of interest,
it requires the complete Ig and TCR reference sequences, which is often lacking due to the extreme
complexity of Ig and TCR loci. Recently, we obtained a complete reference set for the constant regions of
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. rhesus Ig and TCR isotypes and chain types using long read transcriptome sequencing16, which bypasses
the need of rhesus specific germline reference sequences. This reference set enabled us to design and
validate primers targeting rhesus V(D)J constant region sequences to achieve complete and unbiased
repertoire analysis, whereas primer pools targeting the highly diverse variable regions can miss as much as
50% of repertoire sequences.15 Here, we used de novo assembly to reconstruct captured Ig/TCR transcripts
with full-length productive V(D)J regions instead of using germline reference sequence based approaches. Still, the availability of comprehensive rhesus germline reference sequences could greatly facilitate the
analysis and the biological interpretation of the obtained Ig/TCR repertoire data. Building rhesus germline
reference sequence collections would require future developments. In summary, these results demonstrate that scIRS assays we established here for rhesus macaques can be
used to recover every known Ig and TCR isotype and chain type in an unbiased fashion, pair VDJ and VJ
transcripts in the same cells, detect clonally expanded lineages and dual expressing cels, and integrate V(D)J
repertoires with cellular gene expression profiles.
|
These scIRS assays will greatly facilitate the analysis of
the adaptive immune responses in rhesus macaques. Acknowledgments
Cryopreserved splenocyte samples were provided by Dr. Sallie R. Permar at Duke University. Rhesus B
cell line BLCL was provided by the NIH Nonhuman Primate Reagent Resource. Rhesus T cell lines RH447
and RH444 were provided by Dr. Vanessa M. Hirsch at NIAID/NIH. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. References
1. Schroeder, H. W., & Cavacini, L. (2010). Structure and Function of Immunoglobulins. The Journal of
Allergy and Clinical Immunology, 125(2 0 2), S41-S52. 10.1016/j.jaci.2009.09.046
2. Davis, M. M., & Bjorkman, P. J. (1988). T-cell antigen receptor genes and T-cell
recognition. Nature, 334(6181), 395-402. 10.1038/334395a0
3. Charles A Janeway, J., Travers, P., Walport, M., & Shlomchik, M. J. (2001). The Generation of
Lymphocyte Antigen Receptors. Immunobiology: The Immune System in Health and Disease. 5th Edition,
2001. https://www.ncbi.nlm.nih.gov/books/NBK10774/. Accessed Apr 12, 2021. 4. Li, F., Luo, M., Zhou, W., Li, J., Jin, X., Xu, Z., Juan, L., Zhang, Z., Li, Y., Liu, R., Li, Y., Xu, C., Ma,
K., Cao, H., Wang, J., Wang, P., Bu, Z., & Jiang, Q. (2020). Single cell RNA and immune repertoire
profiling of COVID-19 patients reveal novel neutralizing antibody. Protein & Cell, 10.1007/s13238-020-
00807-6
5. Mor, M., Werbner, M., Alter, J., Safra, M., Chomsky, E., Lee, J. C., Hada-Neeman, S., Polonsky, K.,
Nowell, C. J., Clark, A. E., Roitburd-Berman, A., Ben-Shalom, N., Navon, M., Rafael, D., Sharim, H.,
Kiner, E., Griffis, E. R., Gershoni, J. M., Kobiler, O., . . . Freund, N. T. (2021). Multi-clonal SARS-CoV-2
neutralization by antibodies isolated from severe COVID-19 convalescent donors. PLoS Pathogens, 17(2),
e1009165. 10.1371/journal.ppat.1009165
6. Watson, C. T., Glanville, J., & Marasco, W. A. (2017). The Individual and Population Genetics of
Antibody Immunity. Trends in Immunology, 38(7), 459-470. 10.1016/j.it.2017.04.003
7. Williams, K., Lackner, A., & Mallard, J. (2016). Non-human primate models of SIV infection and CNS
neuropathology. Current Opinion in Virology, 19, 92-98. 10.1016/j.coviro.2016.07.012
8. Flynn, J. L., Gideon, H. P., Mattila, J. T., & Lin, P. L. (2015). Immunology studies in non-human primate
models of tuberculosis. Immunological Reviews, 264(1), 60-73. 10.1111/imr.12258
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 9. Meyer, C., Kerns, A., Haberthur, K., Dewane, J., Walker, J., Gray, W., & Messaoudi, I. (2013). Attenuation of the adaptive immune response in rhesus macaques infected with simian varicella virus
lacking open reading frame 61.
|
Journal of Virology, 87(4), 2151-2163. 10.1128/JVI.02369-12
10. Lackner, A. A., & Veazey, R. S. (2007). Current concepts in AIDS pathogenesis: insights from the
SIV/macaque model. Annual Review of Medicine, 58, 461-476. 10.1146/annurev.med.58.082405.094316. 11. Hansen, S. G., Ford, J. C., Lewis, M. S., Ventura, A. B., Hughes, C. M., Coyne-Johnson, L., Whizin,
N., Oswald, K., Shoemaker, R., Swanson, T., Legasse, A. W., Chiuchiolo, M. J., Parks, C. L., Axthelm, M.
K., Nelson, J. A., Jarvis, M. A., Piatak, M., Lifson, J. D., & Picker, L. J. (2011). Profound early control of
highly pathogenic SIV by an effector memory T-cell vaccine. Nature, 473(7348), 523-527. 10.1038/nature10003
12. Hansen, S. G., Piatak, M., Ventura, A. B., Hughes, C. M., Gilbride, R. M., Ford, J. C., Oswald, K.,
Shoemaker, R., Li, Y., Lewis, M. S., Gilliam, A. N., Xu, G., Whizin, N., Burwitz, B. J., Planer, S. L.,
Turner, J. M., Legasse, A. W., Axthelm, M. K., Nelson, J. A., . . . Picker, L. J. (2013). Immune clearance
of highly pathogenic SIV infection. Nature, 502(7469), 100-104. 10.1038/nature12519
13. Munster, V. J., Feldmann, F., Williamson, B. N., van Doremalen, N., Pérez-Pérez, L., Schulz, J., Meade-
White, K., Okumura, A., Callison, J., Brumbaugh, B., Avanzato, V. A., Rosenke, R., Hanley, P. W.,
Saturday, G., Scott, D., Fischer, E. R., & de Wit, E. (2020). Respiratory disease in rhesus macaques
inoculated with SARS-CoV-2. Nature, 585(7824), 268-272. 10.1038/s41586-020-2324-7
14. Knechtle, S. J., Shaw, J. M., Hering, B. J., Kraemer, K., & Madsen, J. C. (2019). Translational impact
of NIH-funded
nonhuman
primate
research
in
transplantation. Science
Translational
Medicine, 11(500)10.1126/scitranslmed.aau0143
15. Miller, W. P., Srinivasan, S., Panoskaltsis-Mortari, A., Singh, K., Sen, S., Hamby, K., Deane, T.,
Stempora, L., Beus, J., Turner, A., Wheeler, C., Anderson, D. C., Sharma, P., Garcia, A., Strobert, E., Elder,
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. E., Crocker, I., Crenshaw, T., Penedo, M. C. T., . . . Kean, L. S. (2010). GVHD after haploidentical
transplantation: a novel, MHC-defined rhesus macaque model identifies CD28- CD8+ T cells as a reservoir
of breakthrough T-cell proliferation during
costimulation blockade
and
sirolimus-based
immunosuppression. Blood, 116(24), 5403-5418. 10.1182/blood-2010-06-289272
16. Brochu, H. N., Tseng, E., Smith, E., Thomas, M. J., Jones, A. M., Diveley, K. R., Law, L., Hansen, S.
G., Picker, L. J., Gale, M., & Peng, X. (2020). Systematic Profiling of Full-Length Ig and TCR Repertoire
Diversity in Rhesus Macaque through Long Read Transcriptome Sequencing. Journal of Immunology
(Baltimore, Md. : 1950), 204(12), 3434-3444. 10.4049/jimmunol.1901256
17. Edgar, R. C. (2010). Search and clustering orders of magnitude faster than BLAST.
|
Bioinformatics
(Oxford, England), 26(19), 2460-2461. 10.1093/bioinformatics/btq461
18. Ye, J., N. Ma, T. L. Madden, and J. M. Ostell. 2013. IgBLAST: an immunoglobulin variable domain
sequence analysis tool. Nucleic Acids Res. 41(W1): W34–W40. 10.1093/nar/gkt382
19. Fu, L., Niu, B., Zhu, Z., Wu, S., & Li, W. (2012). CD-HIT: accelerated for clustering the next-generation
sequencing data. Bioinformatics (Oxford, England), 28(23), 3150-3152. 10.1093/bioinformatics/bts565
20. Hafemeister, C., & Satija, R. (2019). Normalization and variance stabilization of single cell RNA-seq
data using regularized negative binomial regression. Genome Biology, 20(1), 296. 10.1186/s13059-019-
1874-1
21. Goldstein, L. D., Chen, Y. J., Wu, J., Chaudhuri, S., Hsiao, Y., Schneider, K., Hoi, K. H., Lin, Z.,
Guerrero, S., Jaiswal, B. S., Stinson, J., Antony, A., Pahuja, K. B., Seshasayee, D., Modrusan, Z., Hötzel,
I., & Seshagiri, S. (2019). Massively parallel single cell B-cell receptor sequencing enables rapid discovery
of diverse antigen-reactive antibodies. Communications Biology, 2(1), 1-10. 10.1038/s42003-019-0551-y
22. Singh, M., Al-Eryani, G., Carswell, S., Ferguson, J. M., Blackburn, J., Barton, K., Roden, D., Luciani,
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. F., Giang Phan, T., Junankar, S., Jackson, K., Goodnow, C. C., Smith, M. A., & Swarbrick, A. (2019). High-throughput targeted long-read single cell sequencing reveals the clonal and transcriptional landscape
of lymphocytes. Nature Communications, 10(1), 3120. 10.1038/s41467-019-11049-4
23. Jr, Janeway, C. A., Travers, P., Walport, M., & Shlomchik, M. J. (2001). Immunobiology (5th ed.). Garland Science. 24. Tong, P., Granato, A., Zuo, T., Chaudhary, N., Zuiani, A., Han, S. S., Donthula, R., Shrestha, A., Sen,
D., Magee, J. M., Gallagher, M. P., van der Poel, Cees E., Carroll, M. C., & Wesemann, D. R. (2017). IgH
isotype-specific B cell receptor expression influences B cell fate. Proceedings of the National Academy of
Sciences of the United States of America, 114(40), E8411-E8420. 10.1073/pnas.1704962114
25. De Simone, M., Rossetti, G., & Pagani, M. (2018). Single Cell T Cell Receptor Sequencing: Techniques
and Future Challenges. Frontiers in Immunology, 910.3389/fimmu.2018.01638
26. Carter, J. A., Preall, J. B., Grigaityte, K., Goldfless, S. J., Jeffery, E., Briggs, A. W., Vigneault, F., &
Atwal, G. S. (2019). Single T Cell Sequencing Demonstrates the Functional Role of αβ TCR Pairing in Cell
Lineage and Antigen Specificity. Frontiers in Immunology, 1010.3389/fimmu.2019.01516
27. Chien, Y., Meyer, C., & Bonneville, M. (2014). γδ T cells: first line of defense and beyond. Annual
Review of Immunology, 32, 121-155. 10.1146/annurev-immunol-032713-120216
28. Espinosa, J., Herr, F., Tharp, G., Bosinger, S., Song, M., Farris, A.
|
B., George, R., Cheeseman, J.,
Stempora, L., Townsend, R., Durrbach, A., & Kirk, A. D. (2016). CD57(+) CD4 T Cells Underlie
Belatacept-Resistant Allograft Rejection. American Journal of Transplantation: Official Journal of the
American Society of Transplantation and the American Society of Transplant Surgeons, 16(4), 1102-1112. 10.1111/ajt.13613
29. DeWolf, S., Grinshpun, B., Savage, T., Lau, S. P., Obradovic, A., Shonts, B., Yang, S., Morris, H.,
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Zuber, J., Winchester, R., Sykes, M., & Shen, Y. (2018). Quantifying size and diversity of the human T cell
alloresponse. JCI Insight, 3(15)10.1172/jci.insight.121256
30. Edwards, S. C., Sutton, C. E., Ladell, K., Grant, E. J., McLaren, J. E., Roche, F., Dash, P.,
Apiwattanakul, N., Awad, W., Miners, K. L., Lalor, S. J., Ribot, J. C., Baik, S., Moran, B., McGinley, A.,
Pivorunas, V., Dowding, L., Macoritto, M., Paez-Cortez, J., . . . Mills, K. H. G. (2020). A population of
proinflammatory T cells coexpresses αβ and γδ T cell receptors in mice and humans. The Journal of
Experimental Medicine, 217(5)10.1084/jem.20190834
31. Reitermaier, R., Krausgruber, T., Fortelny, N., Ayub, T., Vieyra-Garcia, P. A., Kienzl, P., Wolf, P.,
Scharrer, A., Fiala, C., Kölz, M., Hiess, M., Vierhapper, M., Schuster, C., Spittler, A., Worda, C., Weninger,
W., Bock, C., Eppel, W., & Elbe-Bürger, A. (2021). αβγδ T cells play a vital role in fetal human skin
development and immunity. Journal of Experimental Medicine, 218(e20201189)10.1084/jem.20201189
32. Pellicci, D. G., Uldrich, A. P., Le Nours, J., Ross, F., Chabrol, E., Eckle, S. B. G., de Boer, R., Lim, R.
T., McPherson, K., Besra, G., Howell, A. R., Moretta, L., McCluskey, J., Heemskerk, M. H. M., Gras, S.,
Rossjohn, J., & Godfrey, D. I. (2014). The molecular bases of δ/αβ T cell–mediated antigen
recognition. The Journal of Experimental Medicine, 211(13), 2599-2615. 10.1084/jem.20141764
33. Japp, A. S., Meng, W., Rosenfeld, A. M., Perry, D. J., Thirawatananond, P., Bacher, R. L., Liu, C.,
Gardner, J. S., Atkinson, M. A., Kaestner, K. H., Brusko, T. M., Naji, A., Luning Prak, E. T., & Betts, M.
R. (2021). TCR+/BCR+ dual-expressing cells and their associated public BCR clonotype are not enriched
in type 1 diabetes. Cell, 184(3), 827-839.e14. 10.1016/j.cell.2020.11.035
34. Ahmed, R., Omidian, Z., Giwa, A., Cornwell, B., Majety, N., Bell, D. R., Lee, S., Zhang, H., Michels,
A., Desiderio, S., Sadegh-Nasseri, S., Rabb, H., Gritsch, S., Suva, M. L., Cahan, P., Zhou, R., Jie, C.,
Donner, T., & Hamad, A. R. A. (2019). A Public BCR Present in a Unique Dual-Receptor-Expressing
Lymphocyte from Type 1 Diabetes Patients Encodes a Potent T Cell Autoantigen. Cell, 177(6), 1583-
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021.
|
The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 1599.e16. 10.1016/j.cell.2019.05.007
35. Liu, Y., Ye, S., Guo, X., Li, W., Xia, Y., Wen, X., Yu, J., Jia, Y., Liu, X., Guo, Y., & Zhao, Y. (2020). Discovery and characteristics of B cell-like T cells: A potential novel tumor immune
marker? Immunology Letters, 220, 44-50. 10.1016/j.imlet.2020.01.007
36. Hayes, S., Li, L., & Love, P. (2005). TCR Signal Strength Influences αβ/γδ Lineage
Fate. Immunity, 10.1016/j.immuni.2005.03.014
37. Ishida, I., Verbeek, S., Bonneville, M., Itohara, S., Berns, A., & Tonegawa, S. (1990). T-cell receptor
gamma delta and gamma transgenic mice suggest a role of a gamma gene silencer in the generation of
alpha beta T cells. PNAS, 10.1073/pnas.87.8.3067
38. Bowen, S., Sun, P., Livak, F., Sharrow, S., & Hodes, R., (2014). A Novel T Cell Subset with Trans-
Rearranged Vγ-Cβ TCRs Shows Vβ Expression Is Dispensable for Lineage Choice and MHC Restriction. Journal of Immunology, 10.4049/jimmunol.1302398
39. Hochstenbach, F., & Brenner M. (1989) T-cell receptor δ-chain can substitute for α to form a βδ
heterodimer. https://doi.org/10.1038/340562a0
40. Ichinohe, T., Miyama, T., Kawase, T., Honjo, Y., Kitaura, K., Sato, H., Shin-I, T., & Suzuki, R. (2018). Next-Generation Immune Repertoire Sequencing as a Clue to Elucidate the Landscape of Immune
Modulation
by
Host–Gut
Microbiome
Interactions. Frontiers
in
Immunology, 910.3389/fimmu.2018.00668
41. Gao, F., & Wang, K. (2015). Ligation-anchored PCR unveils immune repertoire of TCR-beta from
whole blood. BMC Biotechnology, 15, 39. 10.1186/s12896-015-0153-9
42. Heather, J. M., Best, K., Oakes, T., Gray, E. R., Roe, J. K., Thomas, N., Friedman, N., Noursadeghi,
M., & Chain, B. (2015). Dynamic Perturbations of the T-Cell Receptor Repertoire in Chronic HIV Infection
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. and following Antiretroviral Therapy. Frontiers in Immunology, 6, 644. 10.3389/fimmu.2015.00644
43. Turchaninova, M. A., Britanova, O. V., Bolotin, D. A., Shugay, M., Putintseva, E. V., Staroverov, D.
B., Sharonov, G., Shcherbo, D., Zvyagin, I. V., Mamedov, I. Z., Linnemann, C., Schumacher, T. N., &
Chudakov, D. M. (2013). Pairing of T-cell receptor chains via emulsion PCR. European Journal of
Immunology, 43(9), 2507-2515. 10.1002/eji.201343453
44. Liu, X., & Wu, J. (2018). History, applications, and challenges of immune repertoire research. Cell
Biology and Toxicology, 34(6), 441-457. 10.1007/s10565-018-9426-0
45. Howie, B., Sherwood, A. M., Berkebile, A. D., Berka, J., Emerson, R. O., Williamson, D. W., Kirsch,
I., Vignali, M., Rieder, M. J., Carlson, C. S., & Robins, H. S. (2015). High-throughput pairing of T cell
receptor
α
and
β
sequences.
|
Science
Translational Medicine, 7(301),
301ra131. 10.1126/scitranslmed.aac5624
Footnotes
This project has been funded in part with Federal funds from the National Institute of Allergy and Infectious
Diseases, National Institutes of Health, Department of Health and Human Services, under Contract No. HHSN272201800008C, and by the Washington National Primate Research Center Core grant P51
OD010425 from the National Institutes of Health, Office of the Director (MG), and grant R21AI120713 (to
X.P.). The nonhuman primate work has also been funded by multiple awards from the National Institute for
Allergy and Infectious Diseases including R38AI40297 (supporting B.I.S.) and U19AI131471 (to A.D.K.). Abbreviations used in this article: NHP, nonhuman primate; GvHD, graft versus host disease; IRS,
Immune repertoire sequencing; RACE, Rapid Amplification of cDNA ends; MPCR, multiplex PCR;
IMGT, the international ImMunoGeneTics information system; Iso-Seq, full-length isoform sequencing;
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. scIRS, single cell immune repertoire sequencing; UMI, unique molecular identifier; NP, non-proliferating,
P, proliferating; UTR, untranslated region; DC, dendritic cell. 5’ gene expression sequencing data presented in this article have been submitted to the National Center for
Biotechnology Information GEO under the accession number GSE179722. Datasets generated during the
current study are available at the NCBI Sequence Read Archive under project number PRJNA746267. Figure Legends
Fig 1. Overview of assay development. a) Droplet-based scRNA-seq was used to generate an initial
barcoded cDNA libraries. To maximize the rate at which cells with paired VDJ and VJ contigs were
recovered, sequencing libraries were generated using pooled and separate primers for VDJ and VJ chain
types. After sequencing and data analysis the chain pairing rate was assessed and compared between the
two configurations using filtered B and T cell barcodes recovered from 5’ gene expression analysis. b)
Targeted enrichment performed on each cDNA sample. The PBMC1 sample is labeled with a * to indicate
that a gene expression library was not constructed for this dataset, so a comparison between split and pooled
library preps was not performed. c) Bar charts indicate the percentage of cell barcodes with either no V(D)J
contigs that pass quality control (QC) filtering, an unpaired contig(s) that does pass, or paired contigs that
do pass. In each sample, the fraction of paired contigs that pass QC increases when VDJ and VJ primers
are split in separate library preps. Fig 2. Characterization of isotypes Ig enrichment libraries. a) Recovery of 3’ of enriched V(D)J transcripts. Tiled plots indicate the number of contigs with respective primer and constant region alignments.
|
b)
Distribution of contig length by Ig isotype across the Ig enrichment datasets. IGHE was omitted because of
the low number of contigs detected. A different IgM inner enrichment primer was used for the PBMC1
library construction which was 59 base pairs 5’ of the primer used for all other datasets. This is consistent
with the differences in the length distributions. c) Distribution of the distances between constant region and
primer alignments and 3’ end of contigs. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fig 3. Characterization of isotypes in TCR enrichment libraries. a) Recovery of 3’ of enriched V(D)J
transcripts. Tiled plots indicate the number of contigs with respective primer and constant region
alignments. b) Distribution of contig length by TCR isotype across the TCR enrichment datasets. c)
Distribution of the distances between constant region and primer alignments and 3’ end of contigs. Fig 4. Detection of clonally expanded lineages in proliferating T cells. a) Number of cells in expanded
lineages in NP and P T cells as a function of the CDR3 sequence identity threshold. b) Lineage expansions
among proliferating (P) sorted T cells. Pie charts indicate percentage of cells belonging to expanded
lineages. c) Same as b), but for non-proliferating (NP) T cells. d) Usage rate differences of variable germline
gene segments between clonally expanded P T cells and singleton NP T cells. Usage rate is defined as the
number of cells a germline gene is present in relative to the total number of cells. Fig 5. Integration of gene expression and V(D)J profiles of single cells. a) UMAP plots of the (left to right)
assignment of cell type, BCR chains, TCR chains, and dual BCR and TCR detection for each cell in PBMC
2 sample. b) Same as a) but the splenocyte sample. c) Normalized expression of Ig and TCR signal
transduction genes, as well as CD3+ B cell gene expression signatures across CD3 B cells (green) and
randomly sampled B (gold) and T cells (blue). d) Distribution of UMIs and unique features in CD3+ B and
B cells. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 1
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder.
|
All rights reserved. No reuse allowed without permission. Figure 2
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 3
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 4
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.17.456682
;
this version posted August 17, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 5
|
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Decoding gene regulation in the fly brain
Jasper Janssens1,2,#, Sara Aibar1,2,#, Ibrahim Ihsan Taskiran1,2,#, Joy N. Ismail1,2, Katina I. Spanier1,2, Carmen Bravo González-Blas1,2, Xiao Jiang Quan1,2, Dafni Papasokrati1,2, Gert Hulselmans1,2, Samira Makhzami1,2, Maxime De Waegeneer1,2, Valerie Christiaens1,2, and Stein Aerts1,2,*
1 VIB Center for Brain & Disease Research, Leuven, Belgium. 2 Department of Human Genetics, KU Leuven, Leuven, Belgium. # These authors contributed equally * Correspondence to [email protected]
Summary
The Drosophila brain is a work horse in neuroscience. Single-cell transcriptome analysis 1–5, 3D morphological classification 6, and detailed EM mapping of the connectome 7–10 have revealed an immense diversity of neuronal and glial cell types that underlie the wide array of functional and behavioral traits in the fruit fly. The identities of these cell types are controlled by – still unknown – gene regulatory networks (GRNs), involving combinations of transcription factors that bind to genomic enhancers to regulate their target genes. To characterize the GRN for each cell type in the Drosophila brain, we profiled chromatin accessibility of 240,919 single cells spanning nine developmental timepoints, and integrated this data with single-cell transcriptomes. We identify more than 95,000 regulatory regions that are used in different neuronal cell types, of which around 70,000 are linked to specific developmental trajectories, involving neurogenesis, reprogramming and maturation. For 40 cell types, their uniquely accessible regions could be associated with their expressed transcription factors and downstream target genes, through a combination of motif discovery, network inference techniques, and deep learning. We illustrate how these “enhancer-GRNs” can be used to reveal enhancer architectures leading to a better understanding of neuronal regulatory diversity. Finally, our atlas of regulatory elements can be used to design genetic driver lines for specific cell types at specific timepoints, facilitating the characterization of brain cell types and the manipulation of brain function. Main
The brain consists of a myriad of different neuronal and glial types, each unique in their morphology and function. The Drosophila brain, which contains around 100,000 cells, is uniquely positioned as a model in which the diversity of brain cell types can be investigated. Recent advances in electron microscopy have allowed the creation of connectome maps of the different regions in the Drosophila brain 7–10 , while the availability of genetic driver lines 11 provides genetic access to many cell types for understanding neuronal function 12. Furthermore, this diversity of cell types has been bolstered by single-cell transcriptomics on the adult brain 1–5, the larval brain 13–15, and the ventral nerve cord 16.
|
The recent development of single-cell assay for transposase accessible chromatin by sequencing (scATAC-seq), makes it possible to measure chromatin accessibility of single cells in high throughput 17,18, providing an additional crucial layer of information underlying neuronal identity: which genomic regions encode the regulatory information to create and maintain each cell type. The integrated analysis of transcriptomics and chromatin accessibility makes it then possible to jointly study enhancers and gene expression to discover precise regulatory programs across cell types 19–21. Cell type identity is defined by the activity of GRNs in which combinations of transcription factors activate or repress target genes. In Davie et al., we showed that for many different cell types in the brain, unique transcription factor (TF) combinations can be identified that govern gene regulatory networks, thus highlighting the unique role TFs play in neuronal fate determination. Neural
1
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. progenitors have been shown to exist along two axes of differentiation, one temporal and one spatial, and both are guided by transcription factor changes 22,23. This patterning of neural progenitors is presumed to give rise to the diverse cell types found in the adult brain. The combinatorial expression of transcription factors also governs key neuronal features by guiding dendritic targeting and neurotransmitter determination 2,24–27 and changing expression of a single TF can lead to a change in neuronal fate 28,29. Whereas transcriptomic studies in Drosophila have led to the inference of TFs and their putative target genes, they suffer from high false positive rates. Moreover, TF activity often cannot be predicted from the transcription level of TFs, as it depends on a large number of variables 30 such as protein activity, protein localization, and the presence of co-binding TFs and co-factors. On the other hand, profiling chromatin accessibility leads to a direct read-out of possible TF binding sites and therefore bypasses these limitations 31. Here, we build a single-cell multi-omics atlas across the development of the fly brain, including transcriptomic and chromatin accessibility profiles from late larva to adult, covering the dynamic processes of neurogenesis, maturation, and maintenance. We identify key regulators of neuronal and glial cell identity, decipher the enhancer code for specific neuronal subtypes, and generate “informed” enhancer driver lines that allow cell type specific manipulation. This information is available as a resource at http://flybrain.aertslab.org to allow users to explore the data in detail for their own cell populations of interest. 2
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021.
|
The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Unique chromatin landscapes underlie neuronal diversity
To study the regulatory programs of neuronal diversity, we profiled chromatin accessibility of 240,919 cells from the entire brain using single-cell ATAC-seq (10x Chromium). We dissected brains at nine time points from third instar larvae to young adult flies, covering the most important stages of neuronal development 26,32 (Fig. 1a). The experiments were carried out with four different wild type polymorphic strains 33, allowing a higher number of nuclei per run, while detecting and removing doublets (see Methods, and Extended Data Table 1). In addition, to enrich for central brain cell types that are often hard to detect 1,3, we performed two additional runs on adult brains without the optic lobes as these contain more than two thirds of all brain cells in numbers, but with more reduced diversity than the central brain. We complement this chromatin atlas with previously published scRNA-seq data from the larval brain (5,054 cells) 15 and we updated our scRNA-seq atlas of the adult brain by expanding our previous atlas of 56,902 cells 1 to 118,687 high quality cells. In this enlarged scRNA-seq dataset, we identified 204 clusters of which 66 could be annotated as known cell types (see Methods). We first analyzed the open chromatin landscape in the adult cell types, as they represent the tips of the developmental manifold. To this end, we combined the 60,624 cells from adult and late stage pupa (72h APF), as these stages are very similar and most neurogenesis and circuit assembly has finished (Extended Data Fig. 1). The analysis with cisTopic 34 resulted in four main categories of cells (Fig. 1b, glia, optic lobe, Kenyon cells, and the remainder of the central brain) that we further subclustered in 79 clusters (Extended Data Fig. 2, and Methods). To link these clusters to specific cell types, we exploited the expanded single-cell transcriptome atlas of the adult brain. We generated a gene accessibility matrix by taking the sum of the accessibility of the regions within the gene body and upstream of its TSS, weighted by distance and variability (Fig. 1c, Extended Data Fig. 3a). Co-clustering the RNA and gene accessibility datasets allowed us to annotate 35 clusters (Extended Data Fig. 3b-d). To match and confirm additional clusters across modalities, we used marker gene enrichment 35 and non-negative least squares (NNLS) regression 36 (Fig. 1d, Extended Data Fig. 3e). The combination of these methods led to a final annotation of 43 of the 79 clusters, which were unambiguously one-to- one linked to RNA. The annotated cell types include six glial subtypes (~10-15%, in blue in Fig. 1b) as well as non-brain cells (~1%, plasmatocytes and photoreceptors, in black); but, as expected, the majority of cells are neurons (~85-90%; optic lobe in green, central brain clusters in red).
|
Interestingly, optic lobe neurons form clear and distinct clusters, while central brain neurons appear as a continuum. This may be explained by the organization of the optic lobe, in which cell types are present in multiple copies, forming repetitive columnar structures that process input from the 800 ommatidia of the compound eye. While many unicolumnar cell types (present in every column) were identified, we were unable to annotate the sparser multicolumnar neurons (connecting multiple columns). In the central brain, we identified the three Kenyon cell (KC) subtypes and two smaller cell types of the central complex: ring neurons of the ellipsoid body, and the protocerebral bridge. This is similar to observations in transcriptome data of the central brain, where central brain cell types and multicolumnar optic lobe neurons were more difficult to identify due to their low cell numbers 1,3. Intriguingly, central brain clusters are split into Imp+ or prospero (pros)+ cells based on the scATAC-seq based gene accessibility, recapitulating differences shown previously by scRNA-seq in the brain 1, ventral nerve cord 16, and larval brain 15 (see also further below). To further validate cell type annotation and their associated regulatory regions, we used split-GAL4 lines to label the three KC subtypes 37 and performed bulk ATAC-seq after FAC-sorting 38 (Fig. 1e). The differential regions from the bulk profiles matched the single-cell clusters (Fig. 1e: cell type specific accessible regions highlight matching clusters in the tSNE). Comparing the scATAC-seq aggregated profile with the sorted bulk data profiles confirms the high concordance at marker regions, as illustrated in Fig. 1e, with the aggregates of αβ-KCs having matching peaks near Gfrl, α’β’-KCs near
3
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. DAT, and γ-KCs near ab, thus confirming the high quality of cell type specific chromatin accessibility based on scATAC-seq. Although not all clusters were mapped to specific cell types, each cluster has a unique chromatin accessibility profile, with a range of 105 to 4,732 differentially accessible regions (DARs) out of a total of 24,543 median accessible regions per cluster or cell type (Fig. 1f-g). Accessibility of intronic and distal intergenic regions correlated better with gene expression compared to the accessibility of the promoter/TSS region, confirming previous observations 39(Extended Data Fig. 4). Indeed, many broadly used and validated marker genes have their TSS ubiquitously accessible and their specificity may be controlled by more distal regions (e.g., bsh in Mi1 neurons, Optix in PB neurons [arrows in Fig. 1g]). Of the 1,017 DARs of the mushroom body Kenyon cells that overlap with tested enhancer- reporter regions in the Janelia FlyLight collection, 588 are reported as active enhancers in the mushroom body (Extended Data Fig.
|
5) suggesting that DARs often function as enhancers. Interestingly, while T4 and T5 neurons are grouped into the same cluster in the transcriptome data their chromatin profiles split them into two distinct clusters, even though there are only 110 differentially accessible regions between T4 and T5. Three of the regions accessible in T4 neurons are located in the locus of TfAP-2, a transcription factor that is specifically expressed in T4 40. Further sub- clustering reveals that a/b and c/d subtypes can also be separated by chromatin accessibility (Extended Data Fig. 6). Given this high resolution in the scATAC-seq data, it is surprising that 60K cells were not sufficient for identification of all cell types in the central brain. For example, the olfactory projection neurons (OPNs) form a separate cluster in the 57K scRNA-seq dataset 1 but not in the 60K scATAC-seq dataset. Further increasing the cell count to 88.3K cells by including the 48h time point, where OPNs are already present, does not resolve this issue (Extended Data Fig. 2e). OPNs and other central brain cell types are thus harder to distinguish at the level of accessible chromatin, compared to the level of the transcriptome. In conclusion, scATAC-seq of the fly brain yields high-quality cell type specific chromatin accessibility profiles that correspond to matching transcriptomes, allowing all large cell types in the optic lobes and multiple cell types in the central brain, including the three KC subtypes, to be characterized by an associated set of differentially accessible regions. 4
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 1: Overview of the chromatin atlas and characterization of the chromatin landscape of the adult cell types
a. Experimental overview showing number of cells per timepoint, covering the spectrum of neurogenesis and synaptogenesis. Runs are demarcated by color switches; central brain only runs are accentuated. b. 2D projections (tSNE) of the adult cell types in scATAC-seq (left, includes adult and 72h APF cells) and scRNA-seq (right, adult cells only). c. Gene accessibility and gene expression display a similar pattern for many genes. d. Heatmap of NNLS regression coefficients after z-normalization showing the correspondence between RNA and ATAC clusters. e. Overview of bulk ATAC-seq on three sorted cell populations. (top) Confocal images of three Kenyon cell subtypes targeted with split-GAL4 lines; (middle) tSNE showing accessibility of cell type specific regions; (bottom) profiles show similarity between bulk ATAC-seq (black) and scATAC-seq. f. Overview of differentially accessible regions per cluster, bar plot shows number of accessible regions per cell type with cell type specific regions in dark. g. Aggregate profiles of differentially accessible regions (max value=50 (30 for rgn and ab)).
|
5
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Note the constitutively accessible regions in Optix, bsh and Gfrl (arrows). Cell type abbreviations: Astrocyte-like glia (AST), Chiasm glia (CSM), Cortex glia (CTX), Ensheathing glia (ENG), Perineurial glia (PNG), Sub-perineurial glia (SUB), Poxn-neurons of the ellipsoid body (PXN), Protocerebral bridge neurons (PB), Photoreceptors (PR), Plasmatocytes (PLM), Kenyon Cells (KC). Numbers are unidentified clusters that match between scRNA- and scATAC-seq. Concordant TF expression, enhancer accessibility, and gene expression yield enhancer-
GRNs
Current descriptions of gene regulatory networks have been mostly focused on imputing transcription factors with their target genes by co-expression, sometimes enhanced with motif detection 35,41–45. The availability of transcriptome and chromatin accessibility profiles of the matched cell types opens up the possibility of exploring their regulatory code in much greater detail. In particular, we aim to unravel high-confidence cell type specific “enhancer-GRNs” (eGRNs), including the key transcription factors of a cell type, as well as their target genes and the enhancers through which they are regulated. To reconstruct eGRNs, we developed a computational strategy that exploits the matched scRNA- and scATAC-seq data to identify candidate transcription factors that are both expressed and have their recognition motif enriched in the open regions of a given cell type (Fig. 2a). As a first step, we defined “cistromes” (i.e., candidate target regions) for each transcription factor: the subset of DARs of each cell type, in which the TF is expressed and its motif is significantly enriched (Fig. 2b-c) (see Methods). We found cistromes for 206 TFs (out of the 251 expressed TFs with a motif in our database), including TFs with pan-neuronal, pan-glial, and cell type-specific activity (89 of them shown in Fig. 2b). The full list of predicted regulators and their cistromes can be downloaded from the web portal (https://flybrain.aertslab.org). These include key regulators for KCs, in which we confirmed Mef2 for γ-KCs and αβ-KCs 46,47, in addition to Ey (Fig. 2d). For ellipsoid body neurons, Grain (grn) and Dichaete (D) emerged as key regulators 48. In T1 neurons, we identified Ets65A 40, and Ocelliless (oc) as main regulators 49 and combinations of Acj6, Fkh, TfAP-2, and SoxN/Sox102F in the other T-neurons 2,4,27,40. Glial cells show repo and Kay expression, and their respective motifs are enriched in glial DARs 50. Interestingly, we also identify TFs with a negative correlation between gene expression and motif enrichment, suggesting a repressive role for instance for Pros, Lola and Cut (ct) (Fig 2b). In the second step, we linked the cistrome regions to their target genes.
|
We approached this by calculating a co-variability score for the expression of each gene and the accessibility of the regulatory regions nearby (using random-forest regression, and “metacells” to match the cell types across data modalities, see methods). Previous work has shown that regulatory interactions can occur over large distances but are mostly confined within chromatin domains, in so-called “genomic regulatory blocks” (127kb median size, 51), a HiC-derived topological associated domain (TAD, 13kb median size, 52), or between two BEAF-32 boundary elements (57kb median distance, 53). Taking this into account, we considered enhancer-gene interactions in a window of >100kbp around each gene (50kbp up- and down-stream, plus introns) (Fig. 2e), which led to an average of 9 linked regions per gene (Fig. 2f), 55% of which lie between BEAF-32 boundaries (Extended Data Fig. 7). These enhancer-gene links provide a set of potential target genes per cistrome. However, to reduce the rate of false positives and obtain a higher-confidence network, as a third and final step, we tested whether the expression of each set of target genes co-varies with the predicted TF, similar to the principles of SCENIC 35 (based on gradient boosted regression and gene set enrichment analysis, see Methods). This confirmed that when the analysis is restricted to links that are within BEAF-32 domains, the correlation with TF expression is of slightly higher quality, allowing us to select the 89 highest- confidence cistromes (highlighted by a dark border in Fig. 2b) together with their top ranked target genes, which form the “enhancer-GRN” for the different cell types (Fig. 2g). 6
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The eGRNs for the 40 available cell types have an average of 6 TFs, collectively regulating 108 target genes through 138 enhancers within BEAF-32 boundaries. Fig. 2g shows the network of T1 neurons, which highlights the combination of 5 TFs (Ets65A, Oc, Ct, Sima and Awh) regulating between 15 and 60 targets, with Oc and Ets65A auto-regulating, and around 50% of the targets being co-regulated by at least two TFs. Similarly, the network for γ-KCs (Fig. 2h) reveals Mef2 and Ey/Toy as auto-regulatory key factors that regulate 77 to 83 genes, alongside new candidate TFs Dif, Sd, and Pan regulating an average of 29 genes. 2/3 of the KC target genes are co-regulated by a minimum of two TFs, specifically revealing a high overlap between Ey/Toy and Mef2. The analysis of the 40 eGRNs also suggests that, while at least 57% of the genes are regulated by several regions within the same cell type, 95% of the regulatory interactions involve multiple TF inputs that occur through a common enhancer. In conclusion, we have developed a new computational strategy and constructed 40 cell type specific “enhancer-GRNs” with key transcription factors, target genes and enhancers (Extended Data Fig.
|
8). 7
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 2: Construction of eGRNs for key brain cell types through multi-omic data integration
8
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. a. A multi-omic approach for the creation of e(nhancer-)GRNs. b. Dotplot showing expression and motif enrichment for the TFs within an eGRN (marked circles show the presence of a cell type-specific eGRN). c. Ey motif enrichment versus expression, per cell type. d. tSNE showing the accessibility of the regions of the Ey-cistrome across cell types in the adult. Note high accessibility in the ey-expressing Kenyon cells. e. Regulatory region selection for Pkc53E; the inset shows the accessibility of the top region versus the gene expression (input to the random-forest), regions with a weight above the threshold are linked to the gene. f. Overview of selected tracks in the Pkc53E locus; only the enhancer-gene links between two BEAF-32 peaks (green bar) are kept. g-h. eGRNs for T1 neurons and γ -KCs (regions are coloured in blue shades, genes in red; regulatory TFs are in the center). Insert heatmaps show Jaccard index between TF target regions. Arrow points to Pkc53E enhancer shown in f.
Combinatorial TF expression is reflected by enhancer architecture
Enhancer-GRNs provide intuitive insight into the regulatory state of a cell type and highlight the combinatorial nature of TF-target interactions. To further investigate how TF ensembles yield highly accurate spatio-temporal gene expression patterns, we examined the sequence of the predicted enhancers in detail using a deep learning model and tested their activity by a selection of in vivo enhancer-reporter assays. We focused on a subset of the data that includes KCs, T-neurons, and glia and we re-analyzed these with cisTopic (Fig. 3a). cisTopic provides both a cell clustering and a region clustering in the form of topics. These topics are sets of regions that can be accessible in specific cell types or in multiple (Fig. 3b). We then trained a convolutional neural network using the sequences of the topics as input in order to predict in which cell types they are accessible 54 (see Methods, Fig. 3c, Extended Data Table 2). Evaluation of the model’s accuracy using cross-validation and left-out test data, yielded accurate classification of promoters, BEAF-32 boundary elements (3 topics, average auPR=0.36), pan-neuronal and pan-glial regions (5 topics with auPR=0.30 and 2 topics with auPR=0.30 respectively), and cell type specific enhancers for KCs, the glial subtypes, and each of the T-neuron classes (Fig.
|
3b, Extended Data Table 3). To reveal the TF motifs that underlie cell type-specific accessibility, DeepExplainer 55, which identifies the importance of each nucleotide on a given sequence for the final prediction, and TF- Modisco 56, which uses the nucleotide importance scores and identifies motifs from reoccurring sequence patterns with high nucleotide importance score, were used to derive key features per topic. KC regions are predicted using motifs of Ey, Onecut, Mef2, Mamo and Dati, matching their high gene expression levels in KC (Fig. 3c). Thus, the deep learning model adds additional TFs, enhancers, and target genes to the KC eGRN. Interestingly, two of these, namely Mamo and Dati, have negative nucleotide importances, meaning that the presence of their motif is correlated with a closing of the peak, which may reflect a repressive function. The most important motifs linked to candidate TFs for T-neurons include Fkh, TfAP-2, Acj6 and Ct, Repo, Zfh2 and Klu for glia (Extended Data Fig. 9). The eGRN framework we presented above, complemented with the deep learning framework, led to the prediction of cell type specific enhancers. We selected 52 genomic regions (of which 27 are nodes in the eGRNs, Extended Data Table 4, Extended Data Fig. 10a-b) for which we created a transgenic GFP reporter line (see Methods). For instance, a region associated with the Bx gene in the γ-KCs eGRN (Fig. 3d) is accurately predicted by the deep learning model, which reveals a candidate Ey binding site and additional motifs that we could match with Hr51 and Sr binding sites. While these two factors are associated with KC cistromes, their target genes did not pass the eGRN association filters, but using deep learning we recovered these TFs. The reporter activity of the enhancer is detected mainly in KCs, matching with the cell type specificity of the peak. Nevertheless, there is some off-target GFP signal in small cell populations in the central brain that are not part of the current clusters due to the lower resolution of cell types there. Another cloned region is predicted to activate CG15117 expression in T1 neurons (Fig. 3e). The model correctly predicts the region as a T1 enhancer, using Oc and Ets65A motifs, and predicts an additional Awh binding site. The reporter activity shows specific activity in T1 neurons, matching the specificity of CG15117 expression. Interestingly, both examples display different expression patterns in the larval brain compared to the adult brain, suggesting a recycling of the same enhancer in different cell types through development. Inspecting
9
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. the conservation of eGRN enhancers using whole-genome alignments across 124 different insect species 57, reveals that the informative nucleotides predicted by the model show a higher conservation in each of the three main classes (KC, glia and T neurons, Wilcoxon p<1e-10), with the highest conservation in T neurons, suggesting a functional importance of the predicted binding sites (Fig.
|
3f). Based on these results, we envision a potential use for this atlas as the starting point for the design of reporter lines to target specific cell populations throughout development. Although we could partially validate this potential by taking advantage of Janelia’s FlyLight 11 and Vienna Drosophila Resource Center (VDRC) enhancer lines (Extended Data Fig. 5) the subsequent interpretation is compromised due to their length, since the Janelia and VDRC regions are 2-3kb long and often contain multiple predicted enhancers. Using our transgenic lines for 52 candidate enhancers (including the examples shown in Fig. 3), where only a 300-1732bp ATAC-seq peak was cloned, we could achieve higher specificity compared to these larger regions. Indeed, we confirmed that for Janelia regions that have only one ATAC-seq peak, the active enhancer resides within the ATAC peak, which recapitulates the correct pattern similar to the larger encompassing Janelia selection (Extended Data Fig. 11a-b). In addition, we found that Janelia lines with multiple ATAC-seq peaks show activity in multiple cell types, while the activity of the individually cloned peaks becomes specific (Extended Data Fig. 11c). Finally, we confirm that split-GAL4 lines, which combine two enhancers in an “AND” logic, can be recapitulated as the intersection of their ATAC-seq signal (Extended Data Fig. 11d). Using complementary approaches of eGRN inference and deep learning is useful, since the eGRN is based on enrichment analyses, region-to-gene mappings, and thresholding, and is biased to TF-target activation. The deep learning model identifies candidate repressors such as Dati and Mamo, predicts additional DARs as enhancers (Fig. 3g-h), and highlights binding sites in T4 and KC enhancers. Many of the cloned enhancers show activity in the larval brain, with some enhancers having a broader expression pattern, suggesting developmental roles for many adult enhancers (Extended Data Fig. 10c-d). Surprisingly, DARs and eGRN regions reach similar levels of activity, although correlation with gene expression is a major factor to determine activity (Extended Data Fig. 10d-f). This suggests that even though the eGRN annotation is a set of active enhancers where both transcription factor and target gene is inferred, many functional enhancers remain to be investigated. Overall our results yield similar rates compared to the embryo 58 and eye-antennal disc 59. In conclusion, we have created a deep learning model that can unravel the logic behind the enhancers in eGRNs and in accessible regions. The binding sites predicted by the model match motifs found by conventional methods, and add additional ones, thereby increasing our recall. Finally, we show how this atlas can be used to create new driver lines starting from the eGRNs or from specifically accessible regions. Existing driver lines that are “noisy” can be separated into smaller functional components that can be used as standalone, more specific driver lines.
|
10
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 3: Deep learning unravels enhancer make-up
a. tSNE of cisTopic analysis on subset of Kenyon cells, T-neurons and glia. b. Topic heatmap showing cell type specific topics. Bar plots show number of regions per topic (cutoff p=0.995) and area under precision-recall curve (auPR) of deep learning model, validated on test dataset. c. Architecture of deep learning model in which regions are assigned to topics through nucleotide contributions. Contributions of the nucleotides lead to de-novo motif discovery, showing motifs of known KC factors, matching their expression. Note that negative nucleotide importance leads to repression of accessibility. d. Selected region from the γ-KC eGRN is accurately predicted to belong to a γ-KC topic. DeepExplainer of a subset of the region view highlights Ey, Hr51 and Sr binding sites. Cloning of the region (black box) leads to KC-specific expression patterns. e. Selected region from the T1 eGRN is accurately predicted to belong to a T1 topic. DeepExplainer of a subset of the region view highlights Ets65A, Awh and Oc binding sites. Cloning of the region leads to T1-specific expression patterns in the adult. f. Histograms showing the conservation of nucleotides in the T-neurons, KC and glia eGRN enhancers: selected by the DL model (colored, n= 12,429 (T), 11,625 (KC), 53,007 (Glia)) against the whole region (grey, n= 486,500 (T), 452,500 (KC), 2,125,500 (Glia)), lines showing the median. Contrast was performed with two-sided Wilcoxon rank-sums test (stars mark pval<1e-10). g. DAR in T4 neurons leads to the creation of a specific T4 driver line. DeepExplainer of a subset of the region view highlights
11
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Fkh, TfAP-2, Pan and Acj6 binding sites. Cloning of the region leads to T4-specific expression patterns. h. DAR in KCs leads to the creation of a specific KC driver line. DeepExplainer of a subset of the region view highlights Mef2, Sr and Ey binding sites. Cloning of the region leads to KC-specific expression patterns. Dynamic changes in chromatin accessibility during brain development
To investigate how adult neuronal diversity is generated, we studied dynamic chromatin accessibility changes during development. We first focused on neurogenesis by separately analyzing the timepoints ranging from third instar larval to 12h after puparium formation (APF), with a total of 135,275 cells. Note that for these time points, we dissected the entire central nervous system (including brain and ventral nerve cord).
|
The analysis of these cells with cisTopic resulted in 200 topics and 54 cell clusters (Fig. 4a). To annotate the cell clusters, we recursively trained an SVM classifier on the adult cell types and used it to label cell types in earlier stages, similar to recent work on RNA datasets 4 (Extended Data Fig. 12a-b). This data shows that adult DARs can be used to identify pupal cells, thus suggesting that neuronal fate is determined early on in development. To elaborate on this, we calculated core sets of regions per cell type that are continuously accessible at each timepoint, specifically characterizing cell types (Extended Data Fig. 12c-d). Similar to RNA-seq where in OPNs and optic lobe neurons, a maximum number of differentially expressed genes was detected in 48h APF and a minimum in the adult 4,32, we find a decrease in DARs over time, and with a relative spike at 48h APF during synaptogenesis (Extended Data Fig. 12e). On the other hand, the cell clusters that cannot be classified using the adult SVM are, as expected, the progenitor cell types. These clusters are characterized by accessible regions near the neuroblast markers deadpan (dpn) and asense (ase) (Extended Data Fig. 13a-b). During larval and pupal neurogenesis, the optic lobe neuroepithelium (ONE) in the outer proliferation center is converted to medulla neuroblasts (NB) and lamina precursor cells (LPCs), while quiescent embryonic central brain neuroblasts are reactivated 60. To reveal the dynamics of the bifurcation of neuroepithelium into either medulla NBs or LPCs, we fitted a trajectory through the UMAP and calculated modules for the different steps, calculating a complementary scRNA-seq trajectory from larval data 15 (Extended Data Fig. 13c-d). In the ONE state, grainyhead (grh) and zelda (zld) are found as major regulators, but are quickly replaced by glial cells missing (gcm) when progressing towards the LPC fate 61. The differentiation into medulla neuroblasts co-occurs with a phase with pointed (pnt) and scute (sc) 60, followed by E(bx) and finally the emergence of a neuro-GGG motif 58,59 and known neuroblast factors Ham and Kr 62,63. To study the differences between neural progenitors in more detail, we performed motif enrichment on differential peaks between LPC, OL NBs, and CB NBs, which highlighted additional regulators like so for lamina precursor cells (LPC) 64 and scro for the optic lobe progenitors 65,66, while the GGG motif is detected in both CB and OL neuroblasts (Extended Data Fig. 13e). One possible candidate for this motif is the transcription factor pros, which is known to bind to GGG motifs and plays a major role in neuroblast/GMC identity 59. The addition of a temporal axis allows us to create driver lines for cell types through development. Selecting a region that is specifically open in the ONE, we created a driver line that shows a matching expression pattern in the larva and no detectable activity in the adult, thereby paralleling the decline of the neuroepithelium (Extended Data Fig.
|
13f). The OL and CB neuroblasts form the roots for each of the two main branches in the developmental UMAPs (more evident in the 3D projections, Fig. 4b), indicating distinct modes of neurogenesis and maturation in each brain region. The OL shows a high amount of branching, while the CB and ventral nerve cord display a continuum, similar to our observations in the adult tSNE (Fig. 1b). Interestingly, the split between pros+ and Imp+ in central brain neurons is already present in these early stages, and, similar to scRNA-seq data, most of the ventral nerve cord neurons belong to the Imp+ neurons (Fig. 4c). It has been speculated that Imp+ neurons are embryonic neurons 16. To test this hypothesis, we used the differential regions belonging to each group and plotted their accessibility in a scATAC-seq dataset of the embryo 58 (Fig. 4c) Indeed, the embryonic CNS shows increased accessibility for the Imp+
12
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. regions, while pros+ regions only become accessible in the larval dataset. To find which transcription factors may be guiding this dichotomy, we performed motif enrichment on the differential regions. Dati and Pros motifs were enriched in the Imp+ cells, most notably in peaks near Imp itself, suggesting a repressive role for these factors, where they may close these regions in pros+ cells and suppress Imp expression. This layer of cell identity, related to birth order, guides most of the clustering in the central brain. Embryonic CB neurons are rewired during early pupal stages to integrate into the adult network. This rewiring consists of a pruning step in which axons retract, followed by axonal regrowth. Both Pros and Imp have been associated with neuronal remodeling in the γ-KCs, with Imp being required for regrowth of pruned axons by promoting transport of ribonucleoprotein granules 67,68, while overexpression of pros inhibits pruning 69. The process of neuronal remodeling is regulated by the ecdysone hormone and Sox14 70. We examined the regions that were specifically accessible during pruning (0h APF to 6h APF) and motif enrichment indeed reveals a spike for both EcR and Sox14 motifs (Fig. 4d). Interestingly, this spike only occurs in the Imp+ neurons, suggesting that only these undergo pruning, consistent with their embryonic origin. In the OL, six distinct branches emerge, each enriched for the motif of one major group of transcription factors (Fig. 4e). These branches show a large number of differential regions, located near genes of the immunoglobulin-like super family (median 22.5, Extended Data Table 5), suggesting a role in axonal development and synaptic partner recognition 71,72. Interestingly, most of the adult OL cell types correspond to one of these branches (Fig. 4f), even though this classification does not respond to the known developmental lineages.
|
For example, T2/T2a/T3 neurons and C2/C3 belong to the same lineage but are put in different branches 73. One difference we observe is in neurotransmitter usage, where cell types from branch 4 and 5 tend to be non-cholinergic. One of the eGRNs that overlaps with these branches is Ets65A, a factor previously hypothesized to be involved in non- cholinergic fate 24. We further note a correlation with neuropils, with most Tm neurons of the medulla and the T-neurons of the lobula plate belonging to separate branches 73. The cell types that have a low score for either of the branches are mostly unknown, except LLPC1 neurons that have a CB origin 8. Therefore, this data-driven trajectory analysis likely reflects a novel regulatory dimension with a potential role in neuronal wiring that is shared across OL neuroblast lineages. During the cloning of eGRN regions we noticed strong differences in driver line expression between larval and adult brains. Therefore, we examined the changes of region accessibility in the eGRNs during development per cell type. Of all eGRN enhancers, 42% (3307) become more accessible in late timepoints, with 28.8% increasing after the ecdysone pulse. We also find 985 regions that undergo an enhancer switch, as their accessibility increases in one cell type while decreasing in another. One of these regions is the T1 enhancer driving CG15117 expression shown in Fig. 3e, which is accessible in glia during development and then switches to T1 at 24h APF, with a major surge at 48h APF (Fig. 4g). Using gene expression data 4, we confirmed that this enhancer switch is accompanied by a gene expression switch, as CG15117 is a glial marker during development and a T1 marker in adult. Interestingly, a small delay can be observed between enhancer accessibility and gene expression changes, similar to what was observed in other studies 74. Measuring the enhancer at different developmental timepoints, we indeed find GFP positive cells with the glial marker repo in development, but they disappear in the adult, coinciding with the closing of the enhancer (Fig. 4h-i). In conclusion, by tracing back all cell types through development, we unravel a highly dynamic chromatin landscape driven by neuronal maturation and ecdysone. Large numbers of adult-specific regions are linked to reprogramming history in the CB and to neurogenesis modes in the OL. These remain accessible through development, providing a link between time points. Finally, this analysis reveals the widespread use of enhancer switching between cell types, in which specific enhancers become accessible in different cell types throughout metamorphosis. 13
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Figure 4: Development of neuronal cell types from neuroblasts through pupation
a. cisTopic UMAP of 92,954 cells for the analysis of early timepoints (third instar larvae to 12h APF, cells>900 FIP).
|
Grey lines connect central brain neuroblasts to the central brain branch. b. Optic lobe and central brain branches in 3D-UMAP. c. Central brain and VNC duality between pros and Imp. d. Trajectory of motif enrichment in Imp+ and pros+ cells through development, showing modules for remodeling and an ecdysone response e. Six branches can be identified in the 3D optic lobe UMAP, each having their own unique accessibility regions. Motif enrichment reveals one transcription factor family per branch. f. Adult optic lobe cell types can be traced back to OL branches by enrichment of branch-specific regions through AUCell. TFs with eGRNs that overlap more than 10 regions with the branch are shown (matching motifs in bold). g. Example of an enhancer-switch between glia and T1 neurons, both in enhancer accessibility (full line) and target gene expression (dotted line). h. Staining of repo (red) and CG15117 enhancer activity (GFP, green), showing switch from glia in early development (top, repo positive, central brain, 15h APF) to T1 neurons (bottom, repo negative, optic lobe, adult)). i. Track showing the dynamics of enhancer-switch between glia and T1 neurons. Cell type abbreviations: Astrocyte-like glia (AST), Cortex glia (CTX), Ensheathing glia (ENG), Surface glia (SUR), Ganglion mother cells (GMC), Lamina precursor cells (LPC), Lamina neurons (LAM), Optic lobe neuroepithelium (ONE), Neuroblasts (NB), Poxn-neurons of the ellipsoid body (PXN), Protocerebral bridge neurons (PB), Photoreceptors (PR), Plasmatocytes (PLM), Kenyon Cells (KC), Ventral nerve cord (VNC). 14
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Discussion
We generated the first single-cell epigenome atlas of the whole fly brain throughout development, tracing neuronal and glial cell types from their birth to their mature state and providing a bridge between the regulatory code in the genome sequence and the transcriptome. Through an integrated multi-omics approach, we identified key regulators per cell type and revealed extensive gene regulatory networks. To depict this information, we introduce the concept of the enhancer-GRNs in which transcription factors are linked to high-confidence binding sites that are in turn linked to target gene expression. Given the strong insurgence of single-cell datasets, the pioneering work in single-cell ATAC-seq in mouse and human 20,21,39,75,76 and the fast development of single-cell multi-omics 74,77,78, eGRNs can soon be derived for many other datasets and may serve as the pinnacle for studying genomic regulatory programs. We showed how deep learning can be integrated with omics data to accurately predict enhancer activity based on the DNA sequence. This “smarter” motif discovery approach revealed a large number of motifs that are missed by conventional algorithms and leads to a base-pair resolution prediction of binding sites.
|
Interestingly, we noticed that some binding sites show mismatches with the canonical motif of that transcription factor, which points towards the fine-tuning of binding affinities through evolution 79–82. It has long been established that neuronal identity is governed by a plethora of transcription factors, yet the study of the working mechanisms underlying these factors, namely their enhancers and target genes, has been obstructed by the complexity of the brain. There are hundreds of cell types in the fly brain, impeding the use of bulk datasets for accurately deciphering diversity. Even more, compared to the embryo 58,83–87 or imaginal discs 59,88–94, the number of ChIP-seq or other epigenomic datasets is very limited in the brain 24,95–97. In contrast, in this study we constructed 40 eGRNs at once, covering 88 transcription factors with 7833 enhancers that are linked to 3776 target genes. Of these 88 transcription factors, 92% have lethal mutations, 72% are linked to known brain phenotypes, and 67% are linked to human diseases. Our analysis reveals that a great number of regions are regulated by multiple transcription factors, and we highlighted Ey and Mef2 as an example in Kenyon cells. The cell type specific chromatin profiles, predicted TF binding sites and interactions, and changes through development, will allow for more insight in the determination of neuronal identity. Genetic access to neuronal cell types through the use of driver lines has revolutionized neurobiology 11,12,98. However, approaches to create driver lines have mostly been uninformed as they were based on selecting regions near genes 99 or randomly bashing genomic regions 11, leading to many unspecific driver lines. Using our scATAC-seq atlas, existing driver lines can be further dissected, and we indeed found many lines that contain multiple enhancers (median 3 and up to 9 enhancers per region in the Janelia FlyLight collection). Subcloning these individual enhancers separately, we created more specific driver lines that target distinct cell types. Furthermore, we identified 96K regions, covering 34.4% of the genome that can be queried as an initial step to create new reporter lines. Both eGRN regions and DARs can be used, the success rate increases if the region has a high correlation with gene expression. Using this system, we created new driver lines for neurons (KCs, T1, T4) and glia, and extrapolated it to development for the optic lobe neuroepithelium. Interestingly, many of the tested enhancers show increased activity during early developmental stages compared to the adult, with enhancer success rates being higher during development (Extended Data Fig. 10c). We show that this is partially caused by developmental dynamics where large numbers of enhancers open/close and switch between cell types, likely as a consequence of the high regulatory density of the Drosophila genome. We highlight one enhancer that switches from glial to neuronal activity, where interestingly only few transcription factors overlap.
|
This points towards a phenotypic convergence phenomenon, where different transcription factors can regulate the same enhancer 2. A second reason for lower success rates in adult brains could be lower expression levels in the adult compared to developmental stages. In the embryo and imaginal discs enhancers linked directly to GFP without using the GAL4
15
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. amplification system, yield higher recall rates compared to the brain. Future enhancer-reporter assays in the brain may thus benefit from using GAL4 as reporter system. Some enhancers may furthermore depend on a specific proximal promoter. Nevertheless, we believe that success rates of >40% are high enough to create a driver line for any cell type. Our atlas also opens up the possibility of creating temporo-controlled driver lines using the split-GAL4 system with enhancers corresponding to different maturation modules. Although for the central brain, we have lower cell type resolution in scATAC-seq compared to scRNA- seq, we detected large overarching programs between pros+ and Imp+ cells. We are able to trace this duality back to the embryo, where only Imp+ cells exist and show that these neurons are reprogrammed during metamorphosis, confirming previous hypotheses 16. Future studies including single-cell multi-ome assays will likely resolve additional cell types in the central brain. Finally, we detect six different modes of neurogenesis in the optic lobe, each dominated by a specific TF family that we can link to neuronal wiring and neurotransmitter expression. We believe that the regulatory atlas of the brain, covering cell types, transcription factors, and enhancer regions, together with their joint representation as eGRNs, will be of great value to the community. Therefore, we have made all the data generated in this project publicly available at http://flybrain.aertslab.org. The website allows users to explore regulatory networks of key SCOPE transcription target Browser (http://scope.aertslab.org/#/Fly_Brain/) (http://genome.ucsc.edu/cgi- bin/hgTracks?db=dm6&hubUrl=http://ucsctracks.aertslab.org/papers/FlyBrain/hub.txt). factors with
regions and
and
genes,
and UCSC
links out Genome
to
the
16
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended data figures
Extended Data Fig. 1: Global analysis of all timepoints reveals major dynamics and optic lobe-central brain split
a. UMAP of global cisTopic (150k cells shown), colored by region accessibilities near elav (red), repo (green) and dpn (blue), matching neurons, glia and neuroblasts respectively b. Overview of regions shown in a. for representative cell types (Kenyon cells for neurons, Astrocytes for glia, optic lobe neuroblasts for neuroblasts) c. Distribution of cells per timepoint in the global UMAP; The border groups timepoints jointly analyzed in the upcoming sections (green: early timepoints, blue: late timepoints) d. Spearman correlation of top 1000 variable regions across timepoints e. UMAP after timepoint correction with Harmony (colored by timepoint: early timepoints in green, late in blue).
|
17
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 2: Sub-clustering of main categories in the adult
a. t-SNE of the 60k cells in the adult cell types analysis (includes adult and 72h pupa). The different colouring schemes illustrate how the enchment of central brain only runs leads to the annotation of the cell clusters according to their location (central brain and optic lobe). Sub-clustering of the cells was performed splitting the neurons based on their location and glia, note that Kenyon cells, plasmatocytes and photoreceptors were not included in these major groups. b. Sub clustering of optic lobe neurons leads to 58 sub clusters, including a further split of T4/T5 neurons. c. Sub clustering of central brain neurons reveals 51 subtypes. Notice how the S-shape of Pros+ cells and the split from Imp+ cells are retained. d. Sub clustering of glia reveals 16 subtypes. e. Clustering including 88k cells from 48h APF to adult, provides three extra clusters including younger cells, but does not increase the resolution of the adult cell types. 18
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 3: Integration of scRNA-seq and snATAC-seq
a. Calculation of gene-accessibility scores using a weighted sum of regions in the gene body and up to 5kb upstream of the TSS. Weights decrease exponentially with distance from the TSS (constant in the gene body), and increase with higher gini coefficients b. Overview of annotation methods used. Main cell types are detected with each method, while low confidence matches are method specific c. Annotated tSNE of the transcriptomes of 118k adult cells d. Integrated tSNE of scRNA-seq
19
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. and snATAC-seq e. Gene set enrichment with of marker genes using AUCell on gene-accessibility matrix, revealing matches per cell type and per major cell type group (glia, optic lobe neurons, central brain neurons, Kenyon cells, photoreceptors and plasmatocytes). 20
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 4: Gene expression correlates with region accessibility
Correlation of gene expression (a.)
|
and TF gene expression (b.) with aggregated accessibility profiles at TSS, averaged gene- accessibility, and averaged accessibility of regions with positive links. Red line shows linear fit, with orange boundaries as the 95% confidence interval. Note the regions near the TSS that have high accessibility but do not lead to gene expression (highlighted in blue) and the increase in performance for the gene-accessibility score in the TF expression, while overall highest correlation is reached with links. 21
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 5: Cell type specific regions can serve as functional enhancers
Kenyon cell DARs overlap with 588 Janelia regions active in Kenyon cells, of which a subset of 110 is shown here. Images
courtesy of the Janelia FlyLight Project. 22
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 6: Identification of T4/T5 subtypes
a. Subclustering of T4 and T5 neurons identifies the a/b and c/d subtypes, with differential regions near marker genes TfAP-2 and bi. b. Locus of TfAP-2, showing differential peaks between T4 and T5 neurons c. Locus of bi shows specific peaks for c/d subtype. 23
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 7: BEAF-32 ChIP-seq peaks can be used to delimit the search space for regulatory regions around
each gene. a. Proportion of BEAF-32 ChIP-seq peaks that have a high-scoring BEAF-32 motif, and are accessible in the fly brain. Despite being performed on whole embryo (0-14 hours, mixed sex), most of the peaks with motif are ubiquitously accessible across the cell types in the brain. b. Distance to the closest BEAF-32 peak with motif upstream and downstream of each gene. Most of the genes (86%) are within 50kbp of a BEAF-32 peak (46% are between two peaks within 50kbp, and 88% within 200kbp). For those genes further than 50kbp, expanding the search space from 50kbp to 200kb adds a median of 2 extra links. c. View of genomic regulatory blocks (GRBs), topological associated domains (TADs) and BEAF-32 ChIP-seq peaks near pros (highlighted by the pink rectangle). For the genes within the defined genomic regulatory blocks, 95% of these associations are contained within the same block, and 25% within the same TADs. Since the current GRBs dataset does not provide enough genome coverage (it only includes 1523 genes, 15% of the 9821 genes with links), and the TADs are very fragmented (only 67% of genes have their biggest transcript within one TAD), we opted to use the BEAF-32 peaks to define the search space per gene.
|
The lowest track shows the search space used for each gene (i.e. the region between the first two BEAF-32 peak within 200kbp of the transcript, skipping the 500bp around the TSS. In case there are no peaks within 200kbp, 50kbp is kept as search space). 24
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 8: eGRN overview
a. eGRNs were determined for different subytpes of Kenyon cells, T-neuron subclasses and glia and are available for exploration on NDEx through https://flybrain.aertslab.org/. b. Link outs from the gene to FlyBase and UCSC allow to explore gene function and chromatin profiles with all nearby predicted enhancers colored. c. Link outs from regions allow to inspect the region with DeepFlyBrain, to visualize nucleotide importances, while also linking to UCSC to view the genomic context with the selected region highlighted. 25
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 9: Deep learning predicts de novo key transcriptional activators and repressors
a. Motifs used in the convolutional filters of the model to classify T-neuron regions predict Fkh, Acj6 and TfAP-2 as activators in cell types matching their expression profiles. b. Motifs identified by the model to classify glial regions reveals activators (Opa, CG42741, Repo and Ct) and repressors (Klu, Ttk, Zfh2), matching expression profiles. 26
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 10: Enhancers selected by DARs or eGRNs generate novel driver lines
a. 52 enhancers were selected and cloned into a construct, flanked by gypsy insulators (GI) driving GFP expression from the Hsp70 promotor. Selected peaks have a median size of 579bp. b. Overview of GFP expression in different cloned enhancers. c. Bar plots showing validation rate for GFP expression with a distinction for high on target expression vs low or off target expression. Success rates are higher in the larval brain compared to the adult (two-sided Fischer’s exact, n=52). d. Off target expression reaches 40 and 20% in larval and adult brains respectively (two-sided Fischer’s exact, n=52). e. Both DARs and eGRN regions can be used to create driver lines (two-sided Fischer’s exact, n shown). f. Precision-recall curve shows accessibility and correlation of region with nearby gene expression are key features to predict GFP activity. 27
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021.
|
The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 11: Breakdown of existing driver lines into functional components
a. Peak in the overlap of two existing KC driver lines recapitulates KC expression. b. DeepExplainer view of the functional element from (a) showing ey and Mef2 binding sites. c. Existing non-specific driver lines can be broken in separate more specific drivers for Kenyon cells and glia using cell-type specific ATAC-peak signals. d. In-silico overlap of ATAC-peak signals resembles that of in-vitro split-GAL4 lines for T4/T5 neurons. Images courtesy of the Janelia FlyLight Project. 28
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 12: Tracking cell types across development reveals presence of core-regions
a. An SVM classifier was used to propagate the adult cell type labels to earlier stages in development. The classifier also included the progenitor cell types –from the developmental analysis– (purple and dark green colors), to provide further confidence. b. Proportion of cells with each label at each developmental timepoint. c. Chromatin landscapes for T1 neurons, shows a highly dynamic opening and closing of peaks during development. A core set remains accessible at all times, of which a subset is specific to T1 neurons. d. Bar plots showing the number of core-regions identified per cell type. Dark colors show specific core regions (core-DARs). e. Number of DARs calculated per cell type (down sampled to 75 cells) for every timepoint shows a decline over time. The arrow notes a small increase at 48h APF during synaptogenesis. Red line highlights the median, 25th and 75th percentiles are shown as the box edges, and data points within 1.5 times the interquartile range from the edge (whiskers) and outliers are shown as data points, n=74,77,78,77,78,77,78,75,76 cell types. 29
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Extended Data Fig. 13: Neural progenitors have unique chromatin profiles
a. Progenitor cell types show specific marker accessibility, while neurons already show accessibility in adult specific regions b. Number of DARs per cell type in the early development dataset, revealing a lower number for progenitors (purple shades) c-d. Trajectory from optic lobe neuroepithelium (ONE) to lamina progenitor cells (LPC) and optic lobe neuroblasts (OL NB) using scATAC-seq (c) and scRNA-seq (d). Heatmap shows dynamic chromatin accessibility modules with enriched motifs (NES score shown) and line plot shows expression profiles for predicted master regulators.
|
e. Specific comparison of different progenitor cell types detects thousands of differential regions, with enrichment of motifs of key transcription factors f. GFP reporter showing activity of ONE specific region in early pupal timepoints, followed by a decrease in signal. 30
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Methods
Data reporting
No statistical methods were used to predetermine sample size. The experiments were not randomized and the investigators were not blinded to allocation during experiments and outcome assessment. Genetics
Flies were raised on a yeast-based medium at 25°C on a 12h/12h day/night light cycle. All Drosophila lines used in the single-cell ATAC-seq experiments are derived from the DGRP collection. One hybrid was created by crossing different DGRP lines, generating genetic diversity. Drosophila line D. melanogaster: DGRP-551
D. melanogaster: DGRP-409
D. melanogaster: DGRP-502
D. melanogaster: DGRP-639
Source Bloomington Drosophila Stock Center
Bloomington Drosophila Stock Center
Bloomington Drosophila Stock Center
Bloomington Drosophila Stock Center
Identifier BDSC: FlyBase: FBsn0000297 BDSC: FlyBase: FBsn0000111 BDSC: FlyBase: FBsn0000124 BDSC: FlyBase: FBsn0000141
55026;
28278;
28204;
25199;
Genetics Wild type
Wild type
Wild type
Wild type
D. melanogaster: (DGRP-551, DGRP-907, DGRP-913) D. melanogaster: MB371B
Hybrid DGRP-360,
This lab
Bloomington Drosophila Stock Center
BDSC: 68383
Wild type
w[1118]; Py[+t7.7]
w[+mC]=R13F02- p65.ADattP40;
Py[+t7.7]
w[+mC]=R85D07- GAL4.DBDattP2
D. melanogaster: MB418B
Bloomington Drosophila Stock Center
BDSC: 68322
w[1118];
Py[+t7.7]
w[+mC]=R26E07- p65.ADattP40/CyO
Py[+t7.7]
w[+mC]=R30F02- GAL4.DBDattP2
D. melanogaster: MB419B
Bloomington Drosophila Stock Center
BDSC: 68323
w[1118];
Py[+t7.7] w[+mC]=R26E07- p65.ADattP40/CyO:
Py[+t7.7]
w[+mC]=R39A11- GAL4.DBDattP2
D. melanogaster: GFP.nls14
UAS-
Bloomington Drosophila Stock Center
BDSC: 4775
w[1118];
Pw[+mC]=UAS- GFP.nls14
D. melanogaster: GFP.nls8
UAS-
Bloomington Drosophila Stock Center
BDSC: 4776
w[1118];
Pw[+mC]=UAS- GFP.nls8
31
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Nuclei isolation
D. melanogaster adult brains were dissected and transferred to a tube containing 100 µl ice cold DPBS solution. After centrifugation at 800 g for 5 min, the supernatant was replaced by 500 µl nuclei lysis buffer composed of 10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% Tween-20, 0.1% Nonidet P40, 0.01% Digitonin, and 1% BSA, in Nuclease-free water. The following procedure was followed to extract the nuclei from the brain tissue: incubation in nuclei lysis buffer on ice for 5 min, transfer to a dounce tissue grinder tube (Merck), 25 strokes with pestle A, incubation on ice for 10 min, 25 strokes with pestle B.
|
The lysis was stopped by adding 1 ml of wash buffer composed of 10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0.1% Tween-20 and 1% BSA, in Nuclease-free water. Nuclei were pelleted by centrifugation at 800 g for 5 min at 4°C and resuspended in 1x Nuclei Buffer (10x Genomics). Nuclei suspensions were passed through a 40 µM Flowmi filter (VWR Bel-Art SP Scienceware). Nuclei concentration was assessed by the LUNA-FL Dual Fluorescence Cell Counter. 10x Genomics
Single-cell libraries were generated using the GemCode Single-Cell Instrument and Single Cell ATAC Library & Gel Bead Kit v1 and ChIP Kit (10x Genomics, US). Briefly, fly brain single nuclei were suspended in 1x nuclei buffer. The single nuclei were incubated for 60 min at 37°C with a transposase that fragments the DNA in open regions of the chromatin and adds adapter sequences to the ends of the DNA fragments. After generation of nanoliter-scale Gel bead-in-Emulsions (GEMs), GEMS were incubated in a C1000 Touch Thermal Cycler (Bio Rad) programmed at 72°C for 5 min, at 98°C for 30 s, 12 cycles of (98°C for 10 s, 59°C for 30 s, 72°C for 1 min), and held at 15°C. After incubation, single-cell droplets were broken and the single-strand DNA was isolated and cleaned with Cleanup Mix containing Silane Dynabeads. Illumina P7 sequence and a sample index were added to the single- strand DNA during library construction via PCR: at 98°C for 45 s, 11-13 cycles of (98°C for 20 s, 67°C for 30 s, 72°C for 20 s), 72°C for 1 min, and hold at 4°C. The sequencing-ready library was cleaned up with SPRIselect beads. Sequencing
Before sequencing, the fragment size of every library was analyzed on a Bioanalyzer high-sensitivity chip. All 10x scATAC libraries were sequenced on NextSeq500 and NovaSeq6000 instruments (Illumina) with the following sequencing parameters: 50 bp read 1 – 8 bp index 1 (i7) – 16 bp index 2 (i5) – 49 bp read 2. Omni-ATAC-seq of FAC-sorted samples
100 GFP-expressing (MB371B, MB418B, MB419B crossed with UAS-nls.GFP) and 15 GFP negative (wild type) fly brains were dissected in PBS on ice. The brains were then centrifuged at 800 g for 5 min, after which the supernatant was replaced by 50 μL of dispase (3 mg/mL, Sigma-Aldrich_D4818-2mg), 75 μl collagenase I (100 mg/mL, Invitrogen_17100-017), and 125 μL trypsin-EDTA (0.05%, Invitrogen_25300054). Brains were dissociated at 25°C in a Thermoshaker (Grant Bio PCMT) for 15 min at 25°C at 1,000 rpm and the solution was mixed by pipette every 5 min. After, cell suspensions were passed through a 10 μM pluriStrainer (ImTec Diagnostics_435001050) and viability was assessed by the LUNA-FL Dual Fluorescence Cell Counter. Next, 4 aliquots were made containing GFP-brains cells with/without PI (10%) and GFP-positive brains with/without PI (10%). FACS was performed on the FACS Aria III (BD Biosciences, US). The GFP-negative brains were used to set the gates on the machine for cell size and viability (PI), the GFP-positive brains for the GFP fluorescence, after which the GFP positive cells with PI were sorted (see Supplementary Data 1).
|
Between 2.6 and 11k GFP positive cells were sorted and 50k cells per negative control. After sorting, regular omni-ATAC-seq was performed as described by Corces et al. 38. 32
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Immunohistochemistry
For immunofluorescence, brains were dissected and transferred to a tube containing 100 μl ice cold DPBS solution. After centrifugation at 800 g for 5 min, the supernatant was replaced by 4% Formaldehyde in PBT 0.3% (DBPS + 0.3% Triton X-100 (Sigma)) and incubated at room temperature with rotation for 15 min. Brains were washed with PBT 0.3% three times, rotating for 10 min at room temperature each time and then blocked in Pax-DG (10 g BSA (Sigma), 3g Deoxycholate Acid (Sigma), 3ml Triton X-100 (Sigma), 50ml Normal Goat Serum (MP Biomedicals), 100ml 10X PBS, 850ml H2O) for 2 hours at room temperature with rotation. Primary antibody mixes were created in Pax-DG (dilutions detailed in the Key Resources Table) and brains were incubated in these mixes overnight at 4°C with rotation. The next day, brains were washed with PBT 0.3% three times, rotating for 10 min at room temperature each time and then stained with secondary antibody mixes in Pax-DG (dilutions detailed in the Key Resources Table) for 2 hours at room temperature with rotation. Brains were washed with PBT 0.3% three times, rotating for 10 min at room temperature each time and mounted in Mowiol mounting medium (Sigma). Imaging was performed using Nikon C2 and Nikon A1 confocal microscopes. Antibody Mouse monoclonal anti-Bruchpilot (1:50 dilution) Rabbit polyclonal anti-GFP (1:1000 dilution) Donkey polyclonal anti-rabbit Alexa Fluor 488 (1:1000 dilution) Donkey polyclonal anti-mouse Alexa Fluor 555 (1:1000 dilution) Goat polyclonal anti-rat Alexa Fluor 647 (1:1000 dilution) Mouse monoclonal anti-repo (1:20 dilution)
Source Developmental Studies Hybridoma Bank Life Technologies
Life Technologies
Life Technologies
Life Technologies
Developmental Studies Hybridoma Bank
Identifier Cat# nc82; RRID: AB_2314866 Cat# A-6455; RRID: AB_221570 Cat# A-21206; RRID: AB_2535792 Cat# A-31570; RRID: AB_2536180 Cat# A-21247; RRID: AB_141778 CAT# 8D12; RRID: AB_528448
10x data processing
The 10x fly brain samples were each processed (alignment, barcode assignment and UMI counting) with CellRangerATAC 1.2.0 count pipeline. The Cell Ranger reference index was built upon the 3rd 2017 FlyBase release (D. melanogaster r6.16) 100. Sequencing saturations were calculated based on Michaelis-Menten kinetics and early pupal timepoints were additionally sequenced and CellRanger aggr was used to aggregate sequencing results. Demuxlet
We used Demuxlet 101 to demultiplex the different genotypes that were used in the DGRP-mixed samples, allowing us to remove doublets of two different genetic backgrounds.
|
The vcf file of the DGRP project (available at http://dgrp2.gnets.ncsu.edu/) was lifted over to dm6 genome and SNPs for DGRP-409 and DGRP-502 were extracted. For DGRP-639 and the DGRP-551 based hybrid we performed bulk ATAC-seq to generate updated SNP profiles. After combining all SNPs, we only retained SNPs that were unique for one line. This vcf file was then used in Demuxlet with default parameters leading to the identification and removal of 43,489 doublets (details in Extended Data Table 1). scATAC topic modelling
After removing doublets, we performed some extra QC filters to select the 240,919 cells that will be used in upcoming analyses (Signac's nucleosome_signal <= 10, global blacklist_ratio < 0.05 and non- outlier blacklist ratio within its own run, number of fragments between 100 and 50k). 33
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. To run cisTopic 34, we created the cell-counts matrix using 129,109 pre-defined regulatory regions (ctx) based on conservation 102 (i.e. counted fragments within these regions). Given the large size of our dataset, we implemented WarpLDA 103 within the cisTopic package as a faster and more efficient alternative to Collapsed Gibbs Sampling (CGS). WarpLDA uses delayed update approach, meaning that topic-region and cell-topic distributions are updated after a number of assignments rather than after each assignment, reducing the number of calculations and memory access. This new faster algorithm is now available in cisTopic version 3 (https://github.com/aertslab/cisTopic). We performed topic modeling on the whole matrix (with between 2 and 500 topics, with 500 iterations, and finally selecting the model with 500 topics). This analysis was used to obtain an overview of the whole dataset, and to perform the analyses across development. However, we noticed that we obtained slightly better region accessibility predictions, and higher clustering resolution, when analyzing subsets of the dataset (e.g. the T4/T5 split is not detected in this global analysis, the TfAP-2 enhancers are not predicted as differential). Therefore, we used independent cisTopic runs to perform the analysis of the adult cell types (including adult and 72h APF, using 200 topics), and developmental stages (larva and 0-12h APF, 200 topics). This split of stages was chosen based on their similarity (e.g. Extended Data Fig. 1c,d). Adult cell clusters were defined based on a two-level analysis: (1) First, on the "Adult + 72h APF" cisTopic run, we clustered the adult cells –with more than 900 fragments in peaks (FIP)– using Louvain- clustering on the cell-topic probability matrix (igraph::cluster_louvain, parameters: k=10, eps=0.1, treetype="bd"). This led to 55 clusters including the main cell types identified in scRNA-seq. Note that we chose this strategy based on several alternative analyses, in which we observed that cisTopic benefits by higher numbers of cells, even if some of them have few reads, while the clustering of only high-count cells (FIP> 900) provided more stable clusters and more concordant with the scRNA-seq.
|
(2) The same process was then applied to each of the major groups of cells: OL, CB and glia, using separate cisTopic runs, and consensus peaks instead of the ctx pre-defined regions (see section below). These sub-clustering analyses provided 130 clusters, which might be over-clustered, as many of them were not matched to scRNA-seq clusters, but it allowed to identify some extra cell types (e.g. ab-cd split of T4/T5 cells, Extended Data Fig. 6). From these analyses, after the scRNA-seq label transfer (see below), we finally selected 79 clusters as main annotation. The clusters for the developmental stages were determined following the equivalent approach on
the "Larva to 12h APF" cisTopic analysis (in this case only one level of clustering was required). Gene accessibility matrix
Gene accessibilities were calculated using the cisTopic probabilities of region accessibility per cell. Next, ctx regions inside the gene body and up to 5kb of its transcription start site were selected. An exponentially decaying function was used to assign distance weights to these regions, were regions further away from the gene have lower weights, similar to ArchR 104. To give higher weights to variable regions, we calculated Gini scores per region, where highly variable regions have a high Gini score. Gini scores were then z-standardized and used as exponent for the variability weight. Final weights were defined as the product of the distance and variability weights, and a weighted sum was calculated to acquire a gene accessibility matrix. 𝑤" = 𝑒
%& ’((( + 𝑒%+
𝑤, = 𝑒-./0/
34
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. scRNA clustering
We used the scRNA-data from the whole ageing fly brain from Davie et al.1, this time using all data from all protocols with updated analysis methods (mostly batch effect correction). Mapping, filtering, normalization, batch effect correction, clustering, marker gene detection and gene regulatory network inference were performed using the VSN pipeline (https://github.com/vib-singlecell-nf/vsn- pipelines), which is a Nextflow DSL2 pipeline using CellRanger (10x), Scanpy 105, Harmony 106 and pySCENIC 45. We used the command nextflow -C nextflow.config run vib-singlecell-nf/vsn- pipelines -entry harmony and a description of the nextflow.config file can be found in Supplementary Data 2. Finally, annotations were transferred from the Davie et al. dataset, by calculating the adjusted rand index between annotations and the different calculated clusterings. The best matching clustering was Leiden resolution 10 (224 clusters) and the clusters were annotated if at least 25% of cells in the cluster had the same annotation. If there was no match, the cluster was retained, ending up with 203 clusters. One modification was made to cluster 15 where a higher resolution (Leiden 12) was chosen in which it split in two (a and b), matching the split detected in the RNA-ATAC co-embedding (see below), leading to the final annotation of 66 clusters out of 204.
|
Subsequently, marker genes were calculated in Seurat’s FindAllMarkers using the Wilcoxon method with min.pct =0.1 and logfc.threshold =0.2). Adult cell type annotation: Label transfer from RNA to ATAC (NNLS, AUCell, Seurat)
To assign cell type identities to scATAC-seq clusters (130) we followed three approaches:
First, we used the non-negative least squares method to compare clusters across modalities, similar to what was used in Domcke et al. 20. We calculated average RNA expression profiles per cell type from the annotated scRNA-seq data and averaged gene accessibility profiles for the scATAC-seq clusters using the top 10 marker genes per cell type as features (sorted by adjusted p-value (Bonferroni corrected)). These were then used as input for the algorithm in which an optimal weighted sum is calculated and the weights resemble cluster similarities. Secondly, we used AUCell 35 to score gene signatures per cell type based on the top marker genes on the gene accessibility matrix. Gene signatures were then averaged per cluster and clusters were assigned to cell types based on their score. Thirdly, Seurat v3 107 was used to integrate the gene accessibility and gene expression data. First separate objects were created for scATAC-seq and scRNA-seq data, with the gene accessibility matrix used as “RNA” assay and the region-accessibility as “peaks” assay in the ATAC-seq object. First, the dimensions of the ATAC-seq object were reduced using RunTFIDF, FnidTopFeatures and RunSVD using latent semantic analysis (LSI) on the “peaks” assay with number of components used 50, 70 and 100. Next, the RNA-seq object was log-normalized with NormalizeData with the median of expressed UMIs as scale factor. FindVariableFeatures was used to find 2500 variable features in the RNA-seq data, to be used as features for integration. Anchors for integration were identified using FindTransferAnchors with the RNA-seq data as reference and the ATAC-seq object as query using canonical component analysis using 50, 70 and 100 components. Then TransferData was used to transfer annotations from scRNA-seq to scATAC-seq using the LSI weights for weight reduction and dimensions ranging 2 to 100. To calculate a co-embedding, we used GetAssayData on the variable genes to get the RNA counts and used this as reference data in a second run of TransferData where we impute RNA counts for the ATAC-seq data, again using the LSI weights for weight reduction. The two objects were subsequently merged, followed by scaling of the data, PCA and UMAP and tSNE calculation (ScaleData, RunPCA, RunUMAP, RunTSNE). Next, we collapsed annotations across all methods and merged non-annotated low-confidence clusters. Tm1/TmY8 and Mi1 matched to the same cluster, but we could separate the sub clusters based on gene accessibility scores of bsh and hth, two markers for Mi1 neurons. 35
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder.
|
All rights reserved. No reuse allowed without permission. omniATAC-seq
Bulk ATAC-seq was performed on 5 samples (three GFP-positive cells from driver lines each targeting one subtype of Kenyon cells (MB371B, MB418B and MB419B) and two negative controls (GFP- negative cells from MB371B and MB419B)). ATAC-seq reads were trimmed using fastq-mcf 108 and a list of sequencing primers. The cleaned reads were then used as input for fastqc for quality control. Next, the reads were mapped to the 3rd 2017 FlyBase release (D. melanogaster r6.16) genome using STAR and SAMtools was used to sort the bam file. Macs2 was then used to call differential peaks between the positive samples and their negative controls and both negative controls were used for the positive sample without its own control using macs2 callpeak -t pos_sample -c neg_sample -g dm –nomodel. Differentially accessible regions and motif analysis
For each of the adult cell clusters (including both clustering resolutions, plus a super-clustering of glia, OL/CB neurons and KCs), we calculated the differentially accessible regions (DARs) based on the predictive distribution from cisTopic (using Wilcoxon rank sum test, run through FindMarkers function in Seurat with the default settings, except logfc.threshold, which was lowered to 0.20, and max.cells.per.ident, which was adjusted to balance the contrasts in some of the analyses). For each of the clusters, the DARs were calculated versus the closest cluster in the tree, and versus all the other clusters in each of the two analyses (i.e. each cluster was compared to rest of the brain, and to the other cells in their same glia/OL/CB/KC category). We then performed motif enrichment analyses for each of the DAR sets with at least 10 regions using RcisTarget (aucMaxRank=0.01 and 0.05, motif collection version 9, and the TF ChIP-seq database). Each of the analyses was performed comparing versus the whole genome (default settings), and using all the regions in topics as background (re-ranking the database). The results from these analyses are available in the website (http://flybrain.aertlab.org). Cistromes
For building cistromes, we focused on cell types linked to a scRNA-seq cluster, re-grouping the CB clusters into CB-Pros and CB-Imp in order to be able to establish the link to their transcriptome (T4 and T5 cells were analyzed as independent clusters from ATAC, but both mapping to the same T4/T5 RNA cluster). For each cell type, the cistromes were built based on the motif enrichment analysis of up-regulated DARs sets with at least 10 regions. Each significantly enriched motif (NES>=3) was annotated to expressed TFs based on cisTarget’s “direct” and “inferred by orthology” annotations (considering as expressed TFs those with expression > 0 in at least 10% of the cells of the given type/cluster). Note that since cisTarget’s annotation include some non-TFs DNA binding proteins, we only kept the 459 TFs listed as such on Flybase and GO MF annotation. For each of those motifs enriched in a DAR-set for a cell type in which the TF is expressed, we retrieved the DARs in which the motif has a significantly high score (i.e.
|
leading edge, see 102, using RcisTarget::getSignificantRegions). The dot-heatmap in Figure 2 shows the average TF expression by cell type (i.e. average of all the cells in the cluster, after normalizing each cell based on its total counts) with max normalization (each gene divided by its maximum value), and the NES of the highest scoring motif (NES capped to 8). 36
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Gene-enhancer links
We calculated the enhancer-to-gene links using the 43 matched clusters between RNA and ATAC plus CB-Pros and CB-Imp (ATAC T4 and T5 clusters were merged to match T4/T5 in RNA). For each cluster and data modality, 200 pseudocells were created as a bootstrap of 5 cells of the cell type. Each transcriptome pseudocell was then matched to a chromatin pseudocell of the same cell type to calculate the Pearson correlation and Random Forest regression (GENIE3) between each gene’s expression and the predicted accessibility (cisTopic cell-region probability) of the regions within 50kbp of its longest transcript (50kbp upstream the TSS and 50kpb downstream the end, plus the introns). The GENIE3 scores were filtered using the Binarize::binarize.BASC function in R. We then created a score, based on the aggregated ranking of these two measures plus the region accessibility, to allow to select the top regulators. The maximum value of this score was scaled to 1000 for compatibility with UCSC Genome browser (where we suggest a threshold of 600 or 800 for link visualization). For the comparison of links between BEAF-32 peaks, we used ChIP-seq on whole Drosophila embryo dataset (mixed https://www.encodeproject.org/files/ENCFF704WGH/ 109). The peaks were filtered based on the enrichment of the BEAF-32 motif (i-cisTarget analysis with default settings), and their accessibility in the adult fly brain (most of the peaks are ubiquitous across cell types, see Extended Data Fig. 7). We then defined the BEAF-32 based search space for each gene, taking the biggest transcript, and extend (up- and down-stream) until the first BEAF32 peak within 200kbp (skipping the 500bp around the TSS). In case there are no peaks within 200kbp, 50kbp is kept as search space. In 82% of the eGRNs, there is a slightly higher GSEA enrichment score with the TF co-expression module (see below) when using only links within BEAF-32 peaks. sex
embryo
of
0-14
hours;
ENCODE
eGRN integration
The regions in each of the cell-type specific cistromes were converted to genes based on the “enhancer-gene links” with score >= 600. Note that this implies that they may include positive and negative associations. We then used GSEA to check whether each of these gene-sets is enriched in each of the TF co-expression modules from SCENIC-GRNboost (the co-expression modules were used as rankings, after adding the TF itself in the first position if the TF is in any cistrome (i.e.
|
not only on the matching TF)). Therefore, we checked the enrichment of each of the 2110 cell-type specific cistromes with at least 5 target genes (corresponding to 200 TFs) and the 199 TFs also available as co- expression module (using 5000 permutations in GSEA). For each TF co-expression module (i.e. ranking) we kept the significant cistromes for the same TF (p-value < 0.01), and selected the genes in the leading edge to build the e-GRNs. To finalize the e-GRNs per cell type, for each of those genes, we then retrieved the linked regions in the cistrome within the BEAF-32 based search space. Thus, obtaining the connections TF – Region -– Gene. eGRN plots
To display the eGRNs as networks in Cytoscape (v3.8.0; 110), we focused on the positive region-target gene links (i.e. correlation over 0.22, which corresponds to the top 1% of all calculated gene-region link correlations), and the genes expressed in at least 15% of the cells of the specific cell type (except T4/T5 neurons, for which we used 5% instead). To reduce the size of the glial networks, they were further filtered by removing TF-region connections with a scaled Cluster-Buster score below 0.25, and regions with a deep learning prediction score below 0.25. The prediction score of each region on the respective deep learning model (Figure 3b; Extended Data Table 2) was calculated with a 500 bp sliding window with a 1 bp shift, and taking the highest score. For each cistrome (i.e. TF and cell type pair), all putative target regions were scored
37
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. with Cluster-Buster 111 using all motifs that had been used to generate the cistrome. For each TF, the highest cluster score per region was kept and scaled between zero and one for the Cytoscape network. In addition, the Cytoscape networks also display differential expression and accessibility for each node (gene or region, respectively). The differential expression was calculated by contrasting the cell type versus all other cells (avg_logFC calculated with the Seurat function FindMarkers), the accessibility in the cell type was calculated by taking the mean over the interval with subsequent RPGC-normalization. Deep learning on topics
Kenyon cells, T-neurons and glia were selected from the adult and 72h APF datasets leading to a total of 17,554 cells. The selected cells were rescored on a set of 207k 150bp peaks (see consensus peaks), which were extended to 300bp for optimal resolution in deep learning. Given the smaller number of cells, we used the conventional Collapsed Gibbs Sampler method in cisTopic using runCGSModels from 1 to 100 topics, with 500 iterations using 250 as burn-in. With selectModel we selected the model with the highest log-likelihood leading to 81 topics. Using runtSNE without PCA on the probability matrix with the cells as target, we acquired the 2D embeddings.
|
We then calculated scores for the topics per region using getRegionsScores with method='NormTop' and scale=TRUE. Finally, we used binarizecisTopics with thrP=0.975 to get 81 sets of peaks. These region sets were annotated to the different cell types based on accessibility per cell type and region features (e.g. promotors, BEAF-32) based on motif enrichment and the annotateRegions function using the Drosophila datasets. These sets of regions were then used as input for a deep learning model, where 500bp DNA sequence are used to predict the topic set to which the region belongs. The architecture of the model was used from an earlier study where they again used the cisTopic clusters as an input for the deep learning model (DeepMEL) 54. The model is a hybrid CNN-RNN multi-class classifier 112, with model architecture details in Extended Data Table 2. In addition to the architecture proposed earlier, we increased the number of filters from 128 to 1024 where 747 of them are initialized as known PWMs representing 212 TFs. Input regions were split into training (80%), validation (10%), and test (10%) sets. The model was trained on the training set, while the validation set was used to do early stopping and to select the best epoch (83rd) as a final model to use. In order to find the nucleotides that are contributing the most for the topic prediction, we used a network explaining tool called DeepExplainer 55. The tool was initialized with 500 random sequences and default parameters were used. The importance score obtained from the DeepExplainer analysis was multiplied by the DNA sequence and visualized as height of the nucleotide letters as in earlier work 113. On top of DeepExplainer plots, we performed in-silico saturation mutagenesis where we calculate the effect of each variant of a region on its model prediction score. The sequences with all possible single mutations were generated and delta model score for each topic was calculated. High nucleotide importances on DeepExplainer plots represent potential binding sites for TFs. We used TF-MoDISco 56 to identify the most common patterns for KCs (Topic 21, 35, 77), T-neurons (Topic 23 ,20, 44, 10, 18, 32), and glia (Topic 68, 25, 56, 34, 36). Default parameters were used to run for each group. For conservation study, nucleotides with a DeepExplainer absolute z-standardized importance in Kipoi larger (https://kipoi.org/models/DeepFlyBrain). than 3 were
selected. The DeepFlyBrain model
is deposited
Cloning and visualization of enhancers
59 enhancers were chosen based on eGRNs, DARs and development, and the ATAC-seq peak in the targeted region in the cell type of interest was selected. Selected enhancers were scored for the
38
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. presence of homopolymers (>10) and GC content and small modifications to the sequence were made if needed.
|
Sequences were ordered at Twist Biosciences (US), and inserted in the pTwist ENTR vector. Gateway cloning was then used to insert the sequence in the pH-Stinger vector containing nuclear GFP, Hsp70 promotor and gypsy insulators 114. Next, the vectors were sent to FlyORF (CH) and divided into 6 pools that were injected in Drosophila embryos. Positive transformants were selected, and PCR was used to determine the identity of the enhancer in each line. This pipeline of pooled injections recovered a transgenic line for 54 of the 59 enhancers. Larval, pupal (24h and 48h) and adult flies were then dissected and stained using the immunohistochemistry protocol for GFP, brp, repo and DAPI. Enhancers were scored using the following system: high off target expression (-2), low off target expression (-1), no expression (0), low on target expression (1) and high on target expression (2). Results can be found in Extended Data Table 4. Tests on the success rate were performed using two- sided Fischer’s exact test. Enhancer ROC curves
We used the scikit-learn 115 framework to fit a roc-curve to separate adult high quality enhancers (score=2) from the other cloned enhancers. As features we used peak height, peak specificity (Z-score, log-fold change, p-value of DAR), motif content from deep learning, DAR and/or eGRN membership and correlation of peak accessibility with gene expression. Annotation of cell types through development
The annotation of cell types through development was performed following two complementary approaches: (1) annotate progenitor cell types based on marker genes near ase, dpn, grh, dac, cas and scro 15 (Extended Data Fig.13 a) and the ventral nerve cord based on abd-A 15, and (2) tracking back the annotated adult cell types. To track back the adult cell types though development we used a Support Vector Machine classifier (SVM): We trained it on the annotated adult cell types, and we used it to iteratively transfer the labels to earlier stages. 1. In the first step, we used the SVM classifier to transfer the labels from the 79 adult clusters (adult cells with more than 900 FIP), to the remaining cells on the adult dataset (adult + 72h APF, the classifiers are trained on the cell-topic matrix). Using cross validation within the adult cells, we estimated that the global accuracy of the classifier is 0.86, with a call rate of 0.97 (it is not forced to assign a class to every cell); having a specificity of over 0.99 for all cell types, and a sensitivity ranging from over 0.90, for many glial, optic lobe and Kenyon cell types, to 0.25-0.50 for the least confident CB-Pros clusters. 2. We then used the adult+72h APF cells to classify the 48h pupa cells using the common cisTopic analysis with these three stages (Extended Data Fig. 2), and Harmony (on the cell-topic matrix) to reduce the effect differences intrinsic to the developmental stage. 3. Finally, we classify the cells in the remaining developmental stages (from larva to 24h AFP). For this we used the global cisTopic analysis (158,116/240,919 cells with more than 900 FIP), with Harmony to correct for developmental stage (Extended Data Fig.
|
1e). In this last training, we noticed that cells on the progenitor clusters remained largely unassigned, so we finally trained a classifier also including the progenitors as training labels (OL Developing neuron 2, CB Developing neuron 1, OL Neuroepithelium, NB Generation, OL Developing neuron 1, OL Type I NB, CB Type I NB, and LPC), and discarding from the training set the few cells from the "new 48h cluster" that had been assigned to a cell type (they seem to be younger cells, and could distort the classification). This way obtained a likely fate for the developmental cells. 39
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Core-set identification
The peaks called per cell type per timepoint for the consensus peaks were used as the basis to identify core regions per cell type. Ctx regions (see topic modelling) that overlapped with the called peaks for that timepoint, were defined as open and the DARs of the timepoint for the cell type were used to get differential accessible regions. All ctx regions that passed the filtering were then taken together as one set of total accessible regions of the cell type. Regions that are accessible in every timepoint are defined as the core-set of regions, with regions that are differentially accessible in every timepoint as core-DARs. Trajectory of Optic Lobe branches
We used Monocle3 116–118 to fit a trajectory in the 3D UMAP of the optic lobe branch from the larval to 12h APF analysis, and assign pseudotimes to the cells. First, we created a cell_data_set with the region probabilities per cell and the 3D UMAP from cisTopic as embeddings. Then we used cluster_cells followed by partition to separate the optic lobe and central brain branches and subset the object to only contain the optic lobe. We performed another cluster_cell, and selected and merged clusters in the same branch. The branch IDs were then used in Seurat’s FindAllMarkers (Wilcoxon, min.pct=0.1, logfc.threshold=0.25) on the predictive distribution matrix to find differentially accessible regions. Subsequently, motif enrichment on the branch DARs was performed using i- cisTarget 119. Regions were linked to genes up to 5kbp up- or downstream and GO was performed using FlyMine (Extended Data Table 5). Trajectory of ONE scATAC-seq
The Monocle3 object created for optic lobe branches was also used to calculate pseudotime using learn_graph fit a principal graph and order_cells to assign pseudotimes. Next, optic lobe neuroepithelium cells were selected together with the tips of the lamina precursor cells and optic lobe neuroblasts (NB generation), focusing on the trajectory between these cell types. The trajectory was split into 15 equal parts that were used in Seurat’s FindAllMarkers (Wilcoxon, min.pct=0.1, logfc.threshold=0.1) to find differentially accessible regions with a two-sided Wilcoxon test.
|
Next, the predictive distribution matrix was subset for DARs and CPM normalized, followed by region-based z- normalisation. DARs were grouped into modules using hierarchical clustering with the Scipy cluster.hierachy module 120 using distance.pdist (Euclidean), linkage (complete) and fcluster (0.85*max distance) leading to 9 modules. RcisTarget was used to identify motifs per module. Trajectory of ONE scRNA-seq
Lamina precursor cells, neuroepithelium cells and optic lobe neuroblasts were selected from a scRNA-seq dataset of the larval brain 15. Monocle3 was used to create a trajectory through the cells and assign pseudotimes. First the data was processed using pre_process_cds with principal component analysis as method, selecting 20 components. Next a batch effect correction was performed to align the two different runs with align_cds. The aligned data was then used for reduce_dimension, followed by learn_graph. Once the principal graph was learned, cells were ordered along it and pseudotimes were assigned. To plot gene expression trajectories over pseudotime, a rolling mean was calculated of the log-normalised CPM counts with a window of 10. Next a 10th degree polynomial was fit through the rolling mean with polyfit using NumPy 121 and plotted. Central brain pros vs Imp
Central brain clusters in the Adult+72hAPF dataset were selected based on enrichment of central brain only runs (Extended Data Fig. 2), with the exception of Kenyon cells. These clusters were assigned to either pros or Imp groups based on their maximal mean gene accessibility. We then used (Wilcoxon, min.pct=0.1, Seurat FindAllMarkers on
the predictive distribution matrix
40
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. logfc.threshold=0.2) to identify 166 regions for pros+ cells and 128 regions for Imp+ cells. Motif enrichment was performed using i-cisTarget. scATAC-seq embryo
We used scATAC-seq from the whole Drosophila embryo 58 to map the different central brain cell types. After data download from GEO, we used cisTopic to map the reads on ctx regions leading to 128,510 regions by 20,594 cells matrix. Given the smaller number of cells, we used the conventional Collapsed Gibbs Sampler method in cisTopic using runCGSModels from 1 to 100 topics, with 500 iterations using 250 as burn-in. We selected the model with the highest log-likelihood leading to 50 topics. Using runtSNE without PCA on the probability matrix with the cells as target, we acquired the 2D embeddings. Annotations were transferred from the dataset, identifying the CNS. We then plotted the mean accessibility of the central brain regions on the tSNE. Enhancer-switch identification
Region accessibilities per cell type were calculated per timepoint using RPGC normalized bigwig files. Next, a linear curve was fit using statsmodels 122 in Python for every region using time as independent variable and region accessibility as dependent variable with 95% confidence intervals calculated for the parameters.
|
Regions with a positive coefficient were assigned to be upregulated and regions with negative coefficients were assigned to be downregulated. Finally, we selected the regions that were upregulated in one cell type while being downregulated in another one, leading to 985 switching regions. Cell type specific bams and bigwigs
We used the annotations calculated by the SVM and extracted cells per cell type per timepoint. Next, we subset the bam files from the runs to only contain reads belonging to the selected cells and created a cell type specific bam file. Then we used SAMtools 123 to remove duplicates (view -F 0x400) and remove regions mapping blacklisted regions 124. This was then used as input for the bamCoverage function from deepTools 125 to create an RPGC normalized bigwig file with the following parameters: -bs 1 -p 8 --normalizeUsing RPGC --effectiveGenomeSize 142573017. Consensus peaks
MACS2 126 was used to call peaks on cell-type specific bam files using the call peak function with following parameters: macs2 callpeak -q 0.05 -g dm --keep-dup all --nolambda --call- summits --nomodel --shift -75 --extsize 150. This was repeated for all the timepoints and for the grouped analyses (Adult+P72, L3-P12, P24, P48). Next, the summits were extended to 500bp (or 150bp) using slopBed from BEDTools (-l 149, -r 150 (or -l 74, -r 75). The extended summits were then merged according to ENCODE standards, with first a normalization of the summit score (CPM) followed by iteratively peak merging until non-overlapping peaks across all timepoints and cell types are retained. This led to a final number of 95,921 (500bp) and 207,325 (150bp) peaks. Code availability
The updated version of cisTopic for scATAC-seq clustering and topic identification including warpLDA can be found at https://github.com/aertslab/cisTopic with set-up instructions and tutorial. Nextflow pipeline for scRNA-seq analysis can be found at https://github.com/vib-singlecell-nf/vsn- pipelines together with example config files and instructions. The DeepFlyBrain model is deposited in Kipoi (https://kipoi.org/models/DeepFlyBrain). Enhancer gene links can be calculated using ScoMAP (https://github.com/aertslab/ScoMAP) and GENIE3 (https://github.com/aertslab/GENIE3). Trajectory analysis was performed using Monocle3 following the package tutorials (http://cole-trapnell- lab.github.io/monocle-release/monocle3). Differential expression, accessibility and integration of RNA- and ATAC-seq was performed using Seurat v3 (with vignettes and install instructions at
41
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Code https://satijalab.org/seurat/). https://github.com/aertslab/FBD_App/. for
the
website
is
available
at
Data availability
The data generated for this study have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE163697.
|
We also provide a dedicated website to browse the results of the analyses and processed data (https://flybrain.aertslab.org), which provides hub link-outs (http://genome.ucsc.edu/cgi- bin/hgTracks?db=dm6&hubUrl=http://ucsctracks.aertslab.org/papers/FlyBrain/hub.txt), the eGRNs in NDEx, the DeepExplainer plots of enhancers, and other information. The following publicly accessible datasets were also used: GSE107451 (scRNA-seq adult brain), GSE157202 (scRNA-seq larval brain), GSE101581 (scATAC-seq embryo). to
the
SCope
session
(http://scope.aertslab.org/#/Fly_Brain/), UCSC
42
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. References
1. Davie, K. et al. A Single-Cell Transcriptome Atlas of the Aging Drosophila Brain. Cell 174, 982-
998.e20 (2018). 2. Konstantinides, N. et al. Phenotypic Convergence: Distinct Transcription Factors Regulate
Common Terminal Features. Cell 174, 622-635.e13 (2018). 3. Croset, V., Treiber, C. D. & Waddell, S. Cellular diversity in the Drosophila midbrain revealed by
single-cell transcriptomics. eLife 7, e34550 (2018). 4. Özel, M. N. et al. Neuronal diversity and convergence in a visual system developmental atlas. Nature 1–8 (2020) doi:10.1038/s41586-020-2879-3. 5. Kurmangaliyev, Y. Z., Yoo, J., Valdes-Aleman, J., Sanfilippo, P. & Zipursky, S. L. Transcriptional
Programs of Circuit Assembly
in
the Drosophila Visual System. Neuron
(2020)
doi:10.1016/j.neuron.2020.10.006. 6. Costa, M., Manton, J. D., Ostrovsky, A. D., Prohaska, S. & Jefferis, G. S. X. E. NBLAST: Rapid,
Sensitive Comparison of Neuronal Structure and Construction of Neuron Family Databases. Neuron 91, 293–311 (2016). 7. Rivera-Alba, M. et al. Wiring economy and volume exclusion determine neuronal placement in
the Drosophila brain. Curr. Biol. 21, 2000–2005 (2011). 8. Scheffer, L. K. et al. A connectome and analysis of the adult Drosophila central brain. eLife 9,
(2020). 9. Xu, C. S. et al. A Connectome of the Adult Drosophila Central Brain. bioRxiv 2020.01.21.911859
(2020) doi:10.1101/2020.01.21.911859. 10. Zheng, Z. et al. A Complete Electron Microscopy Volume of the Brain of Adult Drosophila
melanogaster. Cell 174, 730-743.e22 (2018). 43
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 11. Jenett, A. et al. A GAL4-Driver Line Resource for Drosophila Neurobiology. Cell Rep 2, 991–1001
(2012). 12. Robie, A. A. et al. Mapping the Neural Substrates of Behavior. Cell 170, 393-406.e28 (2017). 13. Brunet Avalos, C., Maier, G. L., Bruggmann, R. & Sprecher, S. G. Single cell transcriptome atlas of
the Drosophila larval brain. Elife 8, (2019). 14.
|
Cocanougher, B. T. et al. Comparative single-cell transcriptomics of complete insect nervous
systems. bioRxiv 785931 (2020) doi:10.1101/785931. 15. Ravenscroft, T. A. et al. Drosophila Voltage-Gated Sodium Channels Are Only Expressed in Active
Neurons and Are Localized to Distal Axonal Initial Segment-like Domains. J. Neurosci. 40, 7999–
8024 (2020). 16. Allen, A. M. et al. A single-cell transcriptomic atlas of the adult Drosophila ventral nerve cord. eLife 9, e54074 (2020). 17. Buenrostro, J. D., Giresi, P. G., Zaba, L. C., Chang, H. Y. & Greenleaf, W. J. Transposition of native
chromatin for fast and sensitive epigenomic profiling of open chromatin, DNA-binding proteins
and nucleosome position. Nat. Methods 10, 1213–1218 (2013). 18. Buenrostro, J. D. et al. Single-cell chromatin accessibility reveals principles of regulatory
variation. Nature 523, 486–490 (2015). 19. Cao, J. et al. A human cell atlas of fetal gene expression. Science 370, (2020). 20. Domcke, S. et al. A human cell atlas of fetal chromatin accessibility. Science 370, (2020). 21. Lake, B. B. et al. Integrative single-cell analysis of transcriptional and epigenetic states in the
human adult brain. Nat. Biotechnol. 36, 70–80 (2018). 22. Doe, C. Q. Temporal Patterning in the Drosophila CNS. Annu. Rev. Cell Dev. Biol. 33, 219–240
(2017). 44
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 23. Erclik, T. et al. Integration of temporal and spatial patterning generates neural diversity. Nature
541, 365–370 (2017). 24. Estacio-Gómez, A., Hassan, A., Walmsley, E., Le, L. W. & Southall, T. D. Dynamic neurotransmitter
specific transcription factor expression profiles during Drosophila development. Biology Open 9,
(2020). 25. Komiyama, T., Johnson, W. A., Luo, L. & Jefferis, G. S. X. E. From lineage to wiring specificity. POU
domain transcription factors control precise connections of Drosophila olfactory projection
neurons. Cell 112, 157–167 (2003). 26. Kurmangaliyev, Y. Z., Yoo, J., LoCascio, S. A. & Zipursky, S. L. Modular transcriptional programs
separately define axon and dendrite connectivity. Elife 8, (2019). 27. Schilling, T., Ali, A. H., Leonhardt, A., Borst, A. & Pujol-Martí, J. Transcriptional control of
morphological properties of direction-selective T4/T5 neurons in Drosophila. Development 146,
(2019). 28. Halder, G., Callaerts, P. & Gehring, W. J. Induction of ectopic eyes by targeted expression of the
eyeless gene in Drosophila. Science 267, 1788–1792 (1995). 29. Masserdotti, G., Gascón, S. & Götz, M. Direct neuronal reprogramming: learning from and for
development. Development 143, 2494–2510 (2016). 30. Hecker, M., Lambeck, S., Toepfer, S., van Someren, E. & Guthke, R. Gene regulatory network
inference: Data integration in dynamic models—A review. Biosystems 96, 86–103 (2009).
|
31. Li, Z. et al. Identification of transcription factor binding sites using ATAC-seq. Genome Biology
20, 45 (2019). 32. Li, H. et al. Classifying Drosophila Olfactory Projection Neuron Subtypes by Single-Cell RNA
Sequencing. Cell 171, 1206-1220.e22 (2017). 45
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 33. Mackay, T. F. C. et al. The Drosophila melanogaster Genetic Reference Panel. Nature 482, 173–
178 (2012). 34. Bravo González-Blas, C. et al. cisTopic: cis-regulatory topic modeling on single-cell ATAC-seq
data. Nature Methods 16, 397–400 (2019). 35. Aibar, S. et al. SCENIC: single-cell regulatory network inference and clustering. Nature Methods
14, 1083–1086 (2017). 36. Stanescu, D. E., Yu, R., Won, K.-J. & Stoffers, D. A. Single cell transcriptomic profiling of mouse
pancreatic progenitors. Physiol Genomics 49, 105–114 (2017). 37. Shih, M.-F. M., Davis, F. P., Henry, G. L. & Dubnau, J. Nuclear Transcriptomes of the Seven
Neuronal Cell Types That Constitute the Drosophila Mushroom Bodies. G3 (Bethesda) 9, 81–94
(2019). 38. Corces, M. R. et al. An improved ATAC-seq protocol reduces background and enables
interrogation of frozen tissues. Nature Methods 14, 959–962 (2017). 39. Trevino, A. E. et al. Chromatin accessibility dynamics in a model of human forebrain
development. Science 367, (2020). 40. Davis, F. P. et al. A genetic, genomic, and computational resource for exploring neural circuit
function. eLife 9, e50901 (2020). 41. Huynh-Thu, V. A., Irrthum, A., Wehenkel, L. & Geurts, P. Inferring Regulatory Networks from
Expression Data Using Tree-Based Methods. PLOS ONE 5, e12776 (2010). 42. Iacono, G., Massoni-Badosa, R. & Heyn, H. Single-cell transcriptomics unveils gene regulatory
network plasticity. Genome Biology 20, 110 (2019). 43. Jackson, C. A., Castro, D. M., Saldi, G.-A., Bonneau, R. & Gresham, D. Gene regulatory network
reconstruction using single-cell RNA sequencing of barcoded genotypes
in diverse
environments. eLife 9, e51254 (2020). 46
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 44. Matsumoto, H. et al. SCODE: an efficient regulatory network inference algorithm from single-
cell RNA-Seq during differentiation. Bioinformatics 33, 2314–2321 (2017). 45. Van de Sande, B. et al. A scalable SCENIC workflow for single-cell gene regulatory network
analysis. Nature Protocols 15, 2247–2276 (2020). 46. Crittenden, J. R., Skoulakis, E. M. C., Goldstein, E. S. & Davis, R. L. Drosophila mef2 is essential
for normal mushroom body and wing development. Biology Open 7, (2018). 47. Schulz, R. A., Chromey, C., Lu, M. F., Zhao, B.
|
& Olson, E. N. Expression of the D-MEF2
transcription in the Drosophila brain suggests a role in neuronal cell differentiation. Oncogene
12, 1827–1831 (1996). 48. Minocha, S., Boll, W. & Noll, M. Crucial roles of Pox neuro in the developing ellipsoid body and
antennal lobes of the Drosophila brain. PLoS One 12, (2017). 49. Naidu, V. G. et al. Temporal progression of Drosophila medulla neuroblasts generates the
transcription factor combination to control T1 neuron morphogenesis. Developmental Biology
464, 35–44 (2020). 50. Avet-Rochex, A., Maierbrugger, K. T. & Bateman, J. M. Glial enriched gene expression profiling
identifies novel factors regulating the proliferation of specific glial subtypes in the Drosophila
brain. Gene Expr Patterns 16, 61–68 (2014). 51. Harmston, N. et al. Topologically associating domains are ancient features that coincide with
Metazoan clusters of extreme noncoding conservation. Nature Communications 8, 441 (2017). 52. Wang, Q., Sun, Q., Czajkowsky, D. M. & Shao, Z. Sub-kb Hi-C in D . melanogaster reveals
conserved characteristics of TADs between insect and mammalian cells. Nature Communications
9, 188 (2018). 53. Yang, J., Ramos, E. & Corces, V. G. The BEAF-32 insulator coordinates genome organization and
function during the evolution of Drosophila species. Genome Res 22, 2199–2207 (2012). 47
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 54. Minnoye, L. et al. Cross-species analysis of enhancer logic using deep learning. Genome Res. gr.260844.120 (2020) doi:10.1101/gr.260844.120. 55. Lundberg, S. M. & Lee, S.-I. A Unified Approach to Interpreting Model Predictions. Advances in
Neural Information Processing Systems 30, 4765–4774 (2017). 56. Shrikumar, A. et al. Technical Note on Transcription Factor Motif Discovery from Importance
Scores (TF-MoDISco) version 0.5.6.5. arXiv:1811.00416 [cs, q-bio, stat] (2020). 57. Hubisz, M. J., Pollard, K. S. & Siepel, A. PHAST and RPHAST: phylogenetic analysis with space/time
models. Brief Bioinform 12, 41–51 (2011). 58. Cusanovich, D. A. et al. The cis -regulatory dynamics of embryonic development at single-cell
resolution. Nature 555, 538–542 (2018). 59. Bravo González-Blas, C. et al. Identification of genomic enhancers through spatial integration of
single-cell transcriptomics and epigenomics. Molecular Systems Biology 16, e9438 (2020). 60. Apitz, H. & Salecker, I. A challenge of numbers and diversity: neurogenesis in the Drosophila
optic lobe. J Neurogenet 28, 233–249 (2014). 61. Chotard, C., Leung, W. & Salecker, I. glial cells missing and gcm2 Cell Autonomously Regulate
Both Glial and Neuronal Development in the Visual System of Drosophila. Neuron 48, 237–251
(2005). 62. Endo, K. et al. Chromatin modification of Notch targets in olfactory receptor neuron
diversification.
|
Nature Neuroscience 15, 224–233 (2012). 63. Eroglu, E. et al. SWI/SNF Complex Prevents Lineage Reversion and Induces Temporal Patterning
in Neural Stem Cells. Cell 156, 1259–1273 (2014). 64. Piñeiro, C., Lopes, C. S. & Casares, F. A conserved transcriptional network regulates lamina
development in the Drosophila visual system. Development 141, 2838–2847 (2014). 48
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 65. Lee, J., Park, S.-Y. & Yoo, S. Roles of Nk2.1/scro homeobox gene in the development of optic lobe
neuroblast in Drosophila melanogaster. IBRO Reports 6, S339–S340 (2019). 66. Yoo, S. et al. Knock-in mutations of scarecrow, a Drosophila homolog of mammalian Nkx2.1,
reveal a novel function required for development of the optic lobe in Drosophila melanogaster. Developmental Biology 461, 145–159 (2020). 67. Medioni, C., Ramialison, M., Ephrussi, A. & Besse, F. Imp Promotes Axonal Remodeling by
Regulating profilin mRNA during Brain Development. Current Biology 24, 793–800 (2014). 68. Vijayakumar, J. et al. The prion-like domain of Drosophila Imp promotes axonal transport of RNP
granules in vivo. Nature Communications 10, 2593 (2019). 69. Alyagor, I. et al. Combining Developmental and Perturbation-Seq Uncovers Transcriptional
Modules Orchestrating Neuronal Remodeling. Developmental Cell 47, 38-52.e6 (2018). 70. Kirilly, D. et al. A genetic pathway composed of Sox14 and Mical governs severing of dendrites
during pruning. Nat. Neurosci. 12, 1497–1505 (2009). 71. Cheng, S. et al. Molecular basis of synaptic specificity by immunoglobulin superfamily receptors
in Drosophila. eLife 8, e41028 (2019). 72. Tan, L. et al. Ig Superfamily Ligand and Receptor Pairs Expressed in Synaptic Partners in
Drosophila. Cell 163, 1756–1769 (2015). 73. Ngo, K. T., Andrade, I. & Hartenstein, V. Spatio-temporal pattern of neuronal differentiation in
the Drosophila visual system: A user’s guide to the dynamic morphology of the developing optic
lobe. Developmental Biology 428, 1–24 (2017). 74. Ma, S. et al. Chromatin Potential Identified by Shared Single-Cell Profiling of RNA and Chromatin. Cell 183, 1103-1116.e20 (2020). 75. Cusanovich, D. A. et al. A Single-Cell Atlas of In Vivo Mammalian Chromatin Accessibility. Cell
174, 1309-1324.e18 (2018). 49
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 76. Preissl, S. et al. Single-nucleus analysis of accessible chromatin in developing mouse forebrain
reveals cell-type-specific transcriptional regulation. Nat Neurosci 21, 432–439 (2018). 77. Chen, S., Lake, B. B. & Zhang, K. High-throughput sequencing of the transcriptome and chromatin
accessibility in the same cell.
|
Nature Biotechnology 37, 1452–1457 (2019). 78. Zhu, C. et al. An ultra high-throughput method for single-cell joint analysis of open chromatin
and transcriptome. Nature Structural & Molecular Biology 26, 1063–1070 (2019). 79. Berg, O. G. & von Hippel, P. H. Selection of DNA binding sites by regulatory proteins. Statistical-
mechanical theory and application to operators and promoters. J Mol Biol 193, 723–750 (1987). 80. Crocker, J., Preger-Ben Noon, E. & Stern, D. L. Chapter Twenty-Seven - The Soft Touch: Low-
Affinity Transcription Factor Binding Sites in Development and Evolution. in Current Topics in
Developmental Biology (ed. Wassarman, P. M.) vol. 117 455–469 (Academic Press, 2016). 81. Kribelbauer, J. F., Rastogi, C., Bussemaker, H. J. & Mann, R. S. Low-Affinity Binding Sites and the
Transcription Factor Specificity Paradox in Eukaryotes. Annu Rev Cell Dev Biol 35, 357–379
(2019). 82. Scardigli, R., Bäumer, N., Gruss, P., Guillemot, F. & Le Roux, I. Direct and concentration-
dependent regulation of the proneural gene Neurogenin2 by Pax6. Development 130, 3269–
3281 (2003). 83. Koromila, T. et al. Odd-paired is a pioneer-like factor that coordinates with Zelda to control gene
expression in embryos. eLife 9, e59610 (2020). 84. Kudron, M. M. et al. The ModERN Resource: Genome-Wide Binding Profiles for Hundreds of
Drosophila and Caenorhabditis elegans Transcription Factors. Genetics 208, 937–949 (2018). 85. Ozdemir, A., Ma, L., White, K. P. & Stathopoulos, A. Su(H)-mediated repression positions gene
boundaries along the dorsal-ventral axis of Drosophila embryos. Dev Cell 31, 100–113 (2014). 50
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 86. Samata, M. et al. Intergenerationally Maintained Histone H4 Lysine 16 Acetylation Is Instructive
for Future Gene Activation. Cell 182, 127-144.e23 (2020). 87. Ye, Y. et al. Chromatin remodeling during the in vivo glial differentiation in early Drosophila
embryos. Scientific Reports 6, 33422 (2016). 88. Brás-Pereira, C. et al. dachshund Potentiates Hedgehog Signaling during Drosophila
Retinogenesis. PLoS Genet 12, (2016). 89. Dardalhon-Cuménal, D. et al. Cyclin G and the Polycomb Repressive complexes PRC1 and PR-
DUB cooperate for developmental stability. PLOS Genetics 14, e1007498 (2018). 90. Donohoe, C. D. et al. Atf3 links loss of epithelial polarity to defects in cell differentiation and
cytoarchitecture. PLoS Genet 14, (2018). 91. Jusiak, B. et al. Regulation of Drosophila Eye Development by the Transcription Factor Sine
oculis. PLOS ONE 9, e89695 (2014). 92. Newcomb, S. et al. cis-regulatory architecture of a short-range EGFR organizing center in the
Drosophila melanogaster leg. PLoS Genet 14, e1007568 (2018). 93. Schertel, C. et al. A large-scale, in vivo transcription factor screen defines bivalent chromatin as
a key property of regulatory factors mediating Drosophila wing development.
|
Genome Res 25,
514–523 (2015). 94. Yeung, K. et al. Integrative genomic analysis reveals novel regulatory mechanisms of eyeless
during Drosophila eye development. Nucleic Acids Res 46, 11743–11758 (2018). 95. Koemans, T. S. et al. Functional convergence of histone methyltransferases EHMT1 and KMT2C
involved in intellectual disability and autism spectrum disorder. PLoS Genet 13, e1006864 (2017). 96. Magadi, S. S. et al. Dissecting Hes-centred transcriptional networks in neural stem cell
maintenance and tumorigenesis in Drosophila. Development 147, (2020). 51
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 97. Ray, P. et al. Combgap contributes to recruitment of Polycomb group proteins in Drosophila. Proc Natl Acad Sci U S A 113, 3826–3831 (2016). 98. Brand, A. H. & Perrimon, N. Targeted gene expression as a means of altering cell fates and
generating dominant phenotypes. Development 118, 401–415 (1993). 99. Boll, W. & Noll, M. The Drosophila Pox neuro gene: control of male courtship behavior and
fertility as revealed by a complete dissection of all enhancers. Development 129, 5667–5681
(2002). 100. Gramates, L. S. et al. FlyBase at 25: looking to the future. Nucleic Acids Res 45, D663–D671
(2017). 101. Kang, H. M. et al. Multiplexed droplet single-cell RNA-sequencing using natural genetic variation. Nature Biotechnology 36, 89–94 (2018). 102. Herrmann, C., Van de Sande, B., Potier, D. & Aerts, S. i-cisTarget: an integrative genomics method
for the prediction of regulatory features and cis-regulatory modules. Nucleic Acids Res 40, e114
(2012). 103. Chen, J., Li, K., Zhu, J. & Chen, W. WarpLDA: a cache efficient O(1) algorithm for latent dirichlet
allocation. Proc. VLDB Endow. 9, 744–755 (2016). 104. Granja, J. M. et al. ArchR: An integrative and scalable software package for single-cell chromatin
accessibility analysis. bioRxiv 2020.04.28.066498 (2020) doi:10.1101/2020.04.28.066498. 105. Wolf, F. A., Angerer, P. & Theis, F. J. SCANPY: large-scale single-cell gene expression data analysis. Genome Biology 19, 15 (2018). 106. Korsunsky, I. et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature Methods 16, 1289–1296 (2019). 107. Stuart, T. et al. Comprehensive Integration of Single-Cell Data. Cell 177, 1888-1902.e21 (2019). 52
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 108. Aronesty et al. ea-utils: ‘Command-line tools for processing biological sequencing data’. (2011). 109. Davis, C. A. et al. The Encyclopedia of DNA elements (ENCODE): data portal update. Nucleic Acids
Res 46, D794–D801 (2018).
|
110. Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular
interaction networks. Genome Res 13, 2498–2504 (2003). 111. Frith, M. C., Li, M. C. & Weng, Z. Cluster-Buster: Finding dense clusters of motifs in DNA
sequences. Nucleic Acids Res 31, 3666–3668 (2003). 112. Quang, D. & Xie, X. DanQ: a hybrid convolutional and recurrent deep neural network for
quantifying the function of DNA sequences. Nucleic Acids Res 44, e107 (2016). 113. Shrikumar, A., Greenside, P. & Kundaje, A. Learning Important Features Through Propagating
Activation Differences. arXiv:1704.02685 [cs] (2019). 114. Aerts, S. et al. Robust Target Gene Discovery through Transcriptome Perturbations and Genome-
Wide Enhancer Predictions in Drosophila Uncovers a Regulatory Basis for Sensory Specification. PLOS Biology 8, e1000435 (2010). 115. Pedregosa, F. et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning
Research 12, 2825–2830 (2011). 116. Cao, J. et al. The single-cell transcriptional landscape of mammalian organogenesis. Nature 566,
496–502 (2019). 117. Qiu, X. et al. Reversed graph embedding resolves complex single-cell trajectories. Nature
Methods 14, 979–982 (2017). 118. Trapnell, C. et al. The dynamics and regulators of cell fate decisions are revealed by
pseudotemporal ordering of single cells. Nature Biotechnology 32, 381–386 (2014). 53
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 119. Imrichová, H., Hulselmans, G., Kalender Atak, Z., Potier, D. & Aerts, S. i-cisTarget 2015 update:
generalized cis-regulatory enrichment analysis in human, mouse and fly. Nucleic Acids Res 43,
W57–W64 (2015). 120. Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in Python. Nature
Methods 17, 261–272 (2020). 121. Harris, C. R. et al. Array programming with NumPy. Nature 585, 357–362 (2020). 122. Seabold, S. & Perktold, J. Statsmodels: Econometric and Statistical Modeling with Python. in 92–
96 (2010). doi:10.25080/Majora-92bf1922-011. 123. Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079
(2009). 124. Amemiya, H. M., Kundaje, A. & Boyle, A. P. The ENCODE Blacklist: Identification of Problematic
Regions of the Genome. Scientific Reports 9, 9354 (2019). 125. Ramírez, F. et al. deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Research 44, W160–W165 (2016). 126. Zhang, Y. et al. Model-based Analysis of ChIP-Seq (MACS). Genome Biology 9, R137 (2008). 54
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Acknowledgements
This work is funded by the following grants to S. Aerts: ERC Consolidator Grant (724226_cis- CONTROL), by the Special Research Fund (BOF) KU Leuven (grant PF/10/016) and F.W.O (grants G.0791.14, G.0C04.17).
|
J.J. and C.B.G.-B are supported by a PhD fellowship of The Research Foundation – Flanders (FWO, 1199518N and 11F1519N). 10x Chromium was partially made available through VIB Tech Watch Funding. Imaging, FAC-sorting and single-cell analyses were supported by the light microscopy, FACS and single-cell expertise units at the VIB-KU Leuven Center for Brain and Disease Research. Computing was performed at the Vlaams Supercomputer Center (VSC). Stocks obtained from the Bloomington Drosophila Stock Center were used in this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. We thank the Janelia FlyLight Project for publicly providing images and reporter lines to assess enhancer activity on the CNS in Drosophila. We also thank Filipe Pinto-Teixeira and members of the Aerts lab for helpful discussions, Friday morning éclairs, and for reviewing the manuscript. Author contributions
S.Ae., J.J., S.Ai. and D.P. conceived the study. J.J., S.Ai., I.I.T. and D.P. performed computational analyses with assistance from K.S., C.B.G.-B, G.H. and M.D. S.M. and V.C. performed scATAC-seq experiments, J.J. and S.M. performed FAC-sorting and omniATAC-seq. J.N.I., J.J, X.J.Q., and S.M. performed antibody staining and visualization. X.J.Q. and V.C. performed the cloning of selected enhancers. S.Ai. created the website with assistance of G.H., D.P. and K.S. S.Ae., J.J., S.Ai. and I.I.T. wrote the manuscript. Competing interest
The authors declare no competing interests. 55
bioRxiv preprint
doi:
https://doi.org/10.1101/2021.08.11.454937
;
this version posted August 12, 2021. The copyright holder for this preprint
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. List of Extended Data Figures
Extended Data Fig. 1: Global analysis of all timepoints reveals major dynamics and optic lobe-central
brain split
Extended Data Fig. 2: Sub-clustering of main categories in the adult
Extended Data Fig. 3: Integration of scRNA-seq and snATAC-seq
Extended Data Fig. 4: Gene expression correlates with region accessibility
Extended Data Fig. 5: Cell type specific regions can serve as functional enhancers
Extended Data Fig. 6: Identification of T4/T5 subtypes
Extended Data Fig. 7: BEAF-32 ChIP-seq peaks can be used to delimit the search space for regulatory
regions around each gene. Extended Data Fig. 8: eGRN overview
Extended Data Fig. 9: Deep learning predicts de novo key transcriptional activators and repressors
Extended Data Fig. 10: Enhancers selected by DARs or eGRNs generate novel driver lines
Extended Data Fig. 11: Breakdown of existing driver lines into functional components
Extended Data Fig. 12: Tracking cell types across development reveals presence of core-regions
Extended Data Fig. 13: Neural progenitors have unique chromatin profiles
List of Extended Data Tables
Extended Data Table 1: CellRanger statistics of the 10x Chromium runs.
|
Extended Data Table 2: Architecture and hyperparameters of DeepFlyBrain. Extended Data Table 3: Topic annotations and performance metrics of DeepFlyBrain. Extended Data Table 4: Cloned enhancers and transgenic lines made for in vivo reporter assays. Extended Data Table 5: GO results for optic lobe branches. List of supplementary files
Supplementary Data 1: FACS Gating Strategy
Supplementary Data 2: VSN config file
Supplementary Data 3: DeepFlyBrain training
Supplementary Data 4: DeepFlyBrain performance
Supplementary Data 5: DeepFlyBrain scoring and DeepExplainer plots
56
|
bioRxiv preprint
April 6, 2021
doi: was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. https://doi.org/10.1101/2021.04.16.440073
;
this version posted April 18, 2021. The copyright holder for this preprint (which
Coupled exoskeleton assistance simplifies control and maintains metabolic benefits: a simulation study
Nicholas A. Bianco1*, Patrick W. Franks1, Jennifer L. Hicks2, Scott L. Delp1,2,3,
1 Department of Mechanical Engineering, Stanford University, Stanford, California, United States of America 2 Department of Bioengineering, Stanford University, Stanford, California, United States of America 3 Department of Orthopaedic Surgery, Stanford University, Stanford, California, United States of America
[email protected]
Abstract
Assistive exoskeletons can reduce the metabolic cost of walking, and recent advances in exoskeleton device design and control have resulted in large metabolic savings. Most exoskeleton devices provide assistance at either the ankle or hip. Exoskeletons that assist multiple joints have the potential to provide greater metabolic savings, but can require many actuators and complicated controllers, making it difficult to design effective assistance. Coupled assistance, when two or more joints are assisted using one actuator or control signal, could reduce control dimensionality while retaining metabolic benefits. However, it is unknown which combinations of assisted joints are most promising and if there are negative consequences associated with coupled assistance. Since designing assistance with human experiments is expensive and time-consuming, we used musculoskeletal simulation to evaluate metabolic savings from multi-joint assistance and identify promising joint combinations. We generated 2D muscle-driven simulations of walking while simultaneously optimizing control strategies for simulated lower-limb exoskeleton assistive devices to minimize metabolic cost. Each device provided assistance either at a single joint or at multiple joints using massless, ideal actuators. To assess if control could be simplified for multi-joint exoskeletons, we simulated different control strategies in which the torque provided at each joint was either controlled independently or coupled between joints. We compared the predicted optimal torque profiles and changes in muscle and whole-body metabolic power consumption across the single joint and multi-joint assistance strategies. We found multi-joint devices–whether independent or coupled–provided 50% greater metabolic savings than single joint devices. The coupled multi-joint devices were able to achieve most of the metabolic savings produced by independently-controlled multi-joint devices. Our results indicate that device designers could simplify multi-joint exoskeleton designs by reducing the number of torque control parameters through coupling, while still maintaining large reductions in metabolic cost. Introduction
Wearable robotic exoskeletons that reduce the metabolic cost of walking could improve mobility for individuals with musculoskeletal or neurological impairments and assist
1/26
1
2
3
bioRxiv preprint
April 6, 2021
doi: was not certified by peer review) is the author/funder.
|
All rights reserved. No reuse allowed without permission. https://doi.org/10.1101/2021.04.16.440073
;
this version posted April 18, 2021. The copyright holder for this preprint (which
soldiers and firefighters carrying heavy loads. Assistance strategies that reduce metabolic cost have only recently been discovered using both powered [1–4] and unpowered [5] devices. Despite these successes, designing controllers for exoskeletons can be counterintuitive and time-consuming. Some exoskeleton designs focused on biomimicry, where assistive devices attempt to emulate biological joint kinematics, kinetics, and power, but these seemingly intuitive approaches have had limited success in reducing metabolic cost [6, 7]. To better understand what aspects of exoskeleton assistance affect metabolic cost, many recent studies have designed assistance by varying the timing and magnitude of assistive torques and powers [8–12]. For example, a recent study showed that optimizing both assistance torque onset timing and average power together produces larger metabolic reductions than when considering each quantity alone [11]. More recent approaches, such as human-in-the-loop optimization experiments, which continuously optimize assistance for a subject based on real-time estimates of metabolic energy, have produced large reductions in metabolic cost [8, 10]. However, since each human-in-the-loop optimization evaluation requires several minutes of human metabolic data from indirect calorimetry, it is time-consuming and expensive to test a large number of devices. For example, a human-in-the-loop optimization may take several days of experimentation to complete. Simulations and experiments suggest that assisting multiple joints at once could deliver larger metabolic savings than from assisting a single joint [12–15]. However, designing assistance for these “multi-joint” exoskeletons can magnify the challenges of optimizing the control, since such devices can include multiple actuators with independent control laws, which increases the number of parameters that must be tested in experiments. For example, the convergence time for human-in-the-loop optimization experiments scales poorly with increasing optimization variables, and therefore may be prohibitively long for multi-joint exoskeletons due to the large number of control variables needed for several assistive torques. As a result, most exoskeleton studies focus on assisting only one degree of freedom to simplify parameter design, usually preferring the hip or the ankle since they produce most of the positive power during walking and running [4, 16–18]. Coupled assistance could greatly simplify the mechanical and control design of exoskeleton devices either by reducing either the number of actuators needed for a device or by simplifying control complexity (i.e., the number of parameters personalized to a subject) and thus reducing the time needed to perform human-in-the-loop optimizations to achieve good reductions in metabolic cost.
|
Assisting two joints at once using one actuator, or “coupling” assistance, showed success in recent exoskeleton studies with an ankle-hip soft exosuit [12, 19–21] and a knee-ankle device [14]. These studies exploit the similar timings of joint moments (e.g., the hip flexion moment and ankle plantarflexion moment reach a maximum at approximately the same point in the gait cycle). Other combinations of assisted joints may be effective but haven’t been tested in experiments, since these experiments are resource-intensive, especially when multiple joints are assisted. Simulations could help us identify which combinations of joints to assist and how control could be coupled across joints, while still achieving significant decreases in metabolic cost. Musculoskeletal simulation has become a valuable tool for examining the complex muscle-level and whole-body metabolic changes produced by exoskeleton devices [22]. Researchers have used simulation to analyze an existing exoskeleton and optimize its mechanical design [23] and to better understand human-device interaction [24]. Other studies have used simulation to help interpret experimental results, for example, to understand how muscle mechanics drive metabolic changes for an ankle exoskeleton [25]. Researchers have also used simulation to model exoskeleton devices as ideal actuators to discover guidelines for designing walking [26] and running [13] exoskeletons. A recent
2/26
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
bioRxiv preprint
April 6, 2021
doi: was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. https://doi.org/10.1101/2021.04.16.440073
;
this version posted April 18, 2021. The copyright holder for this preprint (which
study [27] applied results from assisted running simulations [13] to design assistance for a soft running exoskeleton. The simulation-derived controls provided greater metabolic cost reductions compared to assistance designed based on biological joint moments, demonstrating the ability of simulations to improve exoskeleton design. Another recent study conducted by our group used simulation to design assistance for an experimental hip-knee-ankle exoskeleton, resulting in a large metabolic reduction [28]. While this study and the running simulation study examined multi-joint assistance [13], no study has used simulation to systematically compare different multi-joint assistance strategies for walking. In this study, we examined how simulated multi-joint assistance affects the
metabolic cost of walking. We added ideal, massless assistive devices to a lower-extremity musculoskeletal model and simultaneously optimized muscle activity and device controls to match the net joint moments of normal walking and minimize metabolic cost. Each device assisted a single joint or assisted multiple joints simultaneously. Multi-joint devices could control assistance at joints independently or couple assistance for multiple joints, using the same control with independent peak torque magnitudes.
|
We used the simulations to achieve two goals. First, we sought to estimate the metabolic savings provided by multi-joint exoskeletons during walking as compared to exoskeletons that assist only a single joint. Second, we sought to determine if coupled assistance could achieve similar metabolic savings to independent assistance. To address our second aim, we compared whole-body and muscle metabolic cost savings and optimal device torques between coupled and independent multi-joint assistance. Materials and methods
Experimental data
We used a previously-collected dataset from 5 healthy individuals walking on a treadmill (mean ± s.d. : age: 29.2 ± 6.3 years, height: 1.80 ± 0.03 m, mass: 72.4 ± 5.7 kg) [29]. Subjects in this previous study provided informed consent to a protocol approved by the Stanford Institutional Review Board. The data included marker trajectories, ground reaction forces, and electromyography (EMG) signals. For each subject, we simulated three gait cycles of walking at 1.25 m/s. One gait cycle was used in a model calibration step, and the other two were used for simulations of exoskeleton devices. For validating muscle activation patterns predicted from simulation, we used the processed EMG signals as described in the previous study [29], where signals were normalized by the highest value recorded across all walking speeds (see section “Comparison of simulations with experimental results”). Musculoskeletal model
A generic 29 degree-of-freedom skeletal model was scaled to each subject’s anthropomorphic data based on static marker trials [30]. Nine Hill-type muscle-tendon units, as modeled in a previous simulation study from our group [31], were included on each leg of the model: gluteus maximus, biarticular hamstrings, iliopsoas, rectus femoris, vasti, biceps femoris short head, gastrocnemius, soleus, and tibialis anterior. We used this reduced muscle set since we only simulated sagittal-plane exoskeleton devices and since fewer muscles kept the optimizations tractable. To create the set of nine muscles, we combined muscles (from the model of [30]) that had similar sagittal-plane functions into one muscle with a combined maximum isometric force value. Joint and muscle kinematics and net joint moments were computed through inverse kinematics and inverse dynamics tools using OpenSim 3.3 [32]. 3/26
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
bioRxiv preprint
April 6, 2021
doi: was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. https://doi.org/10.1101/2021.04.16.440073
;
this version posted April 18, 2021. The copyright holder for this preprint (which
Simulation framework
We used a simulation framework [33] based on the GPOPS-II direct collocation optimal control software (Version 2.3) [34] to solve the muscle redundancy problem for unassisted walking. In each simulation, we solved for muscle activity while enforcing muscle activation and tendon compliance dynamics.
|
Muscle kinematics were constrained to match muscle-tendon lengths and velocities obtained from inverse kinematics, and muscle-generated moments were constrained to match net joint moments computed from inverse dynamics. Since we only included sagittal-plane muscles in our model, only sagittal-plane joint moments (hip flexion-extension, knee flexion-extension, and ankle plantarflexion-dorsiflexion) were matched in each optimization. We assumed left-right symmetry of walking and therefore only solved for muscle activity in the right leg. Each problem included reserve torque actuators in addition to muscle-generated moments to help ensure dynamic consistency; these actuators were penalized in the objective function such that the muscles were the primary actuators enforcing the joint moment constraints. Each optimal control problem was solved with the Legendre-Gauss-Radau quadrature collocation method provided by GPOPS-II using an initial mesh of 100 mesh intervals per second. The initial mesh was updated using mesh refinement with a tolerance of 10−3 to reduce muscle activation and tendon compliance dynamic errors in the solution trajectories. The resulting nonlinear programs produced from the collocation method were solved with a convergence tolerance of 10−3 using IPOPT, the non-linear optimization solver [35]. Muscle parameter calibration
We calibrated the model’s muscle parameters so that estimated muscle activations would better match EMG measurements. Our model calibration approach consisted of three main steps. In the first step, we scaled maximum isometric force values based on a previously reported relationship between muscle volume and total body mass [36]. In the second step, we optimized optimal fiber lengths, tendon slack lengths, and passive muscle strain parameters while minimizing the error between model and reported experimental passive muscle moments [37]. We used MATLAB’s fmincon to minimize passive moment errors across a range of static joint positions with a rigid-tendon assumption for computing passive muscle force. In addition to the cost term penalizing deviations from experimental passive muscle moments, secondary cost terms were included to minimize total muscle passive force and prevent deviations from default parameter values which would lead to undesirable solutions with large passive forces in individual muscles. The third step of our model calibration used EMG data to further adjust the model’s muscle parameters. Passive muscle strain parameters were fixed to the values obtained from the first calibration step, and tendon slack length and optimal fiber lengths were again optimized within 25% of their original values, using the first-step calibration values as an initial guess. The error between EMG data and muscle excitations was the primary term minimized in the objective function. Passive muscle forces were also minimized to prevent undesired increases in passive forces due to the readjusted parameters. The muscle activations were also included as a lower-weighted, secondary objective term to aid convergence.
|
The resulting muscle parameters were used in all subsequent simulations. Exoskeleton device simulations
After calibrating the model for a given subject, we simulated unassisted and assisted gait using the subject’s remaining two gait cycles. In both unassisted and assisted gait,
4/26
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
bioRxiv preprint
April 6, 2021
doi: was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. https://doi.org/10.1101/2021.04.16.440073
;
this version posted April 18, 2021. The copyright holder for this preprint (which
the primary objective was to minimize metabolic cost computed from a version of the metabolic energy model developed by Umberger et al. (2003) that was modified to have a continuous first derivative for gradient-based optimization [38, 39]. We included additional secondary objective terms to minimize muscle excitation, muscle activation, and the derivative of tendon force, all of which aided problem convergence. Since our simulation method relied on kinematics obtained from an inverse kinematics solution, the unassisted and assisted simulations used the same healthy walking kinematics (i.e., the simulation did not change the model’s kinematics in response to the assistive device). In the unassisted simulations, the muscles and the heavily-penalized reserve torque actuators were the only actuators available to reproduce the net joint moments. In the assisted simulations, exoskeleton devices were modeled as massless torque actuators and could apply torques to reduce muscle effort, while still matching the net joint moment constraints from inverse dynamics. The actuators had no power limits, but had peak torque limits for hip flexion-extension (1.0 N-m/kg), knee flexion-extension (1.0 N-m/kg), and ankle plantarflexion (2.0 N-m/kg); these peak torque limits were included to speed convergence and were chosen such that optimized device controls never exceeded the optimization bounds. Torques were applied in the following five joint directions: hip flexion, hip extension, knee flexion, knee extension, and ankle plantarflexion. Single-joint exoskeleton devices provided assistive torques in one of the five joint directions. Multi-joint exoskeleton devices provided assistance to the following combinations of joint directions: (1) hip-extension knee-extension, (2) hip-flexion knee-flexion, (3) knee-flexion ankle-plantarflexion, (4) hip-flexion ankle-plantarflexion, and (5) hip-flexion knee-flexion ankle-plantarflexion. The multi-joint exoskeleton devices were actuated by individual control signals (i.e., “independent” control) or with only one control signal applied to all joint directions (i.e., “coupled” control). When using coupled control, additional “gain” variables scaled the applied exoskeleton torques to allow different applied torque magnitudes since net joint moment magnitudes differ between the hip, knee, and ankle.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.