document_id
int64 185
2.68k
| context
stringlengths 2.88k
70.8k
| question
stringlengths 11
194
| id
int64 225
5.32k
| answer
stringlengths 1
933
| answer_start
int64 157
69k
|
|---|---|---|---|---|---|
2,439 |
Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7078228/
SHA: 1cf54d1c77b7f0494ab971429d26e0e199952d09
Authors: Liu, Jia; Cao, Ruiyuan; Xu, Mingyue; Wang, Xi; Zhang, Huanyu; Hu, Hengrui; Li, Yufeng; Hu, Zhihong; Zhong, Wu; Wang, Manli
Date: 2020-03-18
DOI: 10.1038/s41421-020-0156-0
License: cc-by
Abstract: nan
Text: Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro Jia Liu 1 , Ruiyuan Cao 2 , Mingyue Xu 1,3 , Xi Wang 1 , Huanyu Zhang 1,3 , Hengrui Hu 1,3 , Yufeng Li 1,3 , Zhihong Hu 1 , Wu Zhong 2 and Manli Wang 1 Dear Editor, The outbreak of coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2/2019-nCoV) poses a serious threat to global public health and local economies. As of March 3, 2020, over 80,000 cases have been confirmed in China, including 2946 deaths as well as over 10,566 confirmed cases in 72 other countries. Such huge numbers of infected and dead people call for an urgent demand of effective, available, and affordable drugs to control and diminish the epidemic.
We have recently reported that two drugs, remdesivir (GS-5734) and chloroquine (CQ) phosphate, efficiently inhibited SARS-CoV-2 infection in vitro 1 . Remdesivir is a nucleoside analog prodrug developed by Gilead Sciences (USA). A recent case report showed that treatment with remdesivir improved the clinical condition of the first patient infected by SARS-CoV-2 in the United States 2 , and a phase III clinical trial of remdesivir against SARS-CoV-2 was launched in Wuhan on February 4, 2020. However, as an experimental drug, remdesivir is not expected to be largely available for treating a very large number of patients in a timely manner. Therefore, of the two potential drugs, CQ appears to be the drug of choice for large-scale use due to its availability, proven safety record, and a relatively low cost. In light of the preliminary clinical data, CQ has been added to the list of trial drugs in the Guidelines for the Diagnosis and Treatment of COVID-19 (sixth edition) published by National Health Commission of the People's Republic of China. CQ (N4-(7-Chloro-4-quinolinyl)-N1,N1-diethyl-1,4pentanediamine) has long been used to treat malaria and amebiasis. However, Plasmodium falciparum developed widespread resistance to it, and with the development of new antimalarials, it has become a choice for the prophylaxis of malaria. In addition, an overdose of CQ can cause acute poisoning and death 3 . In the past years, due to infrequent utilization of CQ in clinical practice, its production and market supply was greatly reduced, at least in China. Hydroxychloroquine (HCQ) sulfate, a derivative of CQ, was first synthesized in 1946 by introducing a hydroxyl group into CQ and was demonstrated to be much less (~40%) toxic than CQ in animals 4 . More importantly, HCQ is still widely available to treat autoimmune diseases, such as systemic lupus erythematosus and rheumatoid arthritis. Since CQ and HCQ share similar chemical structures and mechanisms of acting as a weak base and immunomodulator, it is easy to conjure up the idea that HCQ may be a potent candidate to treat infection by SARS-CoV-2. Actually, as of February 23, 2020, seven clinical trial registries were found in Chinese Clinical Trial Registry (http://www.chictr.org.cn) for using HCQ to treat COVID-19. Whether HCQ is as efficacious as CQ in treating SARS-CoV-2 infection still lacks the experimental evidence.
To this end, we evaluated the antiviral effect of HCQ against SARS-CoV-2 infection in comparison to CQ in vitro. First, the cytotoxicity of HCQ and CQ in African green monkey kidney VeroE6 cells (ATCC-1586) was measured by standard CCK8 assay, and the result showed © The Author(s) 2020
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. (Fig. 1a) . To better compare the antiviral activity of CQ versus HCQ, the dose-response curves of the two compounds against SARS-CoV-2 were determined at four different multiplicities of infection (MOIs) by quantification of viral RNA copy numbers in the cell supernatant at 48 h post infection (p.i.). The data summarized in Fig. 1a and Supplementary Table S1 show that, at all MOIs (0.01, 0.02, 0.2, and 0.8), the 50% maximal effective concentration (EC 50 ) for CQ (2.71, 3.81, 7.14, and 7.36 μM) was lower than that of HCQ (4.51, 4.06, 17.31, and 12.96 μM). The differences in EC 50 values were statistically significant at an MOI of 0.01 (P < 0.05) and MOI of 0.2 (P < 0.001) (Supplementary Table S1 ). It is worth noting that the EC 50 values of CQ seemed to be a little higher than that in our previous report (1.13 μM at an MOI of 0.05) 1 , which is likely due to the adaptation of the virus in cell culture that significantly increased viral infectivity upon continuous passaging. Consequently, the selectivity index (SI = CC 50 /EC 50 ) of CQ (100.81, 71.71, 38.26, and 37.12) was higher than that of HCQ (55.32, 61.45, 14.41, 19.25) at MOIs of 0.01, 0.02, 0.2, and 0.8, respectively. These results were corroborated by immunofluorescence microscopy as evidenced by different expression levels of virus nucleoprotein (NP) at the indicated drug concentrations at 48 h p.i. (Supplementary Fig. S1 ). Taken together, the data suggest that the anti-SARS-CoV-2 activity of HCQ seems to be less potent compared to CQ, at least at certain MOIs.
Both CQ and HCQ are weak bases that are known to elevate the pH of acidic intracellular organelles, such as endosomes/lysosomes, essential for membrane fusion 5 . In addition, CQ could inhibit SARS-CoV entry through changing the glycosylation of ACE2 receptor and spike protein 6 . Time-of-addition experiment confirmed that HCQ effectively inhibited the entry step, as well as the post-entry stages of SARS-CoV-2, which was also found upon CQ treatment (Supplementary Fig. S2 ). To further explore the detailed mechanism of action of CQ and HCQ in inhibiting virus entry, co-localization of virions with early endosomes (EEs) or endolysosomes (ELs) was analyzed by immunofluorescence analysis (IFA) and confocal microscopy. Quantification analysis showed that, at 90 min p.i. in untreated cells, 16.2% of internalized virions (anti-NP, red) were observed in early endosome antigen 1 (EEA1)-positive EEs (green), while more virions (34.3%) were transported into the late endosomal-lysosomal protein LAMP1 + ELs (green) (n > 30 cells for each group). By contrast, in the presence of CQ or HCQ, significantly more virions (35.3% for CQ and 29.2% for HCQ; P < 0.001) were detected in the EEs, while only very few virions (2.4% for CQ and 0.03% for HCQ; P < 0.001) were found to be co-localized with LAMP1 + ELs (n > 30 cells) (Fig. 1b, c) . This suggested that both CQ and HCQ blocked the transport of SARS-CoV-2 from EEs to ELs, which appears to be a requirement to release the viral genome as in the case of SARS-CoV 7 .
Interestingly, we found that CQ and HCQ treatment caused noticeable changes in the number and size/morphology of EEs and ELs (Fig. 1c) . In the untreated cells, most EEs were much smaller than ELs (Fig. 1c) . In CQand HCQ-treated cells, abnormally enlarged EE vesicles were observed (Fig. 1c , arrows in the upper panels), many of which are even larger than ELs in the untreated cells. This is in agreement with previous report that treatment with CQ induced the formation of expanded cytoplasmic vesicles 8 . Within the EE vesicles, virions (red) were localized around the membrane (green) of the vesicle. CQ treatment did not cause obvious changes in the number and size of ELs; however, the regular vesicle structure seemed to be disrupted, at least partially. By contrast, in HCQ-treated cells, the size and number of ELs increased significantly (Fig. 1c , arrows in the lower panels).
Since acidification is crucial for endosome maturation and function, we surmise that endosome maturation might be blocked at intermediate stages of endocytosis, resulting in failure of further transport of virions to the ultimate releasing site. CQ was reported to elevate the pH (see figure on previous page) Fig. 1 Comparative antiviral efficacy and mechanism of action of CQ and HCQ against SARS-CoV-2 infection in vitro. a Cytotoxicity and antiviral activities of CQ and HCQ. The cytotoxicity of the two drugs in Vero E6 cells was determined by CCK-8 assays. Vero E6 cells were treated with different doses of either compound or with PBS in the controls for 1 h and then infected with SARS-CoV-2 at MOIs of 0.01, 0.02, 0.2, and 0.8. The virus yield in the cell supernatant was quantified by qRT-PCR at 48 h p.i. Y-axis represents the mean of percent inhibition normalized to the PBS group. The experiments were repeated twice. b, c Mechanism of CQ and HCQ in inhibiting virus entry. Vero E6 cells were treated with CQ or HCQ (50 μM) for 1 h, followed by virus binding (MOI = 10) at 4°C for 1 h. Then the unbound virions were removed, and the cells were further supplemented with fresh drug-containing medium at 37°C for 90 min before being fixed and stained with IFA using anti-NP antibody for virions (red) and antibodies against EEA1 for EEs (green) or LAMP1 for ELs (green). The nuclei (blue) were stained with Hoechst dye. The portion of virions that co-localized with EEs or ELs in each group (n > 30 cells) was quantified and is shown in b. Representative confocal microscopic images of viral particles (red), EEA1 + EEs (green), or LAMP1 + ELs (green) in each group are displayed in c. The enlarged images in the boxes indicate a single vesicle-containing virion. The arrows indicated the abnormally enlarged vesicles. Bars, 5 μm. Statistical analysis was performed using a one-way analysis of variance (ANOVA) with GraphPad Prism (F = 102.8, df = 5,182, ***P < 0.001).
of lysosome from about 4.5 to 6.5 at 100 μM 9 . To our knowledge, there is a lack of studies on the impact of HCQ on the morphology and pH values of endosomes/ lysosomes. Our observations suggested that the mode of actions of CQ and HCQ appear to be distinct in certain aspects.
It has been reported that oral absorption of CQ and HCQ in humans is very efficient. In animals, both drugs share similar tissue distribution patterns, with high concentrations in the liver, spleen, kidney, and lung reaching levels of 200-700 times higher than those in the plasma 10 . It was reported that safe dosage (6-6.5 mg/kg per day) of HCQ sulfate could generate serum levels of 1.4-1.5 μM in humans 11 . Therefore, with a safe dosage, HCQ concentration in the above tissues is likely to be achieved to inhibit SARS-CoV-2 infection.
Clinical investigation found that high concentration of cytokines were detected in the plasma of critically ill patients infected with SARS-CoV-2, suggesting that cytokine storm was associated with disease severity 12 . Other than its direct antiviral activity, HCQ is a safe and successful anti-inflammatory agent that has been used extensively in autoimmune diseases and can significantly decrease the production of cytokines and, in particular, pro-inflammatory factors. Therefore, in COVID-19 patients, HCQ may also contribute to attenuating the inflammatory response. In conclusion, our results show that HCQ can efficiently inhibit SARS-CoV-2 infection in vitro. In combination with its anti-inflammatory function, we predict that the drug has a good potential to combat the disease. This possibility awaits confirmation by clinical trials. We need to point out, although HCQ is less toxic than CQ, prolonged and overdose usage can still cause poisoning. And the relatively low SI of HCQ requires careful designing and conducting of clinical trials to achieve efficient and safe control of the SARS-CoV-2 infection.
|
In vitro comparison of antiviral activity of Chloroquine(CQ) and Hydroxychloroquine(HCQ) against COVID-19?
| 655 |
compare the antiviral activity of CQ versus HCQ, the dose-response curves
| 4,794 |
2,439 |
Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7078228/
SHA: 1cf54d1c77b7f0494ab971429d26e0e199952d09
Authors: Liu, Jia; Cao, Ruiyuan; Xu, Mingyue; Wang, Xi; Zhang, Huanyu; Hu, Hengrui; Li, Yufeng; Hu, Zhihong; Zhong, Wu; Wang, Manli
Date: 2020-03-18
DOI: 10.1038/s41421-020-0156-0
License: cc-by
Abstract: nan
Text: Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro Jia Liu 1 , Ruiyuan Cao 2 , Mingyue Xu 1,3 , Xi Wang 1 , Huanyu Zhang 1,3 , Hengrui Hu 1,3 , Yufeng Li 1,3 , Zhihong Hu 1 , Wu Zhong 2 and Manli Wang 1 Dear Editor, The outbreak of coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2/2019-nCoV) poses a serious threat to global public health and local economies. As of March 3, 2020, over 80,000 cases have been confirmed in China, including 2946 deaths as well as over 10,566 confirmed cases in 72 other countries. Such huge numbers of infected and dead people call for an urgent demand of effective, available, and affordable drugs to control and diminish the epidemic.
We have recently reported that two drugs, remdesivir (GS-5734) and chloroquine (CQ) phosphate, efficiently inhibited SARS-CoV-2 infection in vitro 1 . Remdesivir is a nucleoside analog prodrug developed by Gilead Sciences (USA). A recent case report showed that treatment with remdesivir improved the clinical condition of the first patient infected by SARS-CoV-2 in the United States 2 , and a phase III clinical trial of remdesivir against SARS-CoV-2 was launched in Wuhan on February 4, 2020. However, as an experimental drug, remdesivir is not expected to be largely available for treating a very large number of patients in a timely manner. Therefore, of the two potential drugs, CQ appears to be the drug of choice for large-scale use due to its availability, proven safety record, and a relatively low cost. In light of the preliminary clinical data, CQ has been added to the list of trial drugs in the Guidelines for the Diagnosis and Treatment of COVID-19 (sixth edition) published by National Health Commission of the People's Republic of China. CQ (N4-(7-Chloro-4-quinolinyl)-N1,N1-diethyl-1,4pentanediamine) has long been used to treat malaria and amebiasis. However, Plasmodium falciparum developed widespread resistance to it, and with the development of new antimalarials, it has become a choice for the prophylaxis of malaria. In addition, an overdose of CQ can cause acute poisoning and death 3 . In the past years, due to infrequent utilization of CQ in clinical practice, its production and market supply was greatly reduced, at least in China. Hydroxychloroquine (HCQ) sulfate, a derivative of CQ, was first synthesized in 1946 by introducing a hydroxyl group into CQ and was demonstrated to be much less (~40%) toxic than CQ in animals 4 . More importantly, HCQ is still widely available to treat autoimmune diseases, such as systemic lupus erythematosus and rheumatoid arthritis. Since CQ and HCQ share similar chemical structures and mechanisms of acting as a weak base and immunomodulator, it is easy to conjure up the idea that HCQ may be a potent candidate to treat infection by SARS-CoV-2. Actually, as of February 23, 2020, seven clinical trial registries were found in Chinese Clinical Trial Registry (http://www.chictr.org.cn) for using HCQ to treat COVID-19. Whether HCQ is as efficacious as CQ in treating SARS-CoV-2 infection still lacks the experimental evidence.
To this end, we evaluated the antiviral effect of HCQ against SARS-CoV-2 infection in comparison to CQ in vitro. First, the cytotoxicity of HCQ and CQ in African green monkey kidney VeroE6 cells (ATCC-1586) was measured by standard CCK8 assay, and the result showed © The Author(s) 2020
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. (Fig. 1a) . To better compare the antiviral activity of CQ versus HCQ, the dose-response curves of the two compounds against SARS-CoV-2 were determined at four different multiplicities of infection (MOIs) by quantification of viral RNA copy numbers in the cell supernatant at 48 h post infection (p.i.). The data summarized in Fig. 1a and Supplementary Table S1 show that, at all MOIs (0.01, 0.02, 0.2, and 0.8), the 50% maximal effective concentration (EC 50 ) for CQ (2.71, 3.81, 7.14, and 7.36 μM) was lower than that of HCQ (4.51, 4.06, 17.31, and 12.96 μM). The differences in EC 50 values were statistically significant at an MOI of 0.01 (P < 0.05) and MOI of 0.2 (P < 0.001) (Supplementary Table S1 ). It is worth noting that the EC 50 values of CQ seemed to be a little higher than that in our previous report (1.13 μM at an MOI of 0.05) 1 , which is likely due to the adaptation of the virus in cell culture that significantly increased viral infectivity upon continuous passaging. Consequently, the selectivity index (SI = CC 50 /EC 50 ) of CQ (100.81, 71.71, 38.26, and 37.12) was higher than that of HCQ (55.32, 61.45, 14.41, 19.25) at MOIs of 0.01, 0.02, 0.2, and 0.8, respectively. These results were corroborated by immunofluorescence microscopy as evidenced by different expression levels of virus nucleoprotein (NP) at the indicated drug concentrations at 48 h p.i. (Supplementary Fig. S1 ). Taken together, the data suggest that the anti-SARS-CoV-2 activity of HCQ seems to be less potent compared to CQ, at least at certain MOIs.
Both CQ and HCQ are weak bases that are known to elevate the pH of acidic intracellular organelles, such as endosomes/lysosomes, essential for membrane fusion 5 . In addition, CQ could inhibit SARS-CoV entry through changing the glycosylation of ACE2 receptor and spike protein 6 . Time-of-addition experiment confirmed that HCQ effectively inhibited the entry step, as well as the post-entry stages of SARS-CoV-2, which was also found upon CQ treatment (Supplementary Fig. S2 ). To further explore the detailed mechanism of action of CQ and HCQ in inhibiting virus entry, co-localization of virions with early endosomes (EEs) or endolysosomes (ELs) was analyzed by immunofluorescence analysis (IFA) and confocal microscopy. Quantification analysis showed that, at 90 min p.i. in untreated cells, 16.2% of internalized virions (anti-NP, red) were observed in early endosome antigen 1 (EEA1)-positive EEs (green), while more virions (34.3%) were transported into the late endosomal-lysosomal protein LAMP1 + ELs (green) (n > 30 cells for each group). By contrast, in the presence of CQ or HCQ, significantly more virions (35.3% for CQ and 29.2% for HCQ; P < 0.001) were detected in the EEs, while only very few virions (2.4% for CQ and 0.03% for HCQ; P < 0.001) were found to be co-localized with LAMP1 + ELs (n > 30 cells) (Fig. 1b, c) . This suggested that both CQ and HCQ blocked the transport of SARS-CoV-2 from EEs to ELs, which appears to be a requirement to release the viral genome as in the case of SARS-CoV 7 .
Interestingly, we found that CQ and HCQ treatment caused noticeable changes in the number and size/morphology of EEs and ELs (Fig. 1c) . In the untreated cells, most EEs were much smaller than ELs (Fig. 1c) . In CQand HCQ-treated cells, abnormally enlarged EE vesicles were observed (Fig. 1c , arrows in the upper panels), many of which are even larger than ELs in the untreated cells. This is in agreement with previous report that treatment with CQ induced the formation of expanded cytoplasmic vesicles 8 . Within the EE vesicles, virions (red) were localized around the membrane (green) of the vesicle. CQ treatment did not cause obvious changes in the number and size of ELs; however, the regular vesicle structure seemed to be disrupted, at least partially. By contrast, in HCQ-treated cells, the size and number of ELs increased significantly (Fig. 1c , arrows in the lower panels).
Since acidification is crucial for endosome maturation and function, we surmise that endosome maturation might be blocked at intermediate stages of endocytosis, resulting in failure of further transport of virions to the ultimate releasing site. CQ was reported to elevate the pH (see figure on previous page) Fig. 1 Comparative antiviral efficacy and mechanism of action of CQ and HCQ against SARS-CoV-2 infection in vitro. a Cytotoxicity and antiviral activities of CQ and HCQ. The cytotoxicity of the two drugs in Vero E6 cells was determined by CCK-8 assays. Vero E6 cells were treated with different doses of either compound or with PBS in the controls for 1 h and then infected with SARS-CoV-2 at MOIs of 0.01, 0.02, 0.2, and 0.8. The virus yield in the cell supernatant was quantified by qRT-PCR at 48 h p.i. Y-axis represents the mean of percent inhibition normalized to the PBS group. The experiments were repeated twice. b, c Mechanism of CQ and HCQ in inhibiting virus entry. Vero E6 cells were treated with CQ or HCQ (50 μM) for 1 h, followed by virus binding (MOI = 10) at 4°C for 1 h. Then the unbound virions were removed, and the cells were further supplemented with fresh drug-containing medium at 37°C for 90 min before being fixed and stained with IFA using anti-NP antibody for virions (red) and antibodies against EEA1 for EEs (green) or LAMP1 for ELs (green). The nuclei (blue) were stained with Hoechst dye. The portion of virions that co-localized with EEs or ELs in each group (n > 30 cells) was quantified and is shown in b. Representative confocal microscopic images of viral particles (red), EEA1 + EEs (green), or LAMP1 + ELs (green) in each group are displayed in c. The enlarged images in the boxes indicate a single vesicle-containing virion. The arrows indicated the abnormally enlarged vesicles. Bars, 5 μm. Statistical analysis was performed using a one-way analysis of variance (ANOVA) with GraphPad Prism (F = 102.8, df = 5,182, ***P < 0.001).
of lysosome from about 4.5 to 6.5 at 100 μM 9 . To our knowledge, there is a lack of studies on the impact of HCQ on the morphology and pH values of endosomes/ lysosomes. Our observations suggested that the mode of actions of CQ and HCQ appear to be distinct in certain aspects.
It has been reported that oral absorption of CQ and HCQ in humans is very efficient. In animals, both drugs share similar tissue distribution patterns, with high concentrations in the liver, spleen, kidney, and lung reaching levels of 200-700 times higher than those in the plasma 10 . It was reported that safe dosage (6-6.5 mg/kg per day) of HCQ sulfate could generate serum levels of 1.4-1.5 μM in humans 11 . Therefore, with a safe dosage, HCQ concentration in the above tissues is likely to be achieved to inhibit SARS-CoV-2 infection.
Clinical investigation found that high concentration of cytokines were detected in the plasma of critically ill patients infected with SARS-CoV-2, suggesting that cytokine storm was associated with disease severity 12 . Other than its direct antiviral activity, HCQ is a safe and successful anti-inflammatory agent that has been used extensively in autoimmune diseases and can significantly decrease the production of cytokines and, in particular, pro-inflammatory factors. Therefore, in COVID-19 patients, HCQ may also contribute to attenuating the inflammatory response. In conclusion, our results show that HCQ can efficiently inhibit SARS-CoV-2 infection in vitro. In combination with its anti-inflammatory function, we predict that the drug has a good potential to combat the disease. This possibility awaits confirmation by clinical trials. We need to point out, although HCQ is less toxic than CQ, prolonged and overdose usage can still cause poisoning. And the relatively low SI of HCQ requires careful designing and conducting of clinical trials to achieve efficient and safe control of the SARS-CoV-2 infection.
|
What will be the drug of choice for treating COVID-19 between Chloroquine and Remdesivir?
| 651 |
two potential drugs, CQ appears to be the drug of choice for large-scale use due to its availability, proven safety record, and a relatively low cost.
| 1,897 |
2,439 |
Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7078228/
SHA: 1cf54d1c77b7f0494ab971429d26e0e199952d09
Authors: Liu, Jia; Cao, Ruiyuan; Xu, Mingyue; Wang, Xi; Zhang, Huanyu; Hu, Hengrui; Li, Yufeng; Hu, Zhihong; Zhong, Wu; Wang, Manli
Date: 2020-03-18
DOI: 10.1038/s41421-020-0156-0
License: cc-by
Abstract: nan
Text: Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro Jia Liu 1 , Ruiyuan Cao 2 , Mingyue Xu 1,3 , Xi Wang 1 , Huanyu Zhang 1,3 , Hengrui Hu 1,3 , Yufeng Li 1,3 , Zhihong Hu 1 , Wu Zhong 2 and Manli Wang 1 Dear Editor, The outbreak of coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2/2019-nCoV) poses a serious threat to global public health and local economies. As of March 3, 2020, over 80,000 cases have been confirmed in China, including 2946 deaths as well as over 10,566 confirmed cases in 72 other countries. Such huge numbers of infected and dead people call for an urgent demand of effective, available, and affordable drugs to control and diminish the epidemic.
We have recently reported that two drugs, remdesivir (GS-5734) and chloroquine (CQ) phosphate, efficiently inhibited SARS-CoV-2 infection in vitro 1 . Remdesivir is a nucleoside analog prodrug developed by Gilead Sciences (USA). A recent case report showed that treatment with remdesivir improved the clinical condition of the first patient infected by SARS-CoV-2 in the United States 2 , and a phase III clinical trial of remdesivir against SARS-CoV-2 was launched in Wuhan on February 4, 2020. However, as an experimental drug, remdesivir is not expected to be largely available for treating a very large number of patients in a timely manner. Therefore, of the two potential drugs, CQ appears to be the drug of choice for large-scale use due to its availability, proven safety record, and a relatively low cost. In light of the preliminary clinical data, CQ has been added to the list of trial drugs in the Guidelines for the Diagnosis and Treatment of COVID-19 (sixth edition) published by National Health Commission of the People's Republic of China. CQ (N4-(7-Chloro-4-quinolinyl)-N1,N1-diethyl-1,4pentanediamine) has long been used to treat malaria and amebiasis. However, Plasmodium falciparum developed widespread resistance to it, and with the development of new antimalarials, it has become a choice for the prophylaxis of malaria. In addition, an overdose of CQ can cause acute poisoning and death 3 . In the past years, due to infrequent utilization of CQ in clinical practice, its production and market supply was greatly reduced, at least in China. Hydroxychloroquine (HCQ) sulfate, a derivative of CQ, was first synthesized in 1946 by introducing a hydroxyl group into CQ and was demonstrated to be much less (~40%) toxic than CQ in animals 4 . More importantly, HCQ is still widely available to treat autoimmune diseases, such as systemic lupus erythematosus and rheumatoid arthritis. Since CQ and HCQ share similar chemical structures and mechanisms of acting as a weak base and immunomodulator, it is easy to conjure up the idea that HCQ may be a potent candidate to treat infection by SARS-CoV-2. Actually, as of February 23, 2020, seven clinical trial registries were found in Chinese Clinical Trial Registry (http://www.chictr.org.cn) for using HCQ to treat COVID-19. Whether HCQ is as efficacious as CQ in treating SARS-CoV-2 infection still lacks the experimental evidence.
To this end, we evaluated the antiviral effect of HCQ against SARS-CoV-2 infection in comparison to CQ in vitro. First, the cytotoxicity of HCQ and CQ in African green monkey kidney VeroE6 cells (ATCC-1586) was measured by standard CCK8 assay, and the result showed © The Author(s) 2020
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. (Fig. 1a) . To better compare the antiviral activity of CQ versus HCQ, the dose-response curves of the two compounds against SARS-CoV-2 were determined at four different multiplicities of infection (MOIs) by quantification of viral RNA copy numbers in the cell supernatant at 48 h post infection (p.i.). The data summarized in Fig. 1a and Supplementary Table S1 show that, at all MOIs (0.01, 0.02, 0.2, and 0.8), the 50% maximal effective concentration (EC 50 ) for CQ (2.71, 3.81, 7.14, and 7.36 μM) was lower than that of HCQ (4.51, 4.06, 17.31, and 12.96 μM). The differences in EC 50 values were statistically significant at an MOI of 0.01 (P < 0.05) and MOI of 0.2 (P < 0.001) (Supplementary Table S1 ). It is worth noting that the EC 50 values of CQ seemed to be a little higher than that in our previous report (1.13 μM at an MOI of 0.05) 1 , which is likely due to the adaptation of the virus in cell culture that significantly increased viral infectivity upon continuous passaging. Consequently, the selectivity index (SI = CC 50 /EC 50 ) of CQ (100.81, 71.71, 38.26, and 37.12) was higher than that of HCQ (55.32, 61.45, 14.41, 19.25) at MOIs of 0.01, 0.02, 0.2, and 0.8, respectively. These results were corroborated by immunofluorescence microscopy as evidenced by different expression levels of virus nucleoprotein (NP) at the indicated drug concentrations at 48 h p.i. (Supplementary Fig. S1 ). Taken together, the data suggest that the anti-SARS-CoV-2 activity of HCQ seems to be less potent compared to CQ, at least at certain MOIs.
Both CQ and HCQ are weak bases that are known to elevate the pH of acidic intracellular organelles, such as endosomes/lysosomes, essential for membrane fusion 5 . In addition, CQ could inhibit SARS-CoV entry through changing the glycosylation of ACE2 receptor and spike protein 6 . Time-of-addition experiment confirmed that HCQ effectively inhibited the entry step, as well as the post-entry stages of SARS-CoV-2, which was also found upon CQ treatment (Supplementary Fig. S2 ). To further explore the detailed mechanism of action of CQ and HCQ in inhibiting virus entry, co-localization of virions with early endosomes (EEs) or endolysosomes (ELs) was analyzed by immunofluorescence analysis (IFA) and confocal microscopy. Quantification analysis showed that, at 90 min p.i. in untreated cells, 16.2% of internalized virions (anti-NP, red) were observed in early endosome antigen 1 (EEA1)-positive EEs (green), while more virions (34.3%) were transported into the late endosomal-lysosomal protein LAMP1 + ELs (green) (n > 30 cells for each group). By contrast, in the presence of CQ or HCQ, significantly more virions (35.3% for CQ and 29.2% for HCQ; P < 0.001) were detected in the EEs, while only very few virions (2.4% for CQ and 0.03% for HCQ; P < 0.001) were found to be co-localized with LAMP1 + ELs (n > 30 cells) (Fig. 1b, c) . This suggested that both CQ and HCQ blocked the transport of SARS-CoV-2 from EEs to ELs, which appears to be a requirement to release the viral genome as in the case of SARS-CoV 7 .
Interestingly, we found that CQ and HCQ treatment caused noticeable changes in the number and size/morphology of EEs and ELs (Fig. 1c) . In the untreated cells, most EEs were much smaller than ELs (Fig. 1c) . In CQand HCQ-treated cells, abnormally enlarged EE vesicles were observed (Fig. 1c , arrows in the upper panels), many of which are even larger than ELs in the untreated cells. This is in agreement with previous report that treatment with CQ induced the formation of expanded cytoplasmic vesicles 8 . Within the EE vesicles, virions (red) were localized around the membrane (green) of the vesicle. CQ treatment did not cause obvious changes in the number and size of ELs; however, the regular vesicle structure seemed to be disrupted, at least partially. By contrast, in HCQ-treated cells, the size and number of ELs increased significantly (Fig. 1c , arrows in the lower panels).
Since acidification is crucial for endosome maturation and function, we surmise that endosome maturation might be blocked at intermediate stages of endocytosis, resulting in failure of further transport of virions to the ultimate releasing site. CQ was reported to elevate the pH (see figure on previous page) Fig. 1 Comparative antiviral efficacy and mechanism of action of CQ and HCQ against SARS-CoV-2 infection in vitro. a Cytotoxicity and antiviral activities of CQ and HCQ. The cytotoxicity of the two drugs in Vero E6 cells was determined by CCK-8 assays. Vero E6 cells were treated with different doses of either compound or with PBS in the controls for 1 h and then infected with SARS-CoV-2 at MOIs of 0.01, 0.02, 0.2, and 0.8. The virus yield in the cell supernatant was quantified by qRT-PCR at 48 h p.i. Y-axis represents the mean of percent inhibition normalized to the PBS group. The experiments were repeated twice. b, c Mechanism of CQ and HCQ in inhibiting virus entry. Vero E6 cells were treated with CQ or HCQ (50 μM) for 1 h, followed by virus binding (MOI = 10) at 4°C for 1 h. Then the unbound virions were removed, and the cells were further supplemented with fresh drug-containing medium at 37°C for 90 min before being fixed and stained with IFA using anti-NP antibody for virions (red) and antibodies against EEA1 for EEs (green) or LAMP1 for ELs (green). The nuclei (blue) were stained with Hoechst dye. The portion of virions that co-localized with EEs or ELs in each group (n > 30 cells) was quantified and is shown in b. Representative confocal microscopic images of viral particles (red), EEA1 + EEs (green), or LAMP1 + ELs (green) in each group are displayed in c. The enlarged images in the boxes indicate a single vesicle-containing virion. The arrows indicated the abnormally enlarged vesicles. Bars, 5 μm. Statistical analysis was performed using a one-way analysis of variance (ANOVA) with GraphPad Prism (F = 102.8, df = 5,182, ***P < 0.001).
of lysosome from about 4.5 to 6.5 at 100 μM 9 . To our knowledge, there is a lack of studies on the impact of HCQ on the morphology and pH values of endosomes/ lysosomes. Our observations suggested that the mode of actions of CQ and HCQ appear to be distinct in certain aspects.
It has been reported that oral absorption of CQ and HCQ in humans is very efficient. In animals, both drugs share similar tissue distribution patterns, with high concentrations in the liver, spleen, kidney, and lung reaching levels of 200-700 times higher than those in the plasma 10 . It was reported that safe dosage (6-6.5 mg/kg per day) of HCQ sulfate could generate serum levels of 1.4-1.5 μM in humans 11 . Therefore, with a safe dosage, HCQ concentration in the above tissues is likely to be achieved to inhibit SARS-CoV-2 infection.
Clinical investigation found that high concentration of cytokines were detected in the plasma of critically ill patients infected with SARS-CoV-2, suggesting that cytokine storm was associated with disease severity 12 . Other than its direct antiviral activity, HCQ is a safe and successful anti-inflammatory agent that has been used extensively in autoimmune diseases and can significantly decrease the production of cytokines and, in particular, pro-inflammatory factors. Therefore, in COVID-19 patients, HCQ may also contribute to attenuating the inflammatory response. In conclusion, our results show that HCQ can efficiently inhibit SARS-CoV-2 infection in vitro. In combination with its anti-inflammatory function, we predict that the drug has a good potential to combat the disease. This possibility awaits confirmation by clinical trials. We need to point out, although HCQ is less toxic than CQ, prolonged and overdose usage can still cause poisoning. And the relatively low SI of HCQ requires careful designing and conducting of clinical trials to achieve efficient and safe control of the SARS-CoV-2 infection.
|
Mechanism of action of Chloroquine(CQ) and Hydroxychloroquine(HCQ) against COVID-19?
| 658 |
CQ and HCQ are weak bases that are known to elevate the pH of acidic intracellular organelles
| 6,327 |
2,439 |
Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7078228/
SHA: 1cf54d1c77b7f0494ab971429d26e0e199952d09
Authors: Liu, Jia; Cao, Ruiyuan; Xu, Mingyue; Wang, Xi; Zhang, Huanyu; Hu, Hengrui; Li, Yufeng; Hu, Zhihong; Zhong, Wu; Wang, Manli
Date: 2020-03-18
DOI: 10.1038/s41421-020-0156-0
License: cc-by
Abstract: nan
Text: Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro Jia Liu 1 , Ruiyuan Cao 2 , Mingyue Xu 1,3 , Xi Wang 1 , Huanyu Zhang 1,3 , Hengrui Hu 1,3 , Yufeng Li 1,3 , Zhihong Hu 1 , Wu Zhong 2 and Manli Wang 1 Dear Editor, The outbreak of coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2/2019-nCoV) poses a serious threat to global public health and local economies. As of March 3, 2020, over 80,000 cases have been confirmed in China, including 2946 deaths as well as over 10,566 confirmed cases in 72 other countries. Such huge numbers of infected and dead people call for an urgent demand of effective, available, and affordable drugs to control and diminish the epidemic.
We have recently reported that two drugs, remdesivir (GS-5734) and chloroquine (CQ) phosphate, efficiently inhibited SARS-CoV-2 infection in vitro 1 . Remdesivir is a nucleoside analog prodrug developed by Gilead Sciences (USA). A recent case report showed that treatment with remdesivir improved the clinical condition of the first patient infected by SARS-CoV-2 in the United States 2 , and a phase III clinical trial of remdesivir against SARS-CoV-2 was launched in Wuhan on February 4, 2020. However, as an experimental drug, remdesivir is not expected to be largely available for treating a very large number of patients in a timely manner. Therefore, of the two potential drugs, CQ appears to be the drug of choice for large-scale use due to its availability, proven safety record, and a relatively low cost. In light of the preliminary clinical data, CQ has been added to the list of trial drugs in the Guidelines for the Diagnosis and Treatment of COVID-19 (sixth edition) published by National Health Commission of the People's Republic of China. CQ (N4-(7-Chloro-4-quinolinyl)-N1,N1-diethyl-1,4pentanediamine) has long been used to treat malaria and amebiasis. However, Plasmodium falciparum developed widespread resistance to it, and with the development of new antimalarials, it has become a choice for the prophylaxis of malaria. In addition, an overdose of CQ can cause acute poisoning and death 3 . In the past years, due to infrequent utilization of CQ in clinical practice, its production and market supply was greatly reduced, at least in China. Hydroxychloroquine (HCQ) sulfate, a derivative of CQ, was first synthesized in 1946 by introducing a hydroxyl group into CQ and was demonstrated to be much less (~40%) toxic than CQ in animals 4 . More importantly, HCQ is still widely available to treat autoimmune diseases, such as systemic lupus erythematosus and rheumatoid arthritis. Since CQ and HCQ share similar chemical structures and mechanisms of acting as a weak base and immunomodulator, it is easy to conjure up the idea that HCQ may be a potent candidate to treat infection by SARS-CoV-2. Actually, as of February 23, 2020, seven clinical trial registries were found in Chinese Clinical Trial Registry (http://www.chictr.org.cn) for using HCQ to treat COVID-19. Whether HCQ is as efficacious as CQ in treating SARS-CoV-2 infection still lacks the experimental evidence.
To this end, we evaluated the antiviral effect of HCQ against SARS-CoV-2 infection in comparison to CQ in vitro. First, the cytotoxicity of HCQ and CQ in African green monkey kidney VeroE6 cells (ATCC-1586) was measured by standard CCK8 assay, and the result showed © The Author(s) 2020
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. (Fig. 1a) . To better compare the antiviral activity of CQ versus HCQ, the dose-response curves of the two compounds against SARS-CoV-2 were determined at four different multiplicities of infection (MOIs) by quantification of viral RNA copy numbers in the cell supernatant at 48 h post infection (p.i.). The data summarized in Fig. 1a and Supplementary Table S1 show that, at all MOIs (0.01, 0.02, 0.2, and 0.8), the 50% maximal effective concentration (EC 50 ) for CQ (2.71, 3.81, 7.14, and 7.36 μM) was lower than that of HCQ (4.51, 4.06, 17.31, and 12.96 μM). The differences in EC 50 values were statistically significant at an MOI of 0.01 (P < 0.05) and MOI of 0.2 (P < 0.001) (Supplementary Table S1 ). It is worth noting that the EC 50 values of CQ seemed to be a little higher than that in our previous report (1.13 μM at an MOI of 0.05) 1 , which is likely due to the adaptation of the virus in cell culture that significantly increased viral infectivity upon continuous passaging. Consequently, the selectivity index (SI = CC 50 /EC 50 ) of CQ (100.81, 71.71, 38.26, and 37.12) was higher than that of HCQ (55.32, 61.45, 14.41, 19.25) at MOIs of 0.01, 0.02, 0.2, and 0.8, respectively. These results were corroborated by immunofluorescence microscopy as evidenced by different expression levels of virus nucleoprotein (NP) at the indicated drug concentrations at 48 h p.i. (Supplementary Fig. S1 ). Taken together, the data suggest that the anti-SARS-CoV-2 activity of HCQ seems to be less potent compared to CQ, at least at certain MOIs.
Both CQ and HCQ are weak bases that are known to elevate the pH of acidic intracellular organelles, such as endosomes/lysosomes, essential for membrane fusion 5 . In addition, CQ could inhibit SARS-CoV entry through changing the glycosylation of ACE2 receptor and spike protein 6 . Time-of-addition experiment confirmed that HCQ effectively inhibited the entry step, as well as the post-entry stages of SARS-CoV-2, which was also found upon CQ treatment (Supplementary Fig. S2 ). To further explore the detailed mechanism of action of CQ and HCQ in inhibiting virus entry, co-localization of virions with early endosomes (EEs) or endolysosomes (ELs) was analyzed by immunofluorescence analysis (IFA) and confocal microscopy. Quantification analysis showed that, at 90 min p.i. in untreated cells, 16.2% of internalized virions (anti-NP, red) were observed in early endosome antigen 1 (EEA1)-positive EEs (green), while more virions (34.3%) were transported into the late endosomal-lysosomal protein LAMP1 + ELs (green) (n > 30 cells for each group). By contrast, in the presence of CQ or HCQ, significantly more virions (35.3% for CQ and 29.2% for HCQ; P < 0.001) were detected in the EEs, while only very few virions (2.4% for CQ and 0.03% for HCQ; P < 0.001) were found to be co-localized with LAMP1 + ELs (n > 30 cells) (Fig. 1b, c) . This suggested that both CQ and HCQ blocked the transport of SARS-CoV-2 from EEs to ELs, which appears to be a requirement to release the viral genome as in the case of SARS-CoV 7 .
Interestingly, we found that CQ and HCQ treatment caused noticeable changes in the number and size/morphology of EEs and ELs (Fig. 1c) . In the untreated cells, most EEs were much smaller than ELs (Fig. 1c) . In CQand HCQ-treated cells, abnormally enlarged EE vesicles were observed (Fig. 1c , arrows in the upper panels), many of which are even larger than ELs in the untreated cells. This is in agreement with previous report that treatment with CQ induced the formation of expanded cytoplasmic vesicles 8 . Within the EE vesicles, virions (red) were localized around the membrane (green) of the vesicle. CQ treatment did not cause obvious changes in the number and size of ELs; however, the regular vesicle structure seemed to be disrupted, at least partially. By contrast, in HCQ-treated cells, the size and number of ELs increased significantly (Fig. 1c , arrows in the lower panels).
Since acidification is crucial for endosome maturation and function, we surmise that endosome maturation might be blocked at intermediate stages of endocytosis, resulting in failure of further transport of virions to the ultimate releasing site. CQ was reported to elevate the pH (see figure on previous page) Fig. 1 Comparative antiviral efficacy and mechanism of action of CQ and HCQ against SARS-CoV-2 infection in vitro. a Cytotoxicity and antiviral activities of CQ and HCQ. The cytotoxicity of the two drugs in Vero E6 cells was determined by CCK-8 assays. Vero E6 cells were treated with different doses of either compound or with PBS in the controls for 1 h and then infected with SARS-CoV-2 at MOIs of 0.01, 0.02, 0.2, and 0.8. The virus yield in the cell supernatant was quantified by qRT-PCR at 48 h p.i. Y-axis represents the mean of percent inhibition normalized to the PBS group. The experiments were repeated twice. b, c Mechanism of CQ and HCQ in inhibiting virus entry. Vero E6 cells were treated with CQ or HCQ (50 μM) for 1 h, followed by virus binding (MOI = 10) at 4°C for 1 h. Then the unbound virions were removed, and the cells were further supplemented with fresh drug-containing medium at 37°C for 90 min before being fixed and stained with IFA using anti-NP antibody for virions (red) and antibodies against EEA1 for EEs (green) or LAMP1 for ELs (green). The nuclei (blue) were stained with Hoechst dye. The portion of virions that co-localized with EEs or ELs in each group (n > 30 cells) was quantified and is shown in b. Representative confocal microscopic images of viral particles (red), EEA1 + EEs (green), or LAMP1 + ELs (green) in each group are displayed in c. The enlarged images in the boxes indicate a single vesicle-containing virion. The arrows indicated the abnormally enlarged vesicles. Bars, 5 μm. Statistical analysis was performed using a one-way analysis of variance (ANOVA) with GraphPad Prism (F = 102.8, df = 5,182, ***P < 0.001).
of lysosome from about 4.5 to 6.5 at 100 μM 9 . To our knowledge, there is a lack of studies on the impact of HCQ on the morphology and pH values of endosomes/ lysosomes. Our observations suggested that the mode of actions of CQ and HCQ appear to be distinct in certain aspects.
It has been reported that oral absorption of CQ and HCQ in humans is very efficient. In animals, both drugs share similar tissue distribution patterns, with high concentrations in the liver, spleen, kidney, and lung reaching levels of 200-700 times higher than those in the plasma 10 . It was reported that safe dosage (6-6.5 mg/kg per day) of HCQ sulfate could generate serum levels of 1.4-1.5 μM in humans 11 . Therefore, with a safe dosage, HCQ concentration in the above tissues is likely to be achieved to inhibit SARS-CoV-2 infection.
Clinical investigation found that high concentration of cytokines were detected in the plasma of critically ill patients infected with SARS-CoV-2, suggesting that cytokine storm was associated with disease severity 12 . Other than its direct antiviral activity, HCQ is a safe and successful anti-inflammatory agent that has been used extensively in autoimmune diseases and can significantly decrease the production of cytokines and, in particular, pro-inflammatory factors. Therefore, in COVID-19 patients, HCQ may also contribute to attenuating the inflammatory response. In conclusion, our results show that HCQ can efficiently inhibit SARS-CoV-2 infection in vitro. In combination with its anti-inflammatory function, we predict that the drug has a good potential to combat the disease. This possibility awaits confirmation by clinical trials. We need to point out, although HCQ is less toxic than CQ, prolonged and overdose usage can still cause poisoning. And the relatively low SI of HCQ requires careful designing and conducting of clinical trials to achieve efficient and safe control of the SARS-CoV-2 infection.
|
What is the effect of Chloroquine(CQ) and Hydroxychloroquine(HCQ) on endosomal maturation and endocytosis in COVID-19 treatment?
| 661 |
Since acidification is crucial for endosome maturation and function, we surmise that endosome maturation might be blocked at intermediate stages of endocytosis,
| 8,734 |
2,439 |
Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7078228/
SHA: 1cf54d1c77b7f0494ab971429d26e0e199952d09
Authors: Liu, Jia; Cao, Ruiyuan; Xu, Mingyue; Wang, Xi; Zhang, Huanyu; Hu, Hengrui; Li, Yufeng; Hu, Zhihong; Zhong, Wu; Wang, Manli
Date: 2020-03-18
DOI: 10.1038/s41421-020-0156-0
License: cc-by
Abstract: nan
Text: Hydroxychloroquine, a less toxic derivative of chloroquine, is effective in inhibiting SARS-CoV-2 infection in vitro Jia Liu 1 , Ruiyuan Cao 2 , Mingyue Xu 1,3 , Xi Wang 1 , Huanyu Zhang 1,3 , Hengrui Hu 1,3 , Yufeng Li 1,3 , Zhihong Hu 1 , Wu Zhong 2 and Manli Wang 1 Dear Editor, The outbreak of coronavirus disease 2019 (COVID-19) caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2/2019-nCoV) poses a serious threat to global public health and local economies. As of March 3, 2020, over 80,000 cases have been confirmed in China, including 2946 deaths as well as over 10,566 confirmed cases in 72 other countries. Such huge numbers of infected and dead people call for an urgent demand of effective, available, and affordable drugs to control and diminish the epidemic.
We have recently reported that two drugs, remdesivir (GS-5734) and chloroquine (CQ) phosphate, efficiently inhibited SARS-CoV-2 infection in vitro 1 . Remdesivir is a nucleoside analog prodrug developed by Gilead Sciences (USA). A recent case report showed that treatment with remdesivir improved the clinical condition of the first patient infected by SARS-CoV-2 in the United States 2 , and a phase III clinical trial of remdesivir against SARS-CoV-2 was launched in Wuhan on February 4, 2020. However, as an experimental drug, remdesivir is not expected to be largely available for treating a very large number of patients in a timely manner. Therefore, of the two potential drugs, CQ appears to be the drug of choice for large-scale use due to its availability, proven safety record, and a relatively low cost. In light of the preliminary clinical data, CQ has been added to the list of trial drugs in the Guidelines for the Diagnosis and Treatment of COVID-19 (sixth edition) published by National Health Commission of the People's Republic of China. CQ (N4-(7-Chloro-4-quinolinyl)-N1,N1-diethyl-1,4pentanediamine) has long been used to treat malaria and amebiasis. However, Plasmodium falciparum developed widespread resistance to it, and with the development of new antimalarials, it has become a choice for the prophylaxis of malaria. In addition, an overdose of CQ can cause acute poisoning and death 3 . In the past years, due to infrequent utilization of CQ in clinical practice, its production and market supply was greatly reduced, at least in China. Hydroxychloroquine (HCQ) sulfate, a derivative of CQ, was first synthesized in 1946 by introducing a hydroxyl group into CQ and was demonstrated to be much less (~40%) toxic than CQ in animals 4 . More importantly, HCQ is still widely available to treat autoimmune diseases, such as systemic lupus erythematosus and rheumatoid arthritis. Since CQ and HCQ share similar chemical structures and mechanisms of acting as a weak base and immunomodulator, it is easy to conjure up the idea that HCQ may be a potent candidate to treat infection by SARS-CoV-2. Actually, as of February 23, 2020, seven clinical trial registries were found in Chinese Clinical Trial Registry (http://www.chictr.org.cn) for using HCQ to treat COVID-19. Whether HCQ is as efficacious as CQ in treating SARS-CoV-2 infection still lacks the experimental evidence.
To this end, we evaluated the antiviral effect of HCQ against SARS-CoV-2 infection in comparison to CQ in vitro. First, the cytotoxicity of HCQ and CQ in African green monkey kidney VeroE6 cells (ATCC-1586) was measured by standard CCK8 assay, and the result showed © The Author(s) 2020
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. (Fig. 1a) . To better compare the antiviral activity of CQ versus HCQ, the dose-response curves of the two compounds against SARS-CoV-2 were determined at four different multiplicities of infection (MOIs) by quantification of viral RNA copy numbers in the cell supernatant at 48 h post infection (p.i.). The data summarized in Fig. 1a and Supplementary Table S1 show that, at all MOIs (0.01, 0.02, 0.2, and 0.8), the 50% maximal effective concentration (EC 50 ) for CQ (2.71, 3.81, 7.14, and 7.36 μM) was lower than that of HCQ (4.51, 4.06, 17.31, and 12.96 μM). The differences in EC 50 values were statistically significant at an MOI of 0.01 (P < 0.05) and MOI of 0.2 (P < 0.001) (Supplementary Table S1 ). It is worth noting that the EC 50 values of CQ seemed to be a little higher than that in our previous report (1.13 μM at an MOI of 0.05) 1 , which is likely due to the adaptation of the virus in cell culture that significantly increased viral infectivity upon continuous passaging. Consequently, the selectivity index (SI = CC 50 /EC 50 ) of CQ (100.81, 71.71, 38.26, and 37.12) was higher than that of HCQ (55.32, 61.45, 14.41, 19.25) at MOIs of 0.01, 0.02, 0.2, and 0.8, respectively. These results were corroborated by immunofluorescence microscopy as evidenced by different expression levels of virus nucleoprotein (NP) at the indicated drug concentrations at 48 h p.i. (Supplementary Fig. S1 ). Taken together, the data suggest that the anti-SARS-CoV-2 activity of HCQ seems to be less potent compared to CQ, at least at certain MOIs.
Both CQ and HCQ are weak bases that are known to elevate the pH of acidic intracellular organelles, such as endosomes/lysosomes, essential for membrane fusion 5 . In addition, CQ could inhibit SARS-CoV entry through changing the glycosylation of ACE2 receptor and spike protein 6 . Time-of-addition experiment confirmed that HCQ effectively inhibited the entry step, as well as the post-entry stages of SARS-CoV-2, which was also found upon CQ treatment (Supplementary Fig. S2 ). To further explore the detailed mechanism of action of CQ and HCQ in inhibiting virus entry, co-localization of virions with early endosomes (EEs) or endolysosomes (ELs) was analyzed by immunofluorescence analysis (IFA) and confocal microscopy. Quantification analysis showed that, at 90 min p.i. in untreated cells, 16.2% of internalized virions (anti-NP, red) were observed in early endosome antigen 1 (EEA1)-positive EEs (green), while more virions (34.3%) were transported into the late endosomal-lysosomal protein LAMP1 + ELs (green) (n > 30 cells for each group). By contrast, in the presence of CQ or HCQ, significantly more virions (35.3% for CQ and 29.2% for HCQ; P < 0.001) were detected in the EEs, while only very few virions (2.4% for CQ and 0.03% for HCQ; P < 0.001) were found to be co-localized with LAMP1 + ELs (n > 30 cells) (Fig. 1b, c) . This suggested that both CQ and HCQ blocked the transport of SARS-CoV-2 from EEs to ELs, which appears to be a requirement to release the viral genome as in the case of SARS-CoV 7 .
Interestingly, we found that CQ and HCQ treatment caused noticeable changes in the number and size/morphology of EEs and ELs (Fig. 1c) . In the untreated cells, most EEs were much smaller than ELs (Fig. 1c) . In CQand HCQ-treated cells, abnormally enlarged EE vesicles were observed (Fig. 1c , arrows in the upper panels), many of which are even larger than ELs in the untreated cells. This is in agreement with previous report that treatment with CQ induced the formation of expanded cytoplasmic vesicles 8 . Within the EE vesicles, virions (red) were localized around the membrane (green) of the vesicle. CQ treatment did not cause obvious changes in the number and size of ELs; however, the regular vesicle structure seemed to be disrupted, at least partially. By contrast, in HCQ-treated cells, the size and number of ELs increased significantly (Fig. 1c , arrows in the lower panels).
Since acidification is crucial for endosome maturation and function, we surmise that endosome maturation might be blocked at intermediate stages of endocytosis, resulting in failure of further transport of virions to the ultimate releasing site. CQ was reported to elevate the pH (see figure on previous page) Fig. 1 Comparative antiviral efficacy and mechanism of action of CQ and HCQ against SARS-CoV-2 infection in vitro. a Cytotoxicity and antiviral activities of CQ and HCQ. The cytotoxicity of the two drugs in Vero E6 cells was determined by CCK-8 assays. Vero E6 cells were treated with different doses of either compound or with PBS in the controls for 1 h and then infected with SARS-CoV-2 at MOIs of 0.01, 0.02, 0.2, and 0.8. The virus yield in the cell supernatant was quantified by qRT-PCR at 48 h p.i. Y-axis represents the mean of percent inhibition normalized to the PBS group. The experiments were repeated twice. b, c Mechanism of CQ and HCQ in inhibiting virus entry. Vero E6 cells were treated with CQ or HCQ (50 μM) for 1 h, followed by virus binding (MOI = 10) at 4°C for 1 h. Then the unbound virions were removed, and the cells were further supplemented with fresh drug-containing medium at 37°C for 90 min before being fixed and stained with IFA using anti-NP antibody for virions (red) and antibodies against EEA1 for EEs (green) or LAMP1 for ELs (green). The nuclei (blue) were stained with Hoechst dye. The portion of virions that co-localized with EEs or ELs in each group (n > 30 cells) was quantified and is shown in b. Representative confocal microscopic images of viral particles (red), EEA1 + EEs (green), or LAMP1 + ELs (green) in each group are displayed in c. The enlarged images in the boxes indicate a single vesicle-containing virion. The arrows indicated the abnormally enlarged vesicles. Bars, 5 μm. Statistical analysis was performed using a one-way analysis of variance (ANOVA) with GraphPad Prism (F = 102.8, df = 5,182, ***P < 0.001).
of lysosome from about 4.5 to 6.5 at 100 μM 9 . To our knowledge, there is a lack of studies on the impact of HCQ on the morphology and pH values of endosomes/ lysosomes. Our observations suggested that the mode of actions of CQ and HCQ appear to be distinct in certain aspects.
It has been reported that oral absorption of CQ and HCQ in humans is very efficient. In animals, both drugs share similar tissue distribution patterns, with high concentrations in the liver, spleen, kidney, and lung reaching levels of 200-700 times higher than those in the plasma 10 . It was reported that safe dosage (6-6.5 mg/kg per day) of HCQ sulfate could generate serum levels of 1.4-1.5 μM in humans 11 . Therefore, with a safe dosage, HCQ concentration in the above tissues is likely to be achieved to inhibit SARS-CoV-2 infection.
Clinical investigation found that high concentration of cytokines were detected in the plasma of critically ill patients infected with SARS-CoV-2, suggesting that cytokine storm was associated with disease severity 12 . Other than its direct antiviral activity, HCQ is a safe and successful anti-inflammatory agent that has been used extensively in autoimmune diseases and can significantly decrease the production of cytokines and, in particular, pro-inflammatory factors. Therefore, in COVID-19 patients, HCQ may also contribute to attenuating the inflammatory response. In conclusion, our results show that HCQ can efficiently inhibit SARS-CoV-2 infection in vitro. In combination with its anti-inflammatory function, we predict that the drug has a good potential to combat the disease. This possibility awaits confirmation by clinical trials. We need to point out, although HCQ is less toxic than CQ, prolonged and overdose usage can still cause poisoning. And the relatively low SI of HCQ requires careful designing and conducting of clinical trials to achieve efficient and safe control of the SARS-CoV-2 infection.
|
Evidence of Hydroxychloroquine(HCQ) being anti inflammatory in SARS-CoV-2 critically ill patients with elevated plasma cytokines?
| 662 |
HCQ is a safe and successful anti-inflammatory agent
| 11,799 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What proportion of healthcare workers reported symptoms of depression?
| 3,459 |
50.4%
| 2,058 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What proportion of healthcare workers reported symptoms of anxiety?
| 3,460 |
[44.6%
| 2,080 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What proportion of healthcare workers reported symptoms of insomnia?
| 3,461 |
34.0%
| 2,105 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What proportion reported distress?
| 3,462 |
[71.5%
| 2,132 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What were the Generalized Anxiety Disorder scale scores?
| 3,463 |
among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001
| 2,530 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What were the Insomnia Severity Index scores ?
| 3,464 |
among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001
| 2,634 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What were the Impact of Event Scale–Revised scores?
| 3,465 |
21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001).
| 2,844 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What were the results of analysis?
| 3,466 |
participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008)
| 2,999 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What were the results of analysis?
| 3,467 |
Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001).
| 3,192 |
2,432 |
Factors Associated With Mental Health Outcomes Among Health Care Workers Exposed to Coronavirus Disease 2019
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7090843/
SHA: 0a08fddd9dcee1b1254a05b49113521bbc423ccd
Authors: Lai, Jianbo; Ma, Simeng; Wang, Ying; Cai, Zhongxiang; Hu, Jianbo; Wei, Ning; Wu, Jiang; Du, Hui; Chen, Tingting; Li, Ruiting; Tan, Huawei; Kang, Lijun; Yao, Lihua; Huang, Manli; Wang, Huafen; Wang, Gaohua; Liu, Zhongchun; Hu, Shaohua
Date: 2020-03-23
DOI: 10.1001/jamanetworkopen.2020.3976
License: cc-by
Abstract: IMPORTANCE: Health care workers exposed to coronavirus disease 2019 (COVID-19) could be psychologically stressed. OBJECTIVE: To assess the magnitude of mental health outcomes and associated factors among health care workers treating patients exposed to COVID-19 in China. DESIGN, SETTINGS, AND PARTICIPANTS: This cross-sectional, survey-based, region-stratified study collected demographic data and mental health measurements from 1257 health care workers in 34 hospitals from January 29, 2020, to February 3, 2020, in China. Health care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 were eligible. MAIN OUTCOMES AND MEASURES: The degree of symptoms of depression, anxiety, insomnia, and distress was assessed by the Chinese versions of the 9-item Patient Health Questionnaire, the 7-item Generalized Anxiety Disorder scale, the 7-item Insomnia Severity Index, and the 22-item Impact of Event Scale–Revised, respectively. Multivariable logistic regression analysis was performed to identify factors associated with mental health outcomes. RESULTS: A total of 1257 of 1830 contacted individuals completed the survey, with a participation rate of 68.7%. A total of 813 (64.7%) were aged 26 to 40 years, and 964 (76.7%) were women. Of all participants, 764 (60.8%) were nurses, and 493 (39.2%) were physicians; 760 (60.5%) worked in hospitals in Wuhan, and 522 (41.5%) were frontline health care workers. A considerable proportion of participants reported symptoms of depression (634 [50.4%]), anxiety (560 [44.6%]), insomnia (427 [34.0%]), and distress (899 [71.5%]). Nurses, women, frontline health care workers, and those working in Wuhan, China, reported more severe degrees of all measurements of mental health symptoms than other health care workers (eg, median [IQR] Patient Health Questionnaire scores among physicians vs nurses: 4.0 [1.0-7.0] vs 5.0 [2.0-8.0]; P = .007; median [interquartile range {IQR}] Generalized Anxiety Disorder scale scores among men vs women: 2.0 [0-6.0] vs 4.0 [1.0-7.0]; P < .001; median [IQR] Insomnia Severity Index scores among frontline vs second-line workers: 6.0 [2.0-11.0] vs 4.0 [1.0-8.0]; P < .001; median [IQR] Impact of Event Scale–Revised scores among those in Wuhan vs those in Hubei outside Wuhan and those outside Hubei: 21.0 [8.5-34.5] vs 18.0 [6.0-28.0] in Hubei outside Wuhan and 15.0 [4.0-26.0] outside Hubei; P < .001). Multivariable logistic regression analysis showed participants from outside Hubei province were associated with lower risk of experiencing symptoms of distress compared with those in Wuhan (odds ratio [OR], 0.62; 95% CI, 0.43-0.88; P = .008). Frontline health care workers engaged in direct diagnosis, treatment, and care of patients with COVID-19 were associated with a higher risk of symptoms of depression (OR, 1.52; 95% CI, 1.11-2.09; P = .01), anxiety (OR, 1.57; 95% CI, 1.22-2.02; P < .001), insomnia (OR, 2.97; 95% CI, 1.92-4.60; P < .001), and distress (OR, 1.60; 95% CI, 1.25-2.04; P < .001). CONCLUSIONS AND RELEVANCE: In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
Text: Abbreviation: PHQ-9, 9-item Patient Health Questionnaire; GAD-7, 7-item Generalized Anxiety Disorder; ISI, 7-item Insomnia Severity Index; IES-R, 22-item Impact of Event Abbreviation: IES-R, 22-item Impact of Event Scale-Revised; IQR, interquartile range. Hyperarousal, median (IQR) 6.0(2.0, 10.0) 6.0(2.0, 9.0) .29
|
What are the conclusions of this study?
| 3,468 |
In this survey of heath care workers in hospitals equipped with fever clinics or wards for patients with COVID-19 in Wuhan and other regions in China, participants reported experiencing psychological burden, especially nurses, women, those in Wuhan, and frontline health care workers directly engaged in the diagnosis, treatment, and care for patients with COVID-19.
| 3,578 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What are microtubule severing enzymes?
| 924 |
a family of AAA-ATPase proteins
| 2,018 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What genetic mutation is associated with hereditary spastic paraplegia?
| 926 |
SPG4
| 2,332 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What genetic mutation is associated with cerebral malformations?
| 927 |
KATNB1
| 2,338 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What genetic mutation is associated with autism?
| 928 |
KATNAL2
| 2,349 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What is KATNAL1?
| 929 |
microtubule severing enzyme
| 2,548 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What organ is most associated with the KATNAL1 gene?
| 930 |
central nervous system
| 3,332 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What CNS functions can be measured by studying the movement of mice in a T-maze?
| 931 |
working memory and spatial memory
| 11,976 |
1,620 |
A missense mutation in Katnal1 underlies behavioural, neurological and ciliary anomalies
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5761721/
SHA: f4cebabd74b16e710fb41a737d8ef84b7d565d8d
Authors: Banks, G; Lassi, G; Hoerder-Suabedissen, A; Tinarelli, F; Simon, M M; Wilcox, A; Lau, P; Lawson, T N; Johnson, S; Rutman, A; Sweeting, M; Chesham, J E; Barnard, A R; Horner, N; Westerberg, H; Smith, L B; Molnár, Z; Hastings, M H; Hirst, R A; Tucci, V; Nolan, P M
Date: 2017-04-04
DOI: 10.1038/mp.2017.54
License: cc-by
Abstract: Microtubule severing enzymes implement a diverse range of tissue-specific molecular functions throughout development and into adulthood. Although microtubule severing is fundamental to many dynamic neural processes, little is known regarding the role of the family member Katanin p60 subunit A-like 1, KATNAL1, in central nervous system (CNS) function. Recent studies reporting that microdeletions incorporating the KATNAL1 locus in humans result in intellectual disability and microcephaly suggest that KATNAL1 may play a prominent role in the CNS; however, such associations lack the functional data required to highlight potential mechanisms which link the gene to disease symptoms. Here we identify and characterise a mouse line carrying a loss of function allele in Katnal1. We show that mutants express behavioural deficits including in circadian rhythms, sleep, anxiety and learning/memory. Furthermore, in the brains of Katnal1 mutant mice we reveal numerous morphological abnormalities and defects in neuronal migration and morphology. Furthermore we demonstrate defects in the motile cilia of the ventricular ependymal cells of mutants, suggesting a role for Katnal1 in the development of ciliary function. We believe the data we present here are the first to associate KATNAL1 with such phenotypes, demonstrating that the protein plays keys roles in a number of processes integral to the development of neuronal function and behaviour.
Text: Microtubule severing enzymes are a family of AAA-ATPase proteins that participate in fundamental cellular processes such as mitosis, ciliary biogenesis and growth cone motility. In neurons this family is known to control such processes as axonal elongation 1 and synaptic development. 2 In addition, mutations in microtubule severing enzyme genes SPG4, KATNB1 and KATNAL2 are associated with hereditary spastic paraplegia, cerebral malformations and autism, respectively, [3] [4] [5] [6] and mutations in Fign cause a range of phenotypes in mice. 7 Currently the microtubule severing enzyme KATNAL1 is poorly characterised and it is not yet understood how the enzyme functions in the nervous system. Recent evidence from genetic characterisation of human patients suggests that haploinsufficiency of KATNAL1 is linked with a number of symptoms including intellectual disability (ID) and craniofacial dysmorphologies. 8, 9 It is also notable that a very rare KATNAL1 mutation has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/genebook/gene book.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and that Peters syndrome and autism have both been associated with the chromosomal region containing the KATNAL1 locus. 11, 12 Although such association studies strongly suggest that KATNAL1 plays a fundamental role in the central nervous system (CNS), additional studies using cellular or animals models are required to understand how the gene may be causative for disease. Here we present the first study describing neural and behavioural deficits associated with a loss of function allele of Katnal1 in the mouse. This mutant mouse line was independently identified in two parallel phenotyping screens, which demonstrated that mutant mice showed both male sterility and circadian phenotypes. Subsequent behavioural investigations demonstrated that this mutation is associated with anxiety and memory deficits. Underlying these behavioural phenotypes, we identified histopathological abnormalities in the brains of Katnal1 1H/1H mutants, including disordered cellular layers in the hippocampus and cortex and substantially larger ventricles. Further investigations demonstrated that Katnal1 1H/1H mice show neuronal migration and ciliary function deficits suggesting KATNAL1 plays an essential role in these processes. These findings are the first to our knowledge to conclusively show that mutations in Katnal1 lead to behavioural and neuronal disturbances and provide insight regarding the clinical associations that have been linked to the gene. performed on mouse cohorts that were partially or completely congenic on the C57BL/6 J background.
Circadian wheel running was performed as previously described. 14 Sleep assessment by electroencephalography and electromyography Electroencephalography and electromyography was performed as previously described. 15 Behavioural phenotyping Spontaneous alternation. Mice were placed in a walled T-maze (black polyvinyl chloride, lined with sawdust; stem = 88 × 13 cm; arms = 32 × 13 cm) and allowed to enter a goal arm of their choice. The mouse was confined in the goal arm for 30 s, before being allowed a second free choice of goal arm. An alternation was recorded if the second choice differed from that of the first. One trial was performed per day for 10 days.
Open field behaviour. Mice were placed into a walled arena (grey polyvinyl chloride; 45 × 45 cm) and allowed to explore for 20 min. Animals were monitored by EthoVision XT analysis software (Noldus, Wageningen, Netherlands).
Video tracking in the home cage. Activity in the home cage was recorded by video tracking as previously described. 16 Morris water maze and ultrasonic vocalisation. These tests were performed as previously described. 17 Brain histology and immunofluorescence Brains were mounted in OCT (VWR) and 12 μm coronal sections taken. Sections were stained with hematoxylin and eosin, or immunolabelled following standard protocols.
In vivo neuronal migration assessment was performed as previously described 18 using embryos at either E13 or E15 (three mothers per age group) and pups at P9. Cell counts were performed using ImageJ (NIH, Bethesda, MD, USA).
In vitro neuronal migration assessment was performed using a Boyden chamber migration protocol as previously described. 19 Micro-computed tomography scanning Micro-computed tomography was performed using a Skyscan 1172 at 90 kV, 112 μA using an aluminium and copper filter, a rotation step of 0.250 degrees and a pixel size of 4.96 μm.
Segmentation, volume calculation and 3D modelling was performed using ITK-SNAP version 3.0.0 (ref. 20) and 3DSlicer. 21 Golgi-Cox staining of neurons Golgi-Cox neuronal staining was performed using the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Columbia, MD, USA). Neurons were analysed using ImageJ.
Brains from P2 mice were dissected, and the dorsal cerebral half was sectioned (250 μm) through the floor of the lateral and 3rd ventricle using a vibratome. Ciliary beat frequency and pattern was analysed as previously described. 22 Electron microscopy For Scanning Electron Microscopy the ependymal lining of the lateral ventricle was fixed in 2.5% glutaraldehyde, 2% paraformaldehyde in 0.1 M phosphate buffer, incubated in 2% osmium tetroxide, and dehydrated through ethanol solutions. Samples were critical point dried using an Emitech K850 (Quorum Technologies, East Sussex, UK), coated with platinum using a Quorom Q150R S sputter coater (Quorum Technologies). and visualised using a JEOL LSM-6010 scanning electron microscope (Jeol, Herts, UK).
Transmission electron microscopy was performed as previously described. 22 Statistical analysis Data was analysed using two-tailed students T test or AVOVA using SPSS (IBM) or GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA). Significance level for all analysis was set at Po 0.05. All graphs are presented showing mean ± s.e.m.
Additional and more detailed methods can be found in supplementary information.
Identification and cloning of the Katnal1 1H mutation To identify novel gene mutations affecting circadian behaviour we undertook a circadian running wheel screen of pedigrees of N-ethyl-N-nitrosourea mutagenised mice. 13 In one pedigree 17.65% of animals showed a short circadian period in constant darkness (o 23 h observed in 12 out of 68 animals screened). An outcross using an affected female produced no affected animals (33 animals screened). In subsequent intercross screens 15.5% of animals were affected (53 out of 342 animals screened), suggesting that the pedigree carries a mutation causing a recessive circadian phenotype which is 60% penetrant. We found no gender bias in affected animals (proportion of affected animals: male = 47.2%; female = 52.8%).
Concurrently a male sterility phenotype was identified within the same pedigree. 23 Genome-wide SNP linkage analysis mapped the circadian and sterility phenotypes to the same region on chromosome 5 and subsequent sequencing identified the causative mutation as a T to G single point mutation within exon seven of the Katnal1 gene. For full details of mapping and identification of the mutation see reference 23. This mutant allele was designated Katnal1 1H , and results in a leucine to valine substitution at residue 286 of the protein. In vitro functional analysis demonstrated that the mutation is a recessive loss-offunction allele. 23 3D modelling of the protein suggests that this loss of function is due to hydrophobic changes in the AAA domain of the enzyme (Supplementary Figure S1 ). Genotyping confirmed that the mutation was homozygous in affected circadian animals and wild type or heterozygous in unaffected animals, confirming that Katnal1 1H was causative for the circadian phenotype.
Circadian and sleep anomalies in Katnal1 1H/1H mice More extensive circadian phenotyping conducted on Katnal1 homozygotes (Katnal1 1H/1H ) and wild-type littermates (Katnal1 +/+ ) confirmed that Katnal1 1H/1H mice had a shorter free-running circadian period (Figures 1a-c) and furthermore revealed that Katnal1 1H/1H animals were more active in the light phase of the light/dark cycle (Figure 1d ), showed increased anticipation of light to dark transitions and greater shift in activity onset when released from light/dark cycles to constant darkness ( Figure 1e ). Data and cohort details are given in Supplementary Table S1 . Bioluminescence recordings performed using PER2::LUCIFERASE reporter mice carrying the Katnal1 1H mutation revealed that these circadian changes were not due to changes to the core molecular clock of the suprachiasmatic nucleus (the site of the master circadian clock in the brain; Supplementary Figure S2 ).
Circadian disruptions are often associated with deficits in sleep homeostasis. Therefore to complement our circadian studies we conducted wireless electroencephalography recordings over a baseline period of 24 h and following a 6 h period of sleep deprivation. A detailed summary of electroencephalography analysis is given in Supplementary Table S1. Compared to wildtype littermates, the non-REM delta power of Katnal1 1H/1H mice was higher in the dark phase of baseline sleep (mixed ANOVA, interaction factors 'genotype X time, F(1,88) = 8.91, P = 0.0175) ( Figure 1f ) and in both the light and dark phases of recovery sleep (mixed ANOVA, interaction factors 'genotype X time', F(1,136) = 11.93, P = 0.0086; Figure 1g ). All other sleep parameters were unaffected in Katnal1 1H/1H animals.
Katnal1 1H/1H mice display a spectrum of behavioural deficits Human patients carrying a heterozygous deletion incorporating the Katnal1 locus show a number of cognitive deficits including ID and a delay in language acquisition. 8, 9 We therefore investigated whether these deficits were modelled in Katnal1 1H/1H mice by subjecting animal cohorts to a battery of behavioural tests. Data and cohort details are given in Supplementary Table S2 .
Both working memory and spatial memory were significantly poorer in Katnal1 1H/1H mice, as evidenced by reduced spontaneous alternations in a T-maze ( Figure 2a ) and in the Morris water maze where mutants take longer to find the platform in acquisition trials (Figure 2b Compared to wild-type littermates, Katnal1 1H/1H animals have a shorter period (c), are more active in the light phase of the light/dark cycle (d) and show an earlier onset of activity in light/dark transitions and in the transition from light/dark cycles to constant darkness (e). In EEG recordings during sleep, Katnal1 1H/1H mice show increased non-REM delta power in the dark phase of the light/dark cycle (f) and following sleep deprivation (g). *P ⩽ 0.05; **P ⩽ 0.01; ***P ⩽ 0.001. EEG, electroencephalography; DD, constant darkness; LD, light/dark cycle. type = 164 ± 12 m, Katnal1 1H/1H = 243 ± 20 m, P = 0.02; distance travelled in periphery of open field: wild type = 4.3 ± 0.2 m, Katnal1 1H/1H = 6 ± 0.3 m, P = 0.004). Conversely when mouse activity was recorded in the home cage, we found no difference between genotypes (distance travelled over 24 h: wild type = 399 ± 77 m, Katnal1 1H/1H = 418 ± 41 m, P = 0.833) suggesting that the former activity differences were due to the novel environment of the open field rather than generalised hyperactivity in Katnal1 1H/1H animals. Finally, in certain conditions (such as maternal separation) mice emit ultrasonic vocalisations (USVs). To test whether Katnal1 1H/1H animals vocalised differently to wild types, we separated pups at postnatal days 7-8 (the age at which mice show peak of USV emission 24 ) and recorded their USVs. In these tests, compared to wild types, Katnal1 1H/1H pups produced fewer ( Figure 2g ) and shorter (Figure 2h ) vocalisations, containing fewer phrases (Figure 2i ). Gross brain morphological abnormalities in Katnal1 1H/1H mice Since we observed a number of behavioural phenotypes in Katnal1 1H/1H mice, we performed histological analysis to ascertain whether differences in brain histology underlied these behaviours. Data and cohort details are given in Supplementary Table S3 . Analysis of hematoxylin and eosin stained brain sections revealed that, compared to wildtype littermates, Katnal1 1H/1H animals had less tightly packed pyramidal cell layers in the hippocampus (Figures 3a and b) and a narrower cortical layer 1 and wider cortical layer 6 (Figures 3c-e) . To confirm these cortical layer differences, immunofluorescence was performed using the (Figures 3l and m) . Quantification of fluorescence intensity demonstrated that in Katnal1 1H/1H cortex both calbindin and CUX1 labelling was more intense closer to the cortical surface, which is consistent with the reduction in the size of layer 1 (two-way analysis of variance (ANOVA), interaction factors 'genotype X distance of fluorescence from cortical surface', calbindin: F(75,988) = 16.8, P o 0.0005; CUX1: F(93,372 = 2.17, P = 0.001; Figures 3h and k) . Similar quantification revealed that FOXP2 labelling extended further from layer 6b (as labelled by CTGF) in the Katnal1 1H/1H cortex, which is consistent with an increase in the size of layer 6 (two-way ANOVA, interaction factors 'genotype X distance of fluorescence from CTGF labelling:' F(93,372) = 1.32, P = 0.038; Figure 3n ). Finally, three dimensional models of the ventricular system were constructed from brain micro-computed tomography scans (Figures 3o and p) . Volumetric analysis revealed that Katnal1 1H/1H mice had substantially larger ventricles than wild types (Figure 3q ).
Neuronal migration and morphology defects in Katnal1 1H/1H brains The histological phenotypes of Katnal1 1H/1H mouse brains described above are suggestive of neuronal migration defects. 18 We therefore investigated whether Katnal1 1H/1H mice showed abnormal neuronal migration using BrdU labelling of E13 and E15 embryos and quantified labelled cells in the cortex of P9 pups (described in reference 18). At both ages Katnal1 1H/1H animals had greater numbers of labelled neurons in bins close to the cortical surface neurons positioned closer to the cortical surface compared to wild type. To confirm these results we used a Boyden chamber 19 and performed in vitro neuronal migration analysis in E13.5 primary cortical neuronal cultures. Here we found that a greater proportion of Katnal1 1H/1H cortical neurons migrated to the base of the cell culture insert compared to wildtype controls (Supplementary Figure S3) .
Since in both BrdU labelling and the Boyden assay neurons from Katnal1 1H/1H animals migrated further than those of wild-type littermates, these results suggest that Katnal1 1H/1H cortical neurons show defects in the termination of cortical neuronal migration. Given its role in cytoskeletal organisation, we also hypothesised that neuronal morphology is modulated by Katnal1. Analysis of golgi stained neurons from layers 2-3 of the cortex (Figures 4g and i) demonstrated that, compared to wild-type littermates, Katnal1 1H/1H neurons had larger soma (Figure 4k) , and shorter and thinner axons (Figures 4l and m) (data and cohort details are given in Supplementary Table S3 ). Furthermore, analysis at higher magnification (Figures 4h and j) , demonstrated that the number of synaptic spines on Katnal1 1H/1H neurons was significantly reduced compared to wild type (Figure 4n ).
Recent studies have demonstrated that mutations in some microtubule severing enzymes can cause defects in cilia. 5 Since such ciliary defects could underlie the phenotypes described above we studied the motile cilia of the ependymal lining of the lateral ventricle in sections of postnatal day 2 mouse brains from both Katnal1 1H/1H (n = 4) and wild-type animals (n = 3). We found that the ciliary beat frequency (CBF) of Katnal1 1H/1H animals was significantly attenuated compared to wild-type (CBF: wildtype = 22.39 ± 0.94 Hz, Katnal1 1H/1H = 14.25 ± 0.92 Hz, P = 0.0001; Figure 5a , Supplementary Movies S1). This reduction in CBF in Katnal1 1H/1H animals was also associated with an increased proportion of cilia with an abnormal beat pattern (ciliary dyskinesia) (proportion of dyskinetic cilia: wild type = 17%, Katnal1 1H/1H = 75%) (Figure 5b and Supplementary Movies S1). Visual inspection of the cilia identified a number of ciliary abnormalities such as a swollen ciliary tip (Supplementary Movie S3) or extremely long cilia (Supplementary Movie S4) scattered throughout the field of cilia in Katnal1 1H/1H ventricles. These abnormalities were observed in approximately 25% of Katnal1 1H/1H brain slices. The abnormal cilia always showed a dyskinetic beat pattern and lower beat frequency. To further investigate ciliary morphology we performed scanning electron microscopy upon the ependymal lining of the lateral ventricles of both Katnal1 1H/1H (n = 3) and wild-type animals (n = 3; Figures 5c and d) . Cilia measurements showed no significant differences in average cilia length between genotypes (average cilia length: wild type = 6.22 ± 0.86 μm, Katnal1 1H/1H = 6.54 ± 0.94 Hz, P = 0.303). However in Katnal1 1H/1H samples we noted the presence of both long and short cilia (Figures 5e and f ; defined as two standard deviations longer or shorter than the average cilia length) that were not present in wild-type samples. In addition, inspection of Katnal1 1H/1H cilia identified ciliary abnormalities including bifurcated cilia (Figure 5g) , abnormal kinks and bends in the cilia (Figure 5h ) and swellings along the length of the cilia (Figure 5i ). Transmission electron microscopy of ependymal cilia found that vesicular aggre- Katnal1 disruption affects CNS functions G Banks et al gates were present within the ciliary swellings described above (Figure 5j ). Although these abnormalities were present in only a small proportion (o1%) of Katnal1 1H/1H cilia, they were notably absent from wild-type cilia.
Microtubule severing enzymes play diverse roles in the nervous system. 1, 2 However, at present the microtubule severing enzyme Katnal1 is poorly defined in the context of CNS development and function. Here we present a detailed phenotypic analysis of Katnal1 1H and show that the mutation is associated with changes in circadian rhythms, sleep and behaviour. Furthermore we demonstrate that defects in brain histopathology, neuronal migration and neuronal morphology underlie these phenotypes. Finally we also demonstrate that Katnal1 1H causes a range of defects in the motile cilia of ventricular ependymal cells. The data we present here are the first to associate KATNAL1 with such dysfunctions with important implications for clinical association studies.
The Katnal1 1H mutation was initially identified with a circadian deficit including a short free-running period and advanced activity onset. However subsequent ex vivo experiments using SCN slices of animals carrying the PER2::LUC reporter gene demonstrated no defects in SCN cellular rhythms, suggesting that the core circadian clock was unperturbed by the mutation. Phenotypes in circadian running wheel rhythms that are not associated with changes to the core clock mechanism have also been reported in mouse models of schizophrenia. 25 Here it has been suggested that the wheel running changes observed are the result in defects in output pathways from the SCN circadian clock. Similarly, in Katnal1 1H/1H mice we hypothesise that the defects we demonstrate in neuronal anatomy and neuronal morphology may disrupt output signals from the SCN. Alternatively given that various neuropeptides such as oxytocin are secreted in a circadian manner from ependymal cells lining the third ventricle of the brain, 26 altered ventricular morphology and ciliary function in Katnal1 1H/1H mice may disrupt the circulation of factors secreted by the ciliated ventricular ependymal cells and contribute to the disruption of the behavioural rhythms observed.
The behavioural consequences of microtubule severing enzyme dysfunction in mouse models have been poorly characterised. Currently the phenotypes described are limited to motor dysfunction in mice lacking the Spg4 gene 27 and head shaking and circling in the Fign mutant. 7, 28, 29 In contrast, here we demonstrate that loss of function of Katnal1 is associated with a range of behavioural phenotypes, including changes in circadian activity, poor learning and memory, hyperactivity in a novel environment (the open field) and deficits in USVs. Notably the learning and memory, anxiety and vocalisation phenotypes reprise the clinical symptoms of ID, increased anxiety in novel situations and delays in language acquisition reported in human patients who carry microdeletions incorporating haploinsufficiency of KATNAL1. 8, 9 While it is also worth noting that mutant mice spend more time the centre of the open field than wild types (implying that Katnal1 1H/1H animals show reduced anxiety), we suggest that this result is confounded by the hyperactivity in novel environments phenotype we also describe in mutant mice. This observation is backed up by the fact that mutant animals showed increased activity in all regions of the open field rather than just the anxiolytic periphery. Here we also highlight defects in Katnal1 1H/1H mice such as compromised neuronal migration and morphology which may underpin such phenotypes. In Drosophila, the homologue of Katnal1 (kat-60L1) has been demonstrated to play a critical role in neuronal morphology during development, 30 however the data that we present here is the first to demonstrate a similar phenotype in mammals and furthermore suggests how subtle perturbations to KATNAL1 function may contribute to specific neural and behavioural conditions. For example, defects in neuronal migration, synaptic spines and neuronal morphology such as those we have demonstrated here, have been suggested to underpin ID in conditions such as lissencephaly, 18 Down's syndrome 31 and Rett syndrome. 32 While we are not suggesting that Katnal1 is causative for these conditions, similarities in symptoms and neuronal phenotypes between these conditions and those linked to Katnal1 dysfunction should be appreciated. Furthermore a rare mutation in KATNAL1 has been associated with schizophrenia 10 (http://atgu.mgh.harvard.edu/~spurcell/gene book/genebook.cgi?user = guest&cmd = verb-gene&tbox = KATNAL1) and KATNAL1 has been shown to interact with the schizophrenia associated gene DISC1. 33 In line with these observations we note that increases in ventricular volume and reductions in synaptic spines have been reported in schizophrenic patients 34, 35 and our data demonstrates the same phenotypes in Katnal1 1H/1H mice. Thus the range of phenotypes associated with defects in the function of Katnal1 strongly suggests that the gene should be considered in the pathology of disorders such as ID and schizophrenia.
We do note one key genetic difference between the human patients and Katnal1 1H/1H animals. While the human patients were all heterozygous for the Katnal1 deletion, we found no phenotype in heterozygous mutant mice (data not shown) suggesting that while haploinsufficiency is causative for phenotypes in humans, mice require complete loss of KATNAL1 function to show similar effects. A similar discrepancy between humans and mice has also been noted for the intellectual disability candidate gene CTNNB1. 17 While heterozygous loss of function mutations in CTNNB1 are causative for intellectual disability in humans, conditional knock outs for CTNNB1 have no reported behavioural or craniofacial phenotypes. 36, 37 These differences demonstrate that while mouse models of intellectual disability are of great use in our understanding of the causative mechanisms which underlie the condition, there are still genetic and neurodevelopmental differences between species which also must be taken into account. We also note that while the Katnal1 1H mutation shows a loss of catalytic function in both HEK293 cells and Sertoli cells, 23 this loss of function has not been verified in neuronal cells. However, given that our data demonstrates that the Katnal1 1H mutation lies in an essential catalytic domain and that we show neuronal phenotypes in Katnal1 1H/1H mice, we would expect to see the same loss of catalytic function in neurons.
The data we present here also demonstrate defects in motile cilia in Katnal1 1H/1H mice. Ciliary disruptions in humans (ciliopathies) include Bardet-Biedl and Joubert syndrome. 38 While there is currently limited data available regarding the behavioural phenotypes of mouse models of ciliopathies, we note that ciliary dysfunction in mice has been linked with learning and memory 39 and vocalisation phenotypes, 40 both of which were disturbed in the Katnal1 1H/1H mice described here. It is also notable that the neuronal migration and enlarged ventricle phenotypes that we describe in Katnal1 1H/1H mice recapitulate features associated with known ciliopathy gene mutations. [41] [42] [43] [44] Furthermore in Bardet-Biedl syndrome mouse models ciliary defects such as reduced CBF 45 and structural defects such as abnormal lengthening and swellings along their length 41 have been described, that are similar to those we describe in Katnal1 1H/1H mice. There is strong evidence that ciliopathy associated genes play a number of roles in neuronal development by affecting processes such as progenitor proliferation or maintenance of the radial glia scaffold. 43 However it is also clear that defects in microtubule organisation also affect synaptic structure. 2 At present it is difficult to disentangle the relative contributions of defects in microtubule severing and ciliary abnormalities to the overall phenotypes we observe in Katnal1 1H/1H mice. Further investigations are required to clarify the impacts of these two processes. However it is notable that while defects in cilia structure may contribute to the phenotypes we describe in Katnal1 1H/1H mice, they are far less prominent in Katnal1 1H/1H mice than in other mouse ciliopathy models, 41 suggesting that the ciliary component of KATNAL1 dysfunction may be mild compared to other ciliopathies. Similarly while hydrocephalus has been suggested to be a component of some ciliopathy mouse models, 46 Katnal1 1H/1H mice showed only increased ventricle size rather than an increased incidence of hydrocephalus, further suggesting the ciliary defects in these animals are mild compared to other ciliopathies.
In summary the data presented here clearly demonstrate that KATNAL1 plays an important role in a variety of neuronal processes including neuronal migration, neuronal morphology and ependymal ciliary function. The downstream effect of these defects leads in turn to a number of behavioural changes including in learning and memory, reaction to anxiogenic situations and circadian rhythms. These data therefore highlight how perturbations in KATNAL1 may play a role in neuronal dysfunction and demonstrates that the enzyme is a novel candidate in the study of behavioural and neurodevelopmental disorders.
The authors declare no conflict of interest.
|
What CNS functions are changed by mutations in the KATNAL1 gene?
| 932 |
circadian rhythms, sleep and behaviour
| 20,200 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
How is Japanese encephalitis transmitted?
| 1,220 |
arthropod
| 395 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What element is essential to promoting JEV infection?
| 1,222 |
calcium
| 1,085 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
Where is Q130 located in the NS4B protein?
| 1,223 |
transmembrane domain 3
| 1,250 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What is the mechanism of action for manidipine?
| 1,224 |
voltage-gated Ca(2+) channel (VGCC) inhibitor
| 1,409 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
How many different pathogens are members of the Flaviviridae family of virus?
| 1,225 |
70
| 2,672 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What is the size of a flavivirus?
| 1,226 |
11-kb
| 3,311 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
How many open reading frames are in the flavivirus genome?
| 1,227 |
single
| 3,359 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What are the structural protein elements of a flavivirus?
| 1,228 |
capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E)
| 3,504 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What is the function of the nonstructural protein elements of the flavivirus?
| 1,229 |
viral replication, virion assembly, and virus escape from immune surveillance.
| 3,715 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What are RVPs?
| 1,230 |
Recombinant viral particles
| 4,380 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What is HTS?
| 1,231 |
high-throughput screening
| 4,596 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What is the selective index in high throughput screening?
| 1,232 |
the 50% cytotoxic concentration
| 5,400 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What is the structure of a recombiant viral particle?
| 1,233 |
a natural virus-like envelope on the outside and a replicon on the inside
| 7,484 |
2,437 |
Screening of FDA-Approved Drugs for Inhibitors of Japanese Encephalitis Virus Infection
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5640845/
SHA: 1bd2f6497996fc0fccd8dffd7f84846d3d36f964
Authors: Wang, Shaobo; Liu, Yang; Guo, Jiao; Wang, Peilin; Zhang, Leike; Xiao, Gengfu; Wang, Wei
Date: 2017-10-13
DOI: 10.1128/jvi.01055-17
License: cc-by
Abstract: Japanese encephalitis virus (JEV), an arthropod-borne flavivirus, is a major cause of acute viral encephalitis in humans. No approved drug is available for the specific treatment of JEV infections, and the available vaccines are not effective against all clinical JEV isolates. In the study described here, a high-throughput screening of an FDA-approved drug library for inhibitors of JEV was performed. Five hit drugs that inhibited JEV infection with a selective index of >10 were identified. The antiviral activities of these five hit drugs against other flavivirus, including Zika virus, were also validated. As three of the five hit drugs were calcium inhibitors, additional types of calcium inhibitors that confirmed that calcium is essential for JEV infection, most likely during viral replication, were utilized. Adaptive mutant analysis uncovered that replacement of Q130, located in transmembrane domain 3 of the nonstructural NS4B protein, which is relatively conserved in flaviviruses, with R or K conferred JEV resistance to manidipine, a voltage-gated Ca(2+) channel (VGCC) inhibitor, without an apparent loss of the viral growth profile. Furthermore, manidipine was indicated to protect mice against JEV-induced lethality by decreasing the viral load in the brain, while it abrogated the histopathological changes associated with JEV infection. This study provides five antiflavivirus candidates and identifies cytoplasmic calcium to be a novel antiviral target for the treatment of JEV infection. The findings reported here provide therapeutic possibilities for combating infections caused by flaviviruses. IMPORTANCE No approved therapy for the treatment of Japanese encephalitis virus infection is currently available. Repurposing of approved drugs would accelerate the development of a therapeutic stratagem. In this study, we screened a library of FDA-approved drugs and identified five hit drugs, especially calcium inhibitors, exerting antiflavivirus activity that blocked viral replication. The in vivo efficacy and toxicity of manidipine were investigated with a mouse model of JEV infection, and the viral target was identified by generating an adaptive mutant.
Text: F laviviruses are taxonomically classified in the genus Flavivirus and family Flaviviridae.
These viruses comprise over 70 different pathogens, such as Japanese encephalitis virus (JEV), Zika virus (ZIKV), dengue virus (DENV), West Nile virus (WNV), and yellow fever virus (YFV). Most flaviviruses are arthropod borne and cause public health problems worldwide (1) . The development and usage of vaccines against some flaviviruses, such as JEV, YFV, and tick-borne encephalitis virus (TBEV), have decreased the rates of morbidity and mortality from infections caused by these viruses (2) ; however, flavivirus-induced diseases are still pandemic, and few therapies beyond intensive supportive care are currently available.
Flaviviruses have an approximately 11-kb positive-stranded RNA genome containing a single open reading frame (ORF) flanked by untranslated regions (UTRs) at both termini. The ORF encodes three structural proteins, including the capsid (C), membrane (premembrane [prM] and membrane [M] ), and envelope (E), and seven nonstructural proteins (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) (3) . These seven nonstructural proteins participate in viral replication, virion assembly, and virus escape from immune surveillance.
To date, no specific antivirals with activity against flaviviruses are available. To address this, we conducted a screen of a library of 1,018 FDA-approved drugs. Since flaviviruses are similar in structure and pathogenesis, we first utilized JEV as the prototype to screen the drug library and subsequently validated the antiviral activities with ZIKV, WNV, and DENV type 2 (DENV-2). The hit drugs identified in this study offer potential new therapies for the treatment of flavivirus infection and disease.
Screening of an FDA-approved drug library for inhibitors of JEV infection. Recombinant viral particles (RVPs) with the luciferase-reporting replicon enveloped by the JEV structural proteins were used to select inhibitors, with a focus on those that inhibit virus entry and replication, by a high-throughput screening (HTS) assay (4, 5) . The number of genomic RNA copies of RVP was determined to be 8.4 ϫ 10 6 copies/ml by using a standard curve generated with plasmids carrying the infectious clone. The HTS assay conditions, including the seeding cell density and RVP dose, were optimized to be 10,000 cells per 96-well plate and 20 l (16 copies/cell) RVP for the infective dose, respectively. Under the optimized conditions, the signal-to-basal (S/B) ratio, coefficient of variation (CV), and Z= factor were 38,374, 2.8%, and 0.89, respectively, which demonstrated that the assay was robust and suitable for the large-scale screening of compounds.
A schematic of the HTS assay is depicted in Fig. 1B . After three rounds of screening, five hits with a selective index (SI; which is equal to the 50% cytotoxic concentration [CC 50 [/50% inhibitory concentration [IC 50 ]) of Ͼ10 were selected. The CC 50 values of the hit drugs exhibited in Fig. 1B were similar to those previously published for diverse cell systems but determined using different toxicity assays (6) (7) (8) (9) (10) (11) (12) (13) . Three of the hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were dihydropyridine (DHP) voltage-gated Ca 2ϩ channel (VGCC) antagonists, while pimecrolimus is an inhibitor of inflammatory cytokine secretion and nelfinavir mesylate is an HIV-1 protease blocker. All five drugs exhibited a dose-dependent inhibition of JEV RVP infection (Fig. 1C) . To validate the antiviral effect, hit drugs were purchased from other commercial sources and tested. In the reconfirmation screen, all hit drugs showed antiviral and cytotoxic effects similar to those found in the primary screen.
Validation of hit drugs. To verify the results obtained by the luciferase reporter assays, we also investigated the antiviral effect of the five hit drugs on wild-type JEV strain AT31. As expected from the HTS assay, all five drugs robustly inhibited virus production, with a reduction of approximately 4 to 5 log units at the highest concentration and an approximately 1-log-unit decrease with 2.5 M the drugs (Fig. 2B) . A sharp decrease in JEV RNA levels was also detected (Fig. 2C) . The attenuated RNA levels in the high-dose, middle-dose, and low-dose groups were all above 40%. In particular, in the manidipine-treated group, the inhibitory effect was at least 80% compared to that for the control, which showed a strong inhibition of viral replication. Consistent with the inhibition of virus replication and production, expression of the viral structural protein prM was hardly detectable following treatment with the drugs at the high concentration (Fig. 2D) . Overall, the results in Fig. 2 confirmed that the five hit drugs inhibited JEV infection in a dose-dependent manner in vitro.
Drugs inhibit JEV infection during viral RNA synthesis. Because RVPs, which have a natural virus-like envelope on the outside and a replicon on the inside, permitted the quantification of JEV productive entry and replication, a time-of-addition experiment was performed to investigate whether the hit drugs blocked the entry step or the replication step. As shown in Fig. 3B , no suppression of luciferase activity by any of the hit drugs was observed when they were used as treatments before infection or during infection or as a virucide, suggesting that these drugs do not inhibit JEV infection either by inactivating the virus directly or by blocking JEV entry. However, these drugs exerted fully inhibitory effects when they were added at 1 h postinfection, suggesting that viral replication was the stage at which these drugs showed inhibitory activity.
To confirm this suggestion, we investigated the inhibitory effects of these drugs on the JEV replicon. The highest concentration of manidipine and nelfinavir mesylate tested in baby hamster kidney (BHK-21) cells was adjusted to 5 M and 10 M, respectively. It was shown that all five drugs inhibited JEV RNA synthesis in a dosedependent manner, while neither drug inhibited the initial translation of replicon RNA (5, 14) (Fig. 3C) , confirming that these drugs inhibited JEV infection at the stage of replication.
Hit drugs exhibit broad-spectrum antiflavivirus activity. In order to determine whether the antiviral activity of the five hit drugs extended to other flaviviruses, we explored their antiviral effect against ZIKV. Similar to the findings for JEV, the ZIKV titer was decreased by multiple log units when ZIKV was treated with a high concentration of each of the drugs (Fig. 4A) . Moreover, ZIKV exhibited a higher sensitivity to the two calcium channels inhibitors manidipine and cilnidipine than JEV, with no plaque formation being observed at 10 M. Consistent with this result, sharp decreases in the level of replication of ZIKV RNA and the level of expression of viral protein were also detected (Fig. 4A) . Notably, treatment with 5 M manidipine produced a 95% inhibition of viral replication, translation, and viral yields. Taken together, these results indicate that the hit drugs could effectively inhibit ZIKV infection.
Since these drugs exhibited their anti-JEV effects at the stage of viral replication, we further tested the effects against WNV and DENV-2 by using WNV and DENV-2 replicons. Similar to the results for JEV, a dose-dependent reduction in the level of WNV replication was observed with the drug treatments. The same phenotype was observed for DENV-2 for all drugs except nelfinavir mesylate, which showed no effect at the concentrations tested ( Fig. 4B and C). Together, these results indicate that the five hit drugs are excellent candidates for broad-spectrum antiflavivirus treatment. Antiviral effect of calcium inhibitors. Since three hit drugs, manidipine, cilnidipine, and benidipine hydrochloride, were DHP VGCC inhibitors, we asked whether other calcium antagonists could block JEV infection. To address this question, we employed four different classes of inhibitors. Verapamil, a prototype phenylalkylamine (PAA) VGCC inhibitor (15) , exhibited a dose-dependent inhibition of JEV on both African Green monkey kidney (Vero) and human hepatocellular carcinoma (Huh-7) cells (Fig. 5) , which was consistent with the inhibitory effects of the DHP inhibitors, suggesting that calcium channels play an important role in JEV infection. Cyclosporine and 2-aminobiphenyl borate (2-APB), which inhibit the efflux of Ca 2ϩ from the mitochondrial and endoplasmic reticulum (ER) pool, respectively (16) (17) (18) (19) , were also found to block JEV infection effectively. Similarly, treatment with the cell-permeant Ca 2ϩ chelator 1,2-bis-(o-aminophenoxy)-ethane-N,N,N=,N=-tetraacetic acid, tetraacetoxymethyl ester (BAPTA-AM), could also suppress JEV infection. Taken together, we concluded that intracellular Ca 2ϩ is essential for JEV infection and cytoplasmic calcium is a potent target for antiflavivirus treatment.
Selection and characterization of manidipine-resistant JEV. To identify the viral target of the calcium channel inhibitor, we selected a manidipine-resistant virus by serially passaging JEV in the presence of manidipine. Viruses from passage 20 (P20) showed robust resistance compared with the wild type (WT) (Fig. 6A ). When JEV from P20 was treated with 5 M or 10 M manidipine, the viral titer was about 10-and 100-fold higher than that of the WT, respectively. Individual virus clones were isolated, and two isolates were randomly selected and amplified. An amino acid substitution was observed in two isolated clones, resulting in a glutamine (Q)-to-arginine (R) switch at amino acid position 130 in transmembrane domain 3 (TMD3) of NS4B, i.e., position 2401 of the translated polyprotein in the JEV infectious cDNA clone (Fig. 6B ). Sequence alignment of NS4B indicated that Q130 was conserved in all flaviviruses except YFV, which possessed a lysine at that position (Fig. 6B) . The conserved Q130 of NS4B may account for the sensitivity of JEV, ZIKV, WNV, and DENV-2 to manidipine, as described above (Fig. 4) , while YFV showed resistance to the drug (data not shown).
To confirm that the Q130R mutation did confer manidipine resistance and to investigate the role of Q130 in NS4B function, we produced JEV clones with the Q130R, Q130K, Q130E, or Q130A mutation by introducing the desired mutations into the infectious cDNA clone and rescuing the mutant viruses. To investigate the biological properties of the mutant viruses, we first examined the growth kinetics of the rescued viruses. As shown in Fig. 6C , all mutant viruses had an accumulation of infectious virions and reached the highest titer at 60 h postinfection. Infection of the Q130R and Q130K mutant viruses resulted in growth curves similar to the growth curve for the WT (Fig. 6C) , while the Q130E and Q130A mutants produced smaller amounts of viruses between 24 and 60 h. Analysis of the plaque morphology revealed that the plaques of the Q130R, Q130K, and Q130E mutants were similar to the plaques of the WT, whereas the plaques of the Q130A mutant were smaller than those of the WT.
We next investigated the sensitivity of the four mutant viruses to manidipine. As shown in Fig. 6D , the Q130R and Q130K mutant viruses were resistant to manidipine. At a 10 M concentration, manidipine efficiently inhibited WT JEV infection and reduced the viral yields by approximately 4 log units, while the Q130R and Q130K mutant viruses were resistant to manidipine and the viral titer decreased less than 2 log units. The Q130A mutant virus demonstrated moderate resistance and a slightly higher Taken together, it could be concluded that Q130 not only is critical for conferring manidipine sensitivity but also is important for JEV replication. The replacement of glutamine with basic amino acids conferred resistance to manidipine without an apparent loss of growth.
In vivo efficacy of manidipine. As manidipine exhibited the strongest inhibitory activities on JEV replication as well as ZIKV infection when its activities were compared with those of the five hit drugs (Fig. 2 and 4A) , we further examined the protective effect of manidipine against JEV-induced lethality in a mouse model. As anticipated, mice in the JEV-infected vehicle-treated group started to show symptoms, including limb paralysis, restriction of movement, piloerection, body stiffening, and whole-body tremor, from day 5 postinfection. Within 21 days postinfection, most mice in the JEV-infected group succumbed to the infection, with the mortality rate being 73% (4 out of 15 animals survived). Manidipine treatment following JEV infection reduced the mortality rate to 20% (12 out of 15 animals survived) (Fig. 7A ). Mice treated with manidipine alone or treated with manidipine and infected with JEV showed little abnormal behavior, similar to the findings for the mice in the vehicle-treated group. These results suggest that manidipine provided effective protection against JEVinduced mortality.
To further relate these protective effects to the viral load and histopathological changes in the mouse brains, the viral titer was determined and mouse brain sections were collected and assayed at day 5 and day 21 postinfection, since mice started to show symptoms of JEV infection from day 5 postinfection and most of the surviving mice had recovered at day 21. The results indicated that, during the progression of the disease, manidipine treatment significantly reduced the viral load in infected mice compared to that in infected mice not receiving treatment, while no plaques formed in either the manidipine-or vehicle-treated group, and viral loads were undetectable in each group on day 21 postinfection (Fig. 7B) . As JEV was rapidly cleared from the blood after inoculation and was present in the lymphatic system during the preclinical phase, the effects of manidipine on infection of serum and the spleen were evaluated at earlier time points to detect whether the drug reduced the peripheral viral loads (20, 21) . As shown in Fig. 7C , manidipine had little effect on peripheral JEV infection, which indicated that manidipine protected the mice against JEV-induced lethality by decreasing the viral load in the brain. Similarly, apparent damage in the brain, including meningitis, perivascular cuffing, vacuolar degeneration, and glial nodules, was observed in the JEV-infected and vehicle-treated group on day 5 postinfection, while manidipine treatment remarkably alleviated these phenomena (Fig. 7D) . These results indicate that the alleviation of histopathological changes was accompanied by a reduction in the viral load as well as a reduction in the rate of mortality, further confirming the curative effects of manidipine on viral encephalitis.
Among the five hit drugs, manidipine, cilnidipine, and benidipine hydrochloride were VGCC inhibitors. It has been well documented in the literature that Ca 2ϩ inhibitors serve to inhibit virus infection at the stage of either entry (15, 22) or replication (18) and even at the stage of budding (23) . To this end, we first reviewed all 21 calcium inhibitors included in the current library of FDA-approved drugs and found that, in addition to the four DHP VGCC inhibitors listed in Fig. 1B , two other calcium inhibitors, i.e., flunarizine dihydrochloride and lomerizine hydrochloride, were also identified to be primary candidates with levels of inhibition of Ͼ90%. Similarly, three calcium channel antagonists, nisoldipine, felodipine, and nicardipine hydrochloride, showed levels of inhibition of 75%, 72%, and 66%, respectively, in the primary screen. Together, 9 of the 21 calcium inhibitors in the library, accounting for nearly half of the calcium inhibitors, exhibited levels of flavivirus inhibition of greater than 50%, suggesting that calcium, especially the calcium channel, is a potential antiviral target. To address this, another type of VGCC inhibitor, verapamil, an FDA-approved drug not yet included in the drug library used in this study, was investigated. Likewise, a Ca 2ϩ chelator, BAPTA-AM, as well as the Ca 2ϩ inhibitors 2-APB and cyclosporine, targeting ER and the mitochondrial Ca 2ϩ channel, respectively, were employed to investigate the response of JEV infection to the decrease in intracellular Ca 2ϩ levels. In line with the activities of the three hit DHP VGCC inhibitor drugs, the additional Ca 2ϩ inhibitors exerted anti-JEV activity, which indicated that Ca 2ϩ is indispensable for JEV infection. Thus, Ca 2ϩ inhibitors might be utilized as effective treatments for flavivirus infection.
As the hit drugs exerted full inhibitory activity when they were added posttreatment, we believe that Ca 2ϩ is important for flavivirus genome replication. Furthermore, selection and genetic analysis of drug-resistant viruses revealed that NS4B is the viral target of manidipine. NS4B is part of the viral replication complex and is supposed to anchor the viral replicase to the ER membrane (24) . Meanwhile, the N-terminal 125amino-acid domain of DENV NS4B was indicated to be responsible for inhibition of the immune response (25) . Notably, several structurally distinct compounds have been identified to inhibit flavivirus replication by intensively targeting the TMD of NS4B (26) (27) (28) (29) (30) (31) (32) . It is thus conceivable that inhibitors targeting TMD of NS4B would perturb its function, leading to the suppression of viral RNA replication. In this study, the replacement of Q130 of NS4B with a basic amino acid conferred the resistance effect without suppressing JEV replication, suggesting that position 130 could tolerate a basic amino acid and that the basic amino acid might be involved in the interplay of NS4B with host proteins rather than viral proteins.
Moreover, the efficacy and toxicity of manidipine were monitored in vivo, with manidipine demonstrating effective antiviral activity with favorable biocompatibility. However, the dose used in this study was higher than the dose typically used clinically, representing one of the scenarios most commonly encountered in drug repurposing (33, 34) . As manidipine was approved for use for the long-term treatment of hypertension (35, 36) , pulse-dose treatment with manidipine over the shorter period of time required for the treatment of virus infection might be relatively safe. Moreover, use of a combination of manidipine with other Ca 2ϩ inhibitors might improve its therapeutic efficacy, reduce its toxicity, and reduce the risk of resistance development (37) (38) (39) .
Besides the three VGCC inhibitors, two hit drugs, pimecrolimus and nelfinavir mesylate, showed equivalent inhibitory activities on the replication of JEV, ZIKV, WNV, and DENV-2. Although there has been no report on the use of pimecrolimus for the treatment of infectious diseases, we showed that it had a robust effect against JEV with an SI of Ͼ32. The maximum plasma concentration (C max ) of nelfinavir mesylate achieved with an adult dose was 3 to 4 g/ml (40) , which was comparable to the IC 50 reported here. Notably, nelfinavir mesylate was confirmed to inhibit herpes simplex virus 1 (HSV-1) and the replication of several other herpesviruses by interfering directly or indirectly with the later steps of virus formation, such as glycoprotein maturation or virion release, other than functioning in herpesviruses protease (41, 42) . Whether nelfinavir mesylate inhibits flavivirus by interference with the virus protease or by other off-target effects is unknown. Understanding of the mechanism of the antiflavivirus effects of these drugs might uncover novel targets of the drugs, providing further insight into the pathogenesis of flaviviruses.
Above all, the findings reported here provide novel insights into the molecular mechanisms underlying flavivirus infection and offer new and promising therapeutic possibilities for combating infections caused by flaviviruses.
Cells and viruses. BHK-21, SH-SY5Y (human neuroblastoma), Vero, and Huh-7 cells were cultured in Dulbecco modified Eagle medium (HyClone, Logan, UT, USA) supplemented with 10% fetal bovine serum (Gibco, Grand Island, NY, USA). JEV strain AT31, the WNV replicon, and the DENV-2 replicon expressing Renilla luciferase (Rluc) were kindly provided by Bo Zhang, Wuhan Institute of Virology, Chinese Academy of Sciences (CAS), China. JEV replicon recombinant viral particles (RVPs) were generated as previously described (4, 5) . ZIKV strain H/PF/2013, kindly provided by the European Virus Archive Goes Global, was propagated and titrated in Vero cells.
Optimization of HTS assay conditions. The cell density and RVP dose were optimized for the HTS assay. Vero cells at different densities (2,500 to 12,500 cells per well) were infected with from 1.25 to 20 l RVPs (1 to 16 copies per well). The appropriate cell density as well as the RVP dose was selected by comparing the S/B ratio, CV, and Z= values under different conditions as previously described (43) . Methyl--cyclodextrin and dimethyl sulfoxide (DMSO) were used as positive and negative controls, respectively.
HTS assay of an FDA-approved compound library. A library of 1,018 FDA-approved drugs was purchased from Selleck Chemicals (Houston, TX, USA). The compounds were stored as 10 mM stock solutions in DMSO at 4°C until use. The first round of the HTS assay was carried out as shown in Fig. 1A . The criteria used to identify the primary candidates were no apparent cytotoxicity and an average level of inhibition of Ͼ90% in duplicate wells. The criteria of dose-dependent inhibition and cell viability of Ͼ80% were applied for the reconfirmation screen. Furthermore, the CC 50 of each compound was calculated, and those compounds displaying SIs over 10 were considered hits in this study.
Identification of antiviral effects of five hit drugs. The antiviral effects of the drugs were evaluated by quantitative reverse transcription-PCR (qRT-PCR), immunofluorescence assay (IFA), and plaque assay as previously reported (44) (45) (46) (47) . The experimental timeline is depicted in Fig. 2A .
To ensure the effectiveness of the hit drugs in flavivirus replication, BHK-21 cells transfected with the JEV, WNV, or DENV-2 replicon were incubated with each drug at the concentrations indicated above, and the luciferase activities were determined 24 h, 48 h, or 72 h later, respectively.
Time-of-addition experiment. To evaluate which stage of the JEV life cycle was inhibited by each hit, a time-of-addition experiment was performed as previously described (43) . Vero cells were infected with 20 l RVPs for 1 h (0 to 1 h). The test compounds were incubated with the cells for 1 h before infection (Ϫ1 to 0 h), during infection (0 to 1 h), and for 23 h postinfection (1 to 24 h) (Fig. 3A) . To exclude a possible direct inactivating effect of the drugs, RVPs were incubated with each drug at 37°C for 1 h, and the mixtures were diluted 25-fold to infect Vero cells. Twenty-four hours later, the luciferase activities were determined as described above (Fig. 3A) .
Manidipine-resistant virus. Manidipine-resistant virus was generated by passaging of JEV on Vero cells in the presence of manidipine. Passages 1 to 10 used 5 M manidipine, and passages 11 to 20 used 10 M manidipine. As a control, WT virus was passaged in the presence of 2% DMSO in parallel. Passaging was terminated at passage 20, when no further improvement in resistance was detected. Two manidipine-resistant virus isolates were plaque purified and amplified in the presence of manidipine. Viral RNA was extracted, amplified, and purified for sequencing. An infectious cDNA clone of JEV, strain AT31 (pMWJEAT), kindly provided by T. Wakita, Tokyo Metropolitan Institute for Neuroscience, was used to recover WT and mutant viruses as described previously (4) . Virus titers and manidipine sensitivities were determined by plaque assay in Vero cells.
Manidipine administration to JEV-infected mice. Adult female BALB/c mice (age, 4 weeks) were kept in the Laboratory Animal Center of Wuhan Institute of Virology, CAS (Wuhan, China). The mice were randomly divided into four groups (30 mice per group): a JEV-infected and vehicle (2% Tween 80 plus 5% DMSO in phosphate-buffered saline [PBS])-treated group, a manidipine-treated group, a JEV-infected and manidipine-treated group, and a vehicle-treated group. For infection, mice were infected intraperitoneally with 5 ϫ 10 6 PFU of JEV strain AT31. For the manidipine and vehicle treatments, mice were injected intraperitoneally with 25 mg/kg of body weight manidipine or PBS with 2% Tween 80 and 5% DMSO, respectively. Treatments were administered twice a day for the first 2 days and then consecutively administered once a day for up to 21 days. Five mice from each group were sacrificed on days 1, 3, and 5 postinfection. Serum, spleen tissue, and brain tissue samples were collected for viral titer determination and histopathology investigation. Fifteen mice were monitored daily for morbidity and mortality. The mice that showed neurological signs of disease were euthanized according to the Regulations for the Administration of Affairs Concerning Experimental Animals in China. The protocols were reviewed and approved by the Laboratory Animal Care and Use Committee at the Wuhan Institute of Virology, CAS (Wuhan, China).
|
What measure is used in high-throughput screening to identify potential antiviral compounds?
| 1,234 |
selective index of >10
| 812 |
1,625 |
Rotavirus A in wild and domestic animals from areas with environmental degradation in the Brazilian Amazon
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6298726/
SHA: f3c309c596c20f48f493b77e714ce957d877bdcb
Authors: de Barros, Bruno de Cássio Veloso; Chagas, Elaine Nunes; Bezerra, Luna Wanessa; Ribeiro, Laila Graziela; Duarte Júnior, Jose Wandilson Barboza; Pereira, Diego; da Penha Junior, Edvaldo Tavares; Silva, Julia Rezende; Bezerra, Delana Andreza Melo; Bandeira, Renato Silva; Pinheiro, Helder Henrique Costa; Guerra, Sylvia de Fátima dos Santos; Guimarães, Ricardo José de Paula Souza e; Mascarenhas, Joana D'Arc Pereira
Date: 2018-12-18
DOI: 10.1371/journal.pone.0209005
License: cc-by
Abstract: Acute gastroenteritis is one of the main causes of mortality in humans and young animals. Domestic and mainly wild animals such as bats, small rodents and birds are highly diversified animals in relation to their habitats and ecological niches and are widely distributed geographically in environments of forest fragmentation in some areas of the Amazon, being considered important sources for viruses that affect humans and other animals. Due to the anthropical activities, these animals changed their natural habitat and adapted to urbanized environments, thus representing risks to human and animal health. Although the knowledge of the global diversity of enteric viruses is scarce, there are reports demonstrating the detection of rotavirus in domestic animals and animals of productive systems, such as bovines and pigs. The present study investigated the prevalence of Rotavirus A in 648 fecal samples of different animal species from the northeastern mesoregion of the state of Pará, Brazil, which is characterized as an urbanized area with forest fragments. The fecal specimens were collected from October 2014 to April 2016 and subjected to a Qualitative Real-Time Polymerase Chain Reaction (RT-qPCR), using the NSP3 gene as a target. It was observed that 27.5% (178/648) of the samples presented positive results for RVA, with 178 samples distributed in birds (23.6%), canines (21.35%), chiropterans (17.98%), bovines (14.6%), horses (8.43%), small rodents (6.74%), pigs (3.93%) and felines (3.37%), demonstrating the circulation of RVA in domestic animals and suggesting that such proximity could cause transmissions between different species and the occurrence of rearrangements in the genome of RVA as already described in the literature, associated to the traces of environmental degradation in the studied areas.
Text: Emerging and reemerging infectious diseases are increasing each year in several countries, with an impact both on human populations and on domestic and wild animals living in areas with considerable forest remnants [1] . Most of these diseases are of viral origin, suggesting the emergence and reemergence of viruses that are triggered by human activities that modify the environment [2] .
The populations of wild animals that inhabit forest fragments are strategic groups for studies of public health and the transmission of zoonosis, given that they act as indicators in the assistance and intervention in the human populations, aiming at the prevention of outbreaks and epidemics [3] .
Acute gastroenteritis can be caused by infection in the gastrointestinal tract, caused by different infectious or parasite agents [4] [5] [6] [7] . They represent one of the main causes of mortality in humans, and in young animals, counting for about 25% of mortality [8] . Rotavirus is widely distributed in animals, which act as sources of rotavirus emergent strains, with these animals acting in the transmission between species and through reassortment leading to the emergence of new strains which have been reported in human infections [9] [10] [11] [12] .
The rotavirus (RV) belongs to the Reoviridae family and comprises nine species known as Rotavirus group A to I, with a recent proposal of the J species [13, 14] . Rotavirus A (RVA) is widespread worldwide and predominantly infects humans, bovines and other mammal species, as well as birds [15] . They have a double-stranded ribonucleic acid (dsRNA) genome, divided into 11 segments coding for structural proteins (VP1-VP4, VP6 and VP7) and nonstructural (NSP1-NSP5/NSP6) proteins [16, 17] .
There are records of a close relationship between Amazonian wildlife and human populations [18] , and this interaction is the effect of anthropogenic urbanization activities that result in the deforestation of forest areas, causing the degradation of previously isolated sites such as caves and small caves, a continuous and nature progressive process that has led not only to changes in wildlife habitats but also to a greater relationship with human populations in rural and urban environments, contributing to the occurrence and emergence of diseases different from what normally occurs in endemic regions [19] [20] [21] [22] .
Although the results of RVA have already been described globally [12, [23] [24] [25] [26] [27] [28] [29] [30] , in Brazil, the occurrence, diversity and role of rotavirus in these animals are still poorly studied, considering the large number of present species [4, [31] [32] [33] [34] .
In the Brazilian Amazon, especially in the state of Pará, the city of Belém and Northeast metropolitan mesoregions are some of the areas with the highest indexes of environmental changes [35] , which are concentrated, along with the fact that the knowledge of the global diversity of enteric virus in animals is scarce [36] .
Therefore, it is important to monitor the health of domestic and wild animals in their natural habitat, especially in areas with anthropic alterations that have an interface with rural communities and enterprises, in order to investigate the occurrence of RVA in this population. These communities are ecologically complex, because they have multiple hosts and endless pathogens that may eventually circulate in contiguous urban centers, in addition to the fact that it should also be considered that there is still a lack of studies showing the significance of these viruses infecting this population, as in the context of epidemiological surveillance, these animals become important, since they can be considered as natural sources, with the possibility of transmission to humans [37] [38] [39] .
The qualitative real-time polymerase chain reaction (qRT-PCR) used the NSP3 gene and the TaqMan probe from a highly conserved region of the rotavirus non-structural protein 3 (NSP3), which was previously used in samples from human origin and with low viral loads
Precipitation data were obtained from The Brazilian National Institute of Meteorology (Inmethttp://www.inmet.gov.br/) for the years of capture in the Expedito Ribeiro Settlement (2014) and Açailândia (2015) of the Data Collection Platforms (PCDs) of Belém, located 50 km from Santa Bárbara do Pará, and Tracuateua, located 50 km from Peixe-Boi and 100 km from Viseu. Garmin GPSMap 64s Global Positioning System (GPS) coordinates were collected in the field. The municipal boundaries were obtained on the website of the Brazilian Institute of Geography and Statistics (IBGE) (http://www.ibge.gov.br/) and data on deforestation and land use were obtained from the PRODES [43] and TerraClass [44] Projects. PRODES has annual data in digital format since 2000 and TerraClass presents biannual data since 2004.
The satellite image was generated using the sensor Sentinel 2 of the European Space Agency (ESA) (https://sentinel.esa.int/ web/sentinel/user-guides/sentinel-2-msi) with Open Access CC-BY License (http://open.esa.int/) from the years of 2017 and 2018.
All the data obtained was stored in a Geographic Database (BDG). The BDG was imported/ stored in a GIS for the editing of the graphic elements, establishment of topological relations between the elements and their respective attributes, spatial analysis and visualization of the result through thematic maps.
For the present study, forest fragments of similar size, shape and Phyto physiology were chosen, considering an open peri urban matrix with similar soil use. The selected fragments were distributed within the mesoregions studied, and in each selected fragment fecal samples were randomly collected from domestic and wild animals [45] .
Soil use classes were obtained from the TerraClass data mosaic from 2004 to 2016, because the study sites were in an area with a high cloud presence, which prevented observation (the area was not observed).
The data processing, interpretation, visualization and spatial analysis were performed in ArcGIS software (http://www.arcgis.com/). For the analysis of data related to the determination of the richness, composition and abundance of the fauna of the animals studied in the study area, considering the collection methods adopted and the species available in each city, each sample was considered as an independent sample.
The richness of wild fauna and domestic animals was determined by the total number of species including all collection methods, and the similarity of species was made by the chi-square analysis between the samples of the different treatments with the aid of the EstimateS 8.0 software [46] .
For the calculation of the Test T, the Statistica software was used, and the indices of infected animals in the two environments (forest fragment and peridomicile) were calculated for each treatment sampled by collection area, using the software Past 1.92. Aiming at comparing the values of the diversity indexes through the paired test, as well as the descriptive analysis of the anthropic effects [47] .
The data obtained for the occurrence of RVA and the questionnaires was inserted into a database for a descriptive analysis of the epidemiological profile of the animal population in the three forest ecosystems studied. In this analysis, descriptive statistical treatments were carried out, using customized "row-columns" type charts, referring to the data, in order to characterize the sample and quantify the results using absolute frequency values using the chi-square test and the Test T.
Population study, collection of clinical specimens and laboratory methodology. The flying animals (wild birds and chiroptera) were captured using mist nets which were opened at dawn (4:00 a.m.) and closed in the morning (9:00 a.m.) and were inspected every one hour until the closing, with a sampling effort of 15 days. This research was approved by National All procedures with animals were performed by veterinarians, being birds and bats identified and released at the same capture site. The fecal specimens were collected by stimulation of the rectal ampulla with the use of a "Zaragatoa", packed in cryogenic vials, identified, stored in liquid nitrogen, and later sent to the Laboratory.
Wild animals (small non-flying mammals) were trapped within live-containment traps of the Tomahawk cage (size 45x16x16cm) and Sherman type aluminum (size 30x9x8cm). In each sample plot, 61 traps were distributed, 20 Shermans and 41 Tomahawks being baited with a mixture made with peanut paste, sardines, cod liver oil and corn meal, as well as fruit like banana, apple and pineapple. All the traps used were inspected daily in the morning, the baits being exchanged when necessary and later after the capture in bags of cloth and at least five specimens of each species were chosen for the collection of biological material. The wild animals were sedated with a combination of ketamine 20mg/kg and xylazine 2mg/kg intramuscularly and subsequently, euthanized with anesthetic overdose of 2% lidocaine in the foramen magnum, according to the recommendation of the National Council for the Control of Animal Experimentation (CONCEA).
From October 2014 to April 2016, 1,282 fecal samples were collected from wild and domestic animals. Amongthese, 648 (50.5%) samples were randomly selected for RVA research and handled in Level Three Biosafety Laboratory (NB3).
The viral genome was extracted using the TRIZOL LS REAGENT protocol (INVITRO-GEN, USA/KIT QIAGEN), following the manufacturer's recommendations, with minor adaptation according to the protocol described in the supplemental data.
The qRT-PCR was conducted according to Zeng et al. [40] for the detection of RVA using the NSP3 segment of RVA as the target gene sequence. The assay was conducted in a mixture containing: RNAse-free H 2 O, TaqMan RT-PCR Mix (2x), TAqMan RT Enzyme Mix (40x), primers for the NSP3 gene, Primer NSP3 Forward (20mM), Primer NSP3 Reverse (20mM), probe NSP3 S (10nm), Template (RNA) 3μL, having a total reaction volume of 17μL and reverse transcription cycling of 50˚C, 30 minutes, denaturation of 95˚C, 10 minutes, annealing of 45 cycles of 95˚C, 15 seconds and extension of 60˚C, 1 minute.
The analyzes were considered positive when presenting the cycle threshold (CT) � 40. In order to guarantee a reliable test result, the measurements of contamination control were performed with the use of positive animal control (SA11 prototype) and a negative control (ultrapure water).
All RVA-positive samples were subjected to reverse transcription-polymerase chain reaction (RT-PCR) according to Mijatovic et al [41] to genotyping low viral loads samples. First round was performed with consensus primers N-VP4F1/N-VP4R1 and the Nested-PCR was conducted with N-VP4F2/N-VP4R2 primers to amplify VP4 gene. Amplicons were purified and sequencing for VP4 gene using the same primers of Nested-PCR. The sequences were collected from an automated ABI Prism 3130xl DNA sequencer (Applied Biosystems). The sequence fragments were assembled and edited using the Geneious Bioinformatics software platform v.8.1.7. Posteriorly, the data were compared with othersequences from the National Center for Biotechnology Information GenBank database using BLAST alignment tool to elucidate the RVA genotype of the samples.
From October 2014 to April 2016, a total of 648 fecal samples of wild and domestic animals belonging to three forest fragments areas were tested for the NSP3 gene by qualitative qPCR, and 178 (27.5%) were positive for RVA, distributed among the species: birds (23.6%), canines (21.35%), bats (17.98%), cattle (14.6%), horses (8.43%), small rodents (6.74%), swine (3.93%) and felines (3.37%). The CT interval ranged from 28. 47 It was possible to detect viral strains in all genders of animals studied and in the harvesting period none of the animals showed signs of acute infection and / or diarrhea.
Rotavirus A (RVA) detected in the present study of wild and domestic animals belonging to the three areas of forest fragment, according to Fig 2. In relation to the evaluated bovines, only in the city of Viseu, these species were studied because they were created extensively. In addition, most of the animals were young with ages varying from 1 day to 8 yearsold, history of deficient vaccination, lack of technical assistance and raised in the form of subsistence. The animals showed no symptoms of diarrhea, only low weight performance and poor sanitary management status. In relation to chiroptera, 32 (17.98%) positive samples for RVA were distributed among Carollia perspicillata species, with 12 (37.5%) being all adults, 9 (28.12%) Desmodus rotundus samples (4 young and 5 adults), 5 (15.6%) of Uroderma bilobata (15.62%), 3 (9.37%) of Artibeus lituratus and the species Artibeus Planirostus, Diaemus iyoug and Glossophagine with 1 (3.12%) each. These animals came from areas of forest fragments located near bovine and equine farms, in addition to inhabiting small chicken farms. Fig 3 shows the results obtained for all the species of animals investigated in the forest fragment as well as in the peridomicillus area.
The anthropic variables were analyzed for the three cities studied, as well as the use of the soil within the range of the animals, obeying the domicile, the peridomicile and the forest fragment where the traps of small rodents, birds and various species of animals were captured (Fig 4 and Fig 5) .
Considering the factors related to the anthropic activities in the three studied areas within the three cities of the present study, it was observed that the city of Santa Bárbara is the one that has a better area of preserved forest and the city of Viseu a smaller area. However, in the city of Santa Bárbara, a greater concentration of occupations was observed around the area of forest fragment. It was observed in this chosen area of the city, the presence of different families living in a rural settlement, surviving from the exploitation of forest resources and the creation of small animals for subsistence, such as poultry and fish farming, as well as family farming products.
The breeding of animals in native pastures was only observed in the cities of Peixe Boi and Viseu. Extensive livestock farming was practiced with beef cattle, equines for work and small animals (swine and goats). In relation to the most preserved pasture area, the city of Peixe Boi had the largest area, according to the data shown in Fig 5, however, in the city of Viseu, a higher regeneration was observed in the pastures during the period of the study, with significant secondary vegetation.
When comparing the climates of the three areas it was observed that the predominant climate is megathermal and humid with average annual temperature around 27˚C. The months of October, November and December are the hottest, with temperatures between 32˚C and 34˚C and absolute maximums around 41˚C. Annual rainfall is quite high, generally around 2,350 mm, but strongly concentrated from January to June (80%). From September to December, on the contrary, rainfall is rare, about 7%, with a short dry season, of moderate water deficit in those months. The relative humidity of the average air oscillates around 85%, as shown in Fig 6 [48] .
The description of the accumulated precipitation in the year of capture of the fecal specimens compared to the Climatological Normals (CLINO) for the period from 1961-1990 of the PCDs closest to the locations of the Expedito Ribeiro / Santa Bárbara settlement (Belém PCD), Vila Ananim / Peixe-Boi and Açaiteua / Viseu (Tracauateua PCD) show the frequency of rainfall in the regions, which facilitates the renewal of the pastures and the regeneration of the impacted forests, being an important indicator of the reduction of the damages caused by deforestation in the region.
The average deforestation index in the three study areas was calculated from data obtained from INPE information systems. It was observed that in the years of 2013 to 2014 there were no changes in these regions; in the period from 2014 to 2015 about 4.1% of the city of Viseu was changed and 1.6% of the city of Peixe Boi. In relation to the period of 2016, great changes were observed in Peixe Boi (79%) and in the city of Viseu (70%), thus demonstrating that changes in the natural ecosystem may be associated with the frequencies for RVA in the studied areas, according to Fig 7. When assessing the infected animals in relation to the uninfected animals in both the forest fragment and the peridomicile, considering as animals of the forest fragment the birds, the chiroptera and the small rodents and as animals of the peridomicile the canines, bovines, pigs, felines and horses, a percentage of 37.07% infected peri domestic animals (86/232) and 22.12% infected forest fragment animals (92/416) were obtained. Applying the selected statistical analysis, a Pearson x2 Chi-square value was obtained: 16.7159, df = 1 and p <0.001, meaning that the hypothesis was corroborated, that is, the greater the degradation of the environment, the more likely it will be the search for food by wild animals in adjacent areas, or in the edge of the forest or even in the peri domiciliary region. In this sense, the possibility of contagion with other species of animals, even humans, should be considered because of the capacity of the rotavirus to be transmitted via the fecal / oral route or through direct contact with the environment. It is important to point out that the animals detected in this study are important sources of viral strains.
A total of 80 stool samples were selected, reextracted and analyzed using PCR for the VP4 gene. Eight strains (10%) were positive for VP4 gene, being 2 strains bellowed to P [6] genotype and 6 to P[4]-type, according to
In the present study, RVA was detected circulating in 27.5% of the animals; 36% in domestic animals and 64% in wild animals, providing a unique dataset with qRT-PCR detecting a low viral load of RVA in different species, which further correlates with the deforestation index. These data are important because there is a lack of tests for RVA diagnosis in animals, since the current methods of RVA detection does not always detect in these populations [8] . With the advent of real-time PCR (qPCR), there was an exponential growth, compared to conventional PCR essays, since its superior accuracy, sensitivity and specificity is remarkable, and it is Rotavirus A in wild and domestic animals possible to detect RVA in a variety of animal species using NSP3 gene [49] . The sensitivity of RT-qPCR significantly improved the rate of RVA detection in clinical samples from animals and in this context, the present study proposed an interesting study metrics using virus spreading in the wild animals which inhabit forest fragments to indicate human population interventions, with the goal of preventing the virus outbreaks leveraged on the unique geographic characteristics of Brazil and its large number of species in Amazon.
Currently, no data have been described in the literature regarding the RVA detection using real-time qPCR technique in a wide variety of wild animal species. However, a study by Soltan et al. [50] conducted with horses and cattle detected RVA by RT-PCR, commercial RT-PCR and RT-qPCR in 36.7%, 51.4% and 56.9% respectively, differently from the present study that showed higher positivity for chiropterans (17.98%), canines (21.35%), birds (23.6%) and cattle (14.6%).
The first description of RVA in chiroptera was recorded in feces of Eidolon helvum caught in Vihiga, Kenya [51] . Afterwards, several strains of RVA were detected by different molecular techniques involving chiroptera, in several countries, including Kenya (E. helvum), China (Rhinolophus hipposideros and Aselliscus stoliczkanus), France (Myotis mystacinus), Cameroon (E. helvum) and Brazil [31, [51] [52] [53] [54] [55] . The present study shows the occurrence of RVA in 17.98% of the chiroptera, being among the species Carollia perspicillata (37.5%), Desmodus rotundus (28.12%), Uroderma bilobata (15.6%), Artibeus lituratus (9.37%), Artibeus Planirostus (3.12%), Diaemus iyoug and Glossophagine (3.12%).
Barquez et al. [56] reported that Desmodus rotundus is one of the three hematophagous species of the Phyllostomidae family, found throughout South America, Central America and Mexico. Of the positive chiroptera for RVA in the present study, a prevalence of 28.12% was of Desmodus rotundus. This species feeds on birds, can feed on mammals, mostly medium or large, facilitating the dissemination of viral spores among the community within the habitat, as observed in the present study. These findings show the importance of epidemiological data on the studied species due to the lack of studies involving species of neotropical chiroptera, and it is not possible to establish comparative parameters for these animals.
Regarding the circulation of RVA in canines and birds, the prevalence was 53% and 29%, respectively. Although in the Amazon region there are records of RVA, RVD, RVF and RVG that infect birds [57] [58] and RVD in migratory birds [59] , all were detected by RT-PCR assays differently from the present study which detected the RVA by RT-qPCR involving a variety of animal species.
On the other hand, the prevalence in felines (16%) and pigs (22%) was lower, probably because there are few animals of these species in the region, as well as few creations.
The study detected the presence of RVA in different species of animals both in areas near the home and in areas located in fragments of forest, characterized as forest remnants, since they were located in cities that suffered high environmental impacts due to vegetal extractivism, pasture formation for cattle breeding, exploitation of natural resources, and direct reflexes on the habitats of wild animals that can serve as virus sources, thus facilitating the dispersion of RVA among communities of coexisting animals.
It is worth emphasizing that these animals have a greater contact with the human populations of the studied areas since they cohabit with the humans in the region, besides having a high flow of movement between the forest extracts and environments chosen for the present study. However, it is noteworthy that only in the communities of Santa Bárbara and Viseu were collected fecal specimens from asymptomatic humans for diarrhea and tested for RVA, but all were negative. It is notorious yet, the existence of different levels of degradation in the studied environments, considering the presented data. The fragmentation of the forest generates many consequences on the Amazonian biota, being able to alter the diversity and the composition of the animal communities in the fragments and even to interfere in the ecological processes, without considering that the fragments of forest in the Amazon are influenced by the climate, possibly facilitating the dispersion of pathogens by the environment, since the wild animals detected in the present study are asymptomatic and have low viral load for RVA.
The occurrence of RVA in this population of animals may explain the possibility of dispersion of viral strains, since there is a proximity to the human population, besides the biological characteristics of these species that may represent important sources for gastroenteric viruses, along with the fact that all animals were asymptomatic for diarrhea.
Wild birds have unlimited flight capacity, were captured in an interface region between the peridomicile and forest fragments and it is believed that this region has not been influenced by anthropic activities such as those observed in the area of the present study. On the other hand, the breeding method for poultry and canines close to homes and the forest ecosystem, as they are created in the communities surveyed, probably facilitates direct contact with possible sources of contamination, since in the areas the use of septic tanks is deficient and sometimes non-existent, which may facilitate or even increase the risk of viral dispersion throughout the environment.
The high rates of increase and the analysis of land use in the researched areas may be important indicators of how these animals interact, since with deforestation, the populations of wild animals seek refuge in nearby communities facilitating the dispersal of infectious agents and the possible occurrence of carrier animals by direct contact or contamination of the local environment.
To our knowledge, this is the first study in which a real-time PCR assay was applied for the detection of RVA involving a wide variety of domestic and wild animals, facilitating practical utility in epidemiological and molecular studies and assisting in a perspective in the elaboration of sanitary control and monitoring, preventing possible outbreaks in the studied communities. The detection of positive animals was useful to monitor the infection of the agent in the animal population and to provide an early warning signal to predict an impending epidemic and a favorable risk for the human population, given the evidence of RVA circulation in the different forest fragments.
In addition, the RT-qPCR assay may be a useful alternative for the differential diagnosis of RV in possible coexisting mixed infections clinically indistinguishable such as those caused by other viral strains that cause gastroenteritis such as: astrovirus, coronavirus, picobirnavirus, calicivirus, among others as observed in the studies of Jing et al. [60] and Waruhiu et al. [61] .
Diarrhea associated with RVA infections in pigs is an important cause of increased mortality and economic losses in Europe. The most prevalent genotypes isolated from feces of Belgian diarrheal and non-diarrheal piglets in 2012 [62] demonstrate a wide range of combinations of genotypes G / P including; G3P [6] , G4P [6] , G5P [6] , G4P [7] , G5P [7] , G9P [7] , G9P [13] and G9P [23] . On the other hand, in the present study it was possible to detect only P [6] genotype, since majority of samples was asymptomatic for diarrhea.
Finding shows that different P genotypes of RVA strains interact with distinct blood group histological antigens (HBGA, ABOH, Lewis) and sialic acids via VP4 providing insight into the regional prevalence and increased zoonotic potential of some RVA of origin swine [63] . The genotype P [6] was identified in piglets in Brazil [64] and in Italy and Japan resembling genotype P [6] human [65, 66] .
In the population of animals studied the zoonotic transmission can be frequent, since the animals live in contact with humans and in precarious sanitary conditions. In Brazil, this genotype was described in animal and human populations in studies of Luchs et al. [32] ; Honma et al. [67] ; Araújo et al. [68] ; Mascarenhas et al. [69] and Lorenzetti et al. [70] such studies corroborate the importance of continuing to monitor genotypes to verify if uncommon strains or new strains are emerging and can infect animal populations or inter-species transmissions.
Regarding the genotype P [4] , itwas most detected in our samples in bats, dogs, swine and feline. This genotype is not common in animals, being more detected in human and environmental samples in various parts of the world and included our region [71] . It is important to emphasize that the indicators of environmental contamination in Brazil are significant and contribute to the possibility of human-animal transmission [71] . Such data need further investigation in later work to better characterize the interspecies transmission, since the occurrence of enteric viruses in different matrices demonstrates the anthropogenic impact of the exposed population around and points to the potential risk of infection by the possible exposure of individuals susceptible. Our findings may be useful for tracking fecal contamination in the environment using animals as possible sources thus minimizing the risk of infection by exposure to susceptible individuals, in this case different animal species or even human populations.
RVA were detected in wild and domestic animals using a RT-qPCR assay that analyzed samples that had low viral load for RVA. Although the samples are asymptomatic for diarrhea, it is necessary to conduct strategies for the monitoring and control of the animals in the areas studied in the human population as well as in other species of animals, as well as the implementation of preventive measures aimed at future outbreaks in communities animals in the resident population in these impacted areas. Therefore, the present study is unprecedented in the region and in the country in relation to the research of RVA in wild animals. It is noteworthy that, although the quality of the analyzed samples is characterized as low detectable viral load, the technique presented a good analytical response in the detection of the source animals for RVA, facilitating the selection of the samples for future genetic characterization tests.
|
How many known species of Rotavirus exist?
| 430 |
nine species
| 3,868 |
1,625 |
Rotavirus A in wild and domestic animals from areas with environmental degradation in the Brazilian Amazon
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6298726/
SHA: f3c309c596c20f48f493b77e714ce957d877bdcb
Authors: de Barros, Bruno de Cássio Veloso; Chagas, Elaine Nunes; Bezerra, Luna Wanessa; Ribeiro, Laila Graziela; Duarte Júnior, Jose Wandilson Barboza; Pereira, Diego; da Penha Junior, Edvaldo Tavares; Silva, Julia Rezende; Bezerra, Delana Andreza Melo; Bandeira, Renato Silva; Pinheiro, Helder Henrique Costa; Guerra, Sylvia de Fátima dos Santos; Guimarães, Ricardo José de Paula Souza e; Mascarenhas, Joana D'Arc Pereira
Date: 2018-12-18
DOI: 10.1371/journal.pone.0209005
License: cc-by
Abstract: Acute gastroenteritis is one of the main causes of mortality in humans and young animals. Domestic and mainly wild animals such as bats, small rodents and birds are highly diversified animals in relation to their habitats and ecological niches and are widely distributed geographically in environments of forest fragmentation in some areas of the Amazon, being considered important sources for viruses that affect humans and other animals. Due to the anthropical activities, these animals changed their natural habitat and adapted to urbanized environments, thus representing risks to human and animal health. Although the knowledge of the global diversity of enteric viruses is scarce, there are reports demonstrating the detection of rotavirus in domestic animals and animals of productive systems, such as bovines and pigs. The present study investigated the prevalence of Rotavirus A in 648 fecal samples of different animal species from the northeastern mesoregion of the state of Pará, Brazil, which is characterized as an urbanized area with forest fragments. The fecal specimens were collected from October 2014 to April 2016 and subjected to a Qualitative Real-Time Polymerase Chain Reaction (RT-qPCR), using the NSP3 gene as a target. It was observed that 27.5% (178/648) of the samples presented positive results for RVA, with 178 samples distributed in birds (23.6%), canines (21.35%), chiropterans (17.98%), bovines (14.6%), horses (8.43%), small rodents (6.74%), pigs (3.93%) and felines (3.37%), demonstrating the circulation of RVA in domestic animals and suggesting that such proximity could cause transmissions between different species and the occurrence of rearrangements in the genome of RVA as already described in the literature, associated to the traces of environmental degradation in the studied areas.
Text: Emerging and reemerging infectious diseases are increasing each year in several countries, with an impact both on human populations and on domestic and wild animals living in areas with considerable forest remnants [1] . Most of these diseases are of viral origin, suggesting the emergence and reemergence of viruses that are triggered by human activities that modify the environment [2] .
The populations of wild animals that inhabit forest fragments are strategic groups for studies of public health and the transmission of zoonosis, given that they act as indicators in the assistance and intervention in the human populations, aiming at the prevention of outbreaks and epidemics [3] .
Acute gastroenteritis can be caused by infection in the gastrointestinal tract, caused by different infectious or parasite agents [4] [5] [6] [7] . They represent one of the main causes of mortality in humans, and in young animals, counting for about 25% of mortality [8] . Rotavirus is widely distributed in animals, which act as sources of rotavirus emergent strains, with these animals acting in the transmission between species and through reassortment leading to the emergence of new strains which have been reported in human infections [9] [10] [11] [12] .
The rotavirus (RV) belongs to the Reoviridae family and comprises nine species known as Rotavirus group A to I, with a recent proposal of the J species [13, 14] . Rotavirus A (RVA) is widespread worldwide and predominantly infects humans, bovines and other mammal species, as well as birds [15] . They have a double-stranded ribonucleic acid (dsRNA) genome, divided into 11 segments coding for structural proteins (VP1-VP4, VP6 and VP7) and nonstructural (NSP1-NSP5/NSP6) proteins [16, 17] .
There are records of a close relationship between Amazonian wildlife and human populations [18] , and this interaction is the effect of anthropogenic urbanization activities that result in the deforestation of forest areas, causing the degradation of previously isolated sites such as caves and small caves, a continuous and nature progressive process that has led not only to changes in wildlife habitats but also to a greater relationship with human populations in rural and urban environments, contributing to the occurrence and emergence of diseases different from what normally occurs in endemic regions [19] [20] [21] [22] .
Although the results of RVA have already been described globally [12, [23] [24] [25] [26] [27] [28] [29] [30] , in Brazil, the occurrence, diversity and role of rotavirus in these animals are still poorly studied, considering the large number of present species [4, [31] [32] [33] [34] .
In the Brazilian Amazon, especially in the state of Pará, the city of Belém and Northeast metropolitan mesoregions are some of the areas with the highest indexes of environmental changes [35] , which are concentrated, along with the fact that the knowledge of the global diversity of enteric virus in animals is scarce [36] .
Therefore, it is important to monitor the health of domestic and wild animals in their natural habitat, especially in areas with anthropic alterations that have an interface with rural communities and enterprises, in order to investigate the occurrence of RVA in this population. These communities are ecologically complex, because they have multiple hosts and endless pathogens that may eventually circulate in contiguous urban centers, in addition to the fact that it should also be considered that there is still a lack of studies showing the significance of these viruses infecting this population, as in the context of epidemiological surveillance, these animals become important, since they can be considered as natural sources, with the possibility of transmission to humans [37] [38] [39] .
The qualitative real-time polymerase chain reaction (qRT-PCR) used the NSP3 gene and the TaqMan probe from a highly conserved region of the rotavirus non-structural protein 3 (NSP3), which was previously used in samples from human origin and with low viral loads
Precipitation data were obtained from The Brazilian National Institute of Meteorology (Inmethttp://www.inmet.gov.br/) for the years of capture in the Expedito Ribeiro Settlement (2014) and Açailândia (2015) of the Data Collection Platforms (PCDs) of Belém, located 50 km from Santa Bárbara do Pará, and Tracuateua, located 50 km from Peixe-Boi and 100 km from Viseu. Garmin GPSMap 64s Global Positioning System (GPS) coordinates were collected in the field. The municipal boundaries were obtained on the website of the Brazilian Institute of Geography and Statistics (IBGE) (http://www.ibge.gov.br/) and data on deforestation and land use were obtained from the PRODES [43] and TerraClass [44] Projects. PRODES has annual data in digital format since 2000 and TerraClass presents biannual data since 2004.
The satellite image was generated using the sensor Sentinel 2 of the European Space Agency (ESA) (https://sentinel.esa.int/ web/sentinel/user-guides/sentinel-2-msi) with Open Access CC-BY License (http://open.esa.int/) from the years of 2017 and 2018.
All the data obtained was stored in a Geographic Database (BDG). The BDG was imported/ stored in a GIS for the editing of the graphic elements, establishment of topological relations between the elements and their respective attributes, spatial analysis and visualization of the result through thematic maps.
For the present study, forest fragments of similar size, shape and Phyto physiology were chosen, considering an open peri urban matrix with similar soil use. The selected fragments were distributed within the mesoregions studied, and in each selected fragment fecal samples were randomly collected from domestic and wild animals [45] .
Soil use classes were obtained from the TerraClass data mosaic from 2004 to 2016, because the study sites were in an area with a high cloud presence, which prevented observation (the area was not observed).
The data processing, interpretation, visualization and spatial analysis were performed in ArcGIS software (http://www.arcgis.com/). For the analysis of data related to the determination of the richness, composition and abundance of the fauna of the animals studied in the study area, considering the collection methods adopted and the species available in each city, each sample was considered as an independent sample.
The richness of wild fauna and domestic animals was determined by the total number of species including all collection methods, and the similarity of species was made by the chi-square analysis between the samples of the different treatments with the aid of the EstimateS 8.0 software [46] .
For the calculation of the Test T, the Statistica software was used, and the indices of infected animals in the two environments (forest fragment and peridomicile) were calculated for each treatment sampled by collection area, using the software Past 1.92. Aiming at comparing the values of the diversity indexes through the paired test, as well as the descriptive analysis of the anthropic effects [47] .
The data obtained for the occurrence of RVA and the questionnaires was inserted into a database for a descriptive analysis of the epidemiological profile of the animal population in the three forest ecosystems studied. In this analysis, descriptive statistical treatments were carried out, using customized "row-columns" type charts, referring to the data, in order to characterize the sample and quantify the results using absolute frequency values using the chi-square test and the Test T.
Population study, collection of clinical specimens and laboratory methodology. The flying animals (wild birds and chiroptera) were captured using mist nets which were opened at dawn (4:00 a.m.) and closed in the morning (9:00 a.m.) and were inspected every one hour until the closing, with a sampling effort of 15 days. This research was approved by National All procedures with animals were performed by veterinarians, being birds and bats identified and released at the same capture site. The fecal specimens were collected by stimulation of the rectal ampulla with the use of a "Zaragatoa", packed in cryogenic vials, identified, stored in liquid nitrogen, and later sent to the Laboratory.
Wild animals (small non-flying mammals) were trapped within live-containment traps of the Tomahawk cage (size 45x16x16cm) and Sherman type aluminum (size 30x9x8cm). In each sample plot, 61 traps were distributed, 20 Shermans and 41 Tomahawks being baited with a mixture made with peanut paste, sardines, cod liver oil and corn meal, as well as fruit like banana, apple and pineapple. All the traps used were inspected daily in the morning, the baits being exchanged when necessary and later after the capture in bags of cloth and at least five specimens of each species were chosen for the collection of biological material. The wild animals were sedated with a combination of ketamine 20mg/kg and xylazine 2mg/kg intramuscularly and subsequently, euthanized with anesthetic overdose of 2% lidocaine in the foramen magnum, according to the recommendation of the National Council for the Control of Animal Experimentation (CONCEA).
From October 2014 to April 2016, 1,282 fecal samples were collected from wild and domestic animals. Amongthese, 648 (50.5%) samples were randomly selected for RVA research and handled in Level Three Biosafety Laboratory (NB3).
The viral genome was extracted using the TRIZOL LS REAGENT protocol (INVITRO-GEN, USA/KIT QIAGEN), following the manufacturer's recommendations, with minor adaptation according to the protocol described in the supplemental data.
The qRT-PCR was conducted according to Zeng et al. [40] for the detection of RVA using the NSP3 segment of RVA as the target gene sequence. The assay was conducted in a mixture containing: RNAse-free H 2 O, TaqMan RT-PCR Mix (2x), TAqMan RT Enzyme Mix (40x), primers for the NSP3 gene, Primer NSP3 Forward (20mM), Primer NSP3 Reverse (20mM), probe NSP3 S (10nm), Template (RNA) 3μL, having a total reaction volume of 17μL and reverse transcription cycling of 50˚C, 30 minutes, denaturation of 95˚C, 10 minutes, annealing of 45 cycles of 95˚C, 15 seconds and extension of 60˚C, 1 minute.
The analyzes were considered positive when presenting the cycle threshold (CT) � 40. In order to guarantee a reliable test result, the measurements of contamination control were performed with the use of positive animal control (SA11 prototype) and a negative control (ultrapure water).
All RVA-positive samples were subjected to reverse transcription-polymerase chain reaction (RT-PCR) according to Mijatovic et al [41] to genotyping low viral loads samples. First round was performed with consensus primers N-VP4F1/N-VP4R1 and the Nested-PCR was conducted with N-VP4F2/N-VP4R2 primers to amplify VP4 gene. Amplicons were purified and sequencing for VP4 gene using the same primers of Nested-PCR. The sequences were collected from an automated ABI Prism 3130xl DNA sequencer (Applied Biosystems). The sequence fragments were assembled and edited using the Geneious Bioinformatics software platform v.8.1.7. Posteriorly, the data were compared with othersequences from the National Center for Biotechnology Information GenBank database using BLAST alignment tool to elucidate the RVA genotype of the samples.
From October 2014 to April 2016, a total of 648 fecal samples of wild and domestic animals belonging to three forest fragments areas were tested for the NSP3 gene by qualitative qPCR, and 178 (27.5%) were positive for RVA, distributed among the species: birds (23.6%), canines (21.35%), bats (17.98%), cattle (14.6%), horses (8.43%), small rodents (6.74%), swine (3.93%) and felines (3.37%). The CT interval ranged from 28. 47 It was possible to detect viral strains in all genders of animals studied and in the harvesting period none of the animals showed signs of acute infection and / or diarrhea.
Rotavirus A (RVA) detected in the present study of wild and domestic animals belonging to the three areas of forest fragment, according to Fig 2. In relation to the evaluated bovines, only in the city of Viseu, these species were studied because they were created extensively. In addition, most of the animals were young with ages varying from 1 day to 8 yearsold, history of deficient vaccination, lack of technical assistance and raised in the form of subsistence. The animals showed no symptoms of diarrhea, only low weight performance and poor sanitary management status. In relation to chiroptera, 32 (17.98%) positive samples for RVA were distributed among Carollia perspicillata species, with 12 (37.5%) being all adults, 9 (28.12%) Desmodus rotundus samples (4 young and 5 adults), 5 (15.6%) of Uroderma bilobata (15.62%), 3 (9.37%) of Artibeus lituratus and the species Artibeus Planirostus, Diaemus iyoug and Glossophagine with 1 (3.12%) each. These animals came from areas of forest fragments located near bovine and equine farms, in addition to inhabiting small chicken farms. Fig 3 shows the results obtained for all the species of animals investigated in the forest fragment as well as in the peridomicillus area.
The anthropic variables were analyzed for the three cities studied, as well as the use of the soil within the range of the animals, obeying the domicile, the peridomicile and the forest fragment where the traps of small rodents, birds and various species of animals were captured (Fig 4 and Fig 5) .
Considering the factors related to the anthropic activities in the three studied areas within the three cities of the present study, it was observed that the city of Santa Bárbara is the one that has a better area of preserved forest and the city of Viseu a smaller area. However, in the city of Santa Bárbara, a greater concentration of occupations was observed around the area of forest fragment. It was observed in this chosen area of the city, the presence of different families living in a rural settlement, surviving from the exploitation of forest resources and the creation of small animals for subsistence, such as poultry and fish farming, as well as family farming products.
The breeding of animals in native pastures was only observed in the cities of Peixe Boi and Viseu. Extensive livestock farming was practiced with beef cattle, equines for work and small animals (swine and goats). In relation to the most preserved pasture area, the city of Peixe Boi had the largest area, according to the data shown in Fig 5, however, in the city of Viseu, a higher regeneration was observed in the pastures during the period of the study, with significant secondary vegetation.
When comparing the climates of the three areas it was observed that the predominant climate is megathermal and humid with average annual temperature around 27˚C. The months of October, November and December are the hottest, with temperatures between 32˚C and 34˚C and absolute maximums around 41˚C. Annual rainfall is quite high, generally around 2,350 mm, but strongly concentrated from January to June (80%). From September to December, on the contrary, rainfall is rare, about 7%, with a short dry season, of moderate water deficit in those months. The relative humidity of the average air oscillates around 85%, as shown in Fig 6 [48] .
The description of the accumulated precipitation in the year of capture of the fecal specimens compared to the Climatological Normals (CLINO) for the period from 1961-1990 of the PCDs closest to the locations of the Expedito Ribeiro / Santa Bárbara settlement (Belém PCD), Vila Ananim / Peixe-Boi and Açaiteua / Viseu (Tracauateua PCD) show the frequency of rainfall in the regions, which facilitates the renewal of the pastures and the regeneration of the impacted forests, being an important indicator of the reduction of the damages caused by deforestation in the region.
The average deforestation index in the three study areas was calculated from data obtained from INPE information systems. It was observed that in the years of 2013 to 2014 there were no changes in these regions; in the period from 2014 to 2015 about 4.1% of the city of Viseu was changed and 1.6% of the city of Peixe Boi. In relation to the period of 2016, great changes were observed in Peixe Boi (79%) and in the city of Viseu (70%), thus demonstrating that changes in the natural ecosystem may be associated with the frequencies for RVA in the studied areas, according to Fig 7. When assessing the infected animals in relation to the uninfected animals in both the forest fragment and the peridomicile, considering as animals of the forest fragment the birds, the chiroptera and the small rodents and as animals of the peridomicile the canines, bovines, pigs, felines and horses, a percentage of 37.07% infected peri domestic animals (86/232) and 22.12% infected forest fragment animals (92/416) were obtained. Applying the selected statistical analysis, a Pearson x2 Chi-square value was obtained: 16.7159, df = 1 and p <0.001, meaning that the hypothesis was corroborated, that is, the greater the degradation of the environment, the more likely it will be the search for food by wild animals in adjacent areas, or in the edge of the forest or even in the peri domiciliary region. In this sense, the possibility of contagion with other species of animals, even humans, should be considered because of the capacity of the rotavirus to be transmitted via the fecal / oral route or through direct contact with the environment. It is important to point out that the animals detected in this study are important sources of viral strains.
A total of 80 stool samples were selected, reextracted and analyzed using PCR for the VP4 gene. Eight strains (10%) were positive for VP4 gene, being 2 strains bellowed to P [6] genotype and 6 to P[4]-type, according to
In the present study, RVA was detected circulating in 27.5% of the animals; 36% in domestic animals and 64% in wild animals, providing a unique dataset with qRT-PCR detecting a low viral load of RVA in different species, which further correlates with the deforestation index. These data are important because there is a lack of tests for RVA diagnosis in animals, since the current methods of RVA detection does not always detect in these populations [8] . With the advent of real-time PCR (qPCR), there was an exponential growth, compared to conventional PCR essays, since its superior accuracy, sensitivity and specificity is remarkable, and it is Rotavirus A in wild and domestic animals possible to detect RVA in a variety of animal species using NSP3 gene [49] . The sensitivity of RT-qPCR significantly improved the rate of RVA detection in clinical samples from animals and in this context, the present study proposed an interesting study metrics using virus spreading in the wild animals which inhabit forest fragments to indicate human population interventions, with the goal of preventing the virus outbreaks leveraged on the unique geographic characteristics of Brazil and its large number of species in Amazon.
Currently, no data have been described in the literature regarding the RVA detection using real-time qPCR technique in a wide variety of wild animal species. However, a study by Soltan et al. [50] conducted with horses and cattle detected RVA by RT-PCR, commercial RT-PCR and RT-qPCR in 36.7%, 51.4% and 56.9% respectively, differently from the present study that showed higher positivity for chiropterans (17.98%), canines (21.35%), birds (23.6%) and cattle (14.6%).
The first description of RVA in chiroptera was recorded in feces of Eidolon helvum caught in Vihiga, Kenya [51] . Afterwards, several strains of RVA were detected by different molecular techniques involving chiroptera, in several countries, including Kenya (E. helvum), China (Rhinolophus hipposideros and Aselliscus stoliczkanus), France (Myotis mystacinus), Cameroon (E. helvum) and Brazil [31, [51] [52] [53] [54] [55] . The present study shows the occurrence of RVA in 17.98% of the chiroptera, being among the species Carollia perspicillata (37.5%), Desmodus rotundus (28.12%), Uroderma bilobata (15.6%), Artibeus lituratus (9.37%), Artibeus Planirostus (3.12%), Diaemus iyoug and Glossophagine (3.12%).
Barquez et al. [56] reported that Desmodus rotundus is one of the three hematophagous species of the Phyllostomidae family, found throughout South America, Central America and Mexico. Of the positive chiroptera for RVA in the present study, a prevalence of 28.12% was of Desmodus rotundus. This species feeds on birds, can feed on mammals, mostly medium or large, facilitating the dissemination of viral spores among the community within the habitat, as observed in the present study. These findings show the importance of epidemiological data on the studied species due to the lack of studies involving species of neotropical chiroptera, and it is not possible to establish comparative parameters for these animals.
Regarding the circulation of RVA in canines and birds, the prevalence was 53% and 29%, respectively. Although in the Amazon region there are records of RVA, RVD, RVF and RVG that infect birds [57] [58] and RVD in migratory birds [59] , all were detected by RT-PCR assays differently from the present study which detected the RVA by RT-qPCR involving a variety of animal species.
On the other hand, the prevalence in felines (16%) and pigs (22%) was lower, probably because there are few animals of these species in the region, as well as few creations.
The study detected the presence of RVA in different species of animals both in areas near the home and in areas located in fragments of forest, characterized as forest remnants, since they were located in cities that suffered high environmental impacts due to vegetal extractivism, pasture formation for cattle breeding, exploitation of natural resources, and direct reflexes on the habitats of wild animals that can serve as virus sources, thus facilitating the dispersion of RVA among communities of coexisting animals.
It is worth emphasizing that these animals have a greater contact with the human populations of the studied areas since they cohabit with the humans in the region, besides having a high flow of movement between the forest extracts and environments chosen for the present study. However, it is noteworthy that only in the communities of Santa Bárbara and Viseu were collected fecal specimens from asymptomatic humans for diarrhea and tested for RVA, but all were negative. It is notorious yet, the existence of different levels of degradation in the studied environments, considering the presented data. The fragmentation of the forest generates many consequences on the Amazonian biota, being able to alter the diversity and the composition of the animal communities in the fragments and even to interfere in the ecological processes, without considering that the fragments of forest in the Amazon are influenced by the climate, possibly facilitating the dispersion of pathogens by the environment, since the wild animals detected in the present study are asymptomatic and have low viral load for RVA.
The occurrence of RVA in this population of animals may explain the possibility of dispersion of viral strains, since there is a proximity to the human population, besides the biological characteristics of these species that may represent important sources for gastroenteric viruses, along with the fact that all animals were asymptomatic for diarrhea.
Wild birds have unlimited flight capacity, were captured in an interface region between the peridomicile and forest fragments and it is believed that this region has not been influenced by anthropic activities such as those observed in the area of the present study. On the other hand, the breeding method for poultry and canines close to homes and the forest ecosystem, as they are created in the communities surveyed, probably facilitates direct contact with possible sources of contamination, since in the areas the use of septic tanks is deficient and sometimes non-existent, which may facilitate or even increase the risk of viral dispersion throughout the environment.
The high rates of increase and the analysis of land use in the researched areas may be important indicators of how these animals interact, since with deforestation, the populations of wild animals seek refuge in nearby communities facilitating the dispersal of infectious agents and the possible occurrence of carrier animals by direct contact or contamination of the local environment.
To our knowledge, this is the first study in which a real-time PCR assay was applied for the detection of RVA involving a wide variety of domestic and wild animals, facilitating practical utility in epidemiological and molecular studies and assisting in a perspective in the elaboration of sanitary control and monitoring, preventing possible outbreaks in the studied communities. The detection of positive animals was useful to monitor the infection of the agent in the animal population and to provide an early warning signal to predict an impending epidemic and a favorable risk for the human population, given the evidence of RVA circulation in the different forest fragments.
In addition, the RT-qPCR assay may be a useful alternative for the differential diagnosis of RV in possible coexisting mixed infections clinically indistinguishable such as those caused by other viral strains that cause gastroenteritis such as: astrovirus, coronavirus, picobirnavirus, calicivirus, among others as observed in the studies of Jing et al. [60] and Waruhiu et al. [61] .
Diarrhea associated with RVA infections in pigs is an important cause of increased mortality and economic losses in Europe. The most prevalent genotypes isolated from feces of Belgian diarrheal and non-diarrheal piglets in 2012 [62] demonstrate a wide range of combinations of genotypes G / P including; G3P [6] , G4P [6] , G5P [6] , G4P [7] , G5P [7] , G9P [7] , G9P [13] and G9P [23] . On the other hand, in the present study it was possible to detect only P [6] genotype, since majority of samples was asymptomatic for diarrhea.
Finding shows that different P genotypes of RVA strains interact with distinct blood group histological antigens (HBGA, ABOH, Lewis) and sialic acids via VP4 providing insight into the regional prevalence and increased zoonotic potential of some RVA of origin swine [63] . The genotype P [6] was identified in piglets in Brazil [64] and in Italy and Japan resembling genotype P [6] human [65, 66] .
In the population of animals studied the zoonotic transmission can be frequent, since the animals live in contact with humans and in precarious sanitary conditions. In Brazil, this genotype was described in animal and human populations in studies of Luchs et al. [32] ; Honma et al. [67] ; Araújo et al. [68] ; Mascarenhas et al. [69] and Lorenzetti et al. [70] such studies corroborate the importance of continuing to monitor genotypes to verify if uncommon strains or new strains are emerging and can infect animal populations or inter-species transmissions.
Regarding the genotype P [4] , itwas most detected in our samples in bats, dogs, swine and feline. This genotype is not common in animals, being more detected in human and environmental samples in various parts of the world and included our region [71] . It is important to emphasize that the indicators of environmental contamination in Brazil are significant and contribute to the possibility of human-animal transmission [71] . Such data need further investigation in later work to better characterize the interspecies transmission, since the occurrence of enteric viruses in different matrices demonstrates the anthropogenic impact of the exposed population around and points to the potential risk of infection by the possible exposure of individuals susceptible. Our findings may be useful for tracking fecal contamination in the environment using animals as possible sources thus minimizing the risk of infection by exposure to susceptible individuals, in this case different animal species or even human populations.
RVA were detected in wild and domestic animals using a RT-qPCR assay that analyzed samples that had low viral load for RVA. Although the samples are asymptomatic for diarrhea, it is necessary to conduct strategies for the monitoring and control of the animals in the areas studied in the human population as well as in other species of animals, as well as the implementation of preventive measures aimed at future outbreaks in communities animals in the resident population in these impacted areas. Therefore, the present study is unprecedented in the region and in the country in relation to the research of RVA in wild animals. It is noteworthy that, although the quality of the analyzed samples is characterized as low detectable viral load, the technique presented a good analytical response in the detection of the source animals for RVA, facilitating the selection of the samples for future genetic characterization tests.
|
Is Rotavirus single or double-stranded?
| 431 |
double-stranded ribonucleic acid
| 4,111 |
1,625 |
Rotavirus A in wild and domestic animals from areas with environmental degradation in the Brazilian Amazon
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6298726/
SHA: f3c309c596c20f48f493b77e714ce957d877bdcb
Authors: de Barros, Bruno de Cássio Veloso; Chagas, Elaine Nunes; Bezerra, Luna Wanessa; Ribeiro, Laila Graziela; Duarte Júnior, Jose Wandilson Barboza; Pereira, Diego; da Penha Junior, Edvaldo Tavares; Silva, Julia Rezende; Bezerra, Delana Andreza Melo; Bandeira, Renato Silva; Pinheiro, Helder Henrique Costa; Guerra, Sylvia de Fátima dos Santos; Guimarães, Ricardo José de Paula Souza e; Mascarenhas, Joana D'Arc Pereira
Date: 2018-12-18
DOI: 10.1371/journal.pone.0209005
License: cc-by
Abstract: Acute gastroenteritis is one of the main causes of mortality in humans and young animals. Domestic and mainly wild animals such as bats, small rodents and birds are highly diversified animals in relation to their habitats and ecological niches and are widely distributed geographically in environments of forest fragmentation in some areas of the Amazon, being considered important sources for viruses that affect humans and other animals. Due to the anthropical activities, these animals changed their natural habitat and adapted to urbanized environments, thus representing risks to human and animal health. Although the knowledge of the global diversity of enteric viruses is scarce, there are reports demonstrating the detection of rotavirus in domestic animals and animals of productive systems, such as bovines and pigs. The present study investigated the prevalence of Rotavirus A in 648 fecal samples of different animal species from the northeastern mesoregion of the state of Pará, Brazil, which is characterized as an urbanized area with forest fragments. The fecal specimens were collected from October 2014 to April 2016 and subjected to a Qualitative Real-Time Polymerase Chain Reaction (RT-qPCR), using the NSP3 gene as a target. It was observed that 27.5% (178/648) of the samples presented positive results for RVA, with 178 samples distributed in birds (23.6%), canines (21.35%), chiropterans (17.98%), bovines (14.6%), horses (8.43%), small rodents (6.74%), pigs (3.93%) and felines (3.37%), demonstrating the circulation of RVA in domestic animals and suggesting that such proximity could cause transmissions between different species and the occurrence of rearrangements in the genome of RVA as already described in the literature, associated to the traces of environmental degradation in the studied areas.
Text: Emerging and reemerging infectious diseases are increasing each year in several countries, with an impact both on human populations and on domestic and wild animals living in areas with considerable forest remnants [1] . Most of these diseases are of viral origin, suggesting the emergence and reemergence of viruses that are triggered by human activities that modify the environment [2] .
The populations of wild animals that inhabit forest fragments are strategic groups for studies of public health and the transmission of zoonosis, given that they act as indicators in the assistance and intervention in the human populations, aiming at the prevention of outbreaks and epidemics [3] .
Acute gastroenteritis can be caused by infection in the gastrointestinal tract, caused by different infectious or parasite agents [4] [5] [6] [7] . They represent one of the main causes of mortality in humans, and in young animals, counting for about 25% of mortality [8] . Rotavirus is widely distributed in animals, which act as sources of rotavirus emergent strains, with these animals acting in the transmission between species and through reassortment leading to the emergence of new strains which have been reported in human infections [9] [10] [11] [12] .
The rotavirus (RV) belongs to the Reoviridae family and comprises nine species known as Rotavirus group A to I, with a recent proposal of the J species [13, 14] . Rotavirus A (RVA) is widespread worldwide and predominantly infects humans, bovines and other mammal species, as well as birds [15] . They have a double-stranded ribonucleic acid (dsRNA) genome, divided into 11 segments coding for structural proteins (VP1-VP4, VP6 and VP7) and nonstructural (NSP1-NSP5/NSP6) proteins [16, 17] .
There are records of a close relationship between Amazonian wildlife and human populations [18] , and this interaction is the effect of anthropogenic urbanization activities that result in the deforestation of forest areas, causing the degradation of previously isolated sites such as caves and small caves, a continuous and nature progressive process that has led not only to changes in wildlife habitats but also to a greater relationship with human populations in rural and urban environments, contributing to the occurrence and emergence of diseases different from what normally occurs in endemic regions [19] [20] [21] [22] .
Although the results of RVA have already been described globally [12, [23] [24] [25] [26] [27] [28] [29] [30] , in Brazil, the occurrence, diversity and role of rotavirus in these animals are still poorly studied, considering the large number of present species [4, [31] [32] [33] [34] .
In the Brazilian Amazon, especially in the state of Pará, the city of Belém and Northeast metropolitan mesoregions are some of the areas with the highest indexes of environmental changes [35] , which are concentrated, along with the fact that the knowledge of the global diversity of enteric virus in animals is scarce [36] .
Therefore, it is important to monitor the health of domestic and wild animals in their natural habitat, especially in areas with anthropic alterations that have an interface with rural communities and enterprises, in order to investigate the occurrence of RVA in this population. These communities are ecologically complex, because they have multiple hosts and endless pathogens that may eventually circulate in contiguous urban centers, in addition to the fact that it should also be considered that there is still a lack of studies showing the significance of these viruses infecting this population, as in the context of epidemiological surveillance, these animals become important, since they can be considered as natural sources, with the possibility of transmission to humans [37] [38] [39] .
The qualitative real-time polymerase chain reaction (qRT-PCR) used the NSP3 gene and the TaqMan probe from a highly conserved region of the rotavirus non-structural protein 3 (NSP3), which was previously used in samples from human origin and with low viral loads
Precipitation data were obtained from The Brazilian National Institute of Meteorology (Inmethttp://www.inmet.gov.br/) for the years of capture in the Expedito Ribeiro Settlement (2014) and Açailândia (2015) of the Data Collection Platforms (PCDs) of Belém, located 50 km from Santa Bárbara do Pará, and Tracuateua, located 50 km from Peixe-Boi and 100 km from Viseu. Garmin GPSMap 64s Global Positioning System (GPS) coordinates were collected in the field. The municipal boundaries were obtained on the website of the Brazilian Institute of Geography and Statistics (IBGE) (http://www.ibge.gov.br/) and data on deforestation and land use were obtained from the PRODES [43] and TerraClass [44] Projects. PRODES has annual data in digital format since 2000 and TerraClass presents biannual data since 2004.
The satellite image was generated using the sensor Sentinel 2 of the European Space Agency (ESA) (https://sentinel.esa.int/ web/sentinel/user-guides/sentinel-2-msi) with Open Access CC-BY License (http://open.esa.int/) from the years of 2017 and 2018.
All the data obtained was stored in a Geographic Database (BDG). The BDG was imported/ stored in a GIS for the editing of the graphic elements, establishment of topological relations between the elements and their respective attributes, spatial analysis and visualization of the result through thematic maps.
For the present study, forest fragments of similar size, shape and Phyto physiology were chosen, considering an open peri urban matrix with similar soil use. The selected fragments were distributed within the mesoregions studied, and in each selected fragment fecal samples were randomly collected from domestic and wild animals [45] .
Soil use classes were obtained from the TerraClass data mosaic from 2004 to 2016, because the study sites were in an area with a high cloud presence, which prevented observation (the area was not observed).
The data processing, interpretation, visualization and spatial analysis were performed in ArcGIS software (http://www.arcgis.com/). For the analysis of data related to the determination of the richness, composition and abundance of the fauna of the animals studied in the study area, considering the collection methods adopted and the species available in each city, each sample was considered as an independent sample.
The richness of wild fauna and domestic animals was determined by the total number of species including all collection methods, and the similarity of species was made by the chi-square analysis between the samples of the different treatments with the aid of the EstimateS 8.0 software [46] .
For the calculation of the Test T, the Statistica software was used, and the indices of infected animals in the two environments (forest fragment and peridomicile) were calculated for each treatment sampled by collection area, using the software Past 1.92. Aiming at comparing the values of the diversity indexes through the paired test, as well as the descriptive analysis of the anthropic effects [47] .
The data obtained for the occurrence of RVA and the questionnaires was inserted into a database for a descriptive analysis of the epidemiological profile of the animal population in the three forest ecosystems studied. In this analysis, descriptive statistical treatments were carried out, using customized "row-columns" type charts, referring to the data, in order to characterize the sample and quantify the results using absolute frequency values using the chi-square test and the Test T.
Population study, collection of clinical specimens and laboratory methodology. The flying animals (wild birds and chiroptera) were captured using mist nets which were opened at dawn (4:00 a.m.) and closed in the morning (9:00 a.m.) and were inspected every one hour until the closing, with a sampling effort of 15 days. This research was approved by National All procedures with animals were performed by veterinarians, being birds and bats identified and released at the same capture site. The fecal specimens were collected by stimulation of the rectal ampulla with the use of a "Zaragatoa", packed in cryogenic vials, identified, stored in liquid nitrogen, and later sent to the Laboratory.
Wild animals (small non-flying mammals) were trapped within live-containment traps of the Tomahawk cage (size 45x16x16cm) and Sherman type aluminum (size 30x9x8cm). In each sample plot, 61 traps were distributed, 20 Shermans and 41 Tomahawks being baited with a mixture made with peanut paste, sardines, cod liver oil and corn meal, as well as fruit like banana, apple and pineapple. All the traps used were inspected daily in the morning, the baits being exchanged when necessary and later after the capture in bags of cloth and at least five specimens of each species were chosen for the collection of biological material. The wild animals were sedated with a combination of ketamine 20mg/kg and xylazine 2mg/kg intramuscularly and subsequently, euthanized with anesthetic overdose of 2% lidocaine in the foramen magnum, according to the recommendation of the National Council for the Control of Animal Experimentation (CONCEA).
From October 2014 to April 2016, 1,282 fecal samples were collected from wild and domestic animals. Amongthese, 648 (50.5%) samples were randomly selected for RVA research and handled in Level Three Biosafety Laboratory (NB3).
The viral genome was extracted using the TRIZOL LS REAGENT protocol (INVITRO-GEN, USA/KIT QIAGEN), following the manufacturer's recommendations, with minor adaptation according to the protocol described in the supplemental data.
The qRT-PCR was conducted according to Zeng et al. [40] for the detection of RVA using the NSP3 segment of RVA as the target gene sequence. The assay was conducted in a mixture containing: RNAse-free H 2 O, TaqMan RT-PCR Mix (2x), TAqMan RT Enzyme Mix (40x), primers for the NSP3 gene, Primer NSP3 Forward (20mM), Primer NSP3 Reverse (20mM), probe NSP3 S (10nm), Template (RNA) 3μL, having a total reaction volume of 17μL and reverse transcription cycling of 50˚C, 30 minutes, denaturation of 95˚C, 10 minutes, annealing of 45 cycles of 95˚C, 15 seconds and extension of 60˚C, 1 minute.
The analyzes were considered positive when presenting the cycle threshold (CT) � 40. In order to guarantee a reliable test result, the measurements of contamination control were performed with the use of positive animal control (SA11 prototype) and a negative control (ultrapure water).
All RVA-positive samples were subjected to reverse transcription-polymerase chain reaction (RT-PCR) according to Mijatovic et al [41] to genotyping low viral loads samples. First round was performed with consensus primers N-VP4F1/N-VP4R1 and the Nested-PCR was conducted with N-VP4F2/N-VP4R2 primers to amplify VP4 gene. Amplicons were purified and sequencing for VP4 gene using the same primers of Nested-PCR. The sequences were collected from an automated ABI Prism 3130xl DNA sequencer (Applied Biosystems). The sequence fragments were assembled and edited using the Geneious Bioinformatics software platform v.8.1.7. Posteriorly, the data were compared with othersequences from the National Center for Biotechnology Information GenBank database using BLAST alignment tool to elucidate the RVA genotype of the samples.
From October 2014 to April 2016, a total of 648 fecal samples of wild and domestic animals belonging to three forest fragments areas were tested for the NSP3 gene by qualitative qPCR, and 178 (27.5%) were positive for RVA, distributed among the species: birds (23.6%), canines (21.35%), bats (17.98%), cattle (14.6%), horses (8.43%), small rodents (6.74%), swine (3.93%) and felines (3.37%). The CT interval ranged from 28. 47 It was possible to detect viral strains in all genders of animals studied and in the harvesting period none of the animals showed signs of acute infection and / or diarrhea.
Rotavirus A (RVA) detected in the present study of wild and domestic animals belonging to the three areas of forest fragment, according to Fig 2. In relation to the evaluated bovines, only in the city of Viseu, these species were studied because they were created extensively. In addition, most of the animals were young with ages varying from 1 day to 8 yearsold, history of deficient vaccination, lack of technical assistance and raised in the form of subsistence. The animals showed no symptoms of diarrhea, only low weight performance and poor sanitary management status. In relation to chiroptera, 32 (17.98%) positive samples for RVA were distributed among Carollia perspicillata species, with 12 (37.5%) being all adults, 9 (28.12%) Desmodus rotundus samples (4 young and 5 adults), 5 (15.6%) of Uroderma bilobata (15.62%), 3 (9.37%) of Artibeus lituratus and the species Artibeus Planirostus, Diaemus iyoug and Glossophagine with 1 (3.12%) each. These animals came from areas of forest fragments located near bovine and equine farms, in addition to inhabiting small chicken farms. Fig 3 shows the results obtained for all the species of animals investigated in the forest fragment as well as in the peridomicillus area.
The anthropic variables were analyzed for the three cities studied, as well as the use of the soil within the range of the animals, obeying the domicile, the peridomicile and the forest fragment where the traps of small rodents, birds and various species of animals were captured (Fig 4 and Fig 5) .
Considering the factors related to the anthropic activities in the three studied areas within the three cities of the present study, it was observed that the city of Santa Bárbara is the one that has a better area of preserved forest and the city of Viseu a smaller area. However, in the city of Santa Bárbara, a greater concentration of occupations was observed around the area of forest fragment. It was observed in this chosen area of the city, the presence of different families living in a rural settlement, surviving from the exploitation of forest resources and the creation of small animals for subsistence, such as poultry and fish farming, as well as family farming products.
The breeding of animals in native pastures was only observed in the cities of Peixe Boi and Viseu. Extensive livestock farming was practiced with beef cattle, equines for work and small animals (swine and goats). In relation to the most preserved pasture area, the city of Peixe Boi had the largest area, according to the data shown in Fig 5, however, in the city of Viseu, a higher regeneration was observed in the pastures during the period of the study, with significant secondary vegetation.
When comparing the climates of the three areas it was observed that the predominant climate is megathermal and humid with average annual temperature around 27˚C. The months of October, November and December are the hottest, with temperatures between 32˚C and 34˚C and absolute maximums around 41˚C. Annual rainfall is quite high, generally around 2,350 mm, but strongly concentrated from January to June (80%). From September to December, on the contrary, rainfall is rare, about 7%, with a short dry season, of moderate water deficit in those months. The relative humidity of the average air oscillates around 85%, as shown in Fig 6 [48] .
The description of the accumulated precipitation in the year of capture of the fecal specimens compared to the Climatological Normals (CLINO) for the period from 1961-1990 of the PCDs closest to the locations of the Expedito Ribeiro / Santa Bárbara settlement (Belém PCD), Vila Ananim / Peixe-Boi and Açaiteua / Viseu (Tracauateua PCD) show the frequency of rainfall in the regions, which facilitates the renewal of the pastures and the regeneration of the impacted forests, being an important indicator of the reduction of the damages caused by deforestation in the region.
The average deforestation index in the three study areas was calculated from data obtained from INPE information systems. It was observed that in the years of 2013 to 2014 there were no changes in these regions; in the period from 2014 to 2015 about 4.1% of the city of Viseu was changed and 1.6% of the city of Peixe Boi. In relation to the period of 2016, great changes were observed in Peixe Boi (79%) and in the city of Viseu (70%), thus demonstrating that changes in the natural ecosystem may be associated with the frequencies for RVA in the studied areas, according to Fig 7. When assessing the infected animals in relation to the uninfected animals in both the forest fragment and the peridomicile, considering as animals of the forest fragment the birds, the chiroptera and the small rodents and as animals of the peridomicile the canines, bovines, pigs, felines and horses, a percentage of 37.07% infected peri domestic animals (86/232) and 22.12% infected forest fragment animals (92/416) were obtained. Applying the selected statistical analysis, a Pearson x2 Chi-square value was obtained: 16.7159, df = 1 and p <0.001, meaning that the hypothesis was corroborated, that is, the greater the degradation of the environment, the more likely it will be the search for food by wild animals in adjacent areas, or in the edge of the forest or even in the peri domiciliary region. In this sense, the possibility of contagion with other species of animals, even humans, should be considered because of the capacity of the rotavirus to be transmitted via the fecal / oral route or through direct contact with the environment. It is important to point out that the animals detected in this study are important sources of viral strains.
A total of 80 stool samples were selected, reextracted and analyzed using PCR for the VP4 gene. Eight strains (10%) were positive for VP4 gene, being 2 strains bellowed to P [6] genotype and 6 to P[4]-type, according to
In the present study, RVA was detected circulating in 27.5% of the animals; 36% in domestic animals and 64% in wild animals, providing a unique dataset with qRT-PCR detecting a low viral load of RVA in different species, which further correlates with the deforestation index. These data are important because there is a lack of tests for RVA diagnosis in animals, since the current methods of RVA detection does not always detect in these populations [8] . With the advent of real-time PCR (qPCR), there was an exponential growth, compared to conventional PCR essays, since its superior accuracy, sensitivity and specificity is remarkable, and it is Rotavirus A in wild and domestic animals possible to detect RVA in a variety of animal species using NSP3 gene [49] . The sensitivity of RT-qPCR significantly improved the rate of RVA detection in clinical samples from animals and in this context, the present study proposed an interesting study metrics using virus spreading in the wild animals which inhabit forest fragments to indicate human population interventions, with the goal of preventing the virus outbreaks leveraged on the unique geographic characteristics of Brazil and its large number of species in Amazon.
Currently, no data have been described in the literature regarding the RVA detection using real-time qPCR technique in a wide variety of wild animal species. However, a study by Soltan et al. [50] conducted with horses and cattle detected RVA by RT-PCR, commercial RT-PCR and RT-qPCR in 36.7%, 51.4% and 56.9% respectively, differently from the present study that showed higher positivity for chiropterans (17.98%), canines (21.35%), birds (23.6%) and cattle (14.6%).
The first description of RVA in chiroptera was recorded in feces of Eidolon helvum caught in Vihiga, Kenya [51] . Afterwards, several strains of RVA were detected by different molecular techniques involving chiroptera, in several countries, including Kenya (E. helvum), China (Rhinolophus hipposideros and Aselliscus stoliczkanus), France (Myotis mystacinus), Cameroon (E. helvum) and Brazil [31, [51] [52] [53] [54] [55] . The present study shows the occurrence of RVA in 17.98% of the chiroptera, being among the species Carollia perspicillata (37.5%), Desmodus rotundus (28.12%), Uroderma bilobata (15.6%), Artibeus lituratus (9.37%), Artibeus Planirostus (3.12%), Diaemus iyoug and Glossophagine (3.12%).
Barquez et al. [56] reported that Desmodus rotundus is one of the three hematophagous species of the Phyllostomidae family, found throughout South America, Central America and Mexico. Of the positive chiroptera for RVA in the present study, a prevalence of 28.12% was of Desmodus rotundus. This species feeds on birds, can feed on mammals, mostly medium or large, facilitating the dissemination of viral spores among the community within the habitat, as observed in the present study. These findings show the importance of epidemiological data on the studied species due to the lack of studies involving species of neotropical chiroptera, and it is not possible to establish comparative parameters for these animals.
Regarding the circulation of RVA in canines and birds, the prevalence was 53% and 29%, respectively. Although in the Amazon region there are records of RVA, RVD, RVF and RVG that infect birds [57] [58] and RVD in migratory birds [59] , all were detected by RT-PCR assays differently from the present study which detected the RVA by RT-qPCR involving a variety of animal species.
On the other hand, the prevalence in felines (16%) and pigs (22%) was lower, probably because there are few animals of these species in the region, as well as few creations.
The study detected the presence of RVA in different species of animals both in areas near the home and in areas located in fragments of forest, characterized as forest remnants, since they were located in cities that suffered high environmental impacts due to vegetal extractivism, pasture formation for cattle breeding, exploitation of natural resources, and direct reflexes on the habitats of wild animals that can serve as virus sources, thus facilitating the dispersion of RVA among communities of coexisting animals.
It is worth emphasizing that these animals have a greater contact with the human populations of the studied areas since they cohabit with the humans in the region, besides having a high flow of movement between the forest extracts and environments chosen for the present study. However, it is noteworthy that only in the communities of Santa Bárbara and Viseu were collected fecal specimens from asymptomatic humans for diarrhea and tested for RVA, but all were negative. It is notorious yet, the existence of different levels of degradation in the studied environments, considering the presented data. The fragmentation of the forest generates many consequences on the Amazonian biota, being able to alter the diversity and the composition of the animal communities in the fragments and even to interfere in the ecological processes, without considering that the fragments of forest in the Amazon are influenced by the climate, possibly facilitating the dispersion of pathogens by the environment, since the wild animals detected in the present study are asymptomatic and have low viral load for RVA.
The occurrence of RVA in this population of animals may explain the possibility of dispersion of viral strains, since there is a proximity to the human population, besides the biological characteristics of these species that may represent important sources for gastroenteric viruses, along with the fact that all animals were asymptomatic for diarrhea.
Wild birds have unlimited flight capacity, were captured in an interface region between the peridomicile and forest fragments and it is believed that this region has not been influenced by anthropic activities such as those observed in the area of the present study. On the other hand, the breeding method for poultry and canines close to homes and the forest ecosystem, as they are created in the communities surveyed, probably facilitates direct contact with possible sources of contamination, since in the areas the use of septic tanks is deficient and sometimes non-existent, which may facilitate or even increase the risk of viral dispersion throughout the environment.
The high rates of increase and the analysis of land use in the researched areas may be important indicators of how these animals interact, since with deforestation, the populations of wild animals seek refuge in nearby communities facilitating the dispersal of infectious agents and the possible occurrence of carrier animals by direct contact or contamination of the local environment.
To our knowledge, this is the first study in which a real-time PCR assay was applied for the detection of RVA involving a wide variety of domestic and wild animals, facilitating practical utility in epidemiological and molecular studies and assisting in a perspective in the elaboration of sanitary control and monitoring, preventing possible outbreaks in the studied communities. The detection of positive animals was useful to monitor the infection of the agent in the animal population and to provide an early warning signal to predict an impending epidemic and a favorable risk for the human population, given the evidence of RVA circulation in the different forest fragments.
In addition, the RT-qPCR assay may be a useful alternative for the differential diagnosis of RV in possible coexisting mixed infections clinically indistinguishable such as those caused by other viral strains that cause gastroenteritis such as: astrovirus, coronavirus, picobirnavirus, calicivirus, among others as observed in the studies of Jing et al. [60] and Waruhiu et al. [61] .
Diarrhea associated with RVA infections in pigs is an important cause of increased mortality and economic losses in Europe. The most prevalent genotypes isolated from feces of Belgian diarrheal and non-diarrheal piglets in 2012 [62] demonstrate a wide range of combinations of genotypes G / P including; G3P [6] , G4P [6] , G5P [6] , G4P [7] , G5P [7] , G9P [7] , G9P [13] and G9P [23] . On the other hand, in the present study it was possible to detect only P [6] genotype, since majority of samples was asymptomatic for diarrhea.
Finding shows that different P genotypes of RVA strains interact with distinct blood group histological antigens (HBGA, ABOH, Lewis) and sialic acids via VP4 providing insight into the regional prevalence and increased zoonotic potential of some RVA of origin swine [63] . The genotype P [6] was identified in piglets in Brazil [64] and in Italy and Japan resembling genotype P [6] human [65, 66] .
In the population of animals studied the zoonotic transmission can be frequent, since the animals live in contact with humans and in precarious sanitary conditions. In Brazil, this genotype was described in animal and human populations in studies of Luchs et al. [32] ; Honma et al. [67] ; Araújo et al. [68] ; Mascarenhas et al. [69] and Lorenzetti et al. [70] such studies corroborate the importance of continuing to monitor genotypes to verify if uncommon strains or new strains are emerging and can infect animal populations or inter-species transmissions.
Regarding the genotype P [4] , itwas most detected in our samples in bats, dogs, swine and feline. This genotype is not common in animals, being more detected in human and environmental samples in various parts of the world and included our region [71] . It is important to emphasize that the indicators of environmental contamination in Brazil are significant and contribute to the possibility of human-animal transmission [71] . Such data need further investigation in later work to better characterize the interspecies transmission, since the occurrence of enteric viruses in different matrices demonstrates the anthropogenic impact of the exposed population around and points to the potential risk of infection by the possible exposure of individuals susceptible. Our findings may be useful for tracking fecal contamination in the environment using animals as possible sources thus minimizing the risk of infection by exposure to susceptible individuals, in this case different animal species or even human populations.
RVA were detected in wild and domestic animals using a RT-qPCR assay that analyzed samples that had low viral load for RVA. Although the samples are asymptomatic for diarrhea, it is necessary to conduct strategies for the monitoring and control of the animals in the areas studied in the human population as well as in other species of animals, as well as the implementation of preventive measures aimed at future outbreaks in communities animals in the resident population in these impacted areas. Therefore, the present study is unprecedented in the region and in the country in relation to the research of RVA in wild animals. It is noteworthy that, although the quality of the analyzed samples is characterized as low detectable viral load, the technique presented a good analytical response in the detection of the source animals for RVA, facilitating the selection of the samples for future genetic characterization tests.
|
What structural proteins are coded by Rotavirus?
| 432 |
VP1-VP4, VP6 and VP7
| 4,217 |
1,625 |
Rotavirus A in wild and domestic animals from areas with environmental degradation in the Brazilian Amazon
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6298726/
SHA: f3c309c596c20f48f493b77e714ce957d877bdcb
Authors: de Barros, Bruno de Cássio Veloso; Chagas, Elaine Nunes; Bezerra, Luna Wanessa; Ribeiro, Laila Graziela; Duarte Júnior, Jose Wandilson Barboza; Pereira, Diego; da Penha Junior, Edvaldo Tavares; Silva, Julia Rezende; Bezerra, Delana Andreza Melo; Bandeira, Renato Silva; Pinheiro, Helder Henrique Costa; Guerra, Sylvia de Fátima dos Santos; Guimarães, Ricardo José de Paula Souza e; Mascarenhas, Joana D'Arc Pereira
Date: 2018-12-18
DOI: 10.1371/journal.pone.0209005
License: cc-by
Abstract: Acute gastroenteritis is one of the main causes of mortality in humans and young animals. Domestic and mainly wild animals such as bats, small rodents and birds are highly diversified animals in relation to their habitats and ecological niches and are widely distributed geographically in environments of forest fragmentation in some areas of the Amazon, being considered important sources for viruses that affect humans and other animals. Due to the anthropical activities, these animals changed their natural habitat and adapted to urbanized environments, thus representing risks to human and animal health. Although the knowledge of the global diversity of enteric viruses is scarce, there are reports demonstrating the detection of rotavirus in domestic animals and animals of productive systems, such as bovines and pigs. The present study investigated the prevalence of Rotavirus A in 648 fecal samples of different animal species from the northeastern mesoregion of the state of Pará, Brazil, which is characterized as an urbanized area with forest fragments. The fecal specimens were collected from October 2014 to April 2016 and subjected to a Qualitative Real-Time Polymerase Chain Reaction (RT-qPCR), using the NSP3 gene as a target. It was observed that 27.5% (178/648) of the samples presented positive results for RVA, with 178 samples distributed in birds (23.6%), canines (21.35%), chiropterans (17.98%), bovines (14.6%), horses (8.43%), small rodents (6.74%), pigs (3.93%) and felines (3.37%), demonstrating the circulation of RVA in domestic animals and suggesting that such proximity could cause transmissions between different species and the occurrence of rearrangements in the genome of RVA as already described in the literature, associated to the traces of environmental degradation in the studied areas.
Text: Emerging and reemerging infectious diseases are increasing each year in several countries, with an impact both on human populations and on domestic and wild animals living in areas with considerable forest remnants [1] . Most of these diseases are of viral origin, suggesting the emergence and reemergence of viruses that are triggered by human activities that modify the environment [2] .
The populations of wild animals that inhabit forest fragments are strategic groups for studies of public health and the transmission of zoonosis, given that they act as indicators in the assistance and intervention in the human populations, aiming at the prevention of outbreaks and epidemics [3] .
Acute gastroenteritis can be caused by infection in the gastrointestinal tract, caused by different infectious or parasite agents [4] [5] [6] [7] . They represent one of the main causes of mortality in humans, and in young animals, counting for about 25% of mortality [8] . Rotavirus is widely distributed in animals, which act as sources of rotavirus emergent strains, with these animals acting in the transmission between species and through reassortment leading to the emergence of new strains which have been reported in human infections [9] [10] [11] [12] .
The rotavirus (RV) belongs to the Reoviridae family and comprises nine species known as Rotavirus group A to I, with a recent proposal of the J species [13, 14] . Rotavirus A (RVA) is widespread worldwide and predominantly infects humans, bovines and other mammal species, as well as birds [15] . They have a double-stranded ribonucleic acid (dsRNA) genome, divided into 11 segments coding for structural proteins (VP1-VP4, VP6 and VP7) and nonstructural (NSP1-NSP5/NSP6) proteins [16, 17] .
There are records of a close relationship between Amazonian wildlife and human populations [18] , and this interaction is the effect of anthropogenic urbanization activities that result in the deforestation of forest areas, causing the degradation of previously isolated sites such as caves and small caves, a continuous and nature progressive process that has led not only to changes in wildlife habitats but also to a greater relationship with human populations in rural and urban environments, contributing to the occurrence and emergence of diseases different from what normally occurs in endemic regions [19] [20] [21] [22] .
Although the results of RVA have already been described globally [12, [23] [24] [25] [26] [27] [28] [29] [30] , in Brazil, the occurrence, diversity and role of rotavirus in these animals are still poorly studied, considering the large number of present species [4, [31] [32] [33] [34] .
In the Brazilian Amazon, especially in the state of Pará, the city of Belém and Northeast metropolitan mesoregions are some of the areas with the highest indexes of environmental changes [35] , which are concentrated, along with the fact that the knowledge of the global diversity of enteric virus in animals is scarce [36] .
Therefore, it is important to monitor the health of domestic and wild animals in their natural habitat, especially in areas with anthropic alterations that have an interface with rural communities and enterprises, in order to investigate the occurrence of RVA in this population. These communities are ecologically complex, because they have multiple hosts and endless pathogens that may eventually circulate in contiguous urban centers, in addition to the fact that it should also be considered that there is still a lack of studies showing the significance of these viruses infecting this population, as in the context of epidemiological surveillance, these animals become important, since they can be considered as natural sources, with the possibility of transmission to humans [37] [38] [39] .
The qualitative real-time polymerase chain reaction (qRT-PCR) used the NSP3 gene and the TaqMan probe from a highly conserved region of the rotavirus non-structural protein 3 (NSP3), which was previously used in samples from human origin and with low viral loads
Precipitation data were obtained from The Brazilian National Institute of Meteorology (Inmethttp://www.inmet.gov.br/) for the years of capture in the Expedito Ribeiro Settlement (2014) and Açailândia (2015) of the Data Collection Platforms (PCDs) of Belém, located 50 km from Santa Bárbara do Pará, and Tracuateua, located 50 km from Peixe-Boi and 100 km from Viseu. Garmin GPSMap 64s Global Positioning System (GPS) coordinates were collected in the field. The municipal boundaries were obtained on the website of the Brazilian Institute of Geography and Statistics (IBGE) (http://www.ibge.gov.br/) and data on deforestation and land use were obtained from the PRODES [43] and TerraClass [44] Projects. PRODES has annual data in digital format since 2000 and TerraClass presents biannual data since 2004.
The satellite image was generated using the sensor Sentinel 2 of the European Space Agency (ESA) (https://sentinel.esa.int/ web/sentinel/user-guides/sentinel-2-msi) with Open Access CC-BY License (http://open.esa.int/) from the years of 2017 and 2018.
All the data obtained was stored in a Geographic Database (BDG). The BDG was imported/ stored in a GIS for the editing of the graphic elements, establishment of topological relations between the elements and their respective attributes, spatial analysis and visualization of the result through thematic maps.
For the present study, forest fragments of similar size, shape and Phyto physiology were chosen, considering an open peri urban matrix with similar soil use. The selected fragments were distributed within the mesoregions studied, and in each selected fragment fecal samples were randomly collected from domestic and wild animals [45] .
Soil use classes were obtained from the TerraClass data mosaic from 2004 to 2016, because the study sites were in an area with a high cloud presence, which prevented observation (the area was not observed).
The data processing, interpretation, visualization and spatial analysis were performed in ArcGIS software (http://www.arcgis.com/). For the analysis of data related to the determination of the richness, composition and abundance of the fauna of the animals studied in the study area, considering the collection methods adopted and the species available in each city, each sample was considered as an independent sample.
The richness of wild fauna and domestic animals was determined by the total number of species including all collection methods, and the similarity of species was made by the chi-square analysis between the samples of the different treatments with the aid of the EstimateS 8.0 software [46] .
For the calculation of the Test T, the Statistica software was used, and the indices of infected animals in the two environments (forest fragment and peridomicile) were calculated for each treatment sampled by collection area, using the software Past 1.92. Aiming at comparing the values of the diversity indexes through the paired test, as well as the descriptive analysis of the anthropic effects [47] .
The data obtained for the occurrence of RVA and the questionnaires was inserted into a database for a descriptive analysis of the epidemiological profile of the animal population in the three forest ecosystems studied. In this analysis, descriptive statistical treatments were carried out, using customized "row-columns" type charts, referring to the data, in order to characterize the sample and quantify the results using absolute frequency values using the chi-square test and the Test T.
Population study, collection of clinical specimens and laboratory methodology. The flying animals (wild birds and chiroptera) were captured using mist nets which were opened at dawn (4:00 a.m.) and closed in the morning (9:00 a.m.) and were inspected every one hour until the closing, with a sampling effort of 15 days. This research was approved by National All procedures with animals were performed by veterinarians, being birds and bats identified and released at the same capture site. The fecal specimens were collected by stimulation of the rectal ampulla with the use of a "Zaragatoa", packed in cryogenic vials, identified, stored in liquid nitrogen, and later sent to the Laboratory.
Wild animals (small non-flying mammals) were trapped within live-containment traps of the Tomahawk cage (size 45x16x16cm) and Sherman type aluminum (size 30x9x8cm). In each sample plot, 61 traps were distributed, 20 Shermans and 41 Tomahawks being baited with a mixture made with peanut paste, sardines, cod liver oil and corn meal, as well as fruit like banana, apple and pineapple. All the traps used were inspected daily in the morning, the baits being exchanged when necessary and later after the capture in bags of cloth and at least five specimens of each species were chosen for the collection of biological material. The wild animals were sedated with a combination of ketamine 20mg/kg and xylazine 2mg/kg intramuscularly and subsequently, euthanized with anesthetic overdose of 2% lidocaine in the foramen magnum, according to the recommendation of the National Council for the Control of Animal Experimentation (CONCEA).
From October 2014 to April 2016, 1,282 fecal samples were collected from wild and domestic animals. Amongthese, 648 (50.5%) samples were randomly selected for RVA research and handled in Level Three Biosafety Laboratory (NB3).
The viral genome was extracted using the TRIZOL LS REAGENT protocol (INVITRO-GEN, USA/KIT QIAGEN), following the manufacturer's recommendations, with minor adaptation according to the protocol described in the supplemental data.
The qRT-PCR was conducted according to Zeng et al. [40] for the detection of RVA using the NSP3 segment of RVA as the target gene sequence. The assay was conducted in a mixture containing: RNAse-free H 2 O, TaqMan RT-PCR Mix (2x), TAqMan RT Enzyme Mix (40x), primers for the NSP3 gene, Primer NSP3 Forward (20mM), Primer NSP3 Reverse (20mM), probe NSP3 S (10nm), Template (RNA) 3μL, having a total reaction volume of 17μL and reverse transcription cycling of 50˚C, 30 minutes, denaturation of 95˚C, 10 minutes, annealing of 45 cycles of 95˚C, 15 seconds and extension of 60˚C, 1 minute.
The analyzes were considered positive when presenting the cycle threshold (CT) � 40. In order to guarantee a reliable test result, the measurements of contamination control were performed with the use of positive animal control (SA11 prototype) and a negative control (ultrapure water).
All RVA-positive samples were subjected to reverse transcription-polymerase chain reaction (RT-PCR) according to Mijatovic et al [41] to genotyping low viral loads samples. First round was performed with consensus primers N-VP4F1/N-VP4R1 and the Nested-PCR was conducted with N-VP4F2/N-VP4R2 primers to amplify VP4 gene. Amplicons were purified and sequencing for VP4 gene using the same primers of Nested-PCR. The sequences were collected from an automated ABI Prism 3130xl DNA sequencer (Applied Biosystems). The sequence fragments were assembled and edited using the Geneious Bioinformatics software platform v.8.1.7. Posteriorly, the data were compared with othersequences from the National Center for Biotechnology Information GenBank database using BLAST alignment tool to elucidate the RVA genotype of the samples.
From October 2014 to April 2016, a total of 648 fecal samples of wild and domestic animals belonging to three forest fragments areas were tested for the NSP3 gene by qualitative qPCR, and 178 (27.5%) were positive for RVA, distributed among the species: birds (23.6%), canines (21.35%), bats (17.98%), cattle (14.6%), horses (8.43%), small rodents (6.74%), swine (3.93%) and felines (3.37%). The CT interval ranged from 28. 47 It was possible to detect viral strains in all genders of animals studied and in the harvesting period none of the animals showed signs of acute infection and / or diarrhea.
Rotavirus A (RVA) detected in the present study of wild and domestic animals belonging to the three areas of forest fragment, according to Fig 2. In relation to the evaluated bovines, only in the city of Viseu, these species were studied because they were created extensively. In addition, most of the animals were young with ages varying from 1 day to 8 yearsold, history of deficient vaccination, lack of technical assistance and raised in the form of subsistence. The animals showed no symptoms of diarrhea, only low weight performance and poor sanitary management status. In relation to chiroptera, 32 (17.98%) positive samples for RVA were distributed among Carollia perspicillata species, with 12 (37.5%) being all adults, 9 (28.12%) Desmodus rotundus samples (4 young and 5 adults), 5 (15.6%) of Uroderma bilobata (15.62%), 3 (9.37%) of Artibeus lituratus and the species Artibeus Planirostus, Diaemus iyoug and Glossophagine with 1 (3.12%) each. These animals came from areas of forest fragments located near bovine and equine farms, in addition to inhabiting small chicken farms. Fig 3 shows the results obtained for all the species of animals investigated in the forest fragment as well as in the peridomicillus area.
The anthropic variables were analyzed for the three cities studied, as well as the use of the soil within the range of the animals, obeying the domicile, the peridomicile and the forest fragment where the traps of small rodents, birds and various species of animals were captured (Fig 4 and Fig 5) .
Considering the factors related to the anthropic activities in the three studied areas within the three cities of the present study, it was observed that the city of Santa Bárbara is the one that has a better area of preserved forest and the city of Viseu a smaller area. However, in the city of Santa Bárbara, a greater concentration of occupations was observed around the area of forest fragment. It was observed in this chosen area of the city, the presence of different families living in a rural settlement, surviving from the exploitation of forest resources and the creation of small animals for subsistence, such as poultry and fish farming, as well as family farming products.
The breeding of animals in native pastures was only observed in the cities of Peixe Boi and Viseu. Extensive livestock farming was practiced with beef cattle, equines for work and small animals (swine and goats). In relation to the most preserved pasture area, the city of Peixe Boi had the largest area, according to the data shown in Fig 5, however, in the city of Viseu, a higher regeneration was observed in the pastures during the period of the study, with significant secondary vegetation.
When comparing the climates of the three areas it was observed that the predominant climate is megathermal and humid with average annual temperature around 27˚C. The months of October, November and December are the hottest, with temperatures between 32˚C and 34˚C and absolute maximums around 41˚C. Annual rainfall is quite high, generally around 2,350 mm, but strongly concentrated from January to June (80%). From September to December, on the contrary, rainfall is rare, about 7%, with a short dry season, of moderate water deficit in those months. The relative humidity of the average air oscillates around 85%, as shown in Fig 6 [48] .
The description of the accumulated precipitation in the year of capture of the fecal specimens compared to the Climatological Normals (CLINO) for the period from 1961-1990 of the PCDs closest to the locations of the Expedito Ribeiro / Santa Bárbara settlement (Belém PCD), Vila Ananim / Peixe-Boi and Açaiteua / Viseu (Tracauateua PCD) show the frequency of rainfall in the regions, which facilitates the renewal of the pastures and the regeneration of the impacted forests, being an important indicator of the reduction of the damages caused by deforestation in the region.
The average deforestation index in the three study areas was calculated from data obtained from INPE information systems. It was observed that in the years of 2013 to 2014 there were no changes in these regions; in the period from 2014 to 2015 about 4.1% of the city of Viseu was changed and 1.6% of the city of Peixe Boi. In relation to the period of 2016, great changes were observed in Peixe Boi (79%) and in the city of Viseu (70%), thus demonstrating that changes in the natural ecosystem may be associated with the frequencies for RVA in the studied areas, according to Fig 7. When assessing the infected animals in relation to the uninfected animals in both the forest fragment and the peridomicile, considering as animals of the forest fragment the birds, the chiroptera and the small rodents and as animals of the peridomicile the canines, bovines, pigs, felines and horses, a percentage of 37.07% infected peri domestic animals (86/232) and 22.12% infected forest fragment animals (92/416) were obtained. Applying the selected statistical analysis, a Pearson x2 Chi-square value was obtained: 16.7159, df = 1 and p <0.001, meaning that the hypothesis was corroborated, that is, the greater the degradation of the environment, the more likely it will be the search for food by wild animals in adjacent areas, or in the edge of the forest or even in the peri domiciliary region. In this sense, the possibility of contagion with other species of animals, even humans, should be considered because of the capacity of the rotavirus to be transmitted via the fecal / oral route or through direct contact with the environment. It is important to point out that the animals detected in this study are important sources of viral strains.
A total of 80 stool samples were selected, reextracted and analyzed using PCR for the VP4 gene. Eight strains (10%) were positive for VP4 gene, being 2 strains bellowed to P [6] genotype and 6 to P[4]-type, according to
In the present study, RVA was detected circulating in 27.5% of the animals; 36% in domestic animals and 64% in wild animals, providing a unique dataset with qRT-PCR detecting a low viral load of RVA in different species, which further correlates with the deforestation index. These data are important because there is a lack of tests for RVA diagnosis in animals, since the current methods of RVA detection does not always detect in these populations [8] . With the advent of real-time PCR (qPCR), there was an exponential growth, compared to conventional PCR essays, since its superior accuracy, sensitivity and specificity is remarkable, and it is Rotavirus A in wild and domestic animals possible to detect RVA in a variety of animal species using NSP3 gene [49] . The sensitivity of RT-qPCR significantly improved the rate of RVA detection in clinical samples from animals and in this context, the present study proposed an interesting study metrics using virus spreading in the wild animals which inhabit forest fragments to indicate human population interventions, with the goal of preventing the virus outbreaks leveraged on the unique geographic characteristics of Brazil and its large number of species in Amazon.
Currently, no data have been described in the literature regarding the RVA detection using real-time qPCR technique in a wide variety of wild animal species. However, a study by Soltan et al. [50] conducted with horses and cattle detected RVA by RT-PCR, commercial RT-PCR and RT-qPCR in 36.7%, 51.4% and 56.9% respectively, differently from the present study that showed higher positivity for chiropterans (17.98%), canines (21.35%), birds (23.6%) and cattle (14.6%).
The first description of RVA in chiroptera was recorded in feces of Eidolon helvum caught in Vihiga, Kenya [51] . Afterwards, several strains of RVA were detected by different molecular techniques involving chiroptera, in several countries, including Kenya (E. helvum), China (Rhinolophus hipposideros and Aselliscus stoliczkanus), France (Myotis mystacinus), Cameroon (E. helvum) and Brazil [31, [51] [52] [53] [54] [55] . The present study shows the occurrence of RVA in 17.98% of the chiroptera, being among the species Carollia perspicillata (37.5%), Desmodus rotundus (28.12%), Uroderma bilobata (15.6%), Artibeus lituratus (9.37%), Artibeus Planirostus (3.12%), Diaemus iyoug and Glossophagine (3.12%).
Barquez et al. [56] reported that Desmodus rotundus is one of the three hematophagous species of the Phyllostomidae family, found throughout South America, Central America and Mexico. Of the positive chiroptera for RVA in the present study, a prevalence of 28.12% was of Desmodus rotundus. This species feeds on birds, can feed on mammals, mostly medium or large, facilitating the dissemination of viral spores among the community within the habitat, as observed in the present study. These findings show the importance of epidemiological data on the studied species due to the lack of studies involving species of neotropical chiroptera, and it is not possible to establish comparative parameters for these animals.
Regarding the circulation of RVA in canines and birds, the prevalence was 53% and 29%, respectively. Although in the Amazon region there are records of RVA, RVD, RVF and RVG that infect birds [57] [58] and RVD in migratory birds [59] , all were detected by RT-PCR assays differently from the present study which detected the RVA by RT-qPCR involving a variety of animal species.
On the other hand, the prevalence in felines (16%) and pigs (22%) was lower, probably because there are few animals of these species in the region, as well as few creations.
The study detected the presence of RVA in different species of animals both in areas near the home and in areas located in fragments of forest, characterized as forest remnants, since they were located in cities that suffered high environmental impacts due to vegetal extractivism, pasture formation for cattle breeding, exploitation of natural resources, and direct reflexes on the habitats of wild animals that can serve as virus sources, thus facilitating the dispersion of RVA among communities of coexisting animals.
It is worth emphasizing that these animals have a greater contact with the human populations of the studied areas since they cohabit with the humans in the region, besides having a high flow of movement between the forest extracts and environments chosen for the present study. However, it is noteworthy that only in the communities of Santa Bárbara and Viseu were collected fecal specimens from asymptomatic humans for diarrhea and tested for RVA, but all were negative. It is notorious yet, the existence of different levels of degradation in the studied environments, considering the presented data. The fragmentation of the forest generates many consequences on the Amazonian biota, being able to alter the diversity and the composition of the animal communities in the fragments and even to interfere in the ecological processes, without considering that the fragments of forest in the Amazon are influenced by the climate, possibly facilitating the dispersion of pathogens by the environment, since the wild animals detected in the present study are asymptomatic and have low viral load for RVA.
The occurrence of RVA in this population of animals may explain the possibility of dispersion of viral strains, since there is a proximity to the human population, besides the biological characteristics of these species that may represent important sources for gastroenteric viruses, along with the fact that all animals were asymptomatic for diarrhea.
Wild birds have unlimited flight capacity, were captured in an interface region between the peridomicile and forest fragments and it is believed that this region has not been influenced by anthropic activities such as those observed in the area of the present study. On the other hand, the breeding method for poultry and canines close to homes and the forest ecosystem, as they are created in the communities surveyed, probably facilitates direct contact with possible sources of contamination, since in the areas the use of septic tanks is deficient and sometimes non-existent, which may facilitate or even increase the risk of viral dispersion throughout the environment.
The high rates of increase and the analysis of land use in the researched areas may be important indicators of how these animals interact, since with deforestation, the populations of wild animals seek refuge in nearby communities facilitating the dispersal of infectious agents and the possible occurrence of carrier animals by direct contact or contamination of the local environment.
To our knowledge, this is the first study in which a real-time PCR assay was applied for the detection of RVA involving a wide variety of domestic and wild animals, facilitating practical utility in epidemiological and molecular studies and assisting in a perspective in the elaboration of sanitary control and monitoring, preventing possible outbreaks in the studied communities. The detection of positive animals was useful to monitor the infection of the agent in the animal population and to provide an early warning signal to predict an impending epidemic and a favorable risk for the human population, given the evidence of RVA circulation in the different forest fragments.
In addition, the RT-qPCR assay may be a useful alternative for the differential diagnosis of RV in possible coexisting mixed infections clinically indistinguishable such as those caused by other viral strains that cause gastroenteritis such as: astrovirus, coronavirus, picobirnavirus, calicivirus, among others as observed in the studies of Jing et al. [60] and Waruhiu et al. [61] .
Diarrhea associated with RVA infections in pigs is an important cause of increased mortality and economic losses in Europe. The most prevalent genotypes isolated from feces of Belgian diarrheal and non-diarrheal piglets in 2012 [62] demonstrate a wide range of combinations of genotypes G / P including; G3P [6] , G4P [6] , G5P [6] , G4P [7] , G5P [7] , G9P [7] , G9P [13] and G9P [23] . On the other hand, in the present study it was possible to detect only P [6] genotype, since majority of samples was asymptomatic for diarrhea.
Finding shows that different P genotypes of RVA strains interact with distinct blood group histological antigens (HBGA, ABOH, Lewis) and sialic acids via VP4 providing insight into the regional prevalence and increased zoonotic potential of some RVA of origin swine [63] . The genotype P [6] was identified in piglets in Brazil [64] and in Italy and Japan resembling genotype P [6] human [65, 66] .
In the population of animals studied the zoonotic transmission can be frequent, since the animals live in contact with humans and in precarious sanitary conditions. In Brazil, this genotype was described in animal and human populations in studies of Luchs et al. [32] ; Honma et al. [67] ; Araújo et al. [68] ; Mascarenhas et al. [69] and Lorenzetti et al. [70] such studies corroborate the importance of continuing to monitor genotypes to verify if uncommon strains or new strains are emerging and can infect animal populations or inter-species transmissions.
Regarding the genotype P [4] , itwas most detected in our samples in bats, dogs, swine and feline. This genotype is not common in animals, being more detected in human and environmental samples in various parts of the world and included our region [71] . It is important to emphasize that the indicators of environmental contamination in Brazil are significant and contribute to the possibility of human-animal transmission [71] . Such data need further investigation in later work to better characterize the interspecies transmission, since the occurrence of enteric viruses in different matrices demonstrates the anthropogenic impact of the exposed population around and points to the potential risk of infection by the possible exposure of individuals susceptible. Our findings may be useful for tracking fecal contamination in the environment using animals as possible sources thus minimizing the risk of infection by exposure to susceptible individuals, in this case different animal species or even human populations.
RVA were detected in wild and domestic animals using a RT-qPCR assay that analyzed samples that had low viral load for RVA. Although the samples are asymptomatic for diarrhea, it is necessary to conduct strategies for the monitoring and control of the animals in the areas studied in the human population as well as in other species of animals, as well as the implementation of preventive measures aimed at future outbreaks in communities animals in the resident population in these impacted areas. Therefore, the present study is unprecedented in the region and in the country in relation to the research of RVA in wild animals. It is noteworthy that, although the quality of the analyzed samples is characterized as low detectable viral load, the technique presented a good analytical response in the detection of the source animals for RVA, facilitating the selection of the samples for future genetic characterization tests.
|
What non-structural proteins are coded by Rotavirus?
| 433 |
NSP1-NSP5/NSP6
| 4,258 |
1,625 |
Rotavirus A in wild and domestic animals from areas with environmental degradation in the Brazilian Amazon
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6298726/
SHA: f3c309c596c20f48f493b77e714ce957d877bdcb
Authors: de Barros, Bruno de Cássio Veloso; Chagas, Elaine Nunes; Bezerra, Luna Wanessa; Ribeiro, Laila Graziela; Duarte Júnior, Jose Wandilson Barboza; Pereira, Diego; da Penha Junior, Edvaldo Tavares; Silva, Julia Rezende; Bezerra, Delana Andreza Melo; Bandeira, Renato Silva; Pinheiro, Helder Henrique Costa; Guerra, Sylvia de Fátima dos Santos; Guimarães, Ricardo José de Paula Souza e; Mascarenhas, Joana D'Arc Pereira
Date: 2018-12-18
DOI: 10.1371/journal.pone.0209005
License: cc-by
Abstract: Acute gastroenteritis is one of the main causes of mortality in humans and young animals. Domestic and mainly wild animals such as bats, small rodents and birds are highly diversified animals in relation to their habitats and ecological niches and are widely distributed geographically in environments of forest fragmentation in some areas of the Amazon, being considered important sources for viruses that affect humans and other animals. Due to the anthropical activities, these animals changed their natural habitat and adapted to urbanized environments, thus representing risks to human and animal health. Although the knowledge of the global diversity of enteric viruses is scarce, there are reports demonstrating the detection of rotavirus in domestic animals and animals of productive systems, such as bovines and pigs. The present study investigated the prevalence of Rotavirus A in 648 fecal samples of different animal species from the northeastern mesoregion of the state of Pará, Brazil, which is characterized as an urbanized area with forest fragments. The fecal specimens were collected from October 2014 to April 2016 and subjected to a Qualitative Real-Time Polymerase Chain Reaction (RT-qPCR), using the NSP3 gene as a target. It was observed that 27.5% (178/648) of the samples presented positive results for RVA, with 178 samples distributed in birds (23.6%), canines (21.35%), chiropterans (17.98%), bovines (14.6%), horses (8.43%), small rodents (6.74%), pigs (3.93%) and felines (3.37%), demonstrating the circulation of RVA in domestic animals and suggesting that such proximity could cause transmissions between different species and the occurrence of rearrangements in the genome of RVA as already described in the literature, associated to the traces of environmental degradation in the studied areas.
Text: Emerging and reemerging infectious diseases are increasing each year in several countries, with an impact both on human populations and on domestic and wild animals living in areas with considerable forest remnants [1] . Most of these diseases are of viral origin, suggesting the emergence and reemergence of viruses that are triggered by human activities that modify the environment [2] .
The populations of wild animals that inhabit forest fragments are strategic groups for studies of public health and the transmission of zoonosis, given that they act as indicators in the assistance and intervention in the human populations, aiming at the prevention of outbreaks and epidemics [3] .
Acute gastroenteritis can be caused by infection in the gastrointestinal tract, caused by different infectious or parasite agents [4] [5] [6] [7] . They represent one of the main causes of mortality in humans, and in young animals, counting for about 25% of mortality [8] . Rotavirus is widely distributed in animals, which act as sources of rotavirus emergent strains, with these animals acting in the transmission between species and through reassortment leading to the emergence of new strains which have been reported in human infections [9] [10] [11] [12] .
The rotavirus (RV) belongs to the Reoviridae family and comprises nine species known as Rotavirus group A to I, with a recent proposal of the J species [13, 14] . Rotavirus A (RVA) is widespread worldwide and predominantly infects humans, bovines and other mammal species, as well as birds [15] . They have a double-stranded ribonucleic acid (dsRNA) genome, divided into 11 segments coding for structural proteins (VP1-VP4, VP6 and VP7) and nonstructural (NSP1-NSP5/NSP6) proteins [16, 17] .
There are records of a close relationship between Amazonian wildlife and human populations [18] , and this interaction is the effect of anthropogenic urbanization activities that result in the deforestation of forest areas, causing the degradation of previously isolated sites such as caves and small caves, a continuous and nature progressive process that has led not only to changes in wildlife habitats but also to a greater relationship with human populations in rural and urban environments, contributing to the occurrence and emergence of diseases different from what normally occurs in endemic regions [19] [20] [21] [22] .
Although the results of RVA have already been described globally [12, [23] [24] [25] [26] [27] [28] [29] [30] , in Brazil, the occurrence, diversity and role of rotavirus in these animals are still poorly studied, considering the large number of present species [4, [31] [32] [33] [34] .
In the Brazilian Amazon, especially in the state of Pará, the city of Belém and Northeast metropolitan mesoregions are some of the areas with the highest indexes of environmental changes [35] , which are concentrated, along with the fact that the knowledge of the global diversity of enteric virus in animals is scarce [36] .
Therefore, it is important to monitor the health of domestic and wild animals in their natural habitat, especially in areas with anthropic alterations that have an interface with rural communities and enterprises, in order to investigate the occurrence of RVA in this population. These communities are ecologically complex, because they have multiple hosts and endless pathogens that may eventually circulate in contiguous urban centers, in addition to the fact that it should also be considered that there is still a lack of studies showing the significance of these viruses infecting this population, as in the context of epidemiological surveillance, these animals become important, since they can be considered as natural sources, with the possibility of transmission to humans [37] [38] [39] .
The qualitative real-time polymerase chain reaction (qRT-PCR) used the NSP3 gene and the TaqMan probe from a highly conserved region of the rotavirus non-structural protein 3 (NSP3), which was previously used in samples from human origin and with low viral loads
Precipitation data were obtained from The Brazilian National Institute of Meteorology (Inmethttp://www.inmet.gov.br/) for the years of capture in the Expedito Ribeiro Settlement (2014) and Açailândia (2015) of the Data Collection Platforms (PCDs) of Belém, located 50 km from Santa Bárbara do Pará, and Tracuateua, located 50 km from Peixe-Boi and 100 km from Viseu. Garmin GPSMap 64s Global Positioning System (GPS) coordinates were collected in the field. The municipal boundaries were obtained on the website of the Brazilian Institute of Geography and Statistics (IBGE) (http://www.ibge.gov.br/) and data on deforestation and land use were obtained from the PRODES [43] and TerraClass [44] Projects. PRODES has annual data in digital format since 2000 and TerraClass presents biannual data since 2004.
The satellite image was generated using the sensor Sentinel 2 of the European Space Agency (ESA) (https://sentinel.esa.int/ web/sentinel/user-guides/sentinel-2-msi) with Open Access CC-BY License (http://open.esa.int/) from the years of 2017 and 2018.
All the data obtained was stored in a Geographic Database (BDG). The BDG was imported/ stored in a GIS for the editing of the graphic elements, establishment of topological relations between the elements and their respective attributes, spatial analysis and visualization of the result through thematic maps.
For the present study, forest fragments of similar size, shape and Phyto physiology were chosen, considering an open peri urban matrix with similar soil use. The selected fragments were distributed within the mesoregions studied, and in each selected fragment fecal samples were randomly collected from domestic and wild animals [45] .
Soil use classes were obtained from the TerraClass data mosaic from 2004 to 2016, because the study sites were in an area with a high cloud presence, which prevented observation (the area was not observed).
The data processing, interpretation, visualization and spatial analysis were performed in ArcGIS software (http://www.arcgis.com/). For the analysis of data related to the determination of the richness, composition and abundance of the fauna of the animals studied in the study area, considering the collection methods adopted and the species available in each city, each sample was considered as an independent sample.
The richness of wild fauna and domestic animals was determined by the total number of species including all collection methods, and the similarity of species was made by the chi-square analysis between the samples of the different treatments with the aid of the EstimateS 8.0 software [46] .
For the calculation of the Test T, the Statistica software was used, and the indices of infected animals in the two environments (forest fragment and peridomicile) were calculated for each treatment sampled by collection area, using the software Past 1.92. Aiming at comparing the values of the diversity indexes through the paired test, as well as the descriptive analysis of the anthropic effects [47] .
The data obtained for the occurrence of RVA and the questionnaires was inserted into a database for a descriptive analysis of the epidemiological profile of the animal population in the three forest ecosystems studied. In this analysis, descriptive statistical treatments were carried out, using customized "row-columns" type charts, referring to the data, in order to characterize the sample and quantify the results using absolute frequency values using the chi-square test and the Test T.
Population study, collection of clinical specimens and laboratory methodology. The flying animals (wild birds and chiroptera) were captured using mist nets which were opened at dawn (4:00 a.m.) and closed in the morning (9:00 a.m.) and were inspected every one hour until the closing, with a sampling effort of 15 days. This research was approved by National All procedures with animals were performed by veterinarians, being birds and bats identified and released at the same capture site. The fecal specimens were collected by stimulation of the rectal ampulla with the use of a "Zaragatoa", packed in cryogenic vials, identified, stored in liquid nitrogen, and later sent to the Laboratory.
Wild animals (small non-flying mammals) were trapped within live-containment traps of the Tomahawk cage (size 45x16x16cm) and Sherman type aluminum (size 30x9x8cm). In each sample plot, 61 traps were distributed, 20 Shermans and 41 Tomahawks being baited with a mixture made with peanut paste, sardines, cod liver oil and corn meal, as well as fruit like banana, apple and pineapple. All the traps used were inspected daily in the morning, the baits being exchanged when necessary and later after the capture in bags of cloth and at least five specimens of each species were chosen for the collection of biological material. The wild animals were sedated with a combination of ketamine 20mg/kg and xylazine 2mg/kg intramuscularly and subsequently, euthanized with anesthetic overdose of 2% lidocaine in the foramen magnum, according to the recommendation of the National Council for the Control of Animal Experimentation (CONCEA).
From October 2014 to April 2016, 1,282 fecal samples were collected from wild and domestic animals. Amongthese, 648 (50.5%) samples were randomly selected for RVA research and handled in Level Three Biosafety Laboratory (NB3).
The viral genome was extracted using the TRIZOL LS REAGENT protocol (INVITRO-GEN, USA/KIT QIAGEN), following the manufacturer's recommendations, with minor adaptation according to the protocol described in the supplemental data.
The qRT-PCR was conducted according to Zeng et al. [40] for the detection of RVA using the NSP3 segment of RVA as the target gene sequence. The assay was conducted in a mixture containing: RNAse-free H 2 O, TaqMan RT-PCR Mix (2x), TAqMan RT Enzyme Mix (40x), primers for the NSP3 gene, Primer NSP3 Forward (20mM), Primer NSP3 Reverse (20mM), probe NSP3 S (10nm), Template (RNA) 3μL, having a total reaction volume of 17μL and reverse transcription cycling of 50˚C, 30 minutes, denaturation of 95˚C, 10 minutes, annealing of 45 cycles of 95˚C, 15 seconds and extension of 60˚C, 1 minute.
The analyzes were considered positive when presenting the cycle threshold (CT) � 40. In order to guarantee a reliable test result, the measurements of contamination control were performed with the use of positive animal control (SA11 prototype) and a negative control (ultrapure water).
All RVA-positive samples were subjected to reverse transcription-polymerase chain reaction (RT-PCR) according to Mijatovic et al [41] to genotyping low viral loads samples. First round was performed with consensus primers N-VP4F1/N-VP4R1 and the Nested-PCR was conducted with N-VP4F2/N-VP4R2 primers to amplify VP4 gene. Amplicons were purified and sequencing for VP4 gene using the same primers of Nested-PCR. The sequences were collected from an automated ABI Prism 3130xl DNA sequencer (Applied Biosystems). The sequence fragments were assembled and edited using the Geneious Bioinformatics software platform v.8.1.7. Posteriorly, the data were compared with othersequences from the National Center for Biotechnology Information GenBank database using BLAST alignment tool to elucidate the RVA genotype of the samples.
From October 2014 to April 2016, a total of 648 fecal samples of wild and domestic animals belonging to three forest fragments areas were tested for the NSP3 gene by qualitative qPCR, and 178 (27.5%) were positive for RVA, distributed among the species: birds (23.6%), canines (21.35%), bats (17.98%), cattle (14.6%), horses (8.43%), small rodents (6.74%), swine (3.93%) and felines (3.37%). The CT interval ranged from 28. 47 It was possible to detect viral strains in all genders of animals studied and in the harvesting period none of the animals showed signs of acute infection and / or diarrhea.
Rotavirus A (RVA) detected in the present study of wild and domestic animals belonging to the three areas of forest fragment, according to Fig 2. In relation to the evaluated bovines, only in the city of Viseu, these species were studied because they were created extensively. In addition, most of the animals were young with ages varying from 1 day to 8 yearsold, history of deficient vaccination, lack of technical assistance and raised in the form of subsistence. The animals showed no symptoms of diarrhea, only low weight performance and poor sanitary management status. In relation to chiroptera, 32 (17.98%) positive samples for RVA were distributed among Carollia perspicillata species, with 12 (37.5%) being all adults, 9 (28.12%) Desmodus rotundus samples (4 young and 5 adults), 5 (15.6%) of Uroderma bilobata (15.62%), 3 (9.37%) of Artibeus lituratus and the species Artibeus Planirostus, Diaemus iyoug and Glossophagine with 1 (3.12%) each. These animals came from areas of forest fragments located near bovine and equine farms, in addition to inhabiting small chicken farms. Fig 3 shows the results obtained for all the species of animals investigated in the forest fragment as well as in the peridomicillus area.
The anthropic variables were analyzed for the three cities studied, as well as the use of the soil within the range of the animals, obeying the domicile, the peridomicile and the forest fragment where the traps of small rodents, birds and various species of animals were captured (Fig 4 and Fig 5) .
Considering the factors related to the anthropic activities in the three studied areas within the three cities of the present study, it was observed that the city of Santa Bárbara is the one that has a better area of preserved forest and the city of Viseu a smaller area. However, in the city of Santa Bárbara, a greater concentration of occupations was observed around the area of forest fragment. It was observed in this chosen area of the city, the presence of different families living in a rural settlement, surviving from the exploitation of forest resources and the creation of small animals for subsistence, such as poultry and fish farming, as well as family farming products.
The breeding of animals in native pastures was only observed in the cities of Peixe Boi and Viseu. Extensive livestock farming was practiced with beef cattle, equines for work and small animals (swine and goats). In relation to the most preserved pasture area, the city of Peixe Boi had the largest area, according to the data shown in Fig 5, however, in the city of Viseu, a higher regeneration was observed in the pastures during the period of the study, with significant secondary vegetation.
When comparing the climates of the three areas it was observed that the predominant climate is megathermal and humid with average annual temperature around 27˚C. The months of October, November and December are the hottest, with temperatures between 32˚C and 34˚C and absolute maximums around 41˚C. Annual rainfall is quite high, generally around 2,350 mm, but strongly concentrated from January to June (80%). From September to December, on the contrary, rainfall is rare, about 7%, with a short dry season, of moderate water deficit in those months. The relative humidity of the average air oscillates around 85%, as shown in Fig 6 [48] .
The description of the accumulated precipitation in the year of capture of the fecal specimens compared to the Climatological Normals (CLINO) for the period from 1961-1990 of the PCDs closest to the locations of the Expedito Ribeiro / Santa Bárbara settlement (Belém PCD), Vila Ananim / Peixe-Boi and Açaiteua / Viseu (Tracauateua PCD) show the frequency of rainfall in the regions, which facilitates the renewal of the pastures and the regeneration of the impacted forests, being an important indicator of the reduction of the damages caused by deforestation in the region.
The average deforestation index in the three study areas was calculated from data obtained from INPE information systems. It was observed that in the years of 2013 to 2014 there were no changes in these regions; in the period from 2014 to 2015 about 4.1% of the city of Viseu was changed and 1.6% of the city of Peixe Boi. In relation to the period of 2016, great changes were observed in Peixe Boi (79%) and in the city of Viseu (70%), thus demonstrating that changes in the natural ecosystem may be associated with the frequencies for RVA in the studied areas, according to Fig 7. When assessing the infected animals in relation to the uninfected animals in both the forest fragment and the peridomicile, considering as animals of the forest fragment the birds, the chiroptera and the small rodents and as animals of the peridomicile the canines, bovines, pigs, felines and horses, a percentage of 37.07% infected peri domestic animals (86/232) and 22.12% infected forest fragment animals (92/416) were obtained. Applying the selected statistical analysis, a Pearson x2 Chi-square value was obtained: 16.7159, df = 1 and p <0.001, meaning that the hypothesis was corroborated, that is, the greater the degradation of the environment, the more likely it will be the search for food by wild animals in adjacent areas, or in the edge of the forest or even in the peri domiciliary region. In this sense, the possibility of contagion with other species of animals, even humans, should be considered because of the capacity of the rotavirus to be transmitted via the fecal / oral route or through direct contact with the environment. It is important to point out that the animals detected in this study are important sources of viral strains.
A total of 80 stool samples were selected, reextracted and analyzed using PCR for the VP4 gene. Eight strains (10%) were positive for VP4 gene, being 2 strains bellowed to P [6] genotype and 6 to P[4]-type, according to
In the present study, RVA was detected circulating in 27.5% of the animals; 36% in domestic animals and 64% in wild animals, providing a unique dataset with qRT-PCR detecting a low viral load of RVA in different species, which further correlates with the deforestation index. These data are important because there is a lack of tests for RVA diagnosis in animals, since the current methods of RVA detection does not always detect in these populations [8] . With the advent of real-time PCR (qPCR), there was an exponential growth, compared to conventional PCR essays, since its superior accuracy, sensitivity and specificity is remarkable, and it is Rotavirus A in wild and domestic animals possible to detect RVA in a variety of animal species using NSP3 gene [49] . The sensitivity of RT-qPCR significantly improved the rate of RVA detection in clinical samples from animals and in this context, the present study proposed an interesting study metrics using virus spreading in the wild animals which inhabit forest fragments to indicate human population interventions, with the goal of preventing the virus outbreaks leveraged on the unique geographic characteristics of Brazil and its large number of species in Amazon.
Currently, no data have been described in the literature regarding the RVA detection using real-time qPCR technique in a wide variety of wild animal species. However, a study by Soltan et al. [50] conducted with horses and cattle detected RVA by RT-PCR, commercial RT-PCR and RT-qPCR in 36.7%, 51.4% and 56.9% respectively, differently from the present study that showed higher positivity for chiropterans (17.98%), canines (21.35%), birds (23.6%) and cattle (14.6%).
The first description of RVA in chiroptera was recorded in feces of Eidolon helvum caught in Vihiga, Kenya [51] . Afterwards, several strains of RVA were detected by different molecular techniques involving chiroptera, in several countries, including Kenya (E. helvum), China (Rhinolophus hipposideros and Aselliscus stoliczkanus), France (Myotis mystacinus), Cameroon (E. helvum) and Brazil [31, [51] [52] [53] [54] [55] . The present study shows the occurrence of RVA in 17.98% of the chiroptera, being among the species Carollia perspicillata (37.5%), Desmodus rotundus (28.12%), Uroderma bilobata (15.6%), Artibeus lituratus (9.37%), Artibeus Planirostus (3.12%), Diaemus iyoug and Glossophagine (3.12%).
Barquez et al. [56] reported that Desmodus rotundus is one of the three hematophagous species of the Phyllostomidae family, found throughout South America, Central America and Mexico. Of the positive chiroptera for RVA in the present study, a prevalence of 28.12% was of Desmodus rotundus. This species feeds on birds, can feed on mammals, mostly medium or large, facilitating the dissemination of viral spores among the community within the habitat, as observed in the present study. These findings show the importance of epidemiological data on the studied species due to the lack of studies involving species of neotropical chiroptera, and it is not possible to establish comparative parameters for these animals.
Regarding the circulation of RVA in canines and birds, the prevalence was 53% and 29%, respectively. Although in the Amazon region there are records of RVA, RVD, RVF and RVG that infect birds [57] [58] and RVD in migratory birds [59] , all were detected by RT-PCR assays differently from the present study which detected the RVA by RT-qPCR involving a variety of animal species.
On the other hand, the prevalence in felines (16%) and pigs (22%) was lower, probably because there are few animals of these species in the region, as well as few creations.
The study detected the presence of RVA in different species of animals both in areas near the home and in areas located in fragments of forest, characterized as forest remnants, since they were located in cities that suffered high environmental impacts due to vegetal extractivism, pasture formation for cattle breeding, exploitation of natural resources, and direct reflexes on the habitats of wild animals that can serve as virus sources, thus facilitating the dispersion of RVA among communities of coexisting animals.
It is worth emphasizing that these animals have a greater contact with the human populations of the studied areas since they cohabit with the humans in the region, besides having a high flow of movement between the forest extracts and environments chosen for the present study. However, it is noteworthy that only in the communities of Santa Bárbara and Viseu were collected fecal specimens from asymptomatic humans for diarrhea and tested for RVA, but all were negative. It is notorious yet, the existence of different levels of degradation in the studied environments, considering the presented data. The fragmentation of the forest generates many consequences on the Amazonian biota, being able to alter the diversity and the composition of the animal communities in the fragments and even to interfere in the ecological processes, without considering that the fragments of forest in the Amazon are influenced by the climate, possibly facilitating the dispersion of pathogens by the environment, since the wild animals detected in the present study are asymptomatic and have low viral load for RVA.
The occurrence of RVA in this population of animals may explain the possibility of dispersion of viral strains, since there is a proximity to the human population, besides the biological characteristics of these species that may represent important sources for gastroenteric viruses, along with the fact that all animals were asymptomatic for diarrhea.
Wild birds have unlimited flight capacity, were captured in an interface region between the peridomicile and forest fragments and it is believed that this region has not been influenced by anthropic activities such as those observed in the area of the present study. On the other hand, the breeding method for poultry and canines close to homes and the forest ecosystem, as they are created in the communities surveyed, probably facilitates direct contact with possible sources of contamination, since in the areas the use of septic tanks is deficient and sometimes non-existent, which may facilitate or even increase the risk of viral dispersion throughout the environment.
The high rates of increase and the analysis of land use in the researched areas may be important indicators of how these animals interact, since with deforestation, the populations of wild animals seek refuge in nearby communities facilitating the dispersal of infectious agents and the possible occurrence of carrier animals by direct contact or contamination of the local environment.
To our knowledge, this is the first study in which a real-time PCR assay was applied for the detection of RVA involving a wide variety of domestic and wild animals, facilitating practical utility in epidemiological and molecular studies and assisting in a perspective in the elaboration of sanitary control and monitoring, preventing possible outbreaks in the studied communities. The detection of positive animals was useful to monitor the infection of the agent in the animal population and to provide an early warning signal to predict an impending epidemic and a favorable risk for the human population, given the evidence of RVA circulation in the different forest fragments.
In addition, the RT-qPCR assay may be a useful alternative for the differential diagnosis of RV in possible coexisting mixed infections clinically indistinguishable such as those caused by other viral strains that cause gastroenteritis such as: astrovirus, coronavirus, picobirnavirus, calicivirus, among others as observed in the studies of Jing et al. [60] and Waruhiu et al. [61] .
Diarrhea associated with RVA infections in pigs is an important cause of increased mortality and economic losses in Europe. The most prevalent genotypes isolated from feces of Belgian diarrheal and non-diarrheal piglets in 2012 [62] demonstrate a wide range of combinations of genotypes G / P including; G3P [6] , G4P [6] , G5P [6] , G4P [7] , G5P [7] , G9P [7] , G9P [13] and G9P [23] . On the other hand, in the present study it was possible to detect only P [6] genotype, since majority of samples was asymptomatic for diarrhea.
Finding shows that different P genotypes of RVA strains interact with distinct blood group histological antigens (HBGA, ABOH, Lewis) and sialic acids via VP4 providing insight into the regional prevalence and increased zoonotic potential of some RVA of origin swine [63] . The genotype P [6] was identified in piglets in Brazil [64] and in Italy and Japan resembling genotype P [6] human [65, 66] .
In the population of animals studied the zoonotic transmission can be frequent, since the animals live in contact with humans and in precarious sanitary conditions. In Brazil, this genotype was described in animal and human populations in studies of Luchs et al. [32] ; Honma et al. [67] ; Araújo et al. [68] ; Mascarenhas et al. [69] and Lorenzetti et al. [70] such studies corroborate the importance of continuing to monitor genotypes to verify if uncommon strains or new strains are emerging and can infect animal populations or inter-species transmissions.
Regarding the genotype P [4] , itwas most detected in our samples in bats, dogs, swine and feline. This genotype is not common in animals, being more detected in human and environmental samples in various parts of the world and included our region [71] . It is important to emphasize that the indicators of environmental contamination in Brazil are significant and contribute to the possibility of human-animal transmission [71] . Such data need further investigation in later work to better characterize the interspecies transmission, since the occurrence of enteric viruses in different matrices demonstrates the anthropogenic impact of the exposed population around and points to the potential risk of infection by the possible exposure of individuals susceptible. Our findings may be useful for tracking fecal contamination in the environment using animals as possible sources thus minimizing the risk of infection by exposure to susceptible individuals, in this case different animal species or even human populations.
RVA were detected in wild and domestic animals using a RT-qPCR assay that analyzed samples that had low viral load for RVA. Although the samples are asymptomatic for diarrhea, it is necessary to conduct strategies for the monitoring and control of the animals in the areas studied in the human population as well as in other species of animals, as well as the implementation of preventive measures aimed at future outbreaks in communities animals in the resident population in these impacted areas. Therefore, the present study is unprecedented in the region and in the country in relation to the research of RVA in wild animals. It is noteworthy that, although the quality of the analyzed samples is characterized as low detectable viral load, the technique presented a good analytical response in the detection of the source animals for RVA, facilitating the selection of the samples for future genetic characterization tests.
|
What is qRT-PCR?
| 434 |
qualitative real-time polymerase chain reaction
| 6,347 |
1,627 |
Regulation and Maintenance of an Adoptive T-Cell Dependent Memory B Cell Pool
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5120830/
SHA: f4a82ad66962355ffb09e4d1b57fde3e94f0ec53
Authors: Anson, Marie; Amado, Inês; Mailhé, Marie-Pierre; Donnadieu, Emmanuel; Garcia, Sylvie; Huetz, François; Freitas, Antonio A.
Date: 2016-11-23
DOI: 10.1371/journal.pone.0167003
License: cc-by
Abstract: We investigated the ability of monoclonal B cells to restore primary and secondary T-cell dependent antibody responses in adoptive immune-deficient hosts. Priming induced B cell activation and expansion, AID expression, antibody production and the generation of IgM(+)IgG(-) and IgM(-)IgG(+) antigen-experienced B-cell subsets that persisted in the lymphopenic environment by cell division. Upon secondary transfer and recall the IgM(-)IgG(+) cells responded by the production of antigen-specific IgG while the IgM(+) memory cells secreted mainly IgM and little IgG, but generated new B cells expressing germinal center markers. The recall responses were more efficient if the antigenic boost was delayed suggesting that a period of adaptation is necessary before the transferred cells are able to respond. Overall these findings indicate that reconstitution of a functional and complete memory pool requires transfer of all different antigen-experienced B cell subsets. We also found that the size of the memory B cell pool did not rely on the number of the responding naïve B cells, suggesting autonomous homeostatic controls for naïve and memory B cells. By reconstituting a stable memory B cell pool in immune-deficient hosts using a monoclonal high-affinity B cell population we demonstrate the potential value of B cell adoptive immunotherapy.
Text: Immune responses to infectious agents have different out-comes that can either protect or fail to control disease. Protection from re-infection relies on the establishment of efficient secondary immune responses that require the generation of antigen-specific "memory" B and T lymphocytes. The generation and selection of T-cell dependent "memory" B cells involves distinct molecular mechanisms: immunoglobulin isotype recombination and somatic hyper mutation, both dependent on the expression of AID [1] . Therefore, a long-standing paradigm defined memory B cells as IgM -IgG + isotype switched cells [2] . Different lines of evidence indicate that this is not always the case. In humans, it has been shown that some IgM + B cells bear the phenotype of other memory cells, being CD27 + , and carry frequent point mutations in the V region of the Ig genes, suggesting that they must represent highly selected B cell populations [3] . In mice, populations of CD19 + IgM + able to mount secondary responses have been identified [4] [5] [6] [7] . Overall these findings suggest that the T-cell dependent memory B cell pool comprises distinct subsets of memory B cells with different properties and effector functions [4] [5] [6] .
The biological properties that ensure the long-term persistence of memory and efficient secondary antibody responses have not been yet completely established. While initial studies proposed that after transfer memory B cells faded rapidly [8, 9] suggesting that long-lasting memory required the continuous recruitment of new cells [8] and/or antigen persistence [9, 10] , others suggested that memory B cells were able of extended survival without cell division [11] in the absence of antigen [2] . Long-term persistence of antibody responses has also been attributed to populations of long-lived plasma cells mainly resident in the bone marrow following immunization [12, 13] . The demonstration of the compartmentalization of "antibody memory" into different cellular layers suggested that the separate subsets of memory B cells behave differently. Accordingly, it has been reported that IgG + cells that could rapidly respond upon challenge did not persist long, while IgM + cells could generate a second wave of germinal center responses allowing persistence of memory [4] [5] [6] 14] .
Currently, immunotherapy approaches using passive antibody transfer [15, 16] ) is limited by the short half-life of immunoglobulin. Therefore new therapy strategies may require the adoptive transfer of high-affinity memory B cells, ready to respond and able to persist. The development of these new strategies requires a profound understanding of the mechanisms that regulate memory B cell numbers and ensure long persistence upon adoptive transfer. Moreover, knowledge of the mechanisms that determine the size of the memory B cell pool may be also critical to device new reconstitution strategies. So far, studies comparing populations of naïve and memory B cells have been hindered both by the vast clonal heterogeneity of the cells involved and by our inability to generate significant numbers of antigen specific memory B cells. Indeed in a normal laboratory mouse the population of B cells bearing a "memory IgG + phenotype' represent a small fraction of the total B cell pool (<0.5%) and upon immunization the number of the clonal diverse antigen-specific memory B cells generated is generally very limited (<10 3 ) [1, 6] .
To circumvent these limits, we decided to compare the properties of homogeneous populations of naïve and memory B cells of known antigen specificity, belonging to the same clone. We used SW HEL transgenic mice where B cells bear a high-affinity BCR specific for HEL and are capable of class switch recombination and somatic hypermutation (SHM) [17, 18] . To identify "memory B cells" the SW HEL mice were crossed with mice where AID transcription provokes the permanent expression of an YFP reporter in post-germinal center lymphocytes [19] . These mice were in a Rag2-deficient background and therefore contain a pure population of monoclonal HEL-specific B cells. To generate memory cells, purified naïve B cells from the SW HEL .AID/YFP.Rag2 -/mice were transferred into adoptive hosts together with monoclonal OVA-specific CD4 + T cells from OTII.Rag2 -/-TCR transgenic mice. Upon immunization with OVA-HEL complexes, we obtained a significant number of persisting HEL-specific IgM + Ig-G -YFP + and IgM -IgG + YFP + memory B cells, number that did not correlate to the number of precursor naïve cells initially injected suggesting that the memory B cell pool is regulated independently. We characterized the functional capacity of these two memory cell types in immune deficient hosts.
Mice B6 and B6.Rag2-/- [20] mice were kept at the Centre Des Techniques Avancées (CDTA), Centre National de la Recherche Scientifique (CNRS), Orleans, France; SWHEL.AID/YFP.Rag-/-mice, obtained by crossing SWHEL (18)(a gift of Dr. Robert Brink) and AID/YFP [19] (a gift of Dr. Rafael Casellas) with B6.Rag2-/-mice. OTII.Rag-/-mice were kept in our animal facilities at the Pasteur Institute. Experiments were preformed according to Pasteur Institute Safety Committee in accordance with French and European guidelines and the ethics Committee of Paris 1 (permits 2010-0002, -0003 and -0004). Euthanasia of the mice was performed by cervical dislocation. This specific study was approved by the European Research Council (ERC) committee related to the grant AdG09 249740-QSIS. The general status of the mice was controlled daily by monitoring the appearence of obvious pain, distress or suffering (prostration, respiratory issues, loss of weight). The end-point of the experiment was determined by a loss of more than 20% of the weight or as soon as the distress signs appeared. In this case, experiment was stopped and the animals were euthanized.
Single-cell suspensions of B cells from spleens and lymph nodes of SW HEL .AID/YFP.Rag -/mice together with CD4 + T cells from spleens and lymph nodes of OTII.Rag -/mice were transferred intravenously into the retro-orbital sinus of B6.Ly5 a IgH a or B6.Rag2 -/recipient mice. Mice received 10 6 HEL + B cells and 10 6 CD4 + T cells unless stated otherwise. Mice were immunized 24H later with 1 mg of Ovalbumin coupled to Hen Egg Lysozyme (OVA-HEL) in 50μg of Alu-S-Gel (Serva) we determined as the optimal dose of Ag (data not shown).
Naive cells from SW HEL .AID/YFP.Rag -/mice and memory B cells subsets from immunized B6.Rag -/hosts mice were purified from spleens and lymph nodes by flow cytometry sorting. Single-cell suspensions containing 5×10 4 B cells and 10 6 T cells were transferred intravenously into B6.Rag2 -/recipient hosts. The purity of sorted cells was above 98%. 24 h after transfer, mice were immunized with 1 mg of OVA-HEL.
Spleen, bone marrow, inguinal and mesenteric lymph nodes single-cell suspensions were stained for cell surface or intracellular proteins with appropriate combinations of the following monoclonal antibodies conjugated to pacific blue, Qdot-655, Brillant Violet 605, allophycocyanin, peridinin chlorophyll protein-cyanine 5.5, phycoerythrin, phycoerythrin-cyanine7: anti-CD19 (6D5), anti-IgM (R6-60.2), anti-IgG1 (X56), anti-CD138 (281-2), anti-Gl7 (Gl7), anti-CD95 (Jo2), anti-CD62L (MEL-14), anti-CD69 (H1-2F3), anti-BAFFR (7H22-E16), anti-CXCR5 (L138D7), anti-IA b (AF6-120.1), anti-CD80 (16-10A1), anti-CD73 (TY-11-8) and anti-PDL2 (TY25) and anti-Ki-67 (mm1) purchased from Becton Dickinson Pharmingen, Biolegend, Invitrogen and eBioscience. Cells were also stained with HEL (Sigma) coupled with AF594 using Alexa Fluor1 594 Protein Labeling Kit from Life technologies. Before staining, cells were treated with Fc-Block (CD16/CD32, Becton Dickinson Pharmingen). Dead cells were excluded during analysis according to their light-scattering characteristics. For intracellular stainings, cells were first stained with antibodies specific for cell surface antigens. Then, cells were fixed and permeabilized according the manufacturer's recommendations (BD Bisciences). For proliferation assay, mice were injected i.p. with 50 mg/kg of BrdU (Sigma-Aldrich) and were killed 24 or 72 hours later. Incorporated BrdU was detected intracellularly using anti-BrdU APC-conjugated antibodies according to the manufacturer's recommendations (BD Biosciences). All data acquisitions and analyses were performed with LSRFortessa (Becton Dickinson) interfaced with BD FACSDiva (Becton Dickinson) and FlowJo (Tree Star) software. Subsets of memory B cells were sorted as CD19 + HEL + YFP + IgM + or IgG + and naive cells as CD19 + HEL + YFP -IgM + using a FACSAriaIII flow cytometer. The purity of the sorted populations varied from 90-95%.
Sera HEL-specific Ig concentrations were quantified by ELISA. Plates were coated with HEL and saturated with PBS-5% Milk. Dilutions of sera were added. After incubation (2 hours, 37˚C) and washing, HRP-labeled anti-mouse IgM or IgG antibodies were added. After incubation and washing, bound antibodies were revealed with the substrate O-phenylenediamine and H2O2. The reaction was stopped after 10 min. by addition of 10% SDS and the absorbance read at 492nm in a multiscan spectrometer. Ig concentrations were determined by comparing the displacement of the dilution curves in the linear interval between standards at a concentration of 1 mg/ml and the serum samples.
The quantification of IgG or IgM secreting cells was assayed by ELISpot technique. Briefly, plates were coated with HEL. After saturating, the cells were distributed into the micro wells in RPMI1640-2%FCS. The plates were incubated for 12 h at 37˚C, 5% CO2 atmosphere. After extensive wash, plates were incubated with goat anti-mouse IgM or anti-IgG labeled with alkaline phosphatase. After washing, the revealing substrate was added (2,3 mM 5-bromo-4-chloro-3-indolyl phosphate diluted in 2-amino-2-methyl-1-proprenolol buffer).
Spleens from 14 day-immunized mice were initially fixed with paraformaldehyde and embedded in 4% low-gelling-temperature agarose (type VII-A; Sigma-Aldrich) prepared in PBS. 150μm slices were cut with a vibratome (VT 1000S; Leica) in a bath of ice-cold PBS. For immunolabeling, samples were saturated with PBS supplemented with 10% of fetal calf serum, then were labeled with primary antibodies anti-B220-APC (clone RA3-6B2) and anti-IgD-PE (clone 11-26c.2a) and analyzed with a spinning disk confocal microscope equipped with a CoolSnap HQ2 camera (Photometrics) and a 20x objective. Images were acquired and analyzed with MetaMorph 7 imaging software Molecular Devices).
Sample means were compared using the Student's t test. Sample means were considered significantly different at p < 0.05.
During an immune response the complexity of determinants expressed by immunizing antigen and the degeneracy of antigen-specific recognition results in a vast heterogeneity of responding cells rendering impossible the direct comparison of the properties of naïve and memory B cells belonging to the same clone. We have devised an experimental system that permits the comparison between naïve and memory B cells expressing the same antigen receptor and allows marking permanently memory B cells. For that purpose we used SW HEL transgenic mice in a Rag2-deficient background holding a single population of monoclonal B cells, all bearing a high-affinity BCR specific for HEL and capable of class switch recombination and somatic hypermutation (SHM) [17, 18] . To identify antigen-experienced B cells the SW HEL . Rag2 -/mice were crossed with mice where AID transcription induces the permanent expression of an YFP reporter in post-germinal center lymphocytes [19] . Since in intact Tg mice immune responses were not traceable, probably because of the presence of low level pre-existing anti-HEL antibodies that neutralize the immunizing protein, we used an adoptive cell transfer strategy to study the ability of the high affinity monoclonal B cell to reconstitute response in immune-deficient hosts and generate antibody memory. Purified naïve B cells from the SW HEL .AID/YFP.Rag2 -/mice were transferred into Rag2-deficient mice together with monoclonal OVA-specific CD4 + T helper cells from OTII.Rag2 -/-TCR transgenic mice. The day after, host mice were immunized with OVA-HEL complexes (Fig 1A) . In these conditions, antigenic challenge resulted in B cell activation and the development of significant numbers of CD19 + HEL + AID/YFP + B cells, which were not detected in non-immunized mice or in mice immunized in absence of helper T cells (Fig 1B) . We followed the early kinetics of this response. The number of HEL-specific B cells increased from the initial 2x10 6 transferred to about 15x10 6 at day 14 (Fig 1C left) the B cells expressing AID/YFP being the dominant population (Fig 1C right) . A fraction of the HEL-specific B cells underwent class switch recombination and at day 14 we recovered both IgM + IgG -AID/YFP + and IgM -IgG + AID/YFP + cell populations (Fig 1B) . B cell expansion and phenotypic changes were accompanied by the production of IgM and IgG HEL-specific antibodies ( Fig 1D) . Two weeks after antigenic challenge we observed the formation of germinal centers in the spleen of the host mice ( Fig 1E) . Coherently we found that while upon adoptive transfer all B cells expressed CD95, only after antigenic challenge most YFP + B cells expressed the germinal center specific marker GL7 (Fig 1F) .
In conclusion, the adoptive cell transfer strategy allowed the development of a primary immune response with B cell activation and expansion, induction of AID expression, class switch recombination, antigen-specific IgM and IgG antibody production and germinal center formation.
We studied the evolution of the B cell response. From two weeks onwards the total number of B cells contracted and at four weeks we recovered about 2-4x10 6 cells, number that remained stable up to week 20 (Fig 2A) . High titers of HEL-specific IgG were kept from week 3 to 8, declined thereafter, but were still significantly elevated 20 weeks later (Fig 2B) . A population of cells secreting HEL-specific Igs was present in the spleen (Fig 2C) , but not in the BM (not shown) even at the late time points. About 60% of the recovered cells exhibited the phenotype of antigen-experienced ("memory") CD19 + HEL + AID/YFP + expressing either IgM or IgG ( Fig 2D and 2E) . We compared the phenotype of the two AID/YFP + IgM + and AID/YFP + IgM -IgG + memory cell populations recovered with that of the naïve B cells (Fig 2F) . We found that antigen-experience and naïve B cells expressed similar levels of CD62L, CD69 and BAFFR (not shown). Antigen-experienced cells presented sustained expression of CD95 and increased levels of PNA, but the vast majority lost expression of the germinal center marker GL7 present at earlier times post-immunization ( Fig 2F compare to Fig 1F) . Compared to naïve B cells, AID/ YFP + cells expressed higher levels of CD80 and MHC class II and down-regulated expression of CXCR5 (Fig 2F) . These findings indicate that the post-germinal center AID/YFP + B cells express an activated phenotype [5, 21] , have increased antigen-presenting capacity [22] , but may loose the ability to re-enter primary follicles being CXCR5 low [23] . We have also compared the patterns of gene expression (RNAseq) by naïve, activated (YFPcells of immunized mice) and both populations of YFP + memory cells. The data shows a clear discrimination of naïve and activated/memory cells while indicating only minor differences between both subsets of YFP + memory cells (Fig 3) .
mRNA was isolated from sort-purified Naïve (CD19 + HEL + YFP -IgM + ) IgM + IgGor IgM -IgG + HEL + CD19 + YFP + memory B cells from spleen of different recipient mice. Total recommended by the manufacturer. The validated libraries were then subjected to DNA sequencing. The analysis is performed using the R software, Bioconductor packages including DESeq2 and the PF2tools package (version 1.2.9) developed at PF2 (Institut Pasteur). Normalization and differential analysis are carried out according to the DESeq2 model and package (version 1.8.1). Fig 3A shows a representative heat map of the different cells populations. Fig 3B shows
Late in the immune response persistent B cell numbers were kept by active cell division as a significant fraction of the cells were Ki67 + (Fig 2G left) and incorporated BrdU (Fig 2G middle) . The frequency of BrdU + cells was higher among the AID/YFPcells (15%) than in the major AID/YFP + memory population (3%) and similar between the IgM + and IgM -AID/ YFP + populations ( Fig 2G middle and not shown) . Three days after BrdU pulse populations were clear of BrdU + cells (Fig 2G right) attesting their high division rate. In spite of their increased proliferation rate, memory cells numbers were stable indicating that proliferation may be compensated by cell death as suggested by the frequency of caspase3 + cells (Fig 2H) . The frequency of Caspase3 + cells was higher among the AID/YFP + cells suggesting that a fraction of these cells may represent cells undergoing terminal differentiation. Importantly, these findings demonstrate that the transfer strategy allowed the generation of significant numbers of persisting antigen-experienced YFP + cells.
It is not yet known whether the number of antigen-experienced memory B cells correlated to the number of naïve B cells or if it is controlled independently of the initial number of antigenspecific B cells present. To approach this question we transferred different numbers of mature naïve B cells from SW HEL .AID/YFP.Rag2 -/donors (ranging from 10 5 to 5.10 6 ) into Rag2-deficient mice together with an excess of CD4 + T helper cells (10 6 ) and immunize the hosts the day after cell transfer with OVA-HEL in optimal non-limiting quantities. To directly compare the results obtained after the transfer of different all numbers we allowed the responses to reach steady-state eight weeks after antigenic challenge. We studied the amplitude of the immune response by measuring the serum titers of HEL-specific IgG antibodies and enumerating the number of HEL-specific B cells recovered. We found that in the presence of excess T cell help, the levels of the HEL-specific IgGs (Fig 4C) , and both the total number of HEL-specific ( Fig 4A) and of memory YFP + B cells recovered (Fig 4B) , did not correlate to the number of antigen specific naïve B cells initially transferred.
Memory B cells are defined functionally by their ability to induce secondary IgG antibody responses upon secondary antigenic challenge. We investigated whether the subsets of AID/ YFP + IgM + and AID/YFP + IgM -IgG + antigen-experienced (memory) B cells persisting at late time points could mount secondary IgG responses and persist after secondary transfer. For this purpose we followed two different experimental strategies. In the first, 5x10 4 cells of either IgM + or IgM -IgG + memory B cells, were transferred with an excess helper OTII CD4 + T cells into secondary Rag-deficient hosts that were boosted with OVA-HEL the day after cell transfer. In the absence of immunization antibody levels were undetectable (not shown) and three weeks after transfer recovery of both memory B cell subsets was about 10-20% of the initial cell input, exceeding naïve B cell recovery (Fig 5A) , supporting the notion that memory B cells may not require specific ligand recognition to survive (2). One cannot exclude, however, that cross-reactivity of the BCR transgene with environmental antigens may allow signaling sufficient to maintain naïve and memory cell survival in the absence of HEL [24] . Following immunization, the secondarily transferred AID/YFP + IgM -IgG + cells responded promptly with the exclusive production of significant levels HEL-specific IgG thus confirming their memory statute (11) . The AID/YFP + IgM + B cells in response to antigenic boost produced only limited amounts of IgM antibodies (Fig 4B) , little IgG antibodies, but did generate GL7 + B cells more efficiently than the IgG + memory B cell population (Fig 5D) . Thus the IgM + subset may contain precursors able to generate a secondary germinal center reaction and a new progeny of IgG + effectors (4). With time antibody levels decayed rapidly suggesting that the number of transferred memory B cells declined in the secondary hosts after antigenic boost. Indeed, IgM + and IgG + memory B cells failed to expand and 3 weeks after immunization cell recovery was similar to the retrieval observed in the non-immunized hosts (compare Fig 5E and 5A) . In similar experimental conditions, naïve B cells following immunization expanded, acquired AID/ YFP expression and their numbers more than doubled the number initially injected (Figs 5F and 2A). These data suggest that a significant fraction of the memory B cells generated have a reduced expansion capacity being programmed for rapid differentiation for effector functions.
Besides long-term survival memory B cells must maintain functional activity in the absence of nominal antigen to be fully effective. To test this we used an alternative approach where memory cells were parked in secondary Rag-deficient hosts for 30 days before re-immunization. We found that under these conditions antigenic challenge resulted in the production of HEL-specific IgG antibodies and in a 100 fold increase in the number of cells recovered, expansion that largely exceed that observed after immediate challenge (Fig 5G) .
The aim of this study was to characterize the fate of activated B cells and the generation of memory B cells. To do this, we adoptively transferred monoclonal B cells into immune deficient hosts followed by immunization in presence of T cell help. This strategy resulted in the development of different B cell memory subsets, namely IgM + and IgG + , as described for in situ generated memory cells [4, 6, 14] . These findings indicate that distinct memory B cell subsets are not the result of the heterogeneity of initially responding naive cells, but originate from the differentiation of a single B cell clone.
While studying the respective rate of proliferation of both types of memory B cells, we found the same high rate of proliferation for IgM + and IgG + memory B cells. These results contrast with previous published data. First it was been reported that "in situ" memory B cells persist as resting non-dividing cells [11, 25] . However, we have shown that upon adoptive transfer and in absence of competing cells, B cells increase their division rate to occupy the available empty niche [26] , which may explain the higher division rate observed here using this adoptive cell transfer strategy. Secondly, comparing life spans among heterogeneous memory B cell populations it was previously reported a lower division rate among the IgM + subset compared to the IgG + polyclonal subset [6] . Differences in BCR affinity between IgM + and IgG + memory clones may explain the higher division rate previously observed among the IgG + cells [6] . In contrast we compared memory B cell subsets belonging to the same clone bearing the same high affinity BCR. Overall these observations support the notion that lymphocyte division rates and life spans are not an intrinsic cell property, but rather determined by the environment and the presence of competing populations [27] . They demonstrate that upon the correct conditions memory B cells can persist by cell division.
An important question was whether the number of memory B cells depends on the number of initial naïve B cells. We found that, in the presence of an excess of T cell help, that was not the case. However, it was previously reported during polyclonal responses that serum titers of anti-HSA was proportional to the number of cells transferred into irradiated mice [28] . It is possible that limited antigen-specific T-B cell encounters may constraint the number of responding B cells and thus determine linear precursor-progeny between naïve and memory B cells. Our findings indicate that within a single clone the number of precursor naive B cells present in the peripheral B cell pool does determine neither the intensity nor the final number of memory B cells in response to an optimal dose of antigen. They suggest that the size of memory B cell pool may be controlled independently of the number of naïve B cell precursors and that in the absence of clonal competition the memory niche can be filled with a single monoclonal population. Considering diverse polyclonal populations, the limited niche for memory cells will imply strong competition among clones resulting in the selection of best fit (high affinity) cells: rare mutated clones being able to out compete more frequent but less avid clones. In our settings, the transgenic memory B cells are likely to counter select any new mutant clones since they express a very high affinity BCR selected in the course of a secondary immune response [29] . Thus, notwithstanding the expression of AID and proliferation we did not detect any BCR VH and VL Ig-chain nucleotide mutations among the recovered memory B cells (not shown). These findings may have implication for vaccination protocols as they indicate that each new antigenic exposure or unrelated immunization would add extra competing clones supporting the need for repeated antigenic boosts to prevent memory B cell attrition. They also demonstrate that the memory B cell pool can be reconstituted from a relatively small number of antigen-specific cells.
It is likely that the relatively poor memory B cell expansion observed after immediate boost after adoptive transfer could be due to the lack in Rag-deficient hosts of the appropriate environment required for memory B cell survival and function. It should be pointed out that B cell transfer into transgenic ML5 Rag-deficient hosts expressing low levels of HEL [29] resulted in rapid cell loss and recovery suggesting that in these hosts, B cells are trapped by antigen in locations were they are unable to survive (not shown). Nevertheless, it has been shown that B cells can drive the maturation of follicular dendritic cells and the organization of lymphoid follicles [30] . Similarly, transferred helper cells may also modify their immediate environment. Thus, by allowing lymphocytes to adapt and modify their immediate environment we improved their response and more important, we recovered the memory B cell pool size present in the original donor mice.
In this study we show that it is possible to fully reconstitute a primary response and the establishment of antibody memory in immune deficient mice after adoptive transfer of antigen-specific monoclonal B cells together with a population of monoclonal helper T cells. Indeed, it is generally believed that in immune deficiencies, B cell therapy has restricted application due to intrinsic defects of host's lymphoid organs structure that may prevent development of immune responses, germinal center formation, establishment of antibody memory and limit cell survival. In contrast we showed that after adoptive transfer in immune deficient hosts antigen immunization induced B cell activation and expansion, induction of AID expression, class switch recombination, antigen-specific IgM and IgG antibody production, germinal center formation and the generation of two subsets of AID/YFP + IgM + IgGand AID/YFP + Ig-M -IgG + antigen-experienced B cell subsets able to persist in a lymphopenic environment by cell division mimicking responses obtained in intact non-Tg mice [4] . Upon challenge the AID/YFP + IgM -IgG + cells responded promptly with the production of HEL-specific IgG while the AID/YFP + IgM + B cells secreted only limited amounts of IgM antibodies and fail to produce IgG. In contrast the AID/YFP + IgM + B cells could give rise to new GL7 + B cells, suggesting that full reconstitution of the memory B cell pool may require transfer of the different antigen-experienced B cell subsets. Importantly, we found that the recall responses were more efficient if the transferred memory cells were given the required time to adapt to their new environment, suggesting that a period of accommodation is necessary before the transferred cells are fully capable to respond. Our findings also show that different processes can modify the survival conditions of memory B cells. Finally, we found that the generation of the memory B cell pool in response to an optimal dose of Ag did not rely on the number of the initially responding B cells, suggesting autonomous homeostatic controls for naïve and memory B cells a property that may allow reconstitution of the memory pool in immune-deficient hosts using a limited number of precursor naïve B cells. An autonomous control of the memory B cell pool where each antigenic exposure adds new competing clones supports the notion of vaccination strategies using antigenic boosting to prevent memory B cell attrition. Overall the findings reported demonstrate that it is possible to reconstitute the memory B cell pool of an immune deficient host with an artificially induced population of monoclonal high affinity memory B cells.
|
What is required to establish a secondary immune response to a viral infection?
| 4,255 |
the generation of antigen-specific "memory" B and T lymphocytes
| 1,972 |
1,627 |
Regulation and Maintenance of an Adoptive T-Cell Dependent Memory B Cell Pool
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5120830/
SHA: f4a82ad66962355ffb09e4d1b57fde3e94f0ec53
Authors: Anson, Marie; Amado, Inês; Mailhé, Marie-Pierre; Donnadieu, Emmanuel; Garcia, Sylvie; Huetz, François; Freitas, Antonio A.
Date: 2016-11-23
DOI: 10.1371/journal.pone.0167003
License: cc-by
Abstract: We investigated the ability of monoclonal B cells to restore primary and secondary T-cell dependent antibody responses in adoptive immune-deficient hosts. Priming induced B cell activation and expansion, AID expression, antibody production and the generation of IgM(+)IgG(-) and IgM(-)IgG(+) antigen-experienced B-cell subsets that persisted in the lymphopenic environment by cell division. Upon secondary transfer and recall the IgM(-)IgG(+) cells responded by the production of antigen-specific IgG while the IgM(+) memory cells secreted mainly IgM and little IgG, but generated new B cells expressing germinal center markers. The recall responses were more efficient if the antigenic boost was delayed suggesting that a period of adaptation is necessary before the transferred cells are able to respond. Overall these findings indicate that reconstitution of a functional and complete memory pool requires transfer of all different antigen-experienced B cell subsets. We also found that the size of the memory B cell pool did not rely on the number of the responding naïve B cells, suggesting autonomous homeostatic controls for naïve and memory B cells. By reconstituting a stable memory B cell pool in immune-deficient hosts using a monoclonal high-affinity B cell population we demonstrate the potential value of B cell adoptive immunotherapy.
Text: Immune responses to infectious agents have different out-comes that can either protect or fail to control disease. Protection from re-infection relies on the establishment of efficient secondary immune responses that require the generation of antigen-specific "memory" B and T lymphocytes. The generation and selection of T-cell dependent "memory" B cells involves distinct molecular mechanisms: immunoglobulin isotype recombination and somatic hyper mutation, both dependent on the expression of AID [1] . Therefore, a long-standing paradigm defined memory B cells as IgM -IgG + isotype switched cells [2] . Different lines of evidence indicate that this is not always the case. In humans, it has been shown that some IgM + B cells bear the phenotype of other memory cells, being CD27 + , and carry frequent point mutations in the V region of the Ig genes, suggesting that they must represent highly selected B cell populations [3] . In mice, populations of CD19 + IgM + able to mount secondary responses have been identified [4] [5] [6] [7] . Overall these findings suggest that the T-cell dependent memory B cell pool comprises distinct subsets of memory B cells with different properties and effector functions [4] [5] [6] .
The biological properties that ensure the long-term persistence of memory and efficient secondary antibody responses have not been yet completely established. While initial studies proposed that after transfer memory B cells faded rapidly [8, 9] suggesting that long-lasting memory required the continuous recruitment of new cells [8] and/or antigen persistence [9, 10] , others suggested that memory B cells were able of extended survival without cell division [11] in the absence of antigen [2] . Long-term persistence of antibody responses has also been attributed to populations of long-lived plasma cells mainly resident in the bone marrow following immunization [12, 13] . The demonstration of the compartmentalization of "antibody memory" into different cellular layers suggested that the separate subsets of memory B cells behave differently. Accordingly, it has been reported that IgG + cells that could rapidly respond upon challenge did not persist long, while IgM + cells could generate a second wave of germinal center responses allowing persistence of memory [4] [5] [6] 14] .
Currently, immunotherapy approaches using passive antibody transfer [15, 16] ) is limited by the short half-life of immunoglobulin. Therefore new therapy strategies may require the adoptive transfer of high-affinity memory B cells, ready to respond and able to persist. The development of these new strategies requires a profound understanding of the mechanisms that regulate memory B cell numbers and ensure long persistence upon adoptive transfer. Moreover, knowledge of the mechanisms that determine the size of the memory B cell pool may be also critical to device new reconstitution strategies. So far, studies comparing populations of naïve and memory B cells have been hindered both by the vast clonal heterogeneity of the cells involved and by our inability to generate significant numbers of antigen specific memory B cells. Indeed in a normal laboratory mouse the population of B cells bearing a "memory IgG + phenotype' represent a small fraction of the total B cell pool (<0.5%) and upon immunization the number of the clonal diverse antigen-specific memory B cells generated is generally very limited (<10 3 ) [1, 6] .
To circumvent these limits, we decided to compare the properties of homogeneous populations of naïve and memory B cells of known antigen specificity, belonging to the same clone. We used SW HEL transgenic mice where B cells bear a high-affinity BCR specific for HEL and are capable of class switch recombination and somatic hypermutation (SHM) [17, 18] . To identify "memory B cells" the SW HEL mice were crossed with mice where AID transcription provokes the permanent expression of an YFP reporter in post-germinal center lymphocytes [19] . These mice were in a Rag2-deficient background and therefore contain a pure population of monoclonal HEL-specific B cells. To generate memory cells, purified naïve B cells from the SW HEL .AID/YFP.Rag2 -/mice were transferred into adoptive hosts together with monoclonal OVA-specific CD4 + T cells from OTII.Rag2 -/-TCR transgenic mice. Upon immunization with OVA-HEL complexes, we obtained a significant number of persisting HEL-specific IgM + Ig-G -YFP + and IgM -IgG + YFP + memory B cells, number that did not correlate to the number of precursor naïve cells initially injected suggesting that the memory B cell pool is regulated independently. We characterized the functional capacity of these two memory cell types in immune deficient hosts.
Mice B6 and B6.Rag2-/- [20] mice were kept at the Centre Des Techniques Avancées (CDTA), Centre National de la Recherche Scientifique (CNRS), Orleans, France; SWHEL.AID/YFP.Rag-/-mice, obtained by crossing SWHEL (18)(a gift of Dr. Robert Brink) and AID/YFP [19] (a gift of Dr. Rafael Casellas) with B6.Rag2-/-mice. OTII.Rag-/-mice were kept in our animal facilities at the Pasteur Institute. Experiments were preformed according to Pasteur Institute Safety Committee in accordance with French and European guidelines and the ethics Committee of Paris 1 (permits 2010-0002, -0003 and -0004). Euthanasia of the mice was performed by cervical dislocation. This specific study was approved by the European Research Council (ERC) committee related to the grant AdG09 249740-QSIS. The general status of the mice was controlled daily by monitoring the appearence of obvious pain, distress or suffering (prostration, respiratory issues, loss of weight). The end-point of the experiment was determined by a loss of more than 20% of the weight or as soon as the distress signs appeared. In this case, experiment was stopped and the animals were euthanized.
Single-cell suspensions of B cells from spleens and lymph nodes of SW HEL .AID/YFP.Rag -/mice together with CD4 + T cells from spleens and lymph nodes of OTII.Rag -/mice were transferred intravenously into the retro-orbital sinus of B6.Ly5 a IgH a or B6.Rag2 -/recipient mice. Mice received 10 6 HEL + B cells and 10 6 CD4 + T cells unless stated otherwise. Mice were immunized 24H later with 1 mg of Ovalbumin coupled to Hen Egg Lysozyme (OVA-HEL) in 50μg of Alu-S-Gel (Serva) we determined as the optimal dose of Ag (data not shown).
Naive cells from SW HEL .AID/YFP.Rag -/mice and memory B cells subsets from immunized B6.Rag -/hosts mice were purified from spleens and lymph nodes by flow cytometry sorting. Single-cell suspensions containing 5×10 4 B cells and 10 6 T cells were transferred intravenously into B6.Rag2 -/recipient hosts. The purity of sorted cells was above 98%. 24 h after transfer, mice were immunized with 1 mg of OVA-HEL.
Spleen, bone marrow, inguinal and mesenteric lymph nodes single-cell suspensions were stained for cell surface or intracellular proteins with appropriate combinations of the following monoclonal antibodies conjugated to pacific blue, Qdot-655, Brillant Violet 605, allophycocyanin, peridinin chlorophyll protein-cyanine 5.5, phycoerythrin, phycoerythrin-cyanine7: anti-CD19 (6D5), anti-IgM (R6-60.2), anti-IgG1 (X56), anti-CD138 (281-2), anti-Gl7 (Gl7), anti-CD95 (Jo2), anti-CD62L (MEL-14), anti-CD69 (H1-2F3), anti-BAFFR (7H22-E16), anti-CXCR5 (L138D7), anti-IA b (AF6-120.1), anti-CD80 (16-10A1), anti-CD73 (TY-11-8) and anti-PDL2 (TY25) and anti-Ki-67 (mm1) purchased from Becton Dickinson Pharmingen, Biolegend, Invitrogen and eBioscience. Cells were also stained with HEL (Sigma) coupled with AF594 using Alexa Fluor1 594 Protein Labeling Kit from Life technologies. Before staining, cells were treated with Fc-Block (CD16/CD32, Becton Dickinson Pharmingen). Dead cells were excluded during analysis according to their light-scattering characteristics. For intracellular stainings, cells were first stained with antibodies specific for cell surface antigens. Then, cells were fixed and permeabilized according the manufacturer's recommendations (BD Bisciences). For proliferation assay, mice were injected i.p. with 50 mg/kg of BrdU (Sigma-Aldrich) and were killed 24 or 72 hours later. Incorporated BrdU was detected intracellularly using anti-BrdU APC-conjugated antibodies according to the manufacturer's recommendations (BD Biosciences). All data acquisitions and analyses were performed with LSRFortessa (Becton Dickinson) interfaced with BD FACSDiva (Becton Dickinson) and FlowJo (Tree Star) software. Subsets of memory B cells were sorted as CD19 + HEL + YFP + IgM + or IgG + and naive cells as CD19 + HEL + YFP -IgM + using a FACSAriaIII flow cytometer. The purity of the sorted populations varied from 90-95%.
Sera HEL-specific Ig concentrations were quantified by ELISA. Plates were coated with HEL and saturated with PBS-5% Milk. Dilutions of sera were added. After incubation (2 hours, 37˚C) and washing, HRP-labeled anti-mouse IgM or IgG antibodies were added. After incubation and washing, bound antibodies were revealed with the substrate O-phenylenediamine and H2O2. The reaction was stopped after 10 min. by addition of 10% SDS and the absorbance read at 492nm in a multiscan spectrometer. Ig concentrations were determined by comparing the displacement of the dilution curves in the linear interval between standards at a concentration of 1 mg/ml and the serum samples.
The quantification of IgG or IgM secreting cells was assayed by ELISpot technique. Briefly, plates were coated with HEL. After saturating, the cells were distributed into the micro wells in RPMI1640-2%FCS. The plates were incubated for 12 h at 37˚C, 5% CO2 atmosphere. After extensive wash, plates were incubated with goat anti-mouse IgM or anti-IgG labeled with alkaline phosphatase. After washing, the revealing substrate was added (2,3 mM 5-bromo-4-chloro-3-indolyl phosphate diluted in 2-amino-2-methyl-1-proprenolol buffer).
Spleens from 14 day-immunized mice were initially fixed with paraformaldehyde and embedded in 4% low-gelling-temperature agarose (type VII-A; Sigma-Aldrich) prepared in PBS. 150μm slices were cut with a vibratome (VT 1000S; Leica) in a bath of ice-cold PBS. For immunolabeling, samples were saturated with PBS supplemented with 10% of fetal calf serum, then were labeled with primary antibodies anti-B220-APC (clone RA3-6B2) and anti-IgD-PE (clone 11-26c.2a) and analyzed with a spinning disk confocal microscope equipped with a CoolSnap HQ2 camera (Photometrics) and a 20x objective. Images were acquired and analyzed with MetaMorph 7 imaging software Molecular Devices).
Sample means were compared using the Student's t test. Sample means were considered significantly different at p < 0.05.
During an immune response the complexity of determinants expressed by immunizing antigen and the degeneracy of antigen-specific recognition results in a vast heterogeneity of responding cells rendering impossible the direct comparison of the properties of naïve and memory B cells belonging to the same clone. We have devised an experimental system that permits the comparison between naïve and memory B cells expressing the same antigen receptor and allows marking permanently memory B cells. For that purpose we used SW HEL transgenic mice in a Rag2-deficient background holding a single population of monoclonal B cells, all bearing a high-affinity BCR specific for HEL and capable of class switch recombination and somatic hypermutation (SHM) [17, 18] . To identify antigen-experienced B cells the SW HEL . Rag2 -/mice were crossed with mice where AID transcription induces the permanent expression of an YFP reporter in post-germinal center lymphocytes [19] . Since in intact Tg mice immune responses were not traceable, probably because of the presence of low level pre-existing anti-HEL antibodies that neutralize the immunizing protein, we used an adoptive cell transfer strategy to study the ability of the high affinity monoclonal B cell to reconstitute response in immune-deficient hosts and generate antibody memory. Purified naïve B cells from the SW HEL .AID/YFP.Rag2 -/mice were transferred into Rag2-deficient mice together with monoclonal OVA-specific CD4 + T helper cells from OTII.Rag2 -/-TCR transgenic mice. The day after, host mice were immunized with OVA-HEL complexes (Fig 1A) . In these conditions, antigenic challenge resulted in B cell activation and the development of significant numbers of CD19 + HEL + AID/YFP + B cells, which were not detected in non-immunized mice or in mice immunized in absence of helper T cells (Fig 1B) . We followed the early kinetics of this response. The number of HEL-specific B cells increased from the initial 2x10 6 transferred to about 15x10 6 at day 14 (Fig 1C left) the B cells expressing AID/YFP being the dominant population (Fig 1C right) . A fraction of the HEL-specific B cells underwent class switch recombination and at day 14 we recovered both IgM + IgG -AID/YFP + and IgM -IgG + AID/YFP + cell populations (Fig 1B) . B cell expansion and phenotypic changes were accompanied by the production of IgM and IgG HEL-specific antibodies ( Fig 1D) . Two weeks after antigenic challenge we observed the formation of germinal centers in the spleen of the host mice ( Fig 1E) . Coherently we found that while upon adoptive transfer all B cells expressed CD95, only after antigenic challenge most YFP + B cells expressed the germinal center specific marker GL7 (Fig 1F) .
In conclusion, the adoptive cell transfer strategy allowed the development of a primary immune response with B cell activation and expansion, induction of AID expression, class switch recombination, antigen-specific IgM and IgG antibody production and germinal center formation.
We studied the evolution of the B cell response. From two weeks onwards the total number of B cells contracted and at four weeks we recovered about 2-4x10 6 cells, number that remained stable up to week 20 (Fig 2A) . High titers of HEL-specific IgG were kept from week 3 to 8, declined thereafter, but were still significantly elevated 20 weeks later (Fig 2B) . A population of cells secreting HEL-specific Igs was present in the spleen (Fig 2C) , but not in the BM (not shown) even at the late time points. About 60% of the recovered cells exhibited the phenotype of antigen-experienced ("memory") CD19 + HEL + AID/YFP + expressing either IgM or IgG ( Fig 2D and 2E) . We compared the phenotype of the two AID/YFP + IgM + and AID/YFP + IgM -IgG + memory cell populations recovered with that of the naïve B cells (Fig 2F) . We found that antigen-experience and naïve B cells expressed similar levels of CD62L, CD69 and BAFFR (not shown). Antigen-experienced cells presented sustained expression of CD95 and increased levels of PNA, but the vast majority lost expression of the germinal center marker GL7 present at earlier times post-immunization ( Fig 2F compare to Fig 1F) . Compared to naïve B cells, AID/ YFP + cells expressed higher levels of CD80 and MHC class II and down-regulated expression of CXCR5 (Fig 2F) . These findings indicate that the post-germinal center AID/YFP + B cells express an activated phenotype [5, 21] , have increased antigen-presenting capacity [22] , but may loose the ability to re-enter primary follicles being CXCR5 low [23] . We have also compared the patterns of gene expression (RNAseq) by naïve, activated (YFPcells of immunized mice) and both populations of YFP + memory cells. The data shows a clear discrimination of naïve and activated/memory cells while indicating only minor differences between both subsets of YFP + memory cells (Fig 3) .
mRNA was isolated from sort-purified Naïve (CD19 + HEL + YFP -IgM + ) IgM + IgGor IgM -IgG + HEL + CD19 + YFP + memory B cells from spleen of different recipient mice. Total recommended by the manufacturer. The validated libraries were then subjected to DNA sequencing. The analysis is performed using the R software, Bioconductor packages including DESeq2 and the PF2tools package (version 1.2.9) developed at PF2 (Institut Pasteur). Normalization and differential analysis are carried out according to the DESeq2 model and package (version 1.8.1). Fig 3A shows a representative heat map of the different cells populations. Fig 3B shows
Late in the immune response persistent B cell numbers were kept by active cell division as a significant fraction of the cells were Ki67 + (Fig 2G left) and incorporated BrdU (Fig 2G middle) . The frequency of BrdU + cells was higher among the AID/YFPcells (15%) than in the major AID/YFP + memory population (3%) and similar between the IgM + and IgM -AID/ YFP + populations ( Fig 2G middle and not shown) . Three days after BrdU pulse populations were clear of BrdU + cells (Fig 2G right) attesting their high division rate. In spite of their increased proliferation rate, memory cells numbers were stable indicating that proliferation may be compensated by cell death as suggested by the frequency of caspase3 + cells (Fig 2H) . The frequency of Caspase3 + cells was higher among the AID/YFP + cells suggesting that a fraction of these cells may represent cells undergoing terminal differentiation. Importantly, these findings demonstrate that the transfer strategy allowed the generation of significant numbers of persisting antigen-experienced YFP + cells.
It is not yet known whether the number of antigen-experienced memory B cells correlated to the number of naïve B cells or if it is controlled independently of the initial number of antigenspecific B cells present. To approach this question we transferred different numbers of mature naïve B cells from SW HEL .AID/YFP.Rag2 -/donors (ranging from 10 5 to 5.10 6 ) into Rag2-deficient mice together with an excess of CD4 + T helper cells (10 6 ) and immunize the hosts the day after cell transfer with OVA-HEL in optimal non-limiting quantities. To directly compare the results obtained after the transfer of different all numbers we allowed the responses to reach steady-state eight weeks after antigenic challenge. We studied the amplitude of the immune response by measuring the serum titers of HEL-specific IgG antibodies and enumerating the number of HEL-specific B cells recovered. We found that in the presence of excess T cell help, the levels of the HEL-specific IgGs (Fig 4C) , and both the total number of HEL-specific ( Fig 4A) and of memory YFP + B cells recovered (Fig 4B) , did not correlate to the number of antigen specific naïve B cells initially transferred.
Memory B cells are defined functionally by their ability to induce secondary IgG antibody responses upon secondary antigenic challenge. We investigated whether the subsets of AID/ YFP + IgM + and AID/YFP + IgM -IgG + antigen-experienced (memory) B cells persisting at late time points could mount secondary IgG responses and persist after secondary transfer. For this purpose we followed two different experimental strategies. In the first, 5x10 4 cells of either IgM + or IgM -IgG + memory B cells, were transferred with an excess helper OTII CD4 + T cells into secondary Rag-deficient hosts that were boosted with OVA-HEL the day after cell transfer. In the absence of immunization antibody levels were undetectable (not shown) and three weeks after transfer recovery of both memory B cell subsets was about 10-20% of the initial cell input, exceeding naïve B cell recovery (Fig 5A) , supporting the notion that memory B cells may not require specific ligand recognition to survive (2). One cannot exclude, however, that cross-reactivity of the BCR transgene with environmental antigens may allow signaling sufficient to maintain naïve and memory cell survival in the absence of HEL [24] . Following immunization, the secondarily transferred AID/YFP + IgM -IgG + cells responded promptly with the exclusive production of significant levels HEL-specific IgG thus confirming their memory statute (11) . The AID/YFP + IgM + B cells in response to antigenic boost produced only limited amounts of IgM antibodies (Fig 4B) , little IgG antibodies, but did generate GL7 + B cells more efficiently than the IgG + memory B cell population (Fig 5D) . Thus the IgM + subset may contain precursors able to generate a secondary germinal center reaction and a new progeny of IgG + effectors (4). With time antibody levels decayed rapidly suggesting that the number of transferred memory B cells declined in the secondary hosts after antigenic boost. Indeed, IgM + and IgG + memory B cells failed to expand and 3 weeks after immunization cell recovery was similar to the retrieval observed in the non-immunized hosts (compare Fig 5E and 5A) . In similar experimental conditions, naïve B cells following immunization expanded, acquired AID/ YFP expression and their numbers more than doubled the number initially injected (Figs 5F and 2A). These data suggest that a significant fraction of the memory B cells generated have a reduced expansion capacity being programmed for rapid differentiation for effector functions.
Besides long-term survival memory B cells must maintain functional activity in the absence of nominal antigen to be fully effective. To test this we used an alternative approach where memory cells were parked in secondary Rag-deficient hosts for 30 days before re-immunization. We found that under these conditions antigenic challenge resulted in the production of HEL-specific IgG antibodies and in a 100 fold increase in the number of cells recovered, expansion that largely exceed that observed after immediate challenge (Fig 5G) .
The aim of this study was to characterize the fate of activated B cells and the generation of memory B cells. To do this, we adoptively transferred monoclonal B cells into immune deficient hosts followed by immunization in presence of T cell help. This strategy resulted in the development of different B cell memory subsets, namely IgM + and IgG + , as described for in situ generated memory cells [4, 6, 14] . These findings indicate that distinct memory B cell subsets are not the result of the heterogeneity of initially responding naive cells, but originate from the differentiation of a single B cell clone.
While studying the respective rate of proliferation of both types of memory B cells, we found the same high rate of proliferation for IgM + and IgG + memory B cells. These results contrast with previous published data. First it was been reported that "in situ" memory B cells persist as resting non-dividing cells [11, 25] . However, we have shown that upon adoptive transfer and in absence of competing cells, B cells increase their division rate to occupy the available empty niche [26] , which may explain the higher division rate observed here using this adoptive cell transfer strategy. Secondly, comparing life spans among heterogeneous memory B cell populations it was previously reported a lower division rate among the IgM + subset compared to the IgG + polyclonal subset [6] . Differences in BCR affinity between IgM + and IgG + memory clones may explain the higher division rate previously observed among the IgG + cells [6] . In contrast we compared memory B cell subsets belonging to the same clone bearing the same high affinity BCR. Overall these observations support the notion that lymphocyte division rates and life spans are not an intrinsic cell property, but rather determined by the environment and the presence of competing populations [27] . They demonstrate that upon the correct conditions memory B cells can persist by cell division.
An important question was whether the number of memory B cells depends on the number of initial naïve B cells. We found that, in the presence of an excess of T cell help, that was not the case. However, it was previously reported during polyclonal responses that serum titers of anti-HSA was proportional to the number of cells transferred into irradiated mice [28] . It is possible that limited antigen-specific T-B cell encounters may constraint the number of responding B cells and thus determine linear precursor-progeny between naïve and memory B cells. Our findings indicate that within a single clone the number of precursor naive B cells present in the peripheral B cell pool does determine neither the intensity nor the final number of memory B cells in response to an optimal dose of antigen. They suggest that the size of memory B cell pool may be controlled independently of the number of naïve B cell precursors and that in the absence of clonal competition the memory niche can be filled with a single monoclonal population. Considering diverse polyclonal populations, the limited niche for memory cells will imply strong competition among clones resulting in the selection of best fit (high affinity) cells: rare mutated clones being able to out compete more frequent but less avid clones. In our settings, the transgenic memory B cells are likely to counter select any new mutant clones since they express a very high affinity BCR selected in the course of a secondary immune response [29] . Thus, notwithstanding the expression of AID and proliferation we did not detect any BCR VH and VL Ig-chain nucleotide mutations among the recovered memory B cells (not shown). These findings may have implication for vaccination protocols as they indicate that each new antigenic exposure or unrelated immunization would add extra competing clones supporting the need for repeated antigenic boosts to prevent memory B cell attrition. They also demonstrate that the memory B cell pool can be reconstituted from a relatively small number of antigen-specific cells.
It is likely that the relatively poor memory B cell expansion observed after immediate boost after adoptive transfer could be due to the lack in Rag-deficient hosts of the appropriate environment required for memory B cell survival and function. It should be pointed out that B cell transfer into transgenic ML5 Rag-deficient hosts expressing low levels of HEL [29] resulted in rapid cell loss and recovery suggesting that in these hosts, B cells are trapped by antigen in locations were they are unable to survive (not shown). Nevertheless, it has been shown that B cells can drive the maturation of follicular dendritic cells and the organization of lymphoid follicles [30] . Similarly, transferred helper cells may also modify their immediate environment. Thus, by allowing lymphocytes to adapt and modify their immediate environment we improved their response and more important, we recovered the memory B cell pool size present in the original donor mice.
In this study we show that it is possible to fully reconstitute a primary response and the establishment of antibody memory in immune deficient mice after adoptive transfer of antigen-specific monoclonal B cells together with a population of monoclonal helper T cells. Indeed, it is generally believed that in immune deficiencies, B cell therapy has restricted application due to intrinsic defects of host's lymphoid organs structure that may prevent development of immune responses, germinal center formation, establishment of antibody memory and limit cell survival. In contrast we showed that after adoptive transfer in immune deficient hosts antigen immunization induced B cell activation and expansion, induction of AID expression, class switch recombination, antigen-specific IgM and IgG antibody production, germinal center formation and the generation of two subsets of AID/YFP + IgM + IgGand AID/YFP + Ig-M -IgG + antigen-experienced B cell subsets able to persist in a lymphopenic environment by cell division mimicking responses obtained in intact non-Tg mice [4] . Upon challenge the AID/YFP + IgM -IgG + cells responded promptly with the production of HEL-specific IgG while the AID/YFP + IgM + B cells secreted only limited amounts of IgM antibodies and fail to produce IgG. In contrast the AID/YFP + IgM + B cells could give rise to new GL7 + B cells, suggesting that full reconstitution of the memory B cell pool may require transfer of the different antigen-experienced B cell subsets. Importantly, we found that the recall responses were more efficient if the transferred memory cells were given the required time to adapt to their new environment, suggesting that a period of accommodation is necessary before the transferred cells are fully capable to respond. Our findings also show that different processes can modify the survival conditions of memory B cells. Finally, we found that the generation of the memory B cell pool in response to an optimal dose of Ag did not rely on the number of the initially responding B cells, suggesting autonomous homeostatic controls for naïve and memory B cells a property that may allow reconstitution of the memory pool in immune-deficient hosts using a limited number of precursor naïve B cells. An autonomous control of the memory B cell pool where each antigenic exposure adds new competing clones supports the notion of vaccination strategies using antigenic boosting to prevent memory B cell attrition. Overall the findings reported demonstrate that it is possible to reconstitute the memory B cell pool of an immune deficient host with an artificially induced population of monoclonal high affinity memory B cells.
|
What do immunoglobulin isotype recombination and somatic hyper mutation depend on?
| 4,256 |
expression of AID
| 2,230 |
1,627 |
Regulation and Maintenance of an Adoptive T-Cell Dependent Memory B Cell Pool
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5120830/
SHA: f4a82ad66962355ffb09e4d1b57fde3e94f0ec53
Authors: Anson, Marie; Amado, Inês; Mailhé, Marie-Pierre; Donnadieu, Emmanuel; Garcia, Sylvie; Huetz, François; Freitas, Antonio A.
Date: 2016-11-23
DOI: 10.1371/journal.pone.0167003
License: cc-by
Abstract: We investigated the ability of monoclonal B cells to restore primary and secondary T-cell dependent antibody responses in adoptive immune-deficient hosts. Priming induced B cell activation and expansion, AID expression, antibody production and the generation of IgM(+)IgG(-) and IgM(-)IgG(+) antigen-experienced B-cell subsets that persisted in the lymphopenic environment by cell division. Upon secondary transfer and recall the IgM(-)IgG(+) cells responded by the production of antigen-specific IgG while the IgM(+) memory cells secreted mainly IgM and little IgG, but generated new B cells expressing germinal center markers. The recall responses were more efficient if the antigenic boost was delayed suggesting that a period of adaptation is necessary before the transferred cells are able to respond. Overall these findings indicate that reconstitution of a functional and complete memory pool requires transfer of all different antigen-experienced B cell subsets. We also found that the size of the memory B cell pool did not rely on the number of the responding naïve B cells, suggesting autonomous homeostatic controls for naïve and memory B cells. By reconstituting a stable memory B cell pool in immune-deficient hosts using a monoclonal high-affinity B cell population we demonstrate the potential value of B cell adoptive immunotherapy.
Text: Immune responses to infectious agents have different out-comes that can either protect or fail to control disease. Protection from re-infection relies on the establishment of efficient secondary immune responses that require the generation of antigen-specific "memory" B and T lymphocytes. The generation and selection of T-cell dependent "memory" B cells involves distinct molecular mechanisms: immunoglobulin isotype recombination and somatic hyper mutation, both dependent on the expression of AID [1] . Therefore, a long-standing paradigm defined memory B cells as IgM -IgG + isotype switched cells [2] . Different lines of evidence indicate that this is not always the case. In humans, it has been shown that some IgM + B cells bear the phenotype of other memory cells, being CD27 + , and carry frequent point mutations in the V region of the Ig genes, suggesting that they must represent highly selected B cell populations [3] . In mice, populations of CD19 + IgM + able to mount secondary responses have been identified [4] [5] [6] [7] . Overall these findings suggest that the T-cell dependent memory B cell pool comprises distinct subsets of memory B cells with different properties and effector functions [4] [5] [6] .
The biological properties that ensure the long-term persistence of memory and efficient secondary antibody responses have not been yet completely established. While initial studies proposed that after transfer memory B cells faded rapidly [8, 9] suggesting that long-lasting memory required the continuous recruitment of new cells [8] and/or antigen persistence [9, 10] , others suggested that memory B cells were able of extended survival without cell division [11] in the absence of antigen [2] . Long-term persistence of antibody responses has also been attributed to populations of long-lived plasma cells mainly resident in the bone marrow following immunization [12, 13] . The demonstration of the compartmentalization of "antibody memory" into different cellular layers suggested that the separate subsets of memory B cells behave differently. Accordingly, it has been reported that IgG + cells that could rapidly respond upon challenge did not persist long, while IgM + cells could generate a second wave of germinal center responses allowing persistence of memory [4] [5] [6] 14] .
Currently, immunotherapy approaches using passive antibody transfer [15, 16] ) is limited by the short half-life of immunoglobulin. Therefore new therapy strategies may require the adoptive transfer of high-affinity memory B cells, ready to respond and able to persist. The development of these new strategies requires a profound understanding of the mechanisms that regulate memory B cell numbers and ensure long persistence upon adoptive transfer. Moreover, knowledge of the mechanisms that determine the size of the memory B cell pool may be also critical to device new reconstitution strategies. So far, studies comparing populations of naïve and memory B cells have been hindered both by the vast clonal heterogeneity of the cells involved and by our inability to generate significant numbers of antigen specific memory B cells. Indeed in a normal laboratory mouse the population of B cells bearing a "memory IgG + phenotype' represent a small fraction of the total B cell pool (<0.5%) and upon immunization the number of the clonal diverse antigen-specific memory B cells generated is generally very limited (<10 3 ) [1, 6] .
To circumvent these limits, we decided to compare the properties of homogeneous populations of naïve and memory B cells of known antigen specificity, belonging to the same clone. We used SW HEL transgenic mice where B cells bear a high-affinity BCR specific for HEL and are capable of class switch recombination and somatic hypermutation (SHM) [17, 18] . To identify "memory B cells" the SW HEL mice were crossed with mice where AID transcription provokes the permanent expression of an YFP reporter in post-germinal center lymphocytes [19] . These mice were in a Rag2-deficient background and therefore contain a pure population of monoclonal HEL-specific B cells. To generate memory cells, purified naïve B cells from the SW HEL .AID/YFP.Rag2 -/mice were transferred into adoptive hosts together with monoclonal OVA-specific CD4 + T cells from OTII.Rag2 -/-TCR transgenic mice. Upon immunization with OVA-HEL complexes, we obtained a significant number of persisting HEL-specific IgM + Ig-G -YFP + and IgM -IgG + YFP + memory B cells, number that did not correlate to the number of precursor naïve cells initially injected suggesting that the memory B cell pool is regulated independently. We characterized the functional capacity of these two memory cell types in immune deficient hosts.
Mice B6 and B6.Rag2-/- [20] mice were kept at the Centre Des Techniques Avancées (CDTA), Centre National de la Recherche Scientifique (CNRS), Orleans, France; SWHEL.AID/YFP.Rag-/-mice, obtained by crossing SWHEL (18)(a gift of Dr. Robert Brink) and AID/YFP [19] (a gift of Dr. Rafael Casellas) with B6.Rag2-/-mice. OTII.Rag-/-mice were kept in our animal facilities at the Pasteur Institute. Experiments were preformed according to Pasteur Institute Safety Committee in accordance with French and European guidelines and the ethics Committee of Paris 1 (permits 2010-0002, -0003 and -0004). Euthanasia of the mice was performed by cervical dislocation. This specific study was approved by the European Research Council (ERC) committee related to the grant AdG09 249740-QSIS. The general status of the mice was controlled daily by monitoring the appearence of obvious pain, distress or suffering (prostration, respiratory issues, loss of weight). The end-point of the experiment was determined by a loss of more than 20% of the weight or as soon as the distress signs appeared. In this case, experiment was stopped and the animals were euthanized.
Single-cell suspensions of B cells from spleens and lymph nodes of SW HEL .AID/YFP.Rag -/mice together with CD4 + T cells from spleens and lymph nodes of OTII.Rag -/mice were transferred intravenously into the retro-orbital sinus of B6.Ly5 a IgH a or B6.Rag2 -/recipient mice. Mice received 10 6 HEL + B cells and 10 6 CD4 + T cells unless stated otherwise. Mice were immunized 24H later with 1 mg of Ovalbumin coupled to Hen Egg Lysozyme (OVA-HEL) in 50μg of Alu-S-Gel (Serva) we determined as the optimal dose of Ag (data not shown).
Naive cells from SW HEL .AID/YFP.Rag -/mice and memory B cells subsets from immunized B6.Rag -/hosts mice were purified from spleens and lymph nodes by flow cytometry sorting. Single-cell suspensions containing 5×10 4 B cells and 10 6 T cells were transferred intravenously into B6.Rag2 -/recipient hosts. The purity of sorted cells was above 98%. 24 h after transfer, mice were immunized with 1 mg of OVA-HEL.
Spleen, bone marrow, inguinal and mesenteric lymph nodes single-cell suspensions were stained for cell surface or intracellular proteins with appropriate combinations of the following monoclonal antibodies conjugated to pacific blue, Qdot-655, Brillant Violet 605, allophycocyanin, peridinin chlorophyll protein-cyanine 5.5, phycoerythrin, phycoerythrin-cyanine7: anti-CD19 (6D5), anti-IgM (R6-60.2), anti-IgG1 (X56), anti-CD138 (281-2), anti-Gl7 (Gl7), anti-CD95 (Jo2), anti-CD62L (MEL-14), anti-CD69 (H1-2F3), anti-BAFFR (7H22-E16), anti-CXCR5 (L138D7), anti-IA b (AF6-120.1), anti-CD80 (16-10A1), anti-CD73 (TY-11-8) and anti-PDL2 (TY25) and anti-Ki-67 (mm1) purchased from Becton Dickinson Pharmingen, Biolegend, Invitrogen and eBioscience. Cells were also stained with HEL (Sigma) coupled with AF594 using Alexa Fluor1 594 Protein Labeling Kit from Life technologies. Before staining, cells were treated with Fc-Block (CD16/CD32, Becton Dickinson Pharmingen). Dead cells were excluded during analysis according to their light-scattering characteristics. For intracellular stainings, cells were first stained with antibodies specific for cell surface antigens. Then, cells were fixed and permeabilized according the manufacturer's recommendations (BD Bisciences). For proliferation assay, mice were injected i.p. with 50 mg/kg of BrdU (Sigma-Aldrich) and were killed 24 or 72 hours later. Incorporated BrdU was detected intracellularly using anti-BrdU APC-conjugated antibodies according to the manufacturer's recommendations (BD Biosciences). All data acquisitions and analyses were performed with LSRFortessa (Becton Dickinson) interfaced with BD FACSDiva (Becton Dickinson) and FlowJo (Tree Star) software. Subsets of memory B cells were sorted as CD19 + HEL + YFP + IgM + or IgG + and naive cells as CD19 + HEL + YFP -IgM + using a FACSAriaIII flow cytometer. The purity of the sorted populations varied from 90-95%.
Sera HEL-specific Ig concentrations were quantified by ELISA. Plates were coated with HEL and saturated with PBS-5% Milk. Dilutions of sera were added. After incubation (2 hours, 37˚C) and washing, HRP-labeled anti-mouse IgM or IgG antibodies were added. After incubation and washing, bound antibodies were revealed with the substrate O-phenylenediamine and H2O2. The reaction was stopped after 10 min. by addition of 10% SDS and the absorbance read at 492nm in a multiscan spectrometer. Ig concentrations were determined by comparing the displacement of the dilution curves in the linear interval between standards at a concentration of 1 mg/ml and the serum samples.
The quantification of IgG or IgM secreting cells was assayed by ELISpot technique. Briefly, plates were coated with HEL. After saturating, the cells were distributed into the micro wells in RPMI1640-2%FCS. The plates were incubated for 12 h at 37˚C, 5% CO2 atmosphere. After extensive wash, plates were incubated with goat anti-mouse IgM or anti-IgG labeled with alkaline phosphatase. After washing, the revealing substrate was added (2,3 mM 5-bromo-4-chloro-3-indolyl phosphate diluted in 2-amino-2-methyl-1-proprenolol buffer).
Spleens from 14 day-immunized mice were initially fixed with paraformaldehyde and embedded in 4% low-gelling-temperature agarose (type VII-A; Sigma-Aldrich) prepared in PBS. 150μm slices were cut with a vibratome (VT 1000S; Leica) in a bath of ice-cold PBS. For immunolabeling, samples were saturated with PBS supplemented with 10% of fetal calf serum, then were labeled with primary antibodies anti-B220-APC (clone RA3-6B2) and anti-IgD-PE (clone 11-26c.2a) and analyzed with a spinning disk confocal microscope equipped with a CoolSnap HQ2 camera (Photometrics) and a 20x objective. Images were acquired and analyzed with MetaMorph 7 imaging software Molecular Devices).
Sample means were compared using the Student's t test. Sample means were considered significantly different at p < 0.05.
During an immune response the complexity of determinants expressed by immunizing antigen and the degeneracy of antigen-specific recognition results in a vast heterogeneity of responding cells rendering impossible the direct comparison of the properties of naïve and memory B cells belonging to the same clone. We have devised an experimental system that permits the comparison between naïve and memory B cells expressing the same antigen receptor and allows marking permanently memory B cells. For that purpose we used SW HEL transgenic mice in a Rag2-deficient background holding a single population of monoclonal B cells, all bearing a high-affinity BCR specific for HEL and capable of class switch recombination and somatic hypermutation (SHM) [17, 18] . To identify antigen-experienced B cells the SW HEL . Rag2 -/mice were crossed with mice where AID transcription induces the permanent expression of an YFP reporter in post-germinal center lymphocytes [19] . Since in intact Tg mice immune responses were not traceable, probably because of the presence of low level pre-existing anti-HEL antibodies that neutralize the immunizing protein, we used an adoptive cell transfer strategy to study the ability of the high affinity monoclonal B cell to reconstitute response in immune-deficient hosts and generate antibody memory. Purified naïve B cells from the SW HEL .AID/YFP.Rag2 -/mice were transferred into Rag2-deficient mice together with monoclonal OVA-specific CD4 + T helper cells from OTII.Rag2 -/-TCR transgenic mice. The day after, host mice were immunized with OVA-HEL complexes (Fig 1A) . In these conditions, antigenic challenge resulted in B cell activation and the development of significant numbers of CD19 + HEL + AID/YFP + B cells, which were not detected in non-immunized mice or in mice immunized in absence of helper T cells (Fig 1B) . We followed the early kinetics of this response. The number of HEL-specific B cells increased from the initial 2x10 6 transferred to about 15x10 6 at day 14 (Fig 1C left) the B cells expressing AID/YFP being the dominant population (Fig 1C right) . A fraction of the HEL-specific B cells underwent class switch recombination and at day 14 we recovered both IgM + IgG -AID/YFP + and IgM -IgG + AID/YFP + cell populations (Fig 1B) . B cell expansion and phenotypic changes were accompanied by the production of IgM and IgG HEL-specific antibodies ( Fig 1D) . Two weeks after antigenic challenge we observed the formation of germinal centers in the spleen of the host mice ( Fig 1E) . Coherently we found that while upon adoptive transfer all B cells expressed CD95, only after antigenic challenge most YFP + B cells expressed the germinal center specific marker GL7 (Fig 1F) .
In conclusion, the adoptive cell transfer strategy allowed the development of a primary immune response with B cell activation and expansion, induction of AID expression, class switch recombination, antigen-specific IgM and IgG antibody production and germinal center formation.
We studied the evolution of the B cell response. From two weeks onwards the total number of B cells contracted and at four weeks we recovered about 2-4x10 6 cells, number that remained stable up to week 20 (Fig 2A) . High titers of HEL-specific IgG were kept from week 3 to 8, declined thereafter, but were still significantly elevated 20 weeks later (Fig 2B) . A population of cells secreting HEL-specific Igs was present in the spleen (Fig 2C) , but not in the BM (not shown) even at the late time points. About 60% of the recovered cells exhibited the phenotype of antigen-experienced ("memory") CD19 + HEL + AID/YFP + expressing either IgM or IgG ( Fig 2D and 2E) . We compared the phenotype of the two AID/YFP + IgM + and AID/YFP + IgM -IgG + memory cell populations recovered with that of the naïve B cells (Fig 2F) . We found that antigen-experience and naïve B cells expressed similar levels of CD62L, CD69 and BAFFR (not shown). Antigen-experienced cells presented sustained expression of CD95 and increased levels of PNA, but the vast majority lost expression of the germinal center marker GL7 present at earlier times post-immunization ( Fig 2F compare to Fig 1F) . Compared to naïve B cells, AID/ YFP + cells expressed higher levels of CD80 and MHC class II and down-regulated expression of CXCR5 (Fig 2F) . These findings indicate that the post-germinal center AID/YFP + B cells express an activated phenotype [5, 21] , have increased antigen-presenting capacity [22] , but may loose the ability to re-enter primary follicles being CXCR5 low [23] . We have also compared the patterns of gene expression (RNAseq) by naïve, activated (YFPcells of immunized mice) and both populations of YFP + memory cells. The data shows a clear discrimination of naïve and activated/memory cells while indicating only minor differences between both subsets of YFP + memory cells (Fig 3) .
mRNA was isolated from sort-purified Naïve (CD19 + HEL + YFP -IgM + ) IgM + IgGor IgM -IgG + HEL + CD19 + YFP + memory B cells from spleen of different recipient mice. Total recommended by the manufacturer. The validated libraries were then subjected to DNA sequencing. The analysis is performed using the R software, Bioconductor packages including DESeq2 and the PF2tools package (version 1.2.9) developed at PF2 (Institut Pasteur). Normalization and differential analysis are carried out according to the DESeq2 model and package (version 1.8.1). Fig 3A shows a representative heat map of the different cells populations. Fig 3B shows
Late in the immune response persistent B cell numbers were kept by active cell division as a significant fraction of the cells were Ki67 + (Fig 2G left) and incorporated BrdU (Fig 2G middle) . The frequency of BrdU + cells was higher among the AID/YFPcells (15%) than in the major AID/YFP + memory population (3%) and similar between the IgM + and IgM -AID/ YFP + populations ( Fig 2G middle and not shown) . Three days after BrdU pulse populations were clear of BrdU + cells (Fig 2G right) attesting their high division rate. In spite of their increased proliferation rate, memory cells numbers were stable indicating that proliferation may be compensated by cell death as suggested by the frequency of caspase3 + cells (Fig 2H) . The frequency of Caspase3 + cells was higher among the AID/YFP + cells suggesting that a fraction of these cells may represent cells undergoing terminal differentiation. Importantly, these findings demonstrate that the transfer strategy allowed the generation of significant numbers of persisting antigen-experienced YFP + cells.
It is not yet known whether the number of antigen-experienced memory B cells correlated to the number of naïve B cells or if it is controlled independently of the initial number of antigenspecific B cells present. To approach this question we transferred different numbers of mature naïve B cells from SW HEL .AID/YFP.Rag2 -/donors (ranging from 10 5 to 5.10 6 ) into Rag2-deficient mice together with an excess of CD4 + T helper cells (10 6 ) and immunize the hosts the day after cell transfer with OVA-HEL in optimal non-limiting quantities. To directly compare the results obtained after the transfer of different all numbers we allowed the responses to reach steady-state eight weeks after antigenic challenge. We studied the amplitude of the immune response by measuring the serum titers of HEL-specific IgG antibodies and enumerating the number of HEL-specific B cells recovered. We found that in the presence of excess T cell help, the levels of the HEL-specific IgGs (Fig 4C) , and both the total number of HEL-specific ( Fig 4A) and of memory YFP + B cells recovered (Fig 4B) , did not correlate to the number of antigen specific naïve B cells initially transferred.
Memory B cells are defined functionally by their ability to induce secondary IgG antibody responses upon secondary antigenic challenge. We investigated whether the subsets of AID/ YFP + IgM + and AID/YFP + IgM -IgG + antigen-experienced (memory) B cells persisting at late time points could mount secondary IgG responses and persist after secondary transfer. For this purpose we followed two different experimental strategies. In the first, 5x10 4 cells of either IgM + or IgM -IgG + memory B cells, were transferred with an excess helper OTII CD4 + T cells into secondary Rag-deficient hosts that were boosted with OVA-HEL the day after cell transfer. In the absence of immunization antibody levels were undetectable (not shown) and three weeks after transfer recovery of both memory B cell subsets was about 10-20% of the initial cell input, exceeding naïve B cell recovery (Fig 5A) , supporting the notion that memory B cells may not require specific ligand recognition to survive (2). One cannot exclude, however, that cross-reactivity of the BCR transgene with environmental antigens may allow signaling sufficient to maintain naïve and memory cell survival in the absence of HEL [24] . Following immunization, the secondarily transferred AID/YFP + IgM -IgG + cells responded promptly with the exclusive production of significant levels HEL-specific IgG thus confirming their memory statute (11) . The AID/YFP + IgM + B cells in response to antigenic boost produced only limited amounts of IgM antibodies (Fig 4B) , little IgG antibodies, but did generate GL7 + B cells more efficiently than the IgG + memory B cell population (Fig 5D) . Thus the IgM + subset may contain precursors able to generate a secondary germinal center reaction and a new progeny of IgG + effectors (4). With time antibody levels decayed rapidly suggesting that the number of transferred memory B cells declined in the secondary hosts after antigenic boost. Indeed, IgM + and IgG + memory B cells failed to expand and 3 weeks after immunization cell recovery was similar to the retrieval observed in the non-immunized hosts (compare Fig 5E and 5A) . In similar experimental conditions, naïve B cells following immunization expanded, acquired AID/ YFP expression and their numbers more than doubled the number initially injected (Figs 5F and 2A). These data suggest that a significant fraction of the memory B cells generated have a reduced expansion capacity being programmed for rapid differentiation for effector functions.
Besides long-term survival memory B cells must maintain functional activity in the absence of nominal antigen to be fully effective. To test this we used an alternative approach where memory cells were parked in secondary Rag-deficient hosts for 30 days before re-immunization. We found that under these conditions antigenic challenge resulted in the production of HEL-specific IgG antibodies and in a 100 fold increase in the number of cells recovered, expansion that largely exceed that observed after immediate challenge (Fig 5G) .
The aim of this study was to characterize the fate of activated B cells and the generation of memory B cells. To do this, we adoptively transferred monoclonal B cells into immune deficient hosts followed by immunization in presence of T cell help. This strategy resulted in the development of different B cell memory subsets, namely IgM + and IgG + , as described for in situ generated memory cells [4, 6, 14] . These findings indicate that distinct memory B cell subsets are not the result of the heterogeneity of initially responding naive cells, but originate from the differentiation of a single B cell clone.
While studying the respective rate of proliferation of both types of memory B cells, we found the same high rate of proliferation for IgM + and IgG + memory B cells. These results contrast with previous published data. First it was been reported that "in situ" memory B cells persist as resting non-dividing cells [11, 25] . However, we have shown that upon adoptive transfer and in absence of competing cells, B cells increase their division rate to occupy the available empty niche [26] , which may explain the higher division rate observed here using this adoptive cell transfer strategy. Secondly, comparing life spans among heterogeneous memory B cell populations it was previously reported a lower division rate among the IgM + subset compared to the IgG + polyclonal subset [6] . Differences in BCR affinity between IgM + and IgG + memory clones may explain the higher division rate previously observed among the IgG + cells [6] . In contrast we compared memory B cell subsets belonging to the same clone bearing the same high affinity BCR. Overall these observations support the notion that lymphocyte division rates and life spans are not an intrinsic cell property, but rather determined by the environment and the presence of competing populations [27] . They demonstrate that upon the correct conditions memory B cells can persist by cell division.
An important question was whether the number of memory B cells depends on the number of initial naïve B cells. We found that, in the presence of an excess of T cell help, that was not the case. However, it was previously reported during polyclonal responses that serum titers of anti-HSA was proportional to the number of cells transferred into irradiated mice [28] . It is possible that limited antigen-specific T-B cell encounters may constraint the number of responding B cells and thus determine linear precursor-progeny between naïve and memory B cells. Our findings indicate that within a single clone the number of precursor naive B cells present in the peripheral B cell pool does determine neither the intensity nor the final number of memory B cells in response to an optimal dose of antigen. They suggest that the size of memory B cell pool may be controlled independently of the number of naïve B cell precursors and that in the absence of clonal competition the memory niche can be filled with a single monoclonal population. Considering diverse polyclonal populations, the limited niche for memory cells will imply strong competition among clones resulting in the selection of best fit (high affinity) cells: rare mutated clones being able to out compete more frequent but less avid clones. In our settings, the transgenic memory B cells are likely to counter select any new mutant clones since they express a very high affinity BCR selected in the course of a secondary immune response [29] . Thus, notwithstanding the expression of AID and proliferation we did not detect any BCR VH and VL Ig-chain nucleotide mutations among the recovered memory B cells (not shown). These findings may have implication for vaccination protocols as they indicate that each new antigenic exposure or unrelated immunization would add extra competing clones supporting the need for repeated antigenic boosts to prevent memory B cell attrition. They also demonstrate that the memory B cell pool can be reconstituted from a relatively small number of antigen-specific cells.
It is likely that the relatively poor memory B cell expansion observed after immediate boost after adoptive transfer could be due to the lack in Rag-deficient hosts of the appropriate environment required for memory B cell survival and function. It should be pointed out that B cell transfer into transgenic ML5 Rag-deficient hosts expressing low levels of HEL [29] resulted in rapid cell loss and recovery suggesting that in these hosts, B cells are trapped by antigen in locations were they are unable to survive (not shown). Nevertheless, it has been shown that B cells can drive the maturation of follicular dendritic cells and the organization of lymphoid follicles [30] . Similarly, transferred helper cells may also modify their immediate environment. Thus, by allowing lymphocytes to adapt and modify their immediate environment we improved their response and more important, we recovered the memory B cell pool size present in the original donor mice.
In this study we show that it is possible to fully reconstitute a primary response and the establishment of antibody memory in immune deficient mice after adoptive transfer of antigen-specific monoclonal B cells together with a population of monoclonal helper T cells. Indeed, it is generally believed that in immune deficiencies, B cell therapy has restricted application due to intrinsic defects of host's lymphoid organs structure that may prevent development of immune responses, germinal center formation, establishment of antibody memory and limit cell survival. In contrast we showed that after adoptive transfer in immune deficient hosts antigen immunization induced B cell activation and expansion, induction of AID expression, class switch recombination, antigen-specific IgM and IgG antibody production, germinal center formation and the generation of two subsets of AID/YFP + IgM + IgGand AID/YFP + Ig-M -IgG + antigen-experienced B cell subsets able to persist in a lymphopenic environment by cell division mimicking responses obtained in intact non-Tg mice [4] . Upon challenge the AID/YFP + IgM -IgG + cells responded promptly with the production of HEL-specific IgG while the AID/YFP + IgM + B cells secreted only limited amounts of IgM antibodies and fail to produce IgG. In contrast the AID/YFP + IgM + B cells could give rise to new GL7 + B cells, suggesting that full reconstitution of the memory B cell pool may require transfer of the different antigen-experienced B cell subsets. Importantly, we found that the recall responses were more efficient if the transferred memory cells were given the required time to adapt to their new environment, suggesting that a period of accommodation is necessary before the transferred cells are fully capable to respond. Our findings also show that different processes can modify the survival conditions of memory B cells. Finally, we found that the generation of the memory B cell pool in response to an optimal dose of Ag did not rely on the number of the initially responding B cells, suggesting autonomous homeostatic controls for naïve and memory B cells a property that may allow reconstitution of the memory pool in immune-deficient hosts using a limited number of precursor naïve B cells. An autonomous control of the memory B cell pool where each antigenic exposure adds new competing clones supports the notion of vaccination strategies using antigenic boosting to prevent memory B cell attrition. Overall the findings reported demonstrate that it is possible to reconstitute the memory B cell pool of an immune deficient host with an artificially induced population of monoclonal high affinity memory B cells.
|
What was used to quantify the amount of IgM secreting cells?
| 4,257 |
ELISpot technique
| 11,251 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
Where was the coronavirus discovered?
| 4,393 |
Wuhan, China
| 377 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the number of confirmed cases reached on 8 February 2020?
| 4,394 |
34,598
| 531 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is presented in this study?
| 4,395 |
new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China
| 593 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the proposed model?
| 4,396 |
an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA)
| 788 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
When is SSA generally employed?
| 4,397 |
to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima)
| 968 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the main idea behind the proposed model?
| 4,398 |
to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA.
| 1,112 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the proposed model called?
| 4,399 |
FPASSA-ANFIS
| 1,094 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
How is FPASSA-ANFIS model evaluated?
| 4,400 |
using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days.
| 1,237 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What did the comparison of the FPASSA-ANFIS model with several existing models, show?
| 4,401 |
it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time.
| 1,459 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
How was the proposed model tested?
| 4,402 |
using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China
| 1,722 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What were the outcomes of the test?
| 4,403 |
showed good performances
| 1,850 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What are the large family of viruses, called coronaviruses?
| 4,404 |
pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases.
| 1,942 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
Among whom are the coronaviruses distributed?
| 4,405 |
among humans, birds, livestock, mice, bats, and other wild animals
| 2,069 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
In what the Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied?
| 4,406 |
in time series prediction and forecasting problems
| 3,197 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What does ANFIS offer?
| 4,407 |
flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems.
| 3,322 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
In which applications has it been applied?
| 4,408 |
in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition
| 3,509 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is SI?
| 4,421 |
swarm intelligence
| 4,452 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is PSO?
| 4,422 |
particle swarm optimization
| 4,668 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is SCA?
| 4,423 |
sine-cosine algorithm
| 4,747 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is MVO?
| 4,424 |
multi-verse optimizer
| 4,786 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
For what SCA algorithm was applied to improve the ANFIS model ?
| 4,425 |
to forecast oil consumption in three countries, namely, Canada, Germany, and Japan.
| 4,895 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What may be a problem with individual SI algorithm?
| 4,426 |
may stock at local optima
| 5,219 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is OWA?
| 4,427 |
ordered weighted averaging
| 3,754 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is proposed in the current study?
| 4,428 |
an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA).
| 6,145 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the FPA optimization algorithm inspired by?
| 4,429 |
inspired by the flow pollination process of the flowering plants.
| 6,333 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the SSA optimization algorithm inspired by?
| 4,430 |
inspired by the behavior of salp chains
| 6,716 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What is the proposed FPASSA method?
| 4,431 |
is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA
| 6,983 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
How does the proposed FPASSA start?
| 4,432 |
by receiving the historical COVID-19 dataset.
| 7,100 |
2,440 |
Optimization Method for Forecasting Confirmed Cases of COVID-19 in China
https://doi.org/10.3390/jcm9030674
SHA: 1d7f8850c5244fdc9b387038e7eeae9bcbbde6d2
Authors: Al-Qaness, Mohammed A. A.; Ewees, Ahmed A.; Fan, Hong; Abd El Aziz, Mohamed
Date: 2020
DOI: 10.3390/jcm9030674
License: cc-by
Abstract: In December 2019, a novel coronavirus, called COVID-19, was discovered in Wuhan, China, and has spread to different cities in China as well as to 24 other countries. The number of confirmed cases is increasing daily and reached 34,598 on 8 February 2020. In the current study, we present a new forecasting model to estimate and forecast the number of confirmed cases of COVID-19 in the upcoming ten days based on the previously confirmed cases recorded in China. The proposed model is an improved adaptive neuro-fuzzy inference system (ANFIS) using an enhanced flower pollination algorithm (FPA) by using the salp swarm algorithm (SSA). In general, SSA is employed to improve FPA to avoid its drawbacks (i.e., getting trapped at the local optima). The main idea of the proposed model, called FPASSA-ANFIS, is to improve the performance of ANFIS by determining the parameters of ANFIS using FPASSA. The FPASSA-ANFIS model is evaluated using the World Health Organization (WHO) official data of the outbreak of the COVID-19 to forecast the confirmed cases of the upcoming ten days. More so, the FPASSA-ANFIS model is compared to several existing models, and it showed better performance in terms of Mean Absolute Percentage Error (MAPE), Root Mean Squared Relative Error (RMSRE), Root Mean Squared Relative Error (RMSRE), coefficient of determination ( R 2 ), and computing time. Furthermore, we tested the proposed model using two different datasets of weekly influenza confirmed cases in two countries, namely the USA and China. The outcomes also showed good performances.
Text: A large family of viruses, called coronaviruses, are severe pathogens for human beings, which infect respiratory, hepatic, gastrointestinal, and neurologic diseases. They are distributed among humans, birds, livestock, mice, bats, and other wild animals [1] [2] [3] . The outbreaks of two previous coronaviruses, SARS-CoV and MERS-CoV in 2003 and 2012, respectively, have approved the transmission from animal to animal, and human to human [4] . In December 2019, the World Health Organization (WHO) received notifications from China for many cases of respiratory illness that were linked to some people who had visited a seafood market in Wuhan [5] . Currently, Wuhan city suffers from the spreading of a novel coronavirus, called COVID-19 (previously, it was called 2019-nCoV). In [6] , the authors concluded that COVID-19 likely originated in bats, because it is more similar to two bat-derived coronavirus strains. However, the source of the COVID-19 is not confirmed yet, and it communities, Hong Kong and Toronto, were 1.2 and 1.32, respectively. Ong et al. [20] proposed a monitoring and forecasting model for influenza A (H1N1-2009). Furthermore, Nah et al. [21] proposed a probability-based model to predict the spread of the MERS.
The Adaptive Neuro-Fuzzy Inference System (ANFIS) [22] is widely applied in time series prediction and forecasting problems, and it showed good performance in many existing applications. It offers flexibility in determining nonlinearity in the time series data, as well as combining the properties of both artificial neural networks (ANN) and fuzzy logic systems. It has been applied in various forecasting applications, for example, in [23] , a stock price forecasting model was proposed using ANFIS and empirical mode decomposition. Chen et al. [24] proposed a TAIEX time series forecasting model based on a hybrid of ANFIS and ordered weighted averaging (OWA). In [25] , another time series forecasting method was presented for electricity prices based on ANFIS. Svalina et al. [26] proposed an ANFIS based forecasting model for close price indices for a stock market for five days. Ekici and Aksoy [27] presented an ANFIS based building energy consumption forecasting model. More so, ANFIS is also applied to forecast electricity loads [28] . Kumar et al. [29] proposed an ANFIS based model to forecast return products. Ho and Tsai [30] applied ANFIS to forecast product development performance. However, estimating ANFIS parameters is a challenge that needs to be improved. Therefore, in previous studies, some individual swarm intelligence (SI) methods have been applied to the ANFIS parameters to enhance time series forecasting because these parameters have a significant effect on the performance of ANFIS. The SI methods include the particle swarm optimization (PSO) [31, 32] , social-spider optimization [33] , sine-cosine algorithm (SCA) [34] , and multi-verse optimizer (MVO) [35] . For example, in [34] SCA algorithm was applied to improve the ANFIS model to forecast oil consumption in three countries, namely, Canada, Germany, and Japan. In the same context, in [35] , The MVO algorithm was used to enhance the ANFIS model to forecast oil consumption in two countries. In addition, in [36] the PSO was used with ANFIS to predict biochar yield. However, individual SI algorithms may stock at local optima. Therefore, one solution is to apply hybrid SI algorithms to avoid this problem. In [37] , a hybrid of two SI algorithms, namely GA and SSA, was presented to improve the ANFIS model. The proposed new model called GA-SSA-ANFIS was applied to forecast crude oil prices for long-term time series data. However, the previously mentioned methods suffer from some limitations that can affect the performance of the forecasting output such as slow convergence and the ability to balance between exploration and exploitation phases can influence the quality of the final output. This motivated us to propose an alternative forecasting method dependent on the hybridization concept. This concept avoids the limitations of traditional SI techniques by combining the strengths of different techniques, and this produces new SI techniques that are better than traditional ones.
In the current study, we propose an improved ANFIS model based on a modified flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). The FPA is an optimization algorithm proposed by Yang [38] , which was inspired by the flow pollination process of the flowering plants. The FPA was employed in various optimization applications, for example to estimate solar PV parameter [39, 40] , solving sudoku puzzles [41] , feature selection [42] , antenna design [43] , and other applications [44] [45] [46] [47] . Moreover, SSA is also an optimization algorithm proposed by Mirjalili et al. [48] inspired by the behavior of salp chains. In recent years, the SSA was utilized to solve different optimization problems, such as feature selection [49, 50] , data classification [51] , image segmentation [52] , and others [53, 54] .
The proposed method called FPASSA is a hybrid of FPA and SSA, in which the SSA is applied as a local search method for FPA. The proposed FPASSA starts by receiving the historical COVID-19 dataset. Then a set of solutions is generated where each of them represents the value for the parameters of the ANFIS model. Then the quality of each solution is calculated using the fitness value, and the solution that has the best fitness value is chosen to represent the best solution. Then the probability of each solution is computed. Then the current solution will be updated, either using global or local strategy in FPA. However, in the case of local strategy, the operators of SSA or FPA will be used according to the probability of the fitness value for each solution. The process of updating the solutions is repeated until reaching the stop condition, and the best parameter configurations are used to forecast the number of confirmed cases of COVID-19.
The main contribution points of the current study are as follows:
1.
We propose an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
An improved ANFIS model is proposed using a modified FPA algorithm, using SSA.
We compare the proposed model with the original ANFIS and existing modified ANFIS models, such as PSO, GA, ABC, and FPA.
The rest of this study is organized as follows. The preliminaries of ANFIS, FPA, and SSA are described in Section 2. Section 3 presents the proposed FPASSA, and Section 4 presents the experimental setup and results. We conclude this study in Section 5.
The principles of the ANFIS are given in this section. The ANFIS model links the fuzzy logic and neural networks [22] . It generates a mapping between the input and output by applying IF-THEN rules (it is also called Takagi-Sugeno inference model). Figure 1 illustrates the ANFIS model where, y and x define the inputs to Layer 1 whereas, O 1i is its output of node i that is computed as follows:
where µ denotes the generalized Gaussian membership functions. A i and B i define the membership values of µ. α i and ρ i denote the premise parameters set. The output of Layer 2 (it is also known as the firing strength of a rule) is calculated as follows:
Meanwhile, the output of Layer 3 (it is also known as the normalized firing strength) is calculated as follows:
The output of Layer 4 (it is also known as an adaptive node) is calculated as follows:
where r i , q i , and p i define the consequent parameters of the node i. Layer 5 contains only one node; its output is computed as:
Flower Pollination Algorithm is an optimization method proposed by Yang [38] . It simulates the transfer of flowers' pollen by pollinators in nature. This algorithm utilizes the two types of pollination (i.e., self-pollination and cross-pollination). In self-pollination, the pollination occurs with no pollinators, whereas, in cross-pollination, the pollens are moved between different plants. In more detail, the self-pollination can be represented as a local pollination while the cross-pollination can be called global pollination.
The global pollination or cross-pollination can be mathematically formed as follows:
where x t i defines the pollen i at iteration t. L denotes the pollination's strength or the step size. F * is the target position or best solution. In some cases, insects can fly with different distance steps for a long space; therefore, Levy fly distribution is applied to simulate this movement.
where λ = 1.5. Γ(λ) denotes the gamma function. This distribution is available for large steps s > 0. The self-pollination or local pollination can be mathematically formed as follows:
where x t i and x k i represent pollens from different flower in the same plant. in the range [0,1] The process of pollination can be done using cross-pollination or self-pollination. Therefore, the random variable p, in the range [0, 1], is used to determine this process.
SSA is an optimization technique introduced by [48] . It simulates the Salps' behavior in nature. This behavior is called salp chain. The mathematical model of SSA begins by splinting its population into a leader group and followers group. The leader is the front salp, whereas, the followers are the other salps. The search space is determined in n-dimensions with n variables. Equation (10) works to update the salps' positions.
where x 1 j denotes the leader's position in j-th dimension. F j is the target position. ub j and lb j represent the max and min bounds, respectively. c 2 and c 3 denote random numbers in [0, 1]. c 1 is an important parameter; it balances between the exploration and exploitation phases. It is computed as follows:
where the current loop number is t and the max loop' number is t max . Then, the followers' position is updated as follows:
where x i j defines the i-th position of the follower in j-th dimension. i > 1.
This section explains the proposed FPASSA-ANFIS method. It is a time series method for forecasting the confirmed cases of the COVID-19, as given in Figure 2 . The FPASSA-ANFIS utilizes the improved FPA to train the ANFIS model by optimizing its parameters. The FPASSA-ANFIS contains five layers as the classic ANFIS model. Layer 1 contains the input variables (the historical COVID-19 confirmed cases). Whereas Layer 5 produces the forecasted values. In the learning phase, the FPASSA is used to select the best weights between Layer 4 and Layer 5.
The FPASSA-ANFIS starts by formatting the input data in a time series form. In our case, the autocorrelation function (ACF) was considered. ACF is one of the methods applied to find patterns in the data; it presents information about the correlation between points separated by various time lags. Therefore, in this paper, the variables with ACF greater than 0.2 are considered i.e., 5-lags.
Besides, the training data contains 75% of the dataset, whereas the testing data contains 25% of them. The number of clusters is defined by the fuzzy c-mean (FCM) method to construct the ANFIS model.
The parameters of the ANFIS model are prepared by the FPASSA algorithm. In the training phase, the calculation error (as in Equation (13)) between the real data and the predicted data is used to evaluate the parameters' quality.
where T is the real data, and P is the predicted data. N s is the sample length. The smaller values of the objective function indicate good ANFIS's parameter.
On the other hand, the updating phase of the followers' positions in the SSA algorithm is applied to improve the global pollination phase in the FPA algorithm. In this improvement, there is a random variable (r) used to switch between both phases. If r > 0.5, then the operators of the SSA is used; otherwise, the operators of the FPA are used. In general, The FPASSA starts by constructing the population (X); afterward, the objective function is calculated for each solution. The solution with the lowest error value is saved to the next iteration. This sequence is repeated until meeting the stop condition, which in this paper, is the maximum number of iterations. Then the best solution is passed to train the parameters of the ANFIS model.
After finishing the training phase, the testing phase is started with the best solution to compute the final output. The performance of the proposed method is evaluated by comparing the real data with the predicted data using the performance measures. Finally, the FPASSA produces a foretasted value for confirmed cases of COVID-19 in China in the next day. The steps of the proposed FPASSA are presented in Algorithm 1.
Input: Historical COVID-19 dataset, size of population N, total number of iterations t max .
Divide the data into training and testing sets.
Using Fuzzy c-mean method to determine the number of membership functions.
Constructing the ANFIS network.
Set the initial value for N solutions (X). Return the best solution that represents the best configuration for ANFIS.
Apply the testing set to the best ANFIS model.
Forecasting the COVID-19 for the next ten days.
This section presents the description of the used dataset, the performance measures, the parameter setting for all methods, the experiment results, and discussions.
The main dataset of this study is COVID-19 dataset. It was collected from the WHO website (https: //www.who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports/). It contains the daily confirmed cases in China from 21 January 2020 to 18 February 2020, as shown in Table 1 . We used 75% from the dataset to train the model while the rest is used to test it.
Moreover, we evaluated the performance of the proposed method using two datasets of weekly influenza confirmed cases. The first one is called DS1; it was collected from the Centers for Disease Control and Prevention (CDC) (https://www.cdc.gov/flu/weekly/). It starts from week number 40 in 2015 and continues until week number 6 in 2020. Whereas, the second one is called DS2. It was collected from the WHO website (https://www.who.int/influenza). It contains the data of weekly influenza confirmed cases in China from week number 1 in 2016 to week number 8 in 2020.
The quality of the proposed method is evaluated using a set of performance metrics as follows:
• Root Mean Square Error (RMSE):
where Yp and Y are the predicted and original values, respectively. • Mean Absolute Error (MAE):
• Mean Absolute Percentage Error (MAPE):
• Root Mean Squared Relative Error (RMSRE):
N s represents the sample size of the data. • Coefficient of Determination (R 2 ):
where Y represents the average of Y.
The lowest value of RMSE, MAE, MAPE, and RMSRE refers to the best method. The higher value of R 2 indicates better correlation for the method.
This paper aims to assess the ability of the FPASSA to forecast the COVID-19 by comparing its performance with other methods, namely the ANFIS and the trained ANFIS models using PSO, GA, ABC, FPA, and FPASSA. The parameters' setting for these models is listed in Table 2 .
The common parameters, such as population size, are set to 25 and 100 iterations are applied. Besides, each algorithm is performed for 30 independent runs to fair comparisons. The selected parameters are chosen because they produced good behavior in previous experiments, such as [34, 35, 55, 56] . Table 2 . Parameters' setting.
Parameters Setting
Max. epochs = 100, Error goal = 0, Initial step = 0.01, Decrease rate = 0.9, Increase rate = 1.
In this section, the performance of the proposed FPASSA to predict the DS1 and DS2 is discussed. It can be concluded from Table 3 that the performance of FPASSA outperformed the compared methods in all measures, whereas the FPA is ranked second. The results of DS2 indicate that the FPASSA is ranked first in terms of RMSE, MAPE, R 2 , and the CPU time. Whereas, the PSO is ranked second, followed by the FPA, GA, then ABC. These results denote that the proposed method can optimize the parameters of the ANFIS model effectively and produce good results in terms of the performance measures. Comparison results between the proposed FPASSA and other models to forecast COVID-19 are given in Table 4 . It can be concluded that the FPASSA outperforms other models. For example, by analyzing the results of RMSE, MAE, MAPE, RMSRE, and CPU time(s) it can be observed that the FPASSA achieves the smallest value among the comparison algorithms, and this indicates the high quality of the FPASSA. Meanwhile, the FPA allocates the second rank, which provides better results than the rest of the methods.
Moreover, the value of R 2 refers to the high correlation between the prediction obtained by the proposed FPASSA method and the original COVID-19, which has nearly 0.97. This can also be noticed from Figure 3 , which depicts the training of the algorithms using the historical data of the COVID-19 as well as their forecasting values for ten days. Table 5 depicts the forecasting value for the confirmed cases of the COVID-19 in China from 19/2/2020 to 28/2/2020. From these results, it can be noticed that the outbreak will reach its highest level on the day 28/2/2020. The average percentage of the increase over the forecasted period is 10%, the highest percentage is 12% on 28/2/2020, and the lowest percentage is 8.7% on 19/2/2020. From the previous results, it can be concluded that the proposed FPASSA-ANFIS has a high ability to forecast the COVID-19 dataset. These results avoid the limitations of traditional ANFIS because of the combination with the modified FPA method. Moreover, the operators of SSA are combined with the local strategy of FPA to enhance their exploitation ability. However, the time computational of the proposed FPASSA method still requires more improvements.
This paper proposed a modified version for the flower pollination algorithm (FPA) using the salp swarm algorithm (SSA). This modified version, called FPASSA, is applied to improve the performance of the ANFIS through determining the optimal value for its parameters. The developed FPASSA-ANFIS model is applied as a forecasting technique for a novel coronavirus, called COVID-19, that was discovered in Wuhan, China at the end of last year and January of the current year. The proposed FPASSA-ANFIS model has a high ability to predict the number of confirmed cases within ten days. Besides, FPASSA-ANFIS outperforms other forecasting models in terms of RMSE, MAE, MAPE, RMSRE, and R 2 . Furthermore, two datasets of weekly influenza confirmed cases in the USA and China were used to evaluate the proposed method, and the evaluation outcomes showed its good performance. According to the promising results obtained by the proposed FPASSA-ANFIS, it can be applied in different forecasting applications.
|
What do the authors propose?
| 4,433 |
an efficient forecasting model to forecast the confirmed cases of the COVID-19 in China for the upcoming ten days based on previously confirmed cases.
| 7,987 |
1,633 |
The vacuolar-type ATPase inhibitor archazolid increases tumor cell adhesion to endothelial cells by accumulating extracellular collagen
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6133348/
SHA: f1b81916fac1ca3d50dde774df2e1bb26bf0fb39
Authors: Luong, Betty; Schwenk, Rebecca; Bräutigam, Jacqueline; Müller, Rolf; Menche, Dirk; Bischoff, Iris; Fürst, Robert
Date: 2018-09-11
DOI: 10.1371/journal.pone.0203053
License: cc-by
Abstract: The vacuolar-type H(+)-ATPase (v-ATPase) is the major proton pump that acidifies intracellular compartments of eukaryotic cells. Since the inhibition of v-ATPase resulted in anti-tumor and anti-metastatic effects in different tumor models, this enzyme has emerged as promising strategy against cancer. Here, we used the well-established v-ATPase inhibitor archazolid, a natural product first isolated from the myxobacterium Archangium gephyra, to study the consequences of v-ATPase inhibition in endothelial cells (ECs), in particular on the interaction between ECs and cancer cells, which has been neglected so far. Human endothelial cells treated with archazolid showed an increased adhesion of tumor cells, whereas the transendothelial migration of tumor cells was reduced. The adhesion process was independent from the EC adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin. Instead, the adhesion was mediated by β1-integrins expressed on tumor cells, as blocking of the integrin β1 subunit reversed this process. Tumor cells preferentially adhered to the β1-integrin ligand collagen and archazolid led to an increase in the amount of collagen on the surface of ECs. The accumulation of collagen was accompanied by a strong decrease of the expression and activity of the protease cathepsin B. Overexpression of cathepsin B in ECs prevented the capability of archazolid to increase the adhesion of tumor cells onto ECs. Our study demonstrates that the inhibition of v-ATPase by archazolid induces a pro-adhesive phenotype in endothelial cells that promotes their interaction with cancer cells, whereas the transmigration of tumor cells was reduced. These findings further support archazolid as a promising anti-metastatic compound.
Text: The vacuolar-type H + -ATPase (v-ATPase) is the major proton pump responsible for acidification of intracellular compartments in eukaryotic cells [1] . The enzyme consists of two multi-subunit complexes, the soluble V 1 transmembrane V o subcomplex required for the proton transport across membranes [1, 2] . In most cell types v-ATPases are only expressed in the endomembrane system to regulate and maintain the acidic pH of intracellular compartments such as lysosomes, endosomes, the Golgi apparatus, secretory granules and coated vesicles [3] . The function of v-ATPases is essential for cellular processes such as vesicular trafficking, receptor-mediated endocytosis and protein degradation and processing. In specialized cell types including osteoclasts and renal epithelial cells, v-ATPases can also be expressed on the plasma membrane, where they pump protons into the extracellular space [2] [3] [4] . In cancer cells v-ATPases are expressed on the plasma membrane in order to eliminate toxic cytosolic H + . Most importantly, v-ATPases contribute to the acidic tumor microenvironment, which leads to the activation of proteases, thus facilitating tumor cell migration, invasion and angiogenesis [5] [6] [7] . Since the inhibition of v-ATPase was shown to reduce the invasiveness of cancer cells and metastasis formation [8, 9] , this enzyme has emerged as a promising drug target in the recent years. Archazolid A and B are highly potent and specific inhibitors of v-ATPases [10] . They were first isolated from the myxobacterium Archangium gephyra [11] . These compounds inhibit v-ATPase at low nanomolar concentrations [10, 12] by binding to the subunit c of the V o complex. As their biological activity is comparable to the v-ATPase inhibitors bafilomycin and concanamycin [10, 11] , archazolids are natural compounds of high interest that can be used both as a tool to study the consequences of v-ATPase inhibition and as a lead for drug development. Archazolids can be either produced by fermentation [11] or by total synthesis [13, 14] .
In the field of cancer research several studies reported on interesting pharmacological effects of archazolid: It reduced the migration of different invasive tumor cells in vitro and cancer cell metastasis in vivo in a breast tumor mouse model [15] . Furthermore, archazolid activated pathways of cellular stress response and apoptosis in highly invasive tumor cells [16] . In classically activated macrophages, archazolid selectively induced the generation of tumor necrosis factor α (TNFα), which may indirectly promote tumor suppression [17] .
Up to now, the role of v-ATPases in endothelial cells has only rarely been investigated. The endothelium plays a crucial role in the pathogenesis and progression of cancer: The metastatic cascade includes local angiogenesis at the site of the primary tumor and adhesion of tumor cells at the site of metastasis [18] . Angiogenesis, the development of new blood vessels out of existing ones, depends on the proliferation, migration and differentiation of endothelial cells [19] . This process ensures the nutrient supply of the tumor and its growth [20] . Circulating cancer cells can adhere to the endothelium at distant sites. This adhesive interaction is mediated by receptors and corresponding ligands expressed on tumor and endothelial cells [18, 21] . V-ATPases have been reported to regulate intracellular pH and cell migration in microvascular endothelial cells [22, 23] . A recent study showed that the inhibition of v-ATPase by concanamycin prevented proliferation, reduced migration and impaired angiogenesis-related signaling in endothelial cells [24] . So far, there are no investigations on the role of endothelial v-ATPases for the process of tumor cell adhesion onto the endothelium. Thus, we were interested in the consequences of the inhibition of endothelial v-ATPase by archazolid on the interaction between endothelial and cancer cells. Various cell adhesion molecules on the endothelium, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion protein (VCAM-1), E-selectin or N-cadherin [21] as well as integrins expressed on cancer cells have been reported to mediate cell adhesion of cancer cells onto endothelial cells [25] [26] [27] . Accordingly, we focused on these cell adhesion molecules and integrins. For the first time, our study revealed a link between the function of v-ATPases and the adhesion and transmigration properties of endothelial cells.
CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany) was performed according to the manufacturer's protocol for determining the cell viability of cells after treatment with archazolid. This assay is based on the ability of metabolically active cells to reduce resazurin which results in fluorescent resorufin. The CellTiter-Blue Reagent was added to the cells 4 h before the endpoint of treatment. Fluorescence was measured with an Infinite F200 pro microplate reader (Tecan, Männedorf, Switzerland) at 560 nm (excitation) and 590 nm (emission).
CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) was performed according to the manufacturer's instructions for determining the lactate dehydrogenase (LDH) release after treatment with archazolid. Lysis buffer was added to untreated cells 45 min before the end of treatment to induce the release of this enzyme. LDH is a cytosolic enzyme that is released by leaky cells. Released LDH catalyzes the enzymatic conversion of lactate to pyruvate which provides NADH for the conversion of iodonitrotetrazolium violet into a red formazan product in the presence of diaphorase. The absorbance was measured with a Varioskan Flash microplate reader (Thermo Fisher Scientific) at 490 nm.
LysoTracker Red DND-99 (Life Technologies, Thermo Fisher Scientific) is a dye to measure pH values in viable cells. HUVECs were cultured to confluence on collagen G-coated μ-slides (80826, ibidi, Martinsried, Germany) before they were treated with archazolid for 24 h. 1 μg/ ml Hoechst 33342 (Sigma-Aldrich, Munich, Germany) was used to visualize the nuclei and 50 nM LysoTracker Red DND-99 was used to visualize the acidic compartments which correspond to the lysosomes. Both dyes were incubated for 10 min at 37˚C before acquisition of single images by a Leica DMI6000 B fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
HUVECs were seeded in collagen G-coated 24-well plates and grown to confluence for two days before treatment. The cells were incubated with indicated concentrations of archazolid for 24 h. Untreated MDA-MB-231 or PC-3 cells were labeled with CellTracker Green CMFDA Dye (5 μM in serum-free DMEM, 37˚C) for 30 min before 100,000 cells per well were added to HUVECs and were allowed to adhere for various time points at 37˚C. Non-adherent tumor cells were washed off three times with PBS containing Ca 2+ and Mg 2+ . Tumor cell adhesion was determined by fluorescence measurements with an Infinite F200 pro microplate reader (Tecan) at 485 nm (excitation) and 535 nm (emission).
For blocking the integrin β1 subunit on MDA-MB-231 or PC-3 cells, CellTracker Greenlabeled MDA-MB-231 or PC-3 cells were incubated with an anti-integrin β1 antibody (P5D2, ab24693, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg antibody per one million cells in 1 ml DMEM. Before adding to archazolid-treated HUVECs, MDA-MB-231 or PC-3 cells were washed once with DMEM. For blocking the integrin β1 subunit on HUVECs, the cells were incubated with the anti-integrin β1 antibody (0.1 μg/well in ECGM). HUVECs were washed once with ECGM before untreated MDA-MB-231 or PC-3 cells were added to HUVECs.
For the adhesion of MDA-MB-231 or PC-3 cells onto extracellular matrix (ECM) components 24-well plates were coated with collagen G (10 μg/ml in PBS), human plasma fibronectin (10 μg/ml PBS) or laminin-411 (10 μg/ml in Dulbecco's PBS [DPBS] containing Ca 2+ and Mg 2+ ) at 4˚C overnight. The adhesion of MDA-MB-231 and PC-3 cells onto these three most prominent ECM components was carried out as described above (10 min adhesion at 37˚C).
HUVECs were grown on a porous filter membrane (Transwell insert, polycarbonate membrane, 8 μm pores; Corning, New York, USA) for 48 h and were treated as indicated. Untreated MDA-MB-231 cells were labeled with CellTracker Green CMFDA Dye (as described in the section cell adhesion assay) and resuspended in medium 199 (PAN-Biotech) containing 0.1% BSA. HUVECs were washed twice with medium 199 containing 0.1% BSA before MDA-MB-231 cells were allowed to transmigrate through the endothelial monolayer for 24 h. Medium 199 containing 0.1% BSA was used as negative control and medium 199 containing 20% FCS was used as chemoattractant for transmigration in the lower compartment. Non-migrated cells remaining in the upper compartment were carefully removed using a cotton swab. Transmigrated cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and transmigration was quantified by measuring the fluorescence signal at 485 nm (excitation) and 535 nm (emission).
HUVECs were grown to confluence on 6-well plates before they were treated with archazolid for 12 h. The cells were induced to upregulate the gene expression of cell adhesion molecules by TNFα. RNA was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) according to the manufacturer's protocol. On-column DNase digestion was performed to remove genomic DNA. RNA was transcribed into cDNA by Superscript II (Life Technologies, Thermo Fisher Scientific). qPCR experiments were performed using a StepOnePlus System (Applied Biosystems, Thermo Fisher Scientific) and data was analyzed by the StepOne and Ste-pOnePlus Software v2.3. Power SYBR Green PCR Master Mix (Life Technologies) and the comparative C T quantitation method (2 -ΔΔCT ) were used.
HUVECs were grown to confluence on 12-well plates before they were treated with archazolid for 24 h. Cells were treated with TNFα for 24 h to induce the expression of cell adhesion molecules. Subsequently, the cells were detached with HyClone HyQTase (GE Healthcare, Freiburg, Germany). In the case of ICAM-1 the detached cells were fixed with 4% formaldehyde (Polysciences, Hirschberg an der Bergstraße, Germany) in PBS for 10 min and washed once with PBS before incubating with the fluorescein isothiocyanate (FITC)-labeled anti-human CD54 (mouse, ICAM-1) antibody (MCA1615F, Biozol, Eching, Germany) at room temperature for 45 min. For all other proteins, the cells were not fixed and washed once with PBS before incubating with the antibodies phycoerythrin (PE)-labeled anti-human CD106 (mouse, VCAM-1), PE-labeled anti-human CD62E (mouse, E-selectin) and PE-labeled anti-human CD325 (mouse, N-cadherin) from Becton Dickinson on ice for 45 min. These antibodies were diluted in PBS containing 0.2% BSA. The surface expression of cell adhesion molecules was measured by flow cytometry (FACSVerse, Becton Dickinson, Heidelberg, Germany).
To stain the surface collagen on HUVECs, cells were incubated with an anti-human collagen antibody (rabbit, 1:40, ab36064, Abcam) on ice for 30 min. The staining procedure was performed on ice to ensure that surface proteins or antibodies are not endocytosed. The cells were washed once with PBS containing Ca 2+ and Mg 2+ before they were fixed with Roti-Histofix (Carl Roth). Alexa Fluor 488-conjugated anti-rabbit antibody (goat, 1:400, A11008, Life Technologies) was used as secondary antibody and Hoechst 33342 (1 μg/ml, Sigma-Aldrich) was used to visualize nuclei.
Confluent HUVECs in 6-well plates were treated as indicated. Cells were washed with ice-cold PBS and lysed with RIPA buffer supplemented with protease inhibitors (Complete Mini EDTA-free; Roche, Mannheim, Germany), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride, β-glycerophosphate, sodium pyrophosphate and H 2 O 2 . Protein determination was performed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins (10-20 μg) were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad Laboratories, Munich, Germany). Separated proteins were transferred onto polyvinylidene difluoride membranes by tank blotting (Bio-Rad Laboratories) for immunodetection. Membranes were blocked with 5% boltinggrade milk powder (Carl Roth) in TBS containing 0.1% Tween 20 (Sigma-Aldrich). The following antibodies were used: mouse anti-human cathepsin B antibody (IM27L, Merck) (1:500), mouse anti-β-actin-peroxidase antibody (A3854, Sigma-Aldrich) (1:100,000) and antimouse IgG horse radish peroxidase (HRP)-linked antibody (7076, Cell Signaling, Frankfurt, Germany) (1:5,000). ImageJ version 1.49m was used for densitometric analysis.
Cathepsin B activity assay was performed as described in the publication by Kubisch et al. [28] . Confluent HUVECs or HMEC-1 seeded in 6-well plates were treated as indicated. Cells were washed with PBS and lysed with the non-denaturating M-PER mammalian protein extraction reagent (78501, Thermo Fisher Scientific) supplemented with protease inhibitors (Complete Mini EDTA-free, Roche), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride. The fluorogenic cathepsin B substrate Z-Arg-Arg-7-amido-4-methylcoumarin hydrochloride (C5429, Sigma-Aldrich) was added to 30 μg of the cell lysate diluted in assay buffer supplemented with 2 mM L-cysteine (C7880, Sigma-Aldrich) and incubated for 30 min at 40˚C. Fluorescence was measured at 348 nm (excitation) and 440 nm (emission) with a microplate reader (Varioskan Flash, Thermo Fisher Scientific). The intensity of the fluorescence signal corresponded to the cathepsin B enzyme activity. For background subtraction the cathepsin B inhibitor CA-074Me (Enzo Life Sciences, Lörrach, Germany) was added to an additional reaction.
The HUVEC Nucleofector Kit (Lonza, Cologne, Germany) was used to transfect HUVECs. The transfection was performed according to the manufacturer's protocol using 2.5 μg plasmid DNA for 500,000 cells (Nucleofector 2b Device, Lonza). 48 h after transfection the cells were treated for further experiments. The addgene plasmid #11249 hCathepsin B was kindly provided by Hyeryun Choe [29] . hCathepsin B was digested with PmeI and XbaI and the linear DNA fragment not corresponding to the human CTSB gene was religated to generate the empty pcDNA3.1 (-) delta MCS plasmid that was used for control transfections. The original backbone of hCathepsin B is the pcDNA3.1 (-) from Thermo Fisher Scientific. The control vector pcDNA3.1 (-) delta MCS used for our transfections was cloned on the basis of hCathepsin B and is therefore lacking almost the whole part of the multiple cloning site of the pcDNA3.1 (-).
Statistical analyses were performed using GraphPad Prism 5.0 (San Diego, USA). One-way ANOVA followed by Tukey's post-hoc test or unpaired t-test was used for the evaluation of a minimum of three independent experiments. The numbers of independently performed experiments (n) are stated in the corresponding figure legends. p 0.05 was considered as statistically significant. Data are expressed as mean ± standard error of the mean (SEM).
Since the v-ATPase inhibitor archazolid has never been used for studies in endothelial cells, we first performed cytotoxicity assays. We treated confluent HUVECs with up to 1 nM archazolid for 24 and 48 h and observed that this treatment has neither an influence on the metabolic activity nor on the release of LDH after 24 h (Fig 1A and 1B, left panels) . The metabolic activity and the release of LDH were only slightly affected by the highest concentration of archazolid after 48 h (Fig 1A and 1B, right panels) . Consequently, the following experiments were all carried out after 24 h (or less) of archazolid treatment in order to exclude any cytotoxic effects of archazolid within our experimental settings.
Microscopic analysis revealed that also the integrity of the endothelial monolayer was not affected by archazolid, but the cells showed a slightly different morphology (Fig 2A) : Archazolid-treated cells were swollen compared to control cells, which was not unexpected, as vacuolation of the endoplasmic reticulum (ER) has been described for other cell types and is typical for v-ATPase inhibitors [11, 16, 24, 30] . This effect was obvious both in subconfluent and in confluent cells (Fig 2A) . Inhibition of v-ATPase prevents the acidification of lysosomes [1, 31] . Using the cell-permeable dye LysoTracker Red DND-99, it is possible to label the acidic lysosomes in living cells. Thus, this dye can serve as an indicator of v-ATPase inhibition. To proof that archazolid is also functionally active as a v-ATPase inhibitor in HUVECs, cells were treated with 1 nM archazolid before they were incubated with LysoTracker Red DND-99 and Hoechst 33342. As shown in Fig 2B, the red vesicular staining corresponding to acidified lysosomes in control cells disappeared completely after treatment with archazolid. In summary, archazolid treatment for 24 h was not cytotoxic to quiescent HUVECs, but inhibited the functionality of the v-ATPase.
We analyzed the adhesion of MDA-MB-231 cells onto HUVECs. Confluent HUVECs were treated with up to 1 nM archazolid for 24 h. Untreated MDA-MB-231 cells were labeled with Cell-Tracker Green CMFDA Dye. Interestingly, v-ATPase inhibition strongly increased the attachment of the metastatic breast carcinoma cell line MDA-MB-231 onto HUVECs after 10 and 120 min of adhesion (Fig 3A and 3B) . We also investigated the influence of archazolid on the transendothelial migration of MDA-MB-231 cells. HUVECs seeded in a Boyden chamber were treated with 1 nM archazolid for 24 h. CellTracker Green-labeled MDA-MB-231 cells (not treated with archazolid) were allowed to transmigrate through the endothelial monolayer for 24 h. As shown in Fig 3C, archazolid significantly decreased the transendothelial migration of MDA-MB-231 cells.
The influence of archazolid on tumor cell adhesion was not only studied in HUVECs, which represent macrovascular endothelial cells, but also in microvascular HMEC-1 cells. Moreover, besides the breast cancer cell line MDA-MB-231, also PC-3 prostate cancer cells were used as a second metastatic cancer cell line. Archazolid treatment of endothelial cells increased the attachment of MDA-MB-231 cells onto the HMEC-1 monolayer after 120 min of adhesion ( Fig 4A) and increased the attachment of PC-3 cells onto the HUVEC monolayer after 30 and 60 min of adhesion (Fig 4B) . Of note, the adhesion of non-metastatic Jurkat cells, an acute T cell leukemia cell line, remained unaffected after treatment of HUVECs with archazolid (S1A Fig). Taken together, archazolid treatment augmented the adhesive properties of both micro-and macrovascular endothelial cells for metastatic tumor cells, but not for non-metastatic ones. Of note, cancer cell adhesion onto archazolid-activated endothelial cells increased with the time of adhesion.
The adhesion of tumor cells onto the endothelium is in principle similar to that of leukocytes, but slightly differs in the molecules that mediate the adhesion process. Ligands for the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin were found to be expressed on tumor cells and to mediate tumor-endothelial cell interaction [21] . Inhibition of the v-ATPase might affect the expression of endothelial cell adhesion molecules on mRNA or protein levels. To determine the mRNA expression of ICAM-1, VCAM-1, E-selectin and Ncadherin, HUVECs were treated with archazolid for 12 h. TNFα is known to upregulate the expression of ICAM-1, VCAM-1 and E-selectin [32] and, thus, served as positive control. Quantitative real-time PCR showed that v-ATPase inhibition in HUVECs did not alter the mRNA levels of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5A) . The protein expression of these adhesion molecules on the surface of endothelial cells was analyzed by flow cytometry. Archazolid (1 nM, 24 h) did not affect the cell surface expression of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5B) .
Besides ICAM-1, VCAM-1, E-selectin and N-cadherin, also integrins are able to mediate the process of cell adhesion [33] [34] [35] . Since none of the cell adhesion molecules expressed on HUVECs were regulated upon archazolid treatment, we considered integrins as potential interaction partners. Within this protein family β1-integrins have been reported to mediate tumor cell adhesion onto quiescent endothelial cells [25] . In order to elucidate the role of β1-integrins for the archazolid-induced tumor cell adhesion, the integrin β1-subunit was blocked either on MDA-MB-231 cells, PC-3 cells or on HUVECs. (Of note, as in all experiments throughout this study, only endothelial cells were treated with archazolid.) After blocking β1-integrins on MDA-MB-231 or PC-3 cells, the archazolid-induced tumor cell adhesion was reduced almost to control level (Fig 6A and 6B , left panels), whereas blocking of β1-integrins on HUVECs had no significant effect on the increase of tumor cell adhesion by v-ATPase inhibition (Fig 6A and 6B , right panels).
Depending on their α subunit, β1-integrins have a variety of ligands including extracellular matrix (ECM) components such as collagen, fibronectin and laminin [35] . Therefore, we hypothesized that archazolid treatment of endothelial cells might lead to an upregulation of these components. MDA-MB-231 and PC-3 cells were allowed to adhere onto plastic that was coated with these ECM components. This cell adhesion assay revealed that MDA-MB-231 as well as PC-3 cells favor the interaction with the ECM component collagen, as the adhesion onto collagen is much higher than onto the uncoated plastic control (Fig 7A) . MDA-MB-231 and PC-3 cells also adhered to fibronectin-coated plastic, but to a much lesser extent compared to the collagen coating. Therefore, we focused on the interaction between these two tumor cell lines and collagen. Blocking of the integrin β1 subunit on MDA-MB-231 and PC-3 cells clearly abolished the interaction with collagen (Fig 7B) , indicating that the attachment of these tumor cells to collagen is mediated by β1-integrins.
Since collagen is the major ECM component MDA-MB-231 and PC-3 cells interact with, the next step was to prove whether v-ATPase inhibition influences the amount of collagen expressed by HUVECs as extracellular matrix. To detect collagen on the endothelial surface, archazolid-treated HUVECs were labeled with an antibody against collagen type I-IV on ice to prevent endocytosis and to ensure that the antibody does not bind to intracellular collagen. Interestingly, archazolid increased the amount of surface collagen on HUVECs by about 50% (Fig 7C) . Control stainings were performed using an antibody against the cytosolic p65 subunit of the transcription factor nuclear factor κB (NFκB) to show that intracellular proteins were not detected by this staining method (S2 Fig) .
It was reported that v-ATPase inhibition by archazolid impairs the activity of cathepsin B [28, 36] , a lysosomal enzyme that degrades extracellular matrix components including collagen [37] [38] [39] [40] [41] . As collagen is degraded by cathepsin B and the activation of cathepsin B depends on v-ATPase activity [28, [36] [37] [38] 42] , we suggested that an accumulation of collagen on the surface of endothelial cells might be a consequence of an impaired functionality of cathepsin B. Therefore, an enzyme activity assay based on the proteolysis of a fluorogenic cathepsin B substrate was performed. In archazolid-treated HUVECs and HMEC-1 the activity of cathepsin B was induce both the mRNA (1 ng/ml TNF) and the cell surface expression (10 ng/ml TNF) of ICAM-1, VCAM-1, E-selectin and Ncadherin.
https://doi.org/10.1371/journal.pone.0203053.g005 Inhibition of endothelial vATPase increases tumor cell adhesion to endothelial cells strongly decreased by approximately 50% compared to control cells at an archazolid concentration of 1 nM (Fig 8A) . In line with this result, western blot analysis showed that archazolid (1 nM) reduces the protein expression of the mature, active form of cathepsin B to less than 40% of the control in HUVECs (Fig 8B) . To proof whether the archazolid-induced tumor cell adhesion is a consequence of the decreased amount of cathepsin B, HUVECs were transfected with a plasmid coding for human cathepsin B or with the empty vector as control. After 48 h, the transfected cells were treated with 1 nM archazolid. The level of cathepsin B after transfection and treatment was assessed by western blot analysis (Fig 9A) . Overexpression of cathepsin B strongly diminished both the basal and the archazolid-induced adhesion of MDA-MB-231 cells (Fig 9B) .
Targeting the proton pump v-ATPase for cancer therapy has gained great interest since its inhibition was reported to reduce the invasiveness of cancer cells and, most importantly, also metastasis [8, 9] . Thus, intensive research related to v-ATPases was done in cancer cells, whereas there are only few studies investigating v-ATPases in endothelial cells indicating a role in migration, proliferation and possibly angiogenesis [22] [23] [24] . In the present study we used the myxobacterial natural product archazolid to investigate the consequences of v-ATPase inhibition in the endothelium on tumor-endothelial cell interactions.
For the first time, we were able to show a link between v-ATPase and the adhesion and transmigration properties of the endothelium. Inhibition of the v-ATPase in endothelial cells by archazolid significantly increased the adhesion of metastatic cancer cells and decreased the transendothelial migration of cancer cells which was attributed to augmented collagen levels on the surface on archazolid-treated endothelial cells. Of note, adhesion of the non-metastatic Jurkat cell line onto archazolid-treated endothelial cells remained unaffected. The archazolidinduced adhesion of tumor cells was independent from the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin, as their expression was not regulated by the compound. However, we found that the archazolid-induced tumor cell adhesion was mediated by β1-integrins expressed on MDA-MB-231 breast cancer and PC-3 prostate cancer cells as blocking of the integrin β1 subunit on these tumor cells reversed the pro-adhesive effect of archazolid. In adhesion experiments on plastic coated with extracellular matrix components, we could show that MDA-MB-231 and PC-3 cells clearly favored the interaction with collagen, whereas the adhesion of non-metastatic Jurkat cells was largely independent from extracellular matrix proteins (S1B Fig). The different adhesion properties of metastatic cancer cells and Jurkat cells might be a result of the distinct integrin expression pattern of each cell line. MDA-MB-231 and PC-3 cells express α2β1-and α3β1-integrins, which represent collagen receptors [43, 44] , while Jurkat cells express α4β1-integrins but lack α2β1-, α3β1-integrins [44] . α4β1integrins are receptors for VCAM-1 and fibronectin [35] and it has been shown that Jurkat cells interact with human endothelial cells that express VCAM-1 after cytokine treatment or cells transfected with VCAM-1 [45] . Our results are in line with previous studies showing that α2β1-and α3β1-integrin expressing MDA-MB-231 and PC-3 cells were able to rapidly attach to collagen in the cortical bone matrix. In contrast, Jurkat cells were not able to adhere [44] and might preferentially interact with cell adhesion molecules rather than with ECM proteins. α2β1-and α3β1-integrins can additionally act as laminin receptors [46] and at least α3β1integrins recognize fibronectin [46, 47] . Though expressing receptors for fibronectin and laminin, MDA-MB-231 and PC-3 cells adhered to fibronectin to a much lesser extent and did not adhere to laminin, probably due to lower affinities to these extracellular matrix components.
Importantly, v-ATPase inhibition by archazolid increased the surface levels of the extracellular matrix component collagen, which might explain that the increase of MDA-MB-231 and PC-3 cells onto archazolid-treated HUVECs is independent of endothelial cell adhesion molecules. By performing a live cell proteolysis assay, Cavallo-Medved et al. demonstrated ECM degradation, in particular of gelatin and collagen IV, in association with active cathepsin B in caveolae of endothelial cells during tube formation [40] . In addition, recent studies reported that v-ATPase inhibition impairs the activity of cathepsin B in cancer cells [28, 36] . Therefore, we suggested that the accumulation of collagen on the endothelial surface might be a consequence of impaired cathepsin B activity or expression in endothelial cells. In fact, we confirmed the impairment of cathepsin B activity by archazolid as the expression levels of the mature active form of this enzyme was strongly reduced. Cathepsin B is synthesized as preprocathepsin B on membrane-bound ribosomes. Following transport to the Golgi apparatus, the preprocathepsin B is glycosylated with mannose-containing oligosaccharides. The targeting of procathepsin B to lysosomes is mannose-6-phosphate receptor-dependent and its dissociation from the receptor as well as its proteolytic processing into mature cathepsin B requires acidification of the compartment [48] . In cancer cells v-ATPase inhibition by archazolid impaired the mannose-6-phosphate receptor-mediated trafficking from the trans-Golgi network to prelysosomal compartments resulting in a decrease of active lysosomal proteases like cathepsin B [28] . We assumed that the archazolid-induced decrease in cathepsin B activity and expression was based on the same mechanism. Interestingly, overexpression of cathepsin B attenuated the archazolid-induced adhesion of breast cancer cells onto endothelial cells, indicating that the adhesion negatively correlates with the expression of cathepsin B.
As cathepsin B can also degrade other extracellular matrix components such as fibronectin and laminin [38, 49] , v-ATPase inhibition could lead to an accumulation of these proteins and an increased adhesion of cells expressing fibronectin or laminin receptors. However, we did not focus on these ECM components since they were not relevant for the adhesion of MDA-MB-231 and PC-3 cells. These cells predominantly adhered to collagen, while the adhesion of Jurkat cells is mostly independent from the ECM proteins collagen, fibronectin or laminin (S1B Fig). Interestingly [50] . In hepatic cancer cells, archazolid reduces Ras/Raf/MEK/ERK signaling by altering the membrane composition and fluidity [51] . We assume that archazolid affects endothelial cells in a similar way leading to inhibition of Ras signaling and, therefore, reduced transendothelial migration of MDA-MB-231 cells.
Taken together, our study shows that archazolid reduces the activity and expression of cathepsin B in endothelial cells. As a result, the amount of collagen on the surface of endothelial cells was significantly upregulated, which finally resulted in an increased adhesion of the β1-integrin-expressing metastatic cancer cell lines MDA-MB-231 and PC-3 onto archazolidtreated endothelial cells, whereas the adhesion of non-metastatic Jurkat cells was unaffected. This study shows that the v-ATPase plays an important role in regulating the adhesion of cells expressing receptors for extracellular matrix components. Archazolid represents a promising tool to elucidate the role of v-ATPase in endothelial cells. Moreover, we for the first time linked the function of v-ATPase to the adhesion and transmigration of tumor cells onto endothelial cells as well as to the remodeling of the extracellular matrix on the surface of endothelial cells. The fact that the adhesion of metastatic tumor cells onto endothelial cells is increased while their transendothelial migration is reduced upon inhibition of endothelial v-ATPase by archazolid further supports the view of archazolid as a potential anti-metastatic compound.
|
What is vacuolar-type H(+)-ATPase (v-ATPase)?
| 5,301 |
the major proton pump that acidifies intracellular compartments of eukaryotic cells
| 482 |
1,633 |
The vacuolar-type ATPase inhibitor archazolid increases tumor cell adhesion to endothelial cells by accumulating extracellular collagen
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6133348/
SHA: f1b81916fac1ca3d50dde774df2e1bb26bf0fb39
Authors: Luong, Betty; Schwenk, Rebecca; Bräutigam, Jacqueline; Müller, Rolf; Menche, Dirk; Bischoff, Iris; Fürst, Robert
Date: 2018-09-11
DOI: 10.1371/journal.pone.0203053
License: cc-by
Abstract: The vacuolar-type H(+)-ATPase (v-ATPase) is the major proton pump that acidifies intracellular compartments of eukaryotic cells. Since the inhibition of v-ATPase resulted in anti-tumor and anti-metastatic effects in different tumor models, this enzyme has emerged as promising strategy against cancer. Here, we used the well-established v-ATPase inhibitor archazolid, a natural product first isolated from the myxobacterium Archangium gephyra, to study the consequences of v-ATPase inhibition in endothelial cells (ECs), in particular on the interaction between ECs and cancer cells, which has been neglected so far. Human endothelial cells treated with archazolid showed an increased adhesion of tumor cells, whereas the transendothelial migration of tumor cells was reduced. The adhesion process was independent from the EC adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin. Instead, the adhesion was mediated by β1-integrins expressed on tumor cells, as blocking of the integrin β1 subunit reversed this process. Tumor cells preferentially adhered to the β1-integrin ligand collagen and archazolid led to an increase in the amount of collagen on the surface of ECs. The accumulation of collagen was accompanied by a strong decrease of the expression and activity of the protease cathepsin B. Overexpression of cathepsin B in ECs prevented the capability of archazolid to increase the adhesion of tumor cells onto ECs. Our study demonstrates that the inhibition of v-ATPase by archazolid induces a pro-adhesive phenotype in endothelial cells that promotes their interaction with cancer cells, whereas the transmigration of tumor cells was reduced. These findings further support archazolid as a promising anti-metastatic compound.
Text: The vacuolar-type H + -ATPase (v-ATPase) is the major proton pump responsible for acidification of intracellular compartments in eukaryotic cells [1] . The enzyme consists of two multi-subunit complexes, the soluble V 1 transmembrane V o subcomplex required for the proton transport across membranes [1, 2] . In most cell types v-ATPases are only expressed in the endomembrane system to regulate and maintain the acidic pH of intracellular compartments such as lysosomes, endosomes, the Golgi apparatus, secretory granules and coated vesicles [3] . The function of v-ATPases is essential for cellular processes such as vesicular trafficking, receptor-mediated endocytosis and protein degradation and processing. In specialized cell types including osteoclasts and renal epithelial cells, v-ATPases can also be expressed on the plasma membrane, where they pump protons into the extracellular space [2] [3] [4] . In cancer cells v-ATPases are expressed on the plasma membrane in order to eliminate toxic cytosolic H + . Most importantly, v-ATPases contribute to the acidic tumor microenvironment, which leads to the activation of proteases, thus facilitating tumor cell migration, invasion and angiogenesis [5] [6] [7] . Since the inhibition of v-ATPase was shown to reduce the invasiveness of cancer cells and metastasis formation [8, 9] , this enzyme has emerged as a promising drug target in the recent years. Archazolid A and B are highly potent and specific inhibitors of v-ATPases [10] . They were first isolated from the myxobacterium Archangium gephyra [11] . These compounds inhibit v-ATPase at low nanomolar concentrations [10, 12] by binding to the subunit c of the V o complex. As their biological activity is comparable to the v-ATPase inhibitors bafilomycin and concanamycin [10, 11] , archazolids are natural compounds of high interest that can be used both as a tool to study the consequences of v-ATPase inhibition and as a lead for drug development. Archazolids can be either produced by fermentation [11] or by total synthesis [13, 14] .
In the field of cancer research several studies reported on interesting pharmacological effects of archazolid: It reduced the migration of different invasive tumor cells in vitro and cancer cell metastasis in vivo in a breast tumor mouse model [15] . Furthermore, archazolid activated pathways of cellular stress response and apoptosis in highly invasive tumor cells [16] . In classically activated macrophages, archazolid selectively induced the generation of tumor necrosis factor α (TNFα), which may indirectly promote tumor suppression [17] .
Up to now, the role of v-ATPases in endothelial cells has only rarely been investigated. The endothelium plays a crucial role in the pathogenesis and progression of cancer: The metastatic cascade includes local angiogenesis at the site of the primary tumor and adhesion of tumor cells at the site of metastasis [18] . Angiogenesis, the development of new blood vessels out of existing ones, depends on the proliferation, migration and differentiation of endothelial cells [19] . This process ensures the nutrient supply of the tumor and its growth [20] . Circulating cancer cells can adhere to the endothelium at distant sites. This adhesive interaction is mediated by receptors and corresponding ligands expressed on tumor and endothelial cells [18, 21] . V-ATPases have been reported to regulate intracellular pH and cell migration in microvascular endothelial cells [22, 23] . A recent study showed that the inhibition of v-ATPase by concanamycin prevented proliferation, reduced migration and impaired angiogenesis-related signaling in endothelial cells [24] . So far, there are no investigations on the role of endothelial v-ATPases for the process of tumor cell adhesion onto the endothelium. Thus, we were interested in the consequences of the inhibition of endothelial v-ATPase by archazolid on the interaction between endothelial and cancer cells. Various cell adhesion molecules on the endothelium, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion protein (VCAM-1), E-selectin or N-cadherin [21] as well as integrins expressed on cancer cells have been reported to mediate cell adhesion of cancer cells onto endothelial cells [25] [26] [27] . Accordingly, we focused on these cell adhesion molecules and integrins. For the first time, our study revealed a link between the function of v-ATPases and the adhesion and transmigration properties of endothelial cells.
CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany) was performed according to the manufacturer's protocol for determining the cell viability of cells after treatment with archazolid. This assay is based on the ability of metabolically active cells to reduce resazurin which results in fluorescent resorufin. The CellTiter-Blue Reagent was added to the cells 4 h before the endpoint of treatment. Fluorescence was measured with an Infinite F200 pro microplate reader (Tecan, Männedorf, Switzerland) at 560 nm (excitation) and 590 nm (emission).
CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) was performed according to the manufacturer's instructions for determining the lactate dehydrogenase (LDH) release after treatment with archazolid. Lysis buffer was added to untreated cells 45 min before the end of treatment to induce the release of this enzyme. LDH is a cytosolic enzyme that is released by leaky cells. Released LDH catalyzes the enzymatic conversion of lactate to pyruvate which provides NADH for the conversion of iodonitrotetrazolium violet into a red formazan product in the presence of diaphorase. The absorbance was measured with a Varioskan Flash microplate reader (Thermo Fisher Scientific) at 490 nm.
LysoTracker Red DND-99 (Life Technologies, Thermo Fisher Scientific) is a dye to measure pH values in viable cells. HUVECs were cultured to confluence on collagen G-coated μ-slides (80826, ibidi, Martinsried, Germany) before they were treated with archazolid for 24 h. 1 μg/ ml Hoechst 33342 (Sigma-Aldrich, Munich, Germany) was used to visualize the nuclei and 50 nM LysoTracker Red DND-99 was used to visualize the acidic compartments which correspond to the lysosomes. Both dyes were incubated for 10 min at 37˚C before acquisition of single images by a Leica DMI6000 B fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
HUVECs were seeded in collagen G-coated 24-well plates and grown to confluence for two days before treatment. The cells were incubated with indicated concentrations of archazolid for 24 h. Untreated MDA-MB-231 or PC-3 cells were labeled with CellTracker Green CMFDA Dye (5 μM in serum-free DMEM, 37˚C) for 30 min before 100,000 cells per well were added to HUVECs and were allowed to adhere for various time points at 37˚C. Non-adherent tumor cells were washed off three times with PBS containing Ca 2+ and Mg 2+ . Tumor cell adhesion was determined by fluorescence measurements with an Infinite F200 pro microplate reader (Tecan) at 485 nm (excitation) and 535 nm (emission).
For blocking the integrin β1 subunit on MDA-MB-231 or PC-3 cells, CellTracker Greenlabeled MDA-MB-231 or PC-3 cells were incubated with an anti-integrin β1 antibody (P5D2, ab24693, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg antibody per one million cells in 1 ml DMEM. Before adding to archazolid-treated HUVECs, MDA-MB-231 or PC-3 cells were washed once with DMEM. For blocking the integrin β1 subunit on HUVECs, the cells were incubated with the anti-integrin β1 antibody (0.1 μg/well in ECGM). HUVECs were washed once with ECGM before untreated MDA-MB-231 or PC-3 cells were added to HUVECs.
For the adhesion of MDA-MB-231 or PC-3 cells onto extracellular matrix (ECM) components 24-well plates were coated with collagen G (10 μg/ml in PBS), human plasma fibronectin (10 μg/ml PBS) or laminin-411 (10 μg/ml in Dulbecco's PBS [DPBS] containing Ca 2+ and Mg 2+ ) at 4˚C overnight. The adhesion of MDA-MB-231 and PC-3 cells onto these three most prominent ECM components was carried out as described above (10 min adhesion at 37˚C).
HUVECs were grown on a porous filter membrane (Transwell insert, polycarbonate membrane, 8 μm pores; Corning, New York, USA) for 48 h and were treated as indicated. Untreated MDA-MB-231 cells were labeled with CellTracker Green CMFDA Dye (as described in the section cell adhesion assay) and resuspended in medium 199 (PAN-Biotech) containing 0.1% BSA. HUVECs were washed twice with medium 199 containing 0.1% BSA before MDA-MB-231 cells were allowed to transmigrate through the endothelial monolayer for 24 h. Medium 199 containing 0.1% BSA was used as negative control and medium 199 containing 20% FCS was used as chemoattractant for transmigration in the lower compartment. Non-migrated cells remaining in the upper compartment were carefully removed using a cotton swab. Transmigrated cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and transmigration was quantified by measuring the fluorescence signal at 485 nm (excitation) and 535 nm (emission).
HUVECs were grown to confluence on 6-well plates before they were treated with archazolid for 12 h. The cells were induced to upregulate the gene expression of cell adhesion molecules by TNFα. RNA was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) according to the manufacturer's protocol. On-column DNase digestion was performed to remove genomic DNA. RNA was transcribed into cDNA by Superscript II (Life Technologies, Thermo Fisher Scientific). qPCR experiments were performed using a StepOnePlus System (Applied Biosystems, Thermo Fisher Scientific) and data was analyzed by the StepOne and Ste-pOnePlus Software v2.3. Power SYBR Green PCR Master Mix (Life Technologies) and the comparative C T quantitation method (2 -ΔΔCT ) were used.
HUVECs were grown to confluence on 12-well plates before they were treated with archazolid for 24 h. Cells were treated with TNFα for 24 h to induce the expression of cell adhesion molecules. Subsequently, the cells were detached with HyClone HyQTase (GE Healthcare, Freiburg, Germany). In the case of ICAM-1 the detached cells were fixed with 4% formaldehyde (Polysciences, Hirschberg an der Bergstraße, Germany) in PBS for 10 min and washed once with PBS before incubating with the fluorescein isothiocyanate (FITC)-labeled anti-human CD54 (mouse, ICAM-1) antibody (MCA1615F, Biozol, Eching, Germany) at room temperature for 45 min. For all other proteins, the cells were not fixed and washed once with PBS before incubating with the antibodies phycoerythrin (PE)-labeled anti-human CD106 (mouse, VCAM-1), PE-labeled anti-human CD62E (mouse, E-selectin) and PE-labeled anti-human CD325 (mouse, N-cadherin) from Becton Dickinson on ice for 45 min. These antibodies were diluted in PBS containing 0.2% BSA. The surface expression of cell adhesion molecules was measured by flow cytometry (FACSVerse, Becton Dickinson, Heidelberg, Germany).
To stain the surface collagen on HUVECs, cells were incubated with an anti-human collagen antibody (rabbit, 1:40, ab36064, Abcam) on ice for 30 min. The staining procedure was performed on ice to ensure that surface proteins or antibodies are not endocytosed. The cells were washed once with PBS containing Ca 2+ and Mg 2+ before they were fixed with Roti-Histofix (Carl Roth). Alexa Fluor 488-conjugated anti-rabbit antibody (goat, 1:400, A11008, Life Technologies) was used as secondary antibody and Hoechst 33342 (1 μg/ml, Sigma-Aldrich) was used to visualize nuclei.
Confluent HUVECs in 6-well plates were treated as indicated. Cells were washed with ice-cold PBS and lysed with RIPA buffer supplemented with protease inhibitors (Complete Mini EDTA-free; Roche, Mannheim, Germany), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride, β-glycerophosphate, sodium pyrophosphate and H 2 O 2 . Protein determination was performed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins (10-20 μg) were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad Laboratories, Munich, Germany). Separated proteins were transferred onto polyvinylidene difluoride membranes by tank blotting (Bio-Rad Laboratories) for immunodetection. Membranes were blocked with 5% boltinggrade milk powder (Carl Roth) in TBS containing 0.1% Tween 20 (Sigma-Aldrich). The following antibodies were used: mouse anti-human cathepsin B antibody (IM27L, Merck) (1:500), mouse anti-β-actin-peroxidase antibody (A3854, Sigma-Aldrich) (1:100,000) and antimouse IgG horse radish peroxidase (HRP)-linked antibody (7076, Cell Signaling, Frankfurt, Germany) (1:5,000). ImageJ version 1.49m was used for densitometric analysis.
Cathepsin B activity assay was performed as described in the publication by Kubisch et al. [28] . Confluent HUVECs or HMEC-1 seeded in 6-well plates were treated as indicated. Cells were washed with PBS and lysed with the non-denaturating M-PER mammalian protein extraction reagent (78501, Thermo Fisher Scientific) supplemented with protease inhibitors (Complete Mini EDTA-free, Roche), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride. The fluorogenic cathepsin B substrate Z-Arg-Arg-7-amido-4-methylcoumarin hydrochloride (C5429, Sigma-Aldrich) was added to 30 μg of the cell lysate diluted in assay buffer supplemented with 2 mM L-cysteine (C7880, Sigma-Aldrich) and incubated for 30 min at 40˚C. Fluorescence was measured at 348 nm (excitation) and 440 nm (emission) with a microplate reader (Varioskan Flash, Thermo Fisher Scientific). The intensity of the fluorescence signal corresponded to the cathepsin B enzyme activity. For background subtraction the cathepsin B inhibitor CA-074Me (Enzo Life Sciences, Lörrach, Germany) was added to an additional reaction.
The HUVEC Nucleofector Kit (Lonza, Cologne, Germany) was used to transfect HUVECs. The transfection was performed according to the manufacturer's protocol using 2.5 μg plasmid DNA for 500,000 cells (Nucleofector 2b Device, Lonza). 48 h after transfection the cells were treated for further experiments. The addgene plasmid #11249 hCathepsin B was kindly provided by Hyeryun Choe [29] . hCathepsin B was digested with PmeI and XbaI and the linear DNA fragment not corresponding to the human CTSB gene was religated to generate the empty pcDNA3.1 (-) delta MCS plasmid that was used for control transfections. The original backbone of hCathepsin B is the pcDNA3.1 (-) from Thermo Fisher Scientific. The control vector pcDNA3.1 (-) delta MCS used for our transfections was cloned on the basis of hCathepsin B and is therefore lacking almost the whole part of the multiple cloning site of the pcDNA3.1 (-).
Statistical analyses were performed using GraphPad Prism 5.0 (San Diego, USA). One-way ANOVA followed by Tukey's post-hoc test or unpaired t-test was used for the evaluation of a minimum of three independent experiments. The numbers of independently performed experiments (n) are stated in the corresponding figure legends. p 0.05 was considered as statistically significant. Data are expressed as mean ± standard error of the mean (SEM).
Since the v-ATPase inhibitor archazolid has never been used for studies in endothelial cells, we first performed cytotoxicity assays. We treated confluent HUVECs with up to 1 nM archazolid for 24 and 48 h and observed that this treatment has neither an influence on the metabolic activity nor on the release of LDH after 24 h (Fig 1A and 1B, left panels) . The metabolic activity and the release of LDH were only slightly affected by the highest concentration of archazolid after 48 h (Fig 1A and 1B, right panels) . Consequently, the following experiments were all carried out after 24 h (or less) of archazolid treatment in order to exclude any cytotoxic effects of archazolid within our experimental settings.
Microscopic analysis revealed that also the integrity of the endothelial monolayer was not affected by archazolid, but the cells showed a slightly different morphology (Fig 2A) : Archazolid-treated cells were swollen compared to control cells, which was not unexpected, as vacuolation of the endoplasmic reticulum (ER) has been described for other cell types and is typical for v-ATPase inhibitors [11, 16, 24, 30] . This effect was obvious both in subconfluent and in confluent cells (Fig 2A) . Inhibition of v-ATPase prevents the acidification of lysosomes [1, 31] . Using the cell-permeable dye LysoTracker Red DND-99, it is possible to label the acidic lysosomes in living cells. Thus, this dye can serve as an indicator of v-ATPase inhibition. To proof that archazolid is also functionally active as a v-ATPase inhibitor in HUVECs, cells were treated with 1 nM archazolid before they were incubated with LysoTracker Red DND-99 and Hoechst 33342. As shown in Fig 2B, the red vesicular staining corresponding to acidified lysosomes in control cells disappeared completely after treatment with archazolid. In summary, archazolid treatment for 24 h was not cytotoxic to quiescent HUVECs, but inhibited the functionality of the v-ATPase.
We analyzed the adhesion of MDA-MB-231 cells onto HUVECs. Confluent HUVECs were treated with up to 1 nM archazolid for 24 h. Untreated MDA-MB-231 cells were labeled with Cell-Tracker Green CMFDA Dye. Interestingly, v-ATPase inhibition strongly increased the attachment of the metastatic breast carcinoma cell line MDA-MB-231 onto HUVECs after 10 and 120 min of adhesion (Fig 3A and 3B) . We also investigated the influence of archazolid on the transendothelial migration of MDA-MB-231 cells. HUVECs seeded in a Boyden chamber were treated with 1 nM archazolid for 24 h. CellTracker Green-labeled MDA-MB-231 cells (not treated with archazolid) were allowed to transmigrate through the endothelial monolayer for 24 h. As shown in Fig 3C, archazolid significantly decreased the transendothelial migration of MDA-MB-231 cells.
The influence of archazolid on tumor cell adhesion was not only studied in HUVECs, which represent macrovascular endothelial cells, but also in microvascular HMEC-1 cells. Moreover, besides the breast cancer cell line MDA-MB-231, also PC-3 prostate cancer cells were used as a second metastatic cancer cell line. Archazolid treatment of endothelial cells increased the attachment of MDA-MB-231 cells onto the HMEC-1 monolayer after 120 min of adhesion ( Fig 4A) and increased the attachment of PC-3 cells onto the HUVEC monolayer after 30 and 60 min of adhesion (Fig 4B) . Of note, the adhesion of non-metastatic Jurkat cells, an acute T cell leukemia cell line, remained unaffected after treatment of HUVECs with archazolid (S1A Fig). Taken together, archazolid treatment augmented the adhesive properties of both micro-and macrovascular endothelial cells for metastatic tumor cells, but not for non-metastatic ones. Of note, cancer cell adhesion onto archazolid-activated endothelial cells increased with the time of adhesion.
The adhesion of tumor cells onto the endothelium is in principle similar to that of leukocytes, but slightly differs in the molecules that mediate the adhesion process. Ligands for the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin were found to be expressed on tumor cells and to mediate tumor-endothelial cell interaction [21] . Inhibition of the v-ATPase might affect the expression of endothelial cell adhesion molecules on mRNA or protein levels. To determine the mRNA expression of ICAM-1, VCAM-1, E-selectin and Ncadherin, HUVECs were treated with archazolid for 12 h. TNFα is known to upregulate the expression of ICAM-1, VCAM-1 and E-selectin [32] and, thus, served as positive control. Quantitative real-time PCR showed that v-ATPase inhibition in HUVECs did not alter the mRNA levels of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5A) . The protein expression of these adhesion molecules on the surface of endothelial cells was analyzed by flow cytometry. Archazolid (1 nM, 24 h) did not affect the cell surface expression of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5B) .
Besides ICAM-1, VCAM-1, E-selectin and N-cadherin, also integrins are able to mediate the process of cell adhesion [33] [34] [35] . Since none of the cell adhesion molecules expressed on HUVECs were regulated upon archazolid treatment, we considered integrins as potential interaction partners. Within this protein family β1-integrins have been reported to mediate tumor cell adhesion onto quiescent endothelial cells [25] . In order to elucidate the role of β1-integrins for the archazolid-induced tumor cell adhesion, the integrin β1-subunit was blocked either on MDA-MB-231 cells, PC-3 cells or on HUVECs. (Of note, as in all experiments throughout this study, only endothelial cells were treated with archazolid.) After blocking β1-integrins on MDA-MB-231 or PC-3 cells, the archazolid-induced tumor cell adhesion was reduced almost to control level (Fig 6A and 6B , left panels), whereas blocking of β1-integrins on HUVECs had no significant effect on the increase of tumor cell adhesion by v-ATPase inhibition (Fig 6A and 6B , right panels).
Depending on their α subunit, β1-integrins have a variety of ligands including extracellular matrix (ECM) components such as collagen, fibronectin and laminin [35] . Therefore, we hypothesized that archazolid treatment of endothelial cells might lead to an upregulation of these components. MDA-MB-231 and PC-3 cells were allowed to adhere onto plastic that was coated with these ECM components. This cell adhesion assay revealed that MDA-MB-231 as well as PC-3 cells favor the interaction with the ECM component collagen, as the adhesion onto collagen is much higher than onto the uncoated plastic control (Fig 7A) . MDA-MB-231 and PC-3 cells also adhered to fibronectin-coated plastic, but to a much lesser extent compared to the collagen coating. Therefore, we focused on the interaction between these two tumor cell lines and collagen. Blocking of the integrin β1 subunit on MDA-MB-231 and PC-3 cells clearly abolished the interaction with collagen (Fig 7B) , indicating that the attachment of these tumor cells to collagen is mediated by β1-integrins.
Since collagen is the major ECM component MDA-MB-231 and PC-3 cells interact with, the next step was to prove whether v-ATPase inhibition influences the amount of collagen expressed by HUVECs as extracellular matrix. To detect collagen on the endothelial surface, archazolid-treated HUVECs were labeled with an antibody against collagen type I-IV on ice to prevent endocytosis and to ensure that the antibody does not bind to intracellular collagen. Interestingly, archazolid increased the amount of surface collagen on HUVECs by about 50% (Fig 7C) . Control stainings were performed using an antibody against the cytosolic p65 subunit of the transcription factor nuclear factor κB (NFκB) to show that intracellular proteins were not detected by this staining method (S2 Fig) .
It was reported that v-ATPase inhibition by archazolid impairs the activity of cathepsin B [28, 36] , a lysosomal enzyme that degrades extracellular matrix components including collagen [37] [38] [39] [40] [41] . As collagen is degraded by cathepsin B and the activation of cathepsin B depends on v-ATPase activity [28, [36] [37] [38] 42] , we suggested that an accumulation of collagen on the surface of endothelial cells might be a consequence of an impaired functionality of cathepsin B. Therefore, an enzyme activity assay based on the proteolysis of a fluorogenic cathepsin B substrate was performed. In archazolid-treated HUVECs and HMEC-1 the activity of cathepsin B was induce both the mRNA (1 ng/ml TNF) and the cell surface expression (10 ng/ml TNF) of ICAM-1, VCAM-1, E-selectin and Ncadherin.
https://doi.org/10.1371/journal.pone.0203053.g005 Inhibition of endothelial vATPase increases tumor cell adhesion to endothelial cells strongly decreased by approximately 50% compared to control cells at an archazolid concentration of 1 nM (Fig 8A) . In line with this result, western blot analysis showed that archazolid (1 nM) reduces the protein expression of the mature, active form of cathepsin B to less than 40% of the control in HUVECs (Fig 8B) . To proof whether the archazolid-induced tumor cell adhesion is a consequence of the decreased amount of cathepsin B, HUVECs were transfected with a plasmid coding for human cathepsin B or with the empty vector as control. After 48 h, the transfected cells were treated with 1 nM archazolid. The level of cathepsin B after transfection and treatment was assessed by western blot analysis (Fig 9A) . Overexpression of cathepsin B strongly diminished both the basal and the archazolid-induced adhesion of MDA-MB-231 cells (Fig 9B) .
Targeting the proton pump v-ATPase for cancer therapy has gained great interest since its inhibition was reported to reduce the invasiveness of cancer cells and, most importantly, also metastasis [8, 9] . Thus, intensive research related to v-ATPases was done in cancer cells, whereas there are only few studies investigating v-ATPases in endothelial cells indicating a role in migration, proliferation and possibly angiogenesis [22] [23] [24] . In the present study we used the myxobacterial natural product archazolid to investigate the consequences of v-ATPase inhibition in the endothelium on tumor-endothelial cell interactions.
For the first time, we were able to show a link between v-ATPase and the adhesion and transmigration properties of the endothelium. Inhibition of the v-ATPase in endothelial cells by archazolid significantly increased the adhesion of metastatic cancer cells and decreased the transendothelial migration of cancer cells which was attributed to augmented collagen levels on the surface on archazolid-treated endothelial cells. Of note, adhesion of the non-metastatic Jurkat cell line onto archazolid-treated endothelial cells remained unaffected. The archazolidinduced adhesion of tumor cells was independent from the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin, as their expression was not regulated by the compound. However, we found that the archazolid-induced tumor cell adhesion was mediated by β1-integrins expressed on MDA-MB-231 breast cancer and PC-3 prostate cancer cells as blocking of the integrin β1 subunit on these tumor cells reversed the pro-adhesive effect of archazolid. In adhesion experiments on plastic coated with extracellular matrix components, we could show that MDA-MB-231 and PC-3 cells clearly favored the interaction with collagen, whereas the adhesion of non-metastatic Jurkat cells was largely independent from extracellular matrix proteins (S1B Fig). The different adhesion properties of metastatic cancer cells and Jurkat cells might be a result of the distinct integrin expression pattern of each cell line. MDA-MB-231 and PC-3 cells express α2β1-and α3β1-integrins, which represent collagen receptors [43, 44] , while Jurkat cells express α4β1-integrins but lack α2β1-, α3β1-integrins [44] . α4β1integrins are receptors for VCAM-1 and fibronectin [35] and it has been shown that Jurkat cells interact with human endothelial cells that express VCAM-1 after cytokine treatment or cells transfected with VCAM-1 [45] . Our results are in line with previous studies showing that α2β1-and α3β1-integrin expressing MDA-MB-231 and PC-3 cells were able to rapidly attach to collagen in the cortical bone matrix. In contrast, Jurkat cells were not able to adhere [44] and might preferentially interact with cell adhesion molecules rather than with ECM proteins. α2β1-and α3β1-integrins can additionally act as laminin receptors [46] and at least α3β1integrins recognize fibronectin [46, 47] . Though expressing receptors for fibronectin and laminin, MDA-MB-231 and PC-3 cells adhered to fibronectin to a much lesser extent and did not adhere to laminin, probably due to lower affinities to these extracellular matrix components.
Importantly, v-ATPase inhibition by archazolid increased the surface levels of the extracellular matrix component collagen, which might explain that the increase of MDA-MB-231 and PC-3 cells onto archazolid-treated HUVECs is independent of endothelial cell adhesion molecules. By performing a live cell proteolysis assay, Cavallo-Medved et al. demonstrated ECM degradation, in particular of gelatin and collagen IV, in association with active cathepsin B in caveolae of endothelial cells during tube formation [40] . In addition, recent studies reported that v-ATPase inhibition impairs the activity of cathepsin B in cancer cells [28, 36] . Therefore, we suggested that the accumulation of collagen on the endothelial surface might be a consequence of impaired cathepsin B activity or expression in endothelial cells. In fact, we confirmed the impairment of cathepsin B activity by archazolid as the expression levels of the mature active form of this enzyme was strongly reduced. Cathepsin B is synthesized as preprocathepsin B on membrane-bound ribosomes. Following transport to the Golgi apparatus, the preprocathepsin B is glycosylated with mannose-containing oligosaccharides. The targeting of procathepsin B to lysosomes is mannose-6-phosphate receptor-dependent and its dissociation from the receptor as well as its proteolytic processing into mature cathepsin B requires acidification of the compartment [48] . In cancer cells v-ATPase inhibition by archazolid impaired the mannose-6-phosphate receptor-mediated trafficking from the trans-Golgi network to prelysosomal compartments resulting in a decrease of active lysosomal proteases like cathepsin B [28] . We assumed that the archazolid-induced decrease in cathepsin B activity and expression was based on the same mechanism. Interestingly, overexpression of cathepsin B attenuated the archazolid-induced adhesion of breast cancer cells onto endothelial cells, indicating that the adhesion negatively correlates with the expression of cathepsin B.
As cathepsin B can also degrade other extracellular matrix components such as fibronectin and laminin [38, 49] , v-ATPase inhibition could lead to an accumulation of these proteins and an increased adhesion of cells expressing fibronectin or laminin receptors. However, we did not focus on these ECM components since they were not relevant for the adhesion of MDA-MB-231 and PC-3 cells. These cells predominantly adhered to collagen, while the adhesion of Jurkat cells is mostly independent from the ECM proteins collagen, fibronectin or laminin (S1B Fig). Interestingly [50] . In hepatic cancer cells, archazolid reduces Ras/Raf/MEK/ERK signaling by altering the membrane composition and fluidity [51] . We assume that archazolid affects endothelial cells in a similar way leading to inhibition of Ras signaling and, therefore, reduced transendothelial migration of MDA-MB-231 cells.
Taken together, our study shows that archazolid reduces the activity and expression of cathepsin B in endothelial cells. As a result, the amount of collagen on the surface of endothelial cells was significantly upregulated, which finally resulted in an increased adhesion of the β1-integrin-expressing metastatic cancer cell lines MDA-MB-231 and PC-3 onto archazolidtreated endothelial cells, whereas the adhesion of non-metastatic Jurkat cells was unaffected. This study shows that the v-ATPase plays an important role in regulating the adhesion of cells expressing receptors for extracellular matrix components. Archazolid represents a promising tool to elucidate the role of v-ATPase in endothelial cells. Moreover, we for the first time linked the function of v-ATPase to the adhesion and transmigration of tumor cells onto endothelial cells as well as to the remodeling of the extracellular matrix on the surface of endothelial cells. The fact that the adhesion of metastatic tumor cells onto endothelial cells is increased while their transendothelial migration is reduced upon inhibition of endothelial v-ATPase by archazolid further supports the view of archazolid as a potential anti-metastatic compound.
|
What is archazolid?
| 5,302 |
v-ATPase inhibitor
| 775 |
1,633 |
The vacuolar-type ATPase inhibitor archazolid increases tumor cell adhesion to endothelial cells by accumulating extracellular collagen
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6133348/
SHA: f1b81916fac1ca3d50dde774df2e1bb26bf0fb39
Authors: Luong, Betty; Schwenk, Rebecca; Bräutigam, Jacqueline; Müller, Rolf; Menche, Dirk; Bischoff, Iris; Fürst, Robert
Date: 2018-09-11
DOI: 10.1371/journal.pone.0203053
License: cc-by
Abstract: The vacuolar-type H(+)-ATPase (v-ATPase) is the major proton pump that acidifies intracellular compartments of eukaryotic cells. Since the inhibition of v-ATPase resulted in anti-tumor and anti-metastatic effects in different tumor models, this enzyme has emerged as promising strategy against cancer. Here, we used the well-established v-ATPase inhibitor archazolid, a natural product first isolated from the myxobacterium Archangium gephyra, to study the consequences of v-ATPase inhibition in endothelial cells (ECs), in particular on the interaction between ECs and cancer cells, which has been neglected so far. Human endothelial cells treated with archazolid showed an increased adhesion of tumor cells, whereas the transendothelial migration of tumor cells was reduced. The adhesion process was independent from the EC adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin. Instead, the adhesion was mediated by β1-integrins expressed on tumor cells, as blocking of the integrin β1 subunit reversed this process. Tumor cells preferentially adhered to the β1-integrin ligand collagen and archazolid led to an increase in the amount of collagen on the surface of ECs. The accumulation of collagen was accompanied by a strong decrease of the expression and activity of the protease cathepsin B. Overexpression of cathepsin B in ECs prevented the capability of archazolid to increase the adhesion of tumor cells onto ECs. Our study demonstrates that the inhibition of v-ATPase by archazolid induces a pro-adhesive phenotype in endothelial cells that promotes their interaction with cancer cells, whereas the transmigration of tumor cells was reduced. These findings further support archazolid as a promising anti-metastatic compound.
Text: The vacuolar-type H + -ATPase (v-ATPase) is the major proton pump responsible for acidification of intracellular compartments in eukaryotic cells [1] . The enzyme consists of two multi-subunit complexes, the soluble V 1 transmembrane V o subcomplex required for the proton transport across membranes [1, 2] . In most cell types v-ATPases are only expressed in the endomembrane system to regulate and maintain the acidic pH of intracellular compartments such as lysosomes, endosomes, the Golgi apparatus, secretory granules and coated vesicles [3] . The function of v-ATPases is essential for cellular processes such as vesicular trafficking, receptor-mediated endocytosis and protein degradation and processing. In specialized cell types including osteoclasts and renal epithelial cells, v-ATPases can also be expressed on the plasma membrane, where they pump protons into the extracellular space [2] [3] [4] . In cancer cells v-ATPases are expressed on the plasma membrane in order to eliminate toxic cytosolic H + . Most importantly, v-ATPases contribute to the acidic tumor microenvironment, which leads to the activation of proteases, thus facilitating tumor cell migration, invasion and angiogenesis [5] [6] [7] . Since the inhibition of v-ATPase was shown to reduce the invasiveness of cancer cells and metastasis formation [8, 9] , this enzyme has emerged as a promising drug target in the recent years. Archazolid A and B are highly potent and specific inhibitors of v-ATPases [10] . They were first isolated from the myxobacterium Archangium gephyra [11] . These compounds inhibit v-ATPase at low nanomolar concentrations [10, 12] by binding to the subunit c of the V o complex. As their biological activity is comparable to the v-ATPase inhibitors bafilomycin and concanamycin [10, 11] , archazolids are natural compounds of high interest that can be used both as a tool to study the consequences of v-ATPase inhibition and as a lead for drug development. Archazolids can be either produced by fermentation [11] or by total synthesis [13, 14] .
In the field of cancer research several studies reported on interesting pharmacological effects of archazolid: It reduced the migration of different invasive tumor cells in vitro and cancer cell metastasis in vivo in a breast tumor mouse model [15] . Furthermore, archazolid activated pathways of cellular stress response and apoptosis in highly invasive tumor cells [16] . In classically activated macrophages, archazolid selectively induced the generation of tumor necrosis factor α (TNFα), which may indirectly promote tumor suppression [17] .
Up to now, the role of v-ATPases in endothelial cells has only rarely been investigated. The endothelium plays a crucial role in the pathogenesis and progression of cancer: The metastatic cascade includes local angiogenesis at the site of the primary tumor and adhesion of tumor cells at the site of metastasis [18] . Angiogenesis, the development of new blood vessels out of existing ones, depends on the proliferation, migration and differentiation of endothelial cells [19] . This process ensures the nutrient supply of the tumor and its growth [20] . Circulating cancer cells can adhere to the endothelium at distant sites. This adhesive interaction is mediated by receptors and corresponding ligands expressed on tumor and endothelial cells [18, 21] . V-ATPases have been reported to regulate intracellular pH and cell migration in microvascular endothelial cells [22, 23] . A recent study showed that the inhibition of v-ATPase by concanamycin prevented proliferation, reduced migration and impaired angiogenesis-related signaling in endothelial cells [24] . So far, there are no investigations on the role of endothelial v-ATPases for the process of tumor cell adhesion onto the endothelium. Thus, we were interested in the consequences of the inhibition of endothelial v-ATPase by archazolid on the interaction between endothelial and cancer cells. Various cell adhesion molecules on the endothelium, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion protein (VCAM-1), E-selectin or N-cadherin [21] as well as integrins expressed on cancer cells have been reported to mediate cell adhesion of cancer cells onto endothelial cells [25] [26] [27] . Accordingly, we focused on these cell adhesion molecules and integrins. For the first time, our study revealed a link between the function of v-ATPases and the adhesion and transmigration properties of endothelial cells.
CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany) was performed according to the manufacturer's protocol for determining the cell viability of cells after treatment with archazolid. This assay is based on the ability of metabolically active cells to reduce resazurin which results in fluorescent resorufin. The CellTiter-Blue Reagent was added to the cells 4 h before the endpoint of treatment. Fluorescence was measured with an Infinite F200 pro microplate reader (Tecan, Männedorf, Switzerland) at 560 nm (excitation) and 590 nm (emission).
CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) was performed according to the manufacturer's instructions for determining the lactate dehydrogenase (LDH) release after treatment with archazolid. Lysis buffer was added to untreated cells 45 min before the end of treatment to induce the release of this enzyme. LDH is a cytosolic enzyme that is released by leaky cells. Released LDH catalyzes the enzymatic conversion of lactate to pyruvate which provides NADH for the conversion of iodonitrotetrazolium violet into a red formazan product in the presence of diaphorase. The absorbance was measured with a Varioskan Flash microplate reader (Thermo Fisher Scientific) at 490 nm.
LysoTracker Red DND-99 (Life Technologies, Thermo Fisher Scientific) is a dye to measure pH values in viable cells. HUVECs were cultured to confluence on collagen G-coated μ-slides (80826, ibidi, Martinsried, Germany) before they were treated with archazolid for 24 h. 1 μg/ ml Hoechst 33342 (Sigma-Aldrich, Munich, Germany) was used to visualize the nuclei and 50 nM LysoTracker Red DND-99 was used to visualize the acidic compartments which correspond to the lysosomes. Both dyes were incubated for 10 min at 37˚C before acquisition of single images by a Leica DMI6000 B fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
HUVECs were seeded in collagen G-coated 24-well plates and grown to confluence for two days before treatment. The cells were incubated with indicated concentrations of archazolid for 24 h. Untreated MDA-MB-231 or PC-3 cells were labeled with CellTracker Green CMFDA Dye (5 μM in serum-free DMEM, 37˚C) for 30 min before 100,000 cells per well were added to HUVECs and were allowed to adhere for various time points at 37˚C. Non-adherent tumor cells were washed off three times with PBS containing Ca 2+ and Mg 2+ . Tumor cell adhesion was determined by fluorescence measurements with an Infinite F200 pro microplate reader (Tecan) at 485 nm (excitation) and 535 nm (emission).
For blocking the integrin β1 subunit on MDA-MB-231 or PC-3 cells, CellTracker Greenlabeled MDA-MB-231 or PC-3 cells were incubated with an anti-integrin β1 antibody (P5D2, ab24693, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg antibody per one million cells in 1 ml DMEM. Before adding to archazolid-treated HUVECs, MDA-MB-231 or PC-3 cells were washed once with DMEM. For blocking the integrin β1 subunit on HUVECs, the cells were incubated with the anti-integrin β1 antibody (0.1 μg/well in ECGM). HUVECs were washed once with ECGM before untreated MDA-MB-231 or PC-3 cells were added to HUVECs.
For the adhesion of MDA-MB-231 or PC-3 cells onto extracellular matrix (ECM) components 24-well plates were coated with collagen G (10 μg/ml in PBS), human plasma fibronectin (10 μg/ml PBS) or laminin-411 (10 μg/ml in Dulbecco's PBS [DPBS] containing Ca 2+ and Mg 2+ ) at 4˚C overnight. The adhesion of MDA-MB-231 and PC-3 cells onto these three most prominent ECM components was carried out as described above (10 min adhesion at 37˚C).
HUVECs were grown on a porous filter membrane (Transwell insert, polycarbonate membrane, 8 μm pores; Corning, New York, USA) for 48 h and were treated as indicated. Untreated MDA-MB-231 cells were labeled with CellTracker Green CMFDA Dye (as described in the section cell adhesion assay) and resuspended in medium 199 (PAN-Biotech) containing 0.1% BSA. HUVECs were washed twice with medium 199 containing 0.1% BSA before MDA-MB-231 cells were allowed to transmigrate through the endothelial monolayer for 24 h. Medium 199 containing 0.1% BSA was used as negative control and medium 199 containing 20% FCS was used as chemoattractant for transmigration in the lower compartment. Non-migrated cells remaining in the upper compartment were carefully removed using a cotton swab. Transmigrated cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and transmigration was quantified by measuring the fluorescence signal at 485 nm (excitation) and 535 nm (emission).
HUVECs were grown to confluence on 6-well plates before they were treated with archazolid for 12 h. The cells were induced to upregulate the gene expression of cell adhesion molecules by TNFα. RNA was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) according to the manufacturer's protocol. On-column DNase digestion was performed to remove genomic DNA. RNA was transcribed into cDNA by Superscript II (Life Technologies, Thermo Fisher Scientific). qPCR experiments were performed using a StepOnePlus System (Applied Biosystems, Thermo Fisher Scientific) and data was analyzed by the StepOne and Ste-pOnePlus Software v2.3. Power SYBR Green PCR Master Mix (Life Technologies) and the comparative C T quantitation method (2 -ΔΔCT ) were used.
HUVECs were grown to confluence on 12-well plates before they were treated with archazolid for 24 h. Cells were treated with TNFα for 24 h to induce the expression of cell adhesion molecules. Subsequently, the cells were detached with HyClone HyQTase (GE Healthcare, Freiburg, Germany). In the case of ICAM-1 the detached cells were fixed with 4% formaldehyde (Polysciences, Hirschberg an der Bergstraße, Germany) in PBS for 10 min and washed once with PBS before incubating with the fluorescein isothiocyanate (FITC)-labeled anti-human CD54 (mouse, ICAM-1) antibody (MCA1615F, Biozol, Eching, Germany) at room temperature for 45 min. For all other proteins, the cells were not fixed and washed once with PBS before incubating with the antibodies phycoerythrin (PE)-labeled anti-human CD106 (mouse, VCAM-1), PE-labeled anti-human CD62E (mouse, E-selectin) and PE-labeled anti-human CD325 (mouse, N-cadherin) from Becton Dickinson on ice for 45 min. These antibodies were diluted in PBS containing 0.2% BSA. The surface expression of cell adhesion molecules was measured by flow cytometry (FACSVerse, Becton Dickinson, Heidelberg, Germany).
To stain the surface collagen on HUVECs, cells were incubated with an anti-human collagen antibody (rabbit, 1:40, ab36064, Abcam) on ice for 30 min. The staining procedure was performed on ice to ensure that surface proteins or antibodies are not endocytosed. The cells were washed once with PBS containing Ca 2+ and Mg 2+ before they were fixed with Roti-Histofix (Carl Roth). Alexa Fluor 488-conjugated anti-rabbit antibody (goat, 1:400, A11008, Life Technologies) was used as secondary antibody and Hoechst 33342 (1 μg/ml, Sigma-Aldrich) was used to visualize nuclei.
Confluent HUVECs in 6-well plates were treated as indicated. Cells were washed with ice-cold PBS and lysed with RIPA buffer supplemented with protease inhibitors (Complete Mini EDTA-free; Roche, Mannheim, Germany), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride, β-glycerophosphate, sodium pyrophosphate and H 2 O 2 . Protein determination was performed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins (10-20 μg) were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad Laboratories, Munich, Germany). Separated proteins were transferred onto polyvinylidene difluoride membranes by tank blotting (Bio-Rad Laboratories) for immunodetection. Membranes were blocked with 5% boltinggrade milk powder (Carl Roth) in TBS containing 0.1% Tween 20 (Sigma-Aldrich). The following antibodies were used: mouse anti-human cathepsin B antibody (IM27L, Merck) (1:500), mouse anti-β-actin-peroxidase antibody (A3854, Sigma-Aldrich) (1:100,000) and antimouse IgG horse radish peroxidase (HRP)-linked antibody (7076, Cell Signaling, Frankfurt, Germany) (1:5,000). ImageJ version 1.49m was used for densitometric analysis.
Cathepsin B activity assay was performed as described in the publication by Kubisch et al. [28] . Confluent HUVECs or HMEC-1 seeded in 6-well plates were treated as indicated. Cells were washed with PBS and lysed with the non-denaturating M-PER mammalian protein extraction reagent (78501, Thermo Fisher Scientific) supplemented with protease inhibitors (Complete Mini EDTA-free, Roche), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride. The fluorogenic cathepsin B substrate Z-Arg-Arg-7-amido-4-methylcoumarin hydrochloride (C5429, Sigma-Aldrich) was added to 30 μg of the cell lysate diluted in assay buffer supplemented with 2 mM L-cysteine (C7880, Sigma-Aldrich) and incubated for 30 min at 40˚C. Fluorescence was measured at 348 nm (excitation) and 440 nm (emission) with a microplate reader (Varioskan Flash, Thermo Fisher Scientific). The intensity of the fluorescence signal corresponded to the cathepsin B enzyme activity. For background subtraction the cathepsin B inhibitor CA-074Me (Enzo Life Sciences, Lörrach, Germany) was added to an additional reaction.
The HUVEC Nucleofector Kit (Lonza, Cologne, Germany) was used to transfect HUVECs. The transfection was performed according to the manufacturer's protocol using 2.5 μg plasmid DNA for 500,000 cells (Nucleofector 2b Device, Lonza). 48 h after transfection the cells were treated for further experiments. The addgene plasmid #11249 hCathepsin B was kindly provided by Hyeryun Choe [29] . hCathepsin B was digested with PmeI and XbaI and the linear DNA fragment not corresponding to the human CTSB gene was religated to generate the empty pcDNA3.1 (-) delta MCS plasmid that was used for control transfections. The original backbone of hCathepsin B is the pcDNA3.1 (-) from Thermo Fisher Scientific. The control vector pcDNA3.1 (-) delta MCS used for our transfections was cloned on the basis of hCathepsin B and is therefore lacking almost the whole part of the multiple cloning site of the pcDNA3.1 (-).
Statistical analyses were performed using GraphPad Prism 5.0 (San Diego, USA). One-way ANOVA followed by Tukey's post-hoc test or unpaired t-test was used for the evaluation of a minimum of three independent experiments. The numbers of independently performed experiments (n) are stated in the corresponding figure legends. p 0.05 was considered as statistically significant. Data are expressed as mean ± standard error of the mean (SEM).
Since the v-ATPase inhibitor archazolid has never been used for studies in endothelial cells, we first performed cytotoxicity assays. We treated confluent HUVECs with up to 1 nM archazolid for 24 and 48 h and observed that this treatment has neither an influence on the metabolic activity nor on the release of LDH after 24 h (Fig 1A and 1B, left panels) . The metabolic activity and the release of LDH were only slightly affected by the highest concentration of archazolid after 48 h (Fig 1A and 1B, right panels) . Consequently, the following experiments were all carried out after 24 h (or less) of archazolid treatment in order to exclude any cytotoxic effects of archazolid within our experimental settings.
Microscopic analysis revealed that also the integrity of the endothelial monolayer was not affected by archazolid, but the cells showed a slightly different morphology (Fig 2A) : Archazolid-treated cells were swollen compared to control cells, which was not unexpected, as vacuolation of the endoplasmic reticulum (ER) has been described for other cell types and is typical for v-ATPase inhibitors [11, 16, 24, 30] . This effect was obvious both in subconfluent and in confluent cells (Fig 2A) . Inhibition of v-ATPase prevents the acidification of lysosomes [1, 31] . Using the cell-permeable dye LysoTracker Red DND-99, it is possible to label the acidic lysosomes in living cells. Thus, this dye can serve as an indicator of v-ATPase inhibition. To proof that archazolid is also functionally active as a v-ATPase inhibitor in HUVECs, cells were treated with 1 nM archazolid before they were incubated with LysoTracker Red DND-99 and Hoechst 33342. As shown in Fig 2B, the red vesicular staining corresponding to acidified lysosomes in control cells disappeared completely after treatment with archazolid. In summary, archazolid treatment for 24 h was not cytotoxic to quiescent HUVECs, but inhibited the functionality of the v-ATPase.
We analyzed the adhesion of MDA-MB-231 cells onto HUVECs. Confluent HUVECs were treated with up to 1 nM archazolid for 24 h. Untreated MDA-MB-231 cells were labeled with Cell-Tracker Green CMFDA Dye. Interestingly, v-ATPase inhibition strongly increased the attachment of the metastatic breast carcinoma cell line MDA-MB-231 onto HUVECs after 10 and 120 min of adhesion (Fig 3A and 3B) . We also investigated the influence of archazolid on the transendothelial migration of MDA-MB-231 cells. HUVECs seeded in a Boyden chamber were treated with 1 nM archazolid for 24 h. CellTracker Green-labeled MDA-MB-231 cells (not treated with archazolid) were allowed to transmigrate through the endothelial monolayer for 24 h. As shown in Fig 3C, archazolid significantly decreased the transendothelial migration of MDA-MB-231 cells.
The influence of archazolid on tumor cell adhesion was not only studied in HUVECs, which represent macrovascular endothelial cells, but also in microvascular HMEC-1 cells. Moreover, besides the breast cancer cell line MDA-MB-231, also PC-3 prostate cancer cells were used as a second metastatic cancer cell line. Archazolid treatment of endothelial cells increased the attachment of MDA-MB-231 cells onto the HMEC-1 monolayer after 120 min of adhesion ( Fig 4A) and increased the attachment of PC-3 cells onto the HUVEC monolayer after 30 and 60 min of adhesion (Fig 4B) . Of note, the adhesion of non-metastatic Jurkat cells, an acute T cell leukemia cell line, remained unaffected after treatment of HUVECs with archazolid (S1A Fig). Taken together, archazolid treatment augmented the adhesive properties of both micro-and macrovascular endothelial cells for metastatic tumor cells, but not for non-metastatic ones. Of note, cancer cell adhesion onto archazolid-activated endothelial cells increased with the time of adhesion.
The adhesion of tumor cells onto the endothelium is in principle similar to that of leukocytes, but slightly differs in the molecules that mediate the adhesion process. Ligands for the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin were found to be expressed on tumor cells and to mediate tumor-endothelial cell interaction [21] . Inhibition of the v-ATPase might affect the expression of endothelial cell adhesion molecules on mRNA or protein levels. To determine the mRNA expression of ICAM-1, VCAM-1, E-selectin and Ncadherin, HUVECs were treated with archazolid for 12 h. TNFα is known to upregulate the expression of ICAM-1, VCAM-1 and E-selectin [32] and, thus, served as positive control. Quantitative real-time PCR showed that v-ATPase inhibition in HUVECs did not alter the mRNA levels of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5A) . The protein expression of these adhesion molecules on the surface of endothelial cells was analyzed by flow cytometry. Archazolid (1 nM, 24 h) did not affect the cell surface expression of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5B) .
Besides ICAM-1, VCAM-1, E-selectin and N-cadherin, also integrins are able to mediate the process of cell adhesion [33] [34] [35] . Since none of the cell adhesion molecules expressed on HUVECs were regulated upon archazolid treatment, we considered integrins as potential interaction partners. Within this protein family β1-integrins have been reported to mediate tumor cell adhesion onto quiescent endothelial cells [25] . In order to elucidate the role of β1-integrins for the archazolid-induced tumor cell adhesion, the integrin β1-subunit was blocked either on MDA-MB-231 cells, PC-3 cells or on HUVECs. (Of note, as in all experiments throughout this study, only endothelial cells were treated with archazolid.) After blocking β1-integrins on MDA-MB-231 or PC-3 cells, the archazolid-induced tumor cell adhesion was reduced almost to control level (Fig 6A and 6B , left panels), whereas blocking of β1-integrins on HUVECs had no significant effect on the increase of tumor cell adhesion by v-ATPase inhibition (Fig 6A and 6B , right panels).
Depending on their α subunit, β1-integrins have a variety of ligands including extracellular matrix (ECM) components such as collagen, fibronectin and laminin [35] . Therefore, we hypothesized that archazolid treatment of endothelial cells might lead to an upregulation of these components. MDA-MB-231 and PC-3 cells were allowed to adhere onto plastic that was coated with these ECM components. This cell adhesion assay revealed that MDA-MB-231 as well as PC-3 cells favor the interaction with the ECM component collagen, as the adhesion onto collagen is much higher than onto the uncoated plastic control (Fig 7A) . MDA-MB-231 and PC-3 cells also adhered to fibronectin-coated plastic, but to a much lesser extent compared to the collagen coating. Therefore, we focused on the interaction between these two tumor cell lines and collagen. Blocking of the integrin β1 subunit on MDA-MB-231 and PC-3 cells clearly abolished the interaction with collagen (Fig 7B) , indicating that the attachment of these tumor cells to collagen is mediated by β1-integrins.
Since collagen is the major ECM component MDA-MB-231 and PC-3 cells interact with, the next step was to prove whether v-ATPase inhibition influences the amount of collagen expressed by HUVECs as extracellular matrix. To detect collagen on the endothelial surface, archazolid-treated HUVECs were labeled with an antibody against collagen type I-IV on ice to prevent endocytosis and to ensure that the antibody does not bind to intracellular collagen. Interestingly, archazolid increased the amount of surface collagen on HUVECs by about 50% (Fig 7C) . Control stainings were performed using an antibody against the cytosolic p65 subunit of the transcription factor nuclear factor κB (NFκB) to show that intracellular proteins were not detected by this staining method (S2 Fig) .
It was reported that v-ATPase inhibition by archazolid impairs the activity of cathepsin B [28, 36] , a lysosomal enzyme that degrades extracellular matrix components including collagen [37] [38] [39] [40] [41] . As collagen is degraded by cathepsin B and the activation of cathepsin B depends on v-ATPase activity [28, [36] [37] [38] 42] , we suggested that an accumulation of collagen on the surface of endothelial cells might be a consequence of an impaired functionality of cathepsin B. Therefore, an enzyme activity assay based on the proteolysis of a fluorogenic cathepsin B substrate was performed. In archazolid-treated HUVECs and HMEC-1 the activity of cathepsin B was induce both the mRNA (1 ng/ml TNF) and the cell surface expression (10 ng/ml TNF) of ICAM-1, VCAM-1, E-selectin and Ncadherin.
https://doi.org/10.1371/journal.pone.0203053.g005 Inhibition of endothelial vATPase increases tumor cell adhesion to endothelial cells strongly decreased by approximately 50% compared to control cells at an archazolid concentration of 1 nM (Fig 8A) . In line with this result, western blot analysis showed that archazolid (1 nM) reduces the protein expression of the mature, active form of cathepsin B to less than 40% of the control in HUVECs (Fig 8B) . To proof whether the archazolid-induced tumor cell adhesion is a consequence of the decreased amount of cathepsin B, HUVECs were transfected with a plasmid coding for human cathepsin B or with the empty vector as control. After 48 h, the transfected cells were treated with 1 nM archazolid. The level of cathepsin B after transfection and treatment was assessed by western blot analysis (Fig 9A) . Overexpression of cathepsin B strongly diminished both the basal and the archazolid-induced adhesion of MDA-MB-231 cells (Fig 9B) .
Targeting the proton pump v-ATPase for cancer therapy has gained great interest since its inhibition was reported to reduce the invasiveness of cancer cells and, most importantly, also metastasis [8, 9] . Thus, intensive research related to v-ATPases was done in cancer cells, whereas there are only few studies investigating v-ATPases in endothelial cells indicating a role in migration, proliferation and possibly angiogenesis [22] [23] [24] . In the present study we used the myxobacterial natural product archazolid to investigate the consequences of v-ATPase inhibition in the endothelium on tumor-endothelial cell interactions.
For the first time, we were able to show a link between v-ATPase and the adhesion and transmigration properties of the endothelium. Inhibition of the v-ATPase in endothelial cells by archazolid significantly increased the adhesion of metastatic cancer cells and decreased the transendothelial migration of cancer cells which was attributed to augmented collagen levels on the surface on archazolid-treated endothelial cells. Of note, adhesion of the non-metastatic Jurkat cell line onto archazolid-treated endothelial cells remained unaffected. The archazolidinduced adhesion of tumor cells was independent from the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin, as their expression was not regulated by the compound. However, we found that the archazolid-induced tumor cell adhesion was mediated by β1-integrins expressed on MDA-MB-231 breast cancer and PC-3 prostate cancer cells as blocking of the integrin β1 subunit on these tumor cells reversed the pro-adhesive effect of archazolid. In adhesion experiments on plastic coated with extracellular matrix components, we could show that MDA-MB-231 and PC-3 cells clearly favored the interaction with collagen, whereas the adhesion of non-metastatic Jurkat cells was largely independent from extracellular matrix proteins (S1B Fig). The different adhesion properties of metastatic cancer cells and Jurkat cells might be a result of the distinct integrin expression pattern of each cell line. MDA-MB-231 and PC-3 cells express α2β1-and α3β1-integrins, which represent collagen receptors [43, 44] , while Jurkat cells express α4β1-integrins but lack α2β1-, α3β1-integrins [44] . α4β1integrins are receptors for VCAM-1 and fibronectin [35] and it has been shown that Jurkat cells interact with human endothelial cells that express VCAM-1 after cytokine treatment or cells transfected with VCAM-1 [45] . Our results are in line with previous studies showing that α2β1-and α3β1-integrin expressing MDA-MB-231 and PC-3 cells were able to rapidly attach to collagen in the cortical bone matrix. In contrast, Jurkat cells were not able to adhere [44] and might preferentially interact with cell adhesion molecules rather than with ECM proteins. α2β1-and α3β1-integrins can additionally act as laminin receptors [46] and at least α3β1integrins recognize fibronectin [46, 47] . Though expressing receptors for fibronectin and laminin, MDA-MB-231 and PC-3 cells adhered to fibronectin to a much lesser extent and did not adhere to laminin, probably due to lower affinities to these extracellular matrix components.
Importantly, v-ATPase inhibition by archazolid increased the surface levels of the extracellular matrix component collagen, which might explain that the increase of MDA-MB-231 and PC-3 cells onto archazolid-treated HUVECs is independent of endothelial cell adhesion molecules. By performing a live cell proteolysis assay, Cavallo-Medved et al. demonstrated ECM degradation, in particular of gelatin and collagen IV, in association with active cathepsin B in caveolae of endothelial cells during tube formation [40] . In addition, recent studies reported that v-ATPase inhibition impairs the activity of cathepsin B in cancer cells [28, 36] . Therefore, we suggested that the accumulation of collagen on the endothelial surface might be a consequence of impaired cathepsin B activity or expression in endothelial cells. In fact, we confirmed the impairment of cathepsin B activity by archazolid as the expression levels of the mature active form of this enzyme was strongly reduced. Cathepsin B is synthesized as preprocathepsin B on membrane-bound ribosomes. Following transport to the Golgi apparatus, the preprocathepsin B is glycosylated with mannose-containing oligosaccharides. The targeting of procathepsin B to lysosomes is mannose-6-phosphate receptor-dependent and its dissociation from the receptor as well as its proteolytic processing into mature cathepsin B requires acidification of the compartment [48] . In cancer cells v-ATPase inhibition by archazolid impaired the mannose-6-phosphate receptor-mediated trafficking from the trans-Golgi network to prelysosomal compartments resulting in a decrease of active lysosomal proteases like cathepsin B [28] . We assumed that the archazolid-induced decrease in cathepsin B activity and expression was based on the same mechanism. Interestingly, overexpression of cathepsin B attenuated the archazolid-induced adhesion of breast cancer cells onto endothelial cells, indicating that the adhesion negatively correlates with the expression of cathepsin B.
As cathepsin B can also degrade other extracellular matrix components such as fibronectin and laminin [38, 49] , v-ATPase inhibition could lead to an accumulation of these proteins and an increased adhesion of cells expressing fibronectin or laminin receptors. However, we did not focus on these ECM components since they were not relevant for the adhesion of MDA-MB-231 and PC-3 cells. These cells predominantly adhered to collagen, while the adhesion of Jurkat cells is mostly independent from the ECM proteins collagen, fibronectin or laminin (S1B Fig). Interestingly [50] . In hepatic cancer cells, archazolid reduces Ras/Raf/MEK/ERK signaling by altering the membrane composition and fluidity [51] . We assume that archazolid affects endothelial cells in a similar way leading to inhibition of Ras signaling and, therefore, reduced transendothelial migration of MDA-MB-231 cells.
Taken together, our study shows that archazolid reduces the activity and expression of cathepsin B in endothelial cells. As a result, the amount of collagen on the surface of endothelial cells was significantly upregulated, which finally resulted in an increased adhesion of the β1-integrin-expressing metastatic cancer cell lines MDA-MB-231 and PC-3 onto archazolidtreated endothelial cells, whereas the adhesion of non-metastatic Jurkat cells was unaffected. This study shows that the v-ATPase plays an important role in regulating the adhesion of cells expressing receptors for extracellular matrix components. Archazolid represents a promising tool to elucidate the role of v-ATPase in endothelial cells. Moreover, we for the first time linked the function of v-ATPase to the adhesion and transmigration of tumor cells onto endothelial cells as well as to the remodeling of the extracellular matrix on the surface of endothelial cells. The fact that the adhesion of metastatic tumor cells onto endothelial cells is increased while their transendothelial migration is reduced upon inhibition of endothelial v-ATPase by archazolid further supports the view of archazolid as a potential anti-metastatic compound.
|
How many complexes are within v-ATPase?
| 5,303 |
two
| 2,364 |
1,633 |
The vacuolar-type ATPase inhibitor archazolid increases tumor cell adhesion to endothelial cells by accumulating extracellular collagen
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6133348/
SHA: f1b81916fac1ca3d50dde774df2e1bb26bf0fb39
Authors: Luong, Betty; Schwenk, Rebecca; Bräutigam, Jacqueline; Müller, Rolf; Menche, Dirk; Bischoff, Iris; Fürst, Robert
Date: 2018-09-11
DOI: 10.1371/journal.pone.0203053
License: cc-by
Abstract: The vacuolar-type H(+)-ATPase (v-ATPase) is the major proton pump that acidifies intracellular compartments of eukaryotic cells. Since the inhibition of v-ATPase resulted in anti-tumor and anti-metastatic effects in different tumor models, this enzyme has emerged as promising strategy against cancer. Here, we used the well-established v-ATPase inhibitor archazolid, a natural product first isolated from the myxobacterium Archangium gephyra, to study the consequences of v-ATPase inhibition in endothelial cells (ECs), in particular on the interaction between ECs and cancer cells, which has been neglected so far. Human endothelial cells treated with archazolid showed an increased adhesion of tumor cells, whereas the transendothelial migration of tumor cells was reduced. The adhesion process was independent from the EC adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin. Instead, the adhesion was mediated by β1-integrins expressed on tumor cells, as blocking of the integrin β1 subunit reversed this process. Tumor cells preferentially adhered to the β1-integrin ligand collagen and archazolid led to an increase in the amount of collagen on the surface of ECs. The accumulation of collagen was accompanied by a strong decrease of the expression and activity of the protease cathepsin B. Overexpression of cathepsin B in ECs prevented the capability of archazolid to increase the adhesion of tumor cells onto ECs. Our study demonstrates that the inhibition of v-ATPase by archazolid induces a pro-adhesive phenotype in endothelial cells that promotes their interaction with cancer cells, whereas the transmigration of tumor cells was reduced. These findings further support archazolid as a promising anti-metastatic compound.
Text: The vacuolar-type H + -ATPase (v-ATPase) is the major proton pump responsible for acidification of intracellular compartments in eukaryotic cells [1] . The enzyme consists of two multi-subunit complexes, the soluble V 1 transmembrane V o subcomplex required for the proton transport across membranes [1, 2] . In most cell types v-ATPases are only expressed in the endomembrane system to regulate and maintain the acidic pH of intracellular compartments such as lysosomes, endosomes, the Golgi apparatus, secretory granules and coated vesicles [3] . The function of v-ATPases is essential for cellular processes such as vesicular trafficking, receptor-mediated endocytosis and protein degradation and processing. In specialized cell types including osteoclasts and renal epithelial cells, v-ATPases can also be expressed on the plasma membrane, where they pump protons into the extracellular space [2] [3] [4] . In cancer cells v-ATPases are expressed on the plasma membrane in order to eliminate toxic cytosolic H + . Most importantly, v-ATPases contribute to the acidic tumor microenvironment, which leads to the activation of proteases, thus facilitating tumor cell migration, invasion and angiogenesis [5] [6] [7] . Since the inhibition of v-ATPase was shown to reduce the invasiveness of cancer cells and metastasis formation [8, 9] , this enzyme has emerged as a promising drug target in the recent years. Archazolid A and B are highly potent and specific inhibitors of v-ATPases [10] . They were first isolated from the myxobacterium Archangium gephyra [11] . These compounds inhibit v-ATPase at low nanomolar concentrations [10, 12] by binding to the subunit c of the V o complex. As their biological activity is comparable to the v-ATPase inhibitors bafilomycin and concanamycin [10, 11] , archazolids are natural compounds of high interest that can be used both as a tool to study the consequences of v-ATPase inhibition and as a lead for drug development. Archazolids can be either produced by fermentation [11] or by total synthesis [13, 14] .
In the field of cancer research several studies reported on interesting pharmacological effects of archazolid: It reduced the migration of different invasive tumor cells in vitro and cancer cell metastasis in vivo in a breast tumor mouse model [15] . Furthermore, archazolid activated pathways of cellular stress response and apoptosis in highly invasive tumor cells [16] . In classically activated macrophages, archazolid selectively induced the generation of tumor necrosis factor α (TNFα), which may indirectly promote tumor suppression [17] .
Up to now, the role of v-ATPases in endothelial cells has only rarely been investigated. The endothelium plays a crucial role in the pathogenesis and progression of cancer: The metastatic cascade includes local angiogenesis at the site of the primary tumor and adhesion of tumor cells at the site of metastasis [18] . Angiogenesis, the development of new blood vessels out of existing ones, depends on the proliferation, migration and differentiation of endothelial cells [19] . This process ensures the nutrient supply of the tumor and its growth [20] . Circulating cancer cells can adhere to the endothelium at distant sites. This adhesive interaction is mediated by receptors and corresponding ligands expressed on tumor and endothelial cells [18, 21] . V-ATPases have been reported to regulate intracellular pH and cell migration in microvascular endothelial cells [22, 23] . A recent study showed that the inhibition of v-ATPase by concanamycin prevented proliferation, reduced migration and impaired angiogenesis-related signaling in endothelial cells [24] . So far, there are no investigations on the role of endothelial v-ATPases for the process of tumor cell adhesion onto the endothelium. Thus, we were interested in the consequences of the inhibition of endothelial v-ATPase by archazolid on the interaction between endothelial and cancer cells. Various cell adhesion molecules on the endothelium, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion protein (VCAM-1), E-selectin or N-cadherin [21] as well as integrins expressed on cancer cells have been reported to mediate cell adhesion of cancer cells onto endothelial cells [25] [26] [27] . Accordingly, we focused on these cell adhesion molecules and integrins. For the first time, our study revealed a link between the function of v-ATPases and the adhesion and transmigration properties of endothelial cells.
CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany) was performed according to the manufacturer's protocol for determining the cell viability of cells after treatment with archazolid. This assay is based on the ability of metabolically active cells to reduce resazurin which results in fluorescent resorufin. The CellTiter-Blue Reagent was added to the cells 4 h before the endpoint of treatment. Fluorescence was measured with an Infinite F200 pro microplate reader (Tecan, Männedorf, Switzerland) at 560 nm (excitation) and 590 nm (emission).
CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) was performed according to the manufacturer's instructions for determining the lactate dehydrogenase (LDH) release after treatment with archazolid. Lysis buffer was added to untreated cells 45 min before the end of treatment to induce the release of this enzyme. LDH is a cytosolic enzyme that is released by leaky cells. Released LDH catalyzes the enzymatic conversion of lactate to pyruvate which provides NADH for the conversion of iodonitrotetrazolium violet into a red formazan product in the presence of diaphorase. The absorbance was measured with a Varioskan Flash microplate reader (Thermo Fisher Scientific) at 490 nm.
LysoTracker Red DND-99 (Life Technologies, Thermo Fisher Scientific) is a dye to measure pH values in viable cells. HUVECs were cultured to confluence on collagen G-coated μ-slides (80826, ibidi, Martinsried, Germany) before they were treated with archazolid for 24 h. 1 μg/ ml Hoechst 33342 (Sigma-Aldrich, Munich, Germany) was used to visualize the nuclei and 50 nM LysoTracker Red DND-99 was used to visualize the acidic compartments which correspond to the lysosomes. Both dyes were incubated for 10 min at 37˚C before acquisition of single images by a Leica DMI6000 B fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
HUVECs were seeded in collagen G-coated 24-well plates and grown to confluence for two days before treatment. The cells were incubated with indicated concentrations of archazolid for 24 h. Untreated MDA-MB-231 or PC-3 cells were labeled with CellTracker Green CMFDA Dye (5 μM in serum-free DMEM, 37˚C) for 30 min before 100,000 cells per well were added to HUVECs and were allowed to adhere for various time points at 37˚C. Non-adherent tumor cells were washed off three times with PBS containing Ca 2+ and Mg 2+ . Tumor cell adhesion was determined by fluorescence measurements with an Infinite F200 pro microplate reader (Tecan) at 485 nm (excitation) and 535 nm (emission).
For blocking the integrin β1 subunit on MDA-MB-231 or PC-3 cells, CellTracker Greenlabeled MDA-MB-231 or PC-3 cells were incubated with an anti-integrin β1 antibody (P5D2, ab24693, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg antibody per one million cells in 1 ml DMEM. Before adding to archazolid-treated HUVECs, MDA-MB-231 or PC-3 cells were washed once with DMEM. For blocking the integrin β1 subunit on HUVECs, the cells were incubated with the anti-integrin β1 antibody (0.1 μg/well in ECGM). HUVECs were washed once with ECGM before untreated MDA-MB-231 or PC-3 cells were added to HUVECs.
For the adhesion of MDA-MB-231 or PC-3 cells onto extracellular matrix (ECM) components 24-well plates were coated with collagen G (10 μg/ml in PBS), human plasma fibronectin (10 μg/ml PBS) or laminin-411 (10 μg/ml in Dulbecco's PBS [DPBS] containing Ca 2+ and Mg 2+ ) at 4˚C overnight. The adhesion of MDA-MB-231 and PC-3 cells onto these three most prominent ECM components was carried out as described above (10 min adhesion at 37˚C).
HUVECs were grown on a porous filter membrane (Transwell insert, polycarbonate membrane, 8 μm pores; Corning, New York, USA) for 48 h and were treated as indicated. Untreated MDA-MB-231 cells were labeled with CellTracker Green CMFDA Dye (as described in the section cell adhesion assay) and resuspended in medium 199 (PAN-Biotech) containing 0.1% BSA. HUVECs were washed twice with medium 199 containing 0.1% BSA before MDA-MB-231 cells were allowed to transmigrate through the endothelial monolayer for 24 h. Medium 199 containing 0.1% BSA was used as negative control and medium 199 containing 20% FCS was used as chemoattractant for transmigration in the lower compartment. Non-migrated cells remaining in the upper compartment were carefully removed using a cotton swab. Transmigrated cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and transmigration was quantified by measuring the fluorescence signal at 485 nm (excitation) and 535 nm (emission).
HUVECs were grown to confluence on 6-well plates before they were treated with archazolid for 12 h. The cells were induced to upregulate the gene expression of cell adhesion molecules by TNFα. RNA was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) according to the manufacturer's protocol. On-column DNase digestion was performed to remove genomic DNA. RNA was transcribed into cDNA by Superscript II (Life Technologies, Thermo Fisher Scientific). qPCR experiments were performed using a StepOnePlus System (Applied Biosystems, Thermo Fisher Scientific) and data was analyzed by the StepOne and Ste-pOnePlus Software v2.3. Power SYBR Green PCR Master Mix (Life Technologies) and the comparative C T quantitation method (2 -ΔΔCT ) were used.
HUVECs were grown to confluence on 12-well plates before they were treated with archazolid for 24 h. Cells were treated with TNFα for 24 h to induce the expression of cell adhesion molecules. Subsequently, the cells were detached with HyClone HyQTase (GE Healthcare, Freiburg, Germany). In the case of ICAM-1 the detached cells were fixed with 4% formaldehyde (Polysciences, Hirschberg an der Bergstraße, Germany) in PBS for 10 min and washed once with PBS before incubating with the fluorescein isothiocyanate (FITC)-labeled anti-human CD54 (mouse, ICAM-1) antibody (MCA1615F, Biozol, Eching, Germany) at room temperature for 45 min. For all other proteins, the cells were not fixed and washed once with PBS before incubating with the antibodies phycoerythrin (PE)-labeled anti-human CD106 (mouse, VCAM-1), PE-labeled anti-human CD62E (mouse, E-selectin) and PE-labeled anti-human CD325 (mouse, N-cadherin) from Becton Dickinson on ice for 45 min. These antibodies were diluted in PBS containing 0.2% BSA. The surface expression of cell adhesion molecules was measured by flow cytometry (FACSVerse, Becton Dickinson, Heidelberg, Germany).
To stain the surface collagen on HUVECs, cells were incubated with an anti-human collagen antibody (rabbit, 1:40, ab36064, Abcam) on ice for 30 min. The staining procedure was performed on ice to ensure that surface proteins or antibodies are not endocytosed. The cells were washed once with PBS containing Ca 2+ and Mg 2+ before they were fixed with Roti-Histofix (Carl Roth). Alexa Fluor 488-conjugated anti-rabbit antibody (goat, 1:400, A11008, Life Technologies) was used as secondary antibody and Hoechst 33342 (1 μg/ml, Sigma-Aldrich) was used to visualize nuclei.
Confluent HUVECs in 6-well plates were treated as indicated. Cells were washed with ice-cold PBS and lysed with RIPA buffer supplemented with protease inhibitors (Complete Mini EDTA-free; Roche, Mannheim, Germany), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride, β-glycerophosphate, sodium pyrophosphate and H 2 O 2 . Protein determination was performed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins (10-20 μg) were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad Laboratories, Munich, Germany). Separated proteins were transferred onto polyvinylidene difluoride membranes by tank blotting (Bio-Rad Laboratories) for immunodetection. Membranes were blocked with 5% boltinggrade milk powder (Carl Roth) in TBS containing 0.1% Tween 20 (Sigma-Aldrich). The following antibodies were used: mouse anti-human cathepsin B antibody (IM27L, Merck) (1:500), mouse anti-β-actin-peroxidase antibody (A3854, Sigma-Aldrich) (1:100,000) and antimouse IgG horse radish peroxidase (HRP)-linked antibody (7076, Cell Signaling, Frankfurt, Germany) (1:5,000). ImageJ version 1.49m was used for densitometric analysis.
Cathepsin B activity assay was performed as described in the publication by Kubisch et al. [28] . Confluent HUVECs or HMEC-1 seeded in 6-well plates were treated as indicated. Cells were washed with PBS and lysed with the non-denaturating M-PER mammalian protein extraction reagent (78501, Thermo Fisher Scientific) supplemented with protease inhibitors (Complete Mini EDTA-free, Roche), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride. The fluorogenic cathepsin B substrate Z-Arg-Arg-7-amido-4-methylcoumarin hydrochloride (C5429, Sigma-Aldrich) was added to 30 μg of the cell lysate diluted in assay buffer supplemented with 2 mM L-cysteine (C7880, Sigma-Aldrich) and incubated for 30 min at 40˚C. Fluorescence was measured at 348 nm (excitation) and 440 nm (emission) with a microplate reader (Varioskan Flash, Thermo Fisher Scientific). The intensity of the fluorescence signal corresponded to the cathepsin B enzyme activity. For background subtraction the cathepsin B inhibitor CA-074Me (Enzo Life Sciences, Lörrach, Germany) was added to an additional reaction.
The HUVEC Nucleofector Kit (Lonza, Cologne, Germany) was used to transfect HUVECs. The transfection was performed according to the manufacturer's protocol using 2.5 μg plasmid DNA for 500,000 cells (Nucleofector 2b Device, Lonza). 48 h after transfection the cells were treated for further experiments. The addgene plasmid #11249 hCathepsin B was kindly provided by Hyeryun Choe [29] . hCathepsin B was digested with PmeI and XbaI and the linear DNA fragment not corresponding to the human CTSB gene was religated to generate the empty pcDNA3.1 (-) delta MCS plasmid that was used for control transfections. The original backbone of hCathepsin B is the pcDNA3.1 (-) from Thermo Fisher Scientific. The control vector pcDNA3.1 (-) delta MCS used for our transfections was cloned on the basis of hCathepsin B and is therefore lacking almost the whole part of the multiple cloning site of the pcDNA3.1 (-).
Statistical analyses were performed using GraphPad Prism 5.0 (San Diego, USA). One-way ANOVA followed by Tukey's post-hoc test or unpaired t-test was used for the evaluation of a minimum of three independent experiments. The numbers of independently performed experiments (n) are stated in the corresponding figure legends. p 0.05 was considered as statistically significant. Data are expressed as mean ± standard error of the mean (SEM).
Since the v-ATPase inhibitor archazolid has never been used for studies in endothelial cells, we first performed cytotoxicity assays. We treated confluent HUVECs with up to 1 nM archazolid for 24 and 48 h and observed that this treatment has neither an influence on the metabolic activity nor on the release of LDH after 24 h (Fig 1A and 1B, left panels) . The metabolic activity and the release of LDH were only slightly affected by the highest concentration of archazolid after 48 h (Fig 1A and 1B, right panels) . Consequently, the following experiments were all carried out after 24 h (or less) of archazolid treatment in order to exclude any cytotoxic effects of archazolid within our experimental settings.
Microscopic analysis revealed that also the integrity of the endothelial monolayer was not affected by archazolid, but the cells showed a slightly different morphology (Fig 2A) : Archazolid-treated cells were swollen compared to control cells, which was not unexpected, as vacuolation of the endoplasmic reticulum (ER) has been described for other cell types and is typical for v-ATPase inhibitors [11, 16, 24, 30] . This effect was obvious both in subconfluent and in confluent cells (Fig 2A) . Inhibition of v-ATPase prevents the acidification of lysosomes [1, 31] . Using the cell-permeable dye LysoTracker Red DND-99, it is possible to label the acidic lysosomes in living cells. Thus, this dye can serve as an indicator of v-ATPase inhibition. To proof that archazolid is also functionally active as a v-ATPase inhibitor in HUVECs, cells were treated with 1 nM archazolid before they were incubated with LysoTracker Red DND-99 and Hoechst 33342. As shown in Fig 2B, the red vesicular staining corresponding to acidified lysosomes in control cells disappeared completely after treatment with archazolid. In summary, archazolid treatment for 24 h was not cytotoxic to quiescent HUVECs, but inhibited the functionality of the v-ATPase.
We analyzed the adhesion of MDA-MB-231 cells onto HUVECs. Confluent HUVECs were treated with up to 1 nM archazolid for 24 h. Untreated MDA-MB-231 cells were labeled with Cell-Tracker Green CMFDA Dye. Interestingly, v-ATPase inhibition strongly increased the attachment of the metastatic breast carcinoma cell line MDA-MB-231 onto HUVECs after 10 and 120 min of adhesion (Fig 3A and 3B) . We also investigated the influence of archazolid on the transendothelial migration of MDA-MB-231 cells. HUVECs seeded in a Boyden chamber were treated with 1 nM archazolid for 24 h. CellTracker Green-labeled MDA-MB-231 cells (not treated with archazolid) were allowed to transmigrate through the endothelial monolayer for 24 h. As shown in Fig 3C, archazolid significantly decreased the transendothelial migration of MDA-MB-231 cells.
The influence of archazolid on tumor cell adhesion was not only studied in HUVECs, which represent macrovascular endothelial cells, but also in microvascular HMEC-1 cells. Moreover, besides the breast cancer cell line MDA-MB-231, also PC-3 prostate cancer cells were used as a second metastatic cancer cell line. Archazolid treatment of endothelial cells increased the attachment of MDA-MB-231 cells onto the HMEC-1 monolayer after 120 min of adhesion ( Fig 4A) and increased the attachment of PC-3 cells onto the HUVEC monolayer after 30 and 60 min of adhesion (Fig 4B) . Of note, the adhesion of non-metastatic Jurkat cells, an acute T cell leukemia cell line, remained unaffected after treatment of HUVECs with archazolid (S1A Fig). Taken together, archazolid treatment augmented the adhesive properties of both micro-and macrovascular endothelial cells for metastatic tumor cells, but not for non-metastatic ones. Of note, cancer cell adhesion onto archazolid-activated endothelial cells increased with the time of adhesion.
The adhesion of tumor cells onto the endothelium is in principle similar to that of leukocytes, but slightly differs in the molecules that mediate the adhesion process. Ligands for the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin were found to be expressed on tumor cells and to mediate tumor-endothelial cell interaction [21] . Inhibition of the v-ATPase might affect the expression of endothelial cell adhesion molecules on mRNA or protein levels. To determine the mRNA expression of ICAM-1, VCAM-1, E-selectin and Ncadherin, HUVECs were treated with archazolid for 12 h. TNFα is known to upregulate the expression of ICAM-1, VCAM-1 and E-selectin [32] and, thus, served as positive control. Quantitative real-time PCR showed that v-ATPase inhibition in HUVECs did not alter the mRNA levels of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5A) . The protein expression of these adhesion molecules on the surface of endothelial cells was analyzed by flow cytometry. Archazolid (1 nM, 24 h) did not affect the cell surface expression of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5B) .
Besides ICAM-1, VCAM-1, E-selectin and N-cadherin, also integrins are able to mediate the process of cell adhesion [33] [34] [35] . Since none of the cell adhesion molecules expressed on HUVECs were regulated upon archazolid treatment, we considered integrins as potential interaction partners. Within this protein family β1-integrins have been reported to mediate tumor cell adhesion onto quiescent endothelial cells [25] . In order to elucidate the role of β1-integrins for the archazolid-induced tumor cell adhesion, the integrin β1-subunit was blocked either on MDA-MB-231 cells, PC-3 cells or on HUVECs. (Of note, as in all experiments throughout this study, only endothelial cells were treated with archazolid.) After blocking β1-integrins on MDA-MB-231 or PC-3 cells, the archazolid-induced tumor cell adhesion was reduced almost to control level (Fig 6A and 6B , left panels), whereas blocking of β1-integrins on HUVECs had no significant effect on the increase of tumor cell adhesion by v-ATPase inhibition (Fig 6A and 6B , right panels).
Depending on their α subunit, β1-integrins have a variety of ligands including extracellular matrix (ECM) components such as collagen, fibronectin and laminin [35] . Therefore, we hypothesized that archazolid treatment of endothelial cells might lead to an upregulation of these components. MDA-MB-231 and PC-3 cells were allowed to adhere onto plastic that was coated with these ECM components. This cell adhesion assay revealed that MDA-MB-231 as well as PC-3 cells favor the interaction with the ECM component collagen, as the adhesion onto collagen is much higher than onto the uncoated plastic control (Fig 7A) . MDA-MB-231 and PC-3 cells also adhered to fibronectin-coated plastic, but to a much lesser extent compared to the collagen coating. Therefore, we focused on the interaction between these two tumor cell lines and collagen. Blocking of the integrin β1 subunit on MDA-MB-231 and PC-3 cells clearly abolished the interaction with collagen (Fig 7B) , indicating that the attachment of these tumor cells to collagen is mediated by β1-integrins.
Since collagen is the major ECM component MDA-MB-231 and PC-3 cells interact with, the next step was to prove whether v-ATPase inhibition influences the amount of collagen expressed by HUVECs as extracellular matrix. To detect collagen on the endothelial surface, archazolid-treated HUVECs were labeled with an antibody against collagen type I-IV on ice to prevent endocytosis and to ensure that the antibody does not bind to intracellular collagen. Interestingly, archazolid increased the amount of surface collagen on HUVECs by about 50% (Fig 7C) . Control stainings were performed using an antibody against the cytosolic p65 subunit of the transcription factor nuclear factor κB (NFκB) to show that intracellular proteins were not detected by this staining method (S2 Fig) .
It was reported that v-ATPase inhibition by archazolid impairs the activity of cathepsin B [28, 36] , a lysosomal enzyme that degrades extracellular matrix components including collagen [37] [38] [39] [40] [41] . As collagen is degraded by cathepsin B and the activation of cathepsin B depends on v-ATPase activity [28, [36] [37] [38] 42] , we suggested that an accumulation of collagen on the surface of endothelial cells might be a consequence of an impaired functionality of cathepsin B. Therefore, an enzyme activity assay based on the proteolysis of a fluorogenic cathepsin B substrate was performed. In archazolid-treated HUVECs and HMEC-1 the activity of cathepsin B was induce both the mRNA (1 ng/ml TNF) and the cell surface expression (10 ng/ml TNF) of ICAM-1, VCAM-1, E-selectin and Ncadherin.
https://doi.org/10.1371/journal.pone.0203053.g005 Inhibition of endothelial vATPase increases tumor cell adhesion to endothelial cells strongly decreased by approximately 50% compared to control cells at an archazolid concentration of 1 nM (Fig 8A) . In line with this result, western blot analysis showed that archazolid (1 nM) reduces the protein expression of the mature, active form of cathepsin B to less than 40% of the control in HUVECs (Fig 8B) . To proof whether the archazolid-induced tumor cell adhesion is a consequence of the decreased amount of cathepsin B, HUVECs were transfected with a plasmid coding for human cathepsin B or with the empty vector as control. After 48 h, the transfected cells were treated with 1 nM archazolid. The level of cathepsin B after transfection and treatment was assessed by western blot analysis (Fig 9A) . Overexpression of cathepsin B strongly diminished both the basal and the archazolid-induced adhesion of MDA-MB-231 cells (Fig 9B) .
Targeting the proton pump v-ATPase for cancer therapy has gained great interest since its inhibition was reported to reduce the invasiveness of cancer cells and, most importantly, also metastasis [8, 9] . Thus, intensive research related to v-ATPases was done in cancer cells, whereas there are only few studies investigating v-ATPases in endothelial cells indicating a role in migration, proliferation and possibly angiogenesis [22] [23] [24] . In the present study we used the myxobacterial natural product archazolid to investigate the consequences of v-ATPase inhibition in the endothelium on tumor-endothelial cell interactions.
For the first time, we were able to show a link between v-ATPase and the adhesion and transmigration properties of the endothelium. Inhibition of the v-ATPase in endothelial cells by archazolid significantly increased the adhesion of metastatic cancer cells and decreased the transendothelial migration of cancer cells which was attributed to augmented collagen levels on the surface on archazolid-treated endothelial cells. Of note, adhesion of the non-metastatic Jurkat cell line onto archazolid-treated endothelial cells remained unaffected. The archazolidinduced adhesion of tumor cells was independent from the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin, as their expression was not regulated by the compound. However, we found that the archazolid-induced tumor cell adhesion was mediated by β1-integrins expressed on MDA-MB-231 breast cancer and PC-3 prostate cancer cells as blocking of the integrin β1 subunit on these tumor cells reversed the pro-adhesive effect of archazolid. In adhesion experiments on plastic coated with extracellular matrix components, we could show that MDA-MB-231 and PC-3 cells clearly favored the interaction with collagen, whereas the adhesion of non-metastatic Jurkat cells was largely independent from extracellular matrix proteins (S1B Fig). The different adhesion properties of metastatic cancer cells and Jurkat cells might be a result of the distinct integrin expression pattern of each cell line. MDA-MB-231 and PC-3 cells express α2β1-and α3β1-integrins, which represent collagen receptors [43, 44] , while Jurkat cells express α4β1-integrins but lack α2β1-, α3β1-integrins [44] . α4β1integrins are receptors for VCAM-1 and fibronectin [35] and it has been shown that Jurkat cells interact with human endothelial cells that express VCAM-1 after cytokine treatment or cells transfected with VCAM-1 [45] . Our results are in line with previous studies showing that α2β1-and α3β1-integrin expressing MDA-MB-231 and PC-3 cells were able to rapidly attach to collagen in the cortical bone matrix. In contrast, Jurkat cells were not able to adhere [44] and might preferentially interact with cell adhesion molecules rather than with ECM proteins. α2β1-and α3β1-integrins can additionally act as laminin receptors [46] and at least α3β1integrins recognize fibronectin [46, 47] . Though expressing receptors for fibronectin and laminin, MDA-MB-231 and PC-3 cells adhered to fibronectin to a much lesser extent and did not adhere to laminin, probably due to lower affinities to these extracellular matrix components.
Importantly, v-ATPase inhibition by archazolid increased the surface levels of the extracellular matrix component collagen, which might explain that the increase of MDA-MB-231 and PC-3 cells onto archazolid-treated HUVECs is independent of endothelial cell adhesion molecules. By performing a live cell proteolysis assay, Cavallo-Medved et al. demonstrated ECM degradation, in particular of gelatin and collagen IV, in association with active cathepsin B in caveolae of endothelial cells during tube formation [40] . In addition, recent studies reported that v-ATPase inhibition impairs the activity of cathepsin B in cancer cells [28, 36] . Therefore, we suggested that the accumulation of collagen on the endothelial surface might be a consequence of impaired cathepsin B activity or expression in endothelial cells. In fact, we confirmed the impairment of cathepsin B activity by archazolid as the expression levels of the mature active form of this enzyme was strongly reduced. Cathepsin B is synthesized as preprocathepsin B on membrane-bound ribosomes. Following transport to the Golgi apparatus, the preprocathepsin B is glycosylated with mannose-containing oligosaccharides. The targeting of procathepsin B to lysosomes is mannose-6-phosphate receptor-dependent and its dissociation from the receptor as well as its proteolytic processing into mature cathepsin B requires acidification of the compartment [48] . In cancer cells v-ATPase inhibition by archazolid impaired the mannose-6-phosphate receptor-mediated trafficking from the trans-Golgi network to prelysosomal compartments resulting in a decrease of active lysosomal proteases like cathepsin B [28] . We assumed that the archazolid-induced decrease in cathepsin B activity and expression was based on the same mechanism. Interestingly, overexpression of cathepsin B attenuated the archazolid-induced adhesion of breast cancer cells onto endothelial cells, indicating that the adhesion negatively correlates with the expression of cathepsin B.
As cathepsin B can also degrade other extracellular matrix components such as fibronectin and laminin [38, 49] , v-ATPase inhibition could lead to an accumulation of these proteins and an increased adhesion of cells expressing fibronectin or laminin receptors. However, we did not focus on these ECM components since they were not relevant for the adhesion of MDA-MB-231 and PC-3 cells. These cells predominantly adhered to collagen, while the adhesion of Jurkat cells is mostly independent from the ECM proteins collagen, fibronectin or laminin (S1B Fig). Interestingly [50] . In hepatic cancer cells, archazolid reduces Ras/Raf/MEK/ERK signaling by altering the membrane composition and fluidity [51] . We assume that archazolid affects endothelial cells in a similar way leading to inhibition of Ras signaling and, therefore, reduced transendothelial migration of MDA-MB-231 cells.
Taken together, our study shows that archazolid reduces the activity and expression of cathepsin B in endothelial cells. As a result, the amount of collagen on the surface of endothelial cells was significantly upregulated, which finally resulted in an increased adhesion of the β1-integrin-expressing metastatic cancer cell lines MDA-MB-231 and PC-3 onto archazolidtreated endothelial cells, whereas the adhesion of non-metastatic Jurkat cells was unaffected. This study shows that the v-ATPase plays an important role in regulating the adhesion of cells expressing receptors for extracellular matrix components. Archazolid represents a promising tool to elucidate the role of v-ATPase in endothelial cells. Moreover, we for the first time linked the function of v-ATPase to the adhesion and transmigration of tumor cells onto endothelial cells as well as to the remodeling of the extracellular matrix on the surface of endothelial cells. The fact that the adhesion of metastatic tumor cells onto endothelial cells is increased while their transendothelial migration is reduced upon inhibition of endothelial v-ATPase by archazolid further supports the view of archazolid as a potential anti-metastatic compound.
|
What is the role of v-ATPase in the plasma membrane of osteoclasts and renal epithelial cells?
| 5,304 |
they pump protons into the extracellular space
| 3,039 |
1,633 |
The vacuolar-type ATPase inhibitor archazolid increases tumor cell adhesion to endothelial cells by accumulating extracellular collagen
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6133348/
SHA: f1b81916fac1ca3d50dde774df2e1bb26bf0fb39
Authors: Luong, Betty; Schwenk, Rebecca; Bräutigam, Jacqueline; Müller, Rolf; Menche, Dirk; Bischoff, Iris; Fürst, Robert
Date: 2018-09-11
DOI: 10.1371/journal.pone.0203053
License: cc-by
Abstract: The vacuolar-type H(+)-ATPase (v-ATPase) is the major proton pump that acidifies intracellular compartments of eukaryotic cells. Since the inhibition of v-ATPase resulted in anti-tumor and anti-metastatic effects in different tumor models, this enzyme has emerged as promising strategy against cancer. Here, we used the well-established v-ATPase inhibitor archazolid, a natural product first isolated from the myxobacterium Archangium gephyra, to study the consequences of v-ATPase inhibition in endothelial cells (ECs), in particular on the interaction between ECs and cancer cells, which has been neglected so far. Human endothelial cells treated with archazolid showed an increased adhesion of tumor cells, whereas the transendothelial migration of tumor cells was reduced. The adhesion process was independent from the EC adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin. Instead, the adhesion was mediated by β1-integrins expressed on tumor cells, as blocking of the integrin β1 subunit reversed this process. Tumor cells preferentially adhered to the β1-integrin ligand collagen and archazolid led to an increase in the amount of collagen on the surface of ECs. The accumulation of collagen was accompanied by a strong decrease of the expression and activity of the protease cathepsin B. Overexpression of cathepsin B in ECs prevented the capability of archazolid to increase the adhesion of tumor cells onto ECs. Our study demonstrates that the inhibition of v-ATPase by archazolid induces a pro-adhesive phenotype in endothelial cells that promotes their interaction with cancer cells, whereas the transmigration of tumor cells was reduced. These findings further support archazolid as a promising anti-metastatic compound.
Text: The vacuolar-type H + -ATPase (v-ATPase) is the major proton pump responsible for acidification of intracellular compartments in eukaryotic cells [1] . The enzyme consists of two multi-subunit complexes, the soluble V 1 transmembrane V o subcomplex required for the proton transport across membranes [1, 2] . In most cell types v-ATPases are only expressed in the endomembrane system to regulate and maintain the acidic pH of intracellular compartments such as lysosomes, endosomes, the Golgi apparatus, secretory granules and coated vesicles [3] . The function of v-ATPases is essential for cellular processes such as vesicular trafficking, receptor-mediated endocytosis and protein degradation and processing. In specialized cell types including osteoclasts and renal epithelial cells, v-ATPases can also be expressed on the plasma membrane, where they pump protons into the extracellular space [2] [3] [4] . In cancer cells v-ATPases are expressed on the plasma membrane in order to eliminate toxic cytosolic H + . Most importantly, v-ATPases contribute to the acidic tumor microenvironment, which leads to the activation of proteases, thus facilitating tumor cell migration, invasion and angiogenesis [5] [6] [7] . Since the inhibition of v-ATPase was shown to reduce the invasiveness of cancer cells and metastasis formation [8, 9] , this enzyme has emerged as a promising drug target in the recent years. Archazolid A and B are highly potent and specific inhibitors of v-ATPases [10] . They were first isolated from the myxobacterium Archangium gephyra [11] . These compounds inhibit v-ATPase at low nanomolar concentrations [10, 12] by binding to the subunit c of the V o complex. As their biological activity is comparable to the v-ATPase inhibitors bafilomycin and concanamycin [10, 11] , archazolids are natural compounds of high interest that can be used both as a tool to study the consequences of v-ATPase inhibition and as a lead for drug development. Archazolids can be either produced by fermentation [11] or by total synthesis [13, 14] .
In the field of cancer research several studies reported on interesting pharmacological effects of archazolid: It reduced the migration of different invasive tumor cells in vitro and cancer cell metastasis in vivo in a breast tumor mouse model [15] . Furthermore, archazolid activated pathways of cellular stress response and apoptosis in highly invasive tumor cells [16] . In classically activated macrophages, archazolid selectively induced the generation of tumor necrosis factor α (TNFα), which may indirectly promote tumor suppression [17] .
Up to now, the role of v-ATPases in endothelial cells has only rarely been investigated. The endothelium plays a crucial role in the pathogenesis and progression of cancer: The metastatic cascade includes local angiogenesis at the site of the primary tumor and adhesion of tumor cells at the site of metastasis [18] . Angiogenesis, the development of new blood vessels out of existing ones, depends on the proliferation, migration and differentiation of endothelial cells [19] . This process ensures the nutrient supply of the tumor and its growth [20] . Circulating cancer cells can adhere to the endothelium at distant sites. This adhesive interaction is mediated by receptors and corresponding ligands expressed on tumor and endothelial cells [18, 21] . V-ATPases have been reported to regulate intracellular pH and cell migration in microvascular endothelial cells [22, 23] . A recent study showed that the inhibition of v-ATPase by concanamycin prevented proliferation, reduced migration and impaired angiogenesis-related signaling in endothelial cells [24] . So far, there are no investigations on the role of endothelial v-ATPases for the process of tumor cell adhesion onto the endothelium. Thus, we were interested in the consequences of the inhibition of endothelial v-ATPase by archazolid on the interaction between endothelial and cancer cells. Various cell adhesion molecules on the endothelium, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion protein (VCAM-1), E-selectin or N-cadherin [21] as well as integrins expressed on cancer cells have been reported to mediate cell adhesion of cancer cells onto endothelial cells [25] [26] [27] . Accordingly, we focused on these cell adhesion molecules and integrins. For the first time, our study revealed a link between the function of v-ATPases and the adhesion and transmigration properties of endothelial cells.
CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany) was performed according to the manufacturer's protocol for determining the cell viability of cells after treatment with archazolid. This assay is based on the ability of metabolically active cells to reduce resazurin which results in fluorescent resorufin. The CellTiter-Blue Reagent was added to the cells 4 h before the endpoint of treatment. Fluorescence was measured with an Infinite F200 pro microplate reader (Tecan, Männedorf, Switzerland) at 560 nm (excitation) and 590 nm (emission).
CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) was performed according to the manufacturer's instructions for determining the lactate dehydrogenase (LDH) release after treatment with archazolid. Lysis buffer was added to untreated cells 45 min before the end of treatment to induce the release of this enzyme. LDH is a cytosolic enzyme that is released by leaky cells. Released LDH catalyzes the enzymatic conversion of lactate to pyruvate which provides NADH for the conversion of iodonitrotetrazolium violet into a red formazan product in the presence of diaphorase. The absorbance was measured with a Varioskan Flash microplate reader (Thermo Fisher Scientific) at 490 nm.
LysoTracker Red DND-99 (Life Technologies, Thermo Fisher Scientific) is a dye to measure pH values in viable cells. HUVECs were cultured to confluence on collagen G-coated μ-slides (80826, ibidi, Martinsried, Germany) before they were treated with archazolid for 24 h. 1 μg/ ml Hoechst 33342 (Sigma-Aldrich, Munich, Germany) was used to visualize the nuclei and 50 nM LysoTracker Red DND-99 was used to visualize the acidic compartments which correspond to the lysosomes. Both dyes were incubated for 10 min at 37˚C before acquisition of single images by a Leica DMI6000 B fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
HUVECs were seeded in collagen G-coated 24-well plates and grown to confluence for two days before treatment. The cells were incubated with indicated concentrations of archazolid for 24 h. Untreated MDA-MB-231 or PC-3 cells were labeled with CellTracker Green CMFDA Dye (5 μM in serum-free DMEM, 37˚C) for 30 min before 100,000 cells per well were added to HUVECs and were allowed to adhere for various time points at 37˚C. Non-adherent tumor cells were washed off three times with PBS containing Ca 2+ and Mg 2+ . Tumor cell adhesion was determined by fluorescence measurements with an Infinite F200 pro microplate reader (Tecan) at 485 nm (excitation) and 535 nm (emission).
For blocking the integrin β1 subunit on MDA-MB-231 or PC-3 cells, CellTracker Greenlabeled MDA-MB-231 or PC-3 cells were incubated with an anti-integrin β1 antibody (P5D2, ab24693, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg antibody per one million cells in 1 ml DMEM. Before adding to archazolid-treated HUVECs, MDA-MB-231 or PC-3 cells were washed once with DMEM. For blocking the integrin β1 subunit on HUVECs, the cells were incubated with the anti-integrin β1 antibody (0.1 μg/well in ECGM). HUVECs were washed once with ECGM before untreated MDA-MB-231 or PC-3 cells were added to HUVECs.
For the adhesion of MDA-MB-231 or PC-3 cells onto extracellular matrix (ECM) components 24-well plates were coated with collagen G (10 μg/ml in PBS), human plasma fibronectin (10 μg/ml PBS) or laminin-411 (10 μg/ml in Dulbecco's PBS [DPBS] containing Ca 2+ and Mg 2+ ) at 4˚C overnight. The adhesion of MDA-MB-231 and PC-3 cells onto these three most prominent ECM components was carried out as described above (10 min adhesion at 37˚C).
HUVECs were grown on a porous filter membrane (Transwell insert, polycarbonate membrane, 8 μm pores; Corning, New York, USA) for 48 h and were treated as indicated. Untreated MDA-MB-231 cells were labeled with CellTracker Green CMFDA Dye (as described in the section cell adhesion assay) and resuspended in medium 199 (PAN-Biotech) containing 0.1% BSA. HUVECs were washed twice with medium 199 containing 0.1% BSA before MDA-MB-231 cells were allowed to transmigrate through the endothelial monolayer for 24 h. Medium 199 containing 0.1% BSA was used as negative control and medium 199 containing 20% FCS was used as chemoattractant for transmigration in the lower compartment. Non-migrated cells remaining in the upper compartment were carefully removed using a cotton swab. Transmigrated cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and transmigration was quantified by measuring the fluorescence signal at 485 nm (excitation) and 535 nm (emission).
HUVECs were grown to confluence on 6-well plates before they were treated with archazolid for 12 h. The cells were induced to upregulate the gene expression of cell adhesion molecules by TNFα. RNA was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) according to the manufacturer's protocol. On-column DNase digestion was performed to remove genomic DNA. RNA was transcribed into cDNA by Superscript II (Life Technologies, Thermo Fisher Scientific). qPCR experiments were performed using a StepOnePlus System (Applied Biosystems, Thermo Fisher Scientific) and data was analyzed by the StepOne and Ste-pOnePlus Software v2.3. Power SYBR Green PCR Master Mix (Life Technologies) and the comparative C T quantitation method (2 -ΔΔCT ) were used.
HUVECs were grown to confluence on 12-well plates before they were treated with archazolid for 24 h. Cells were treated with TNFα for 24 h to induce the expression of cell adhesion molecules. Subsequently, the cells were detached with HyClone HyQTase (GE Healthcare, Freiburg, Germany). In the case of ICAM-1 the detached cells were fixed with 4% formaldehyde (Polysciences, Hirschberg an der Bergstraße, Germany) in PBS for 10 min and washed once with PBS before incubating with the fluorescein isothiocyanate (FITC)-labeled anti-human CD54 (mouse, ICAM-1) antibody (MCA1615F, Biozol, Eching, Germany) at room temperature for 45 min. For all other proteins, the cells were not fixed and washed once with PBS before incubating with the antibodies phycoerythrin (PE)-labeled anti-human CD106 (mouse, VCAM-1), PE-labeled anti-human CD62E (mouse, E-selectin) and PE-labeled anti-human CD325 (mouse, N-cadherin) from Becton Dickinson on ice for 45 min. These antibodies were diluted in PBS containing 0.2% BSA. The surface expression of cell adhesion molecules was measured by flow cytometry (FACSVerse, Becton Dickinson, Heidelberg, Germany).
To stain the surface collagen on HUVECs, cells were incubated with an anti-human collagen antibody (rabbit, 1:40, ab36064, Abcam) on ice for 30 min. The staining procedure was performed on ice to ensure that surface proteins or antibodies are not endocytosed. The cells were washed once with PBS containing Ca 2+ and Mg 2+ before they were fixed with Roti-Histofix (Carl Roth). Alexa Fluor 488-conjugated anti-rabbit antibody (goat, 1:400, A11008, Life Technologies) was used as secondary antibody and Hoechst 33342 (1 μg/ml, Sigma-Aldrich) was used to visualize nuclei.
Confluent HUVECs in 6-well plates were treated as indicated. Cells were washed with ice-cold PBS and lysed with RIPA buffer supplemented with protease inhibitors (Complete Mini EDTA-free; Roche, Mannheim, Germany), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride, β-glycerophosphate, sodium pyrophosphate and H 2 O 2 . Protein determination was performed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins (10-20 μg) were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad Laboratories, Munich, Germany). Separated proteins were transferred onto polyvinylidene difluoride membranes by tank blotting (Bio-Rad Laboratories) for immunodetection. Membranes were blocked with 5% boltinggrade milk powder (Carl Roth) in TBS containing 0.1% Tween 20 (Sigma-Aldrich). The following antibodies were used: mouse anti-human cathepsin B antibody (IM27L, Merck) (1:500), mouse anti-β-actin-peroxidase antibody (A3854, Sigma-Aldrich) (1:100,000) and antimouse IgG horse radish peroxidase (HRP)-linked antibody (7076, Cell Signaling, Frankfurt, Germany) (1:5,000). ImageJ version 1.49m was used for densitometric analysis.
Cathepsin B activity assay was performed as described in the publication by Kubisch et al. [28] . Confluent HUVECs or HMEC-1 seeded in 6-well plates were treated as indicated. Cells were washed with PBS and lysed with the non-denaturating M-PER mammalian protein extraction reagent (78501, Thermo Fisher Scientific) supplemented with protease inhibitors (Complete Mini EDTA-free, Roche), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride. The fluorogenic cathepsin B substrate Z-Arg-Arg-7-amido-4-methylcoumarin hydrochloride (C5429, Sigma-Aldrich) was added to 30 μg of the cell lysate diluted in assay buffer supplemented with 2 mM L-cysteine (C7880, Sigma-Aldrich) and incubated for 30 min at 40˚C. Fluorescence was measured at 348 nm (excitation) and 440 nm (emission) with a microplate reader (Varioskan Flash, Thermo Fisher Scientific). The intensity of the fluorescence signal corresponded to the cathepsin B enzyme activity. For background subtraction the cathepsin B inhibitor CA-074Me (Enzo Life Sciences, Lörrach, Germany) was added to an additional reaction.
The HUVEC Nucleofector Kit (Lonza, Cologne, Germany) was used to transfect HUVECs. The transfection was performed according to the manufacturer's protocol using 2.5 μg plasmid DNA for 500,000 cells (Nucleofector 2b Device, Lonza). 48 h after transfection the cells were treated for further experiments. The addgene plasmid #11249 hCathepsin B was kindly provided by Hyeryun Choe [29] . hCathepsin B was digested with PmeI and XbaI and the linear DNA fragment not corresponding to the human CTSB gene was religated to generate the empty pcDNA3.1 (-) delta MCS plasmid that was used for control transfections. The original backbone of hCathepsin B is the pcDNA3.1 (-) from Thermo Fisher Scientific. The control vector pcDNA3.1 (-) delta MCS used for our transfections was cloned on the basis of hCathepsin B and is therefore lacking almost the whole part of the multiple cloning site of the pcDNA3.1 (-).
Statistical analyses were performed using GraphPad Prism 5.0 (San Diego, USA). One-way ANOVA followed by Tukey's post-hoc test or unpaired t-test was used for the evaluation of a minimum of three independent experiments. The numbers of independently performed experiments (n) are stated in the corresponding figure legends. p 0.05 was considered as statistically significant. Data are expressed as mean ± standard error of the mean (SEM).
Since the v-ATPase inhibitor archazolid has never been used for studies in endothelial cells, we first performed cytotoxicity assays. We treated confluent HUVECs with up to 1 nM archazolid for 24 and 48 h and observed that this treatment has neither an influence on the metabolic activity nor on the release of LDH after 24 h (Fig 1A and 1B, left panels) . The metabolic activity and the release of LDH were only slightly affected by the highest concentration of archazolid after 48 h (Fig 1A and 1B, right panels) . Consequently, the following experiments were all carried out after 24 h (or less) of archazolid treatment in order to exclude any cytotoxic effects of archazolid within our experimental settings.
Microscopic analysis revealed that also the integrity of the endothelial monolayer was not affected by archazolid, but the cells showed a slightly different morphology (Fig 2A) : Archazolid-treated cells were swollen compared to control cells, which was not unexpected, as vacuolation of the endoplasmic reticulum (ER) has been described for other cell types and is typical for v-ATPase inhibitors [11, 16, 24, 30] . This effect was obvious both in subconfluent and in confluent cells (Fig 2A) . Inhibition of v-ATPase prevents the acidification of lysosomes [1, 31] . Using the cell-permeable dye LysoTracker Red DND-99, it is possible to label the acidic lysosomes in living cells. Thus, this dye can serve as an indicator of v-ATPase inhibition. To proof that archazolid is also functionally active as a v-ATPase inhibitor in HUVECs, cells were treated with 1 nM archazolid before they were incubated with LysoTracker Red DND-99 and Hoechst 33342. As shown in Fig 2B, the red vesicular staining corresponding to acidified lysosomes in control cells disappeared completely after treatment with archazolid. In summary, archazolid treatment for 24 h was not cytotoxic to quiescent HUVECs, but inhibited the functionality of the v-ATPase.
We analyzed the adhesion of MDA-MB-231 cells onto HUVECs. Confluent HUVECs were treated with up to 1 nM archazolid for 24 h. Untreated MDA-MB-231 cells were labeled with Cell-Tracker Green CMFDA Dye. Interestingly, v-ATPase inhibition strongly increased the attachment of the metastatic breast carcinoma cell line MDA-MB-231 onto HUVECs after 10 and 120 min of adhesion (Fig 3A and 3B) . We also investigated the influence of archazolid on the transendothelial migration of MDA-MB-231 cells. HUVECs seeded in a Boyden chamber were treated with 1 nM archazolid for 24 h. CellTracker Green-labeled MDA-MB-231 cells (not treated with archazolid) were allowed to transmigrate through the endothelial monolayer for 24 h. As shown in Fig 3C, archazolid significantly decreased the transendothelial migration of MDA-MB-231 cells.
The influence of archazolid on tumor cell adhesion was not only studied in HUVECs, which represent macrovascular endothelial cells, but also in microvascular HMEC-1 cells. Moreover, besides the breast cancer cell line MDA-MB-231, also PC-3 prostate cancer cells were used as a second metastatic cancer cell line. Archazolid treatment of endothelial cells increased the attachment of MDA-MB-231 cells onto the HMEC-1 monolayer after 120 min of adhesion ( Fig 4A) and increased the attachment of PC-3 cells onto the HUVEC monolayer after 30 and 60 min of adhesion (Fig 4B) . Of note, the adhesion of non-metastatic Jurkat cells, an acute T cell leukemia cell line, remained unaffected after treatment of HUVECs with archazolid (S1A Fig). Taken together, archazolid treatment augmented the adhesive properties of both micro-and macrovascular endothelial cells for metastatic tumor cells, but not for non-metastatic ones. Of note, cancer cell adhesion onto archazolid-activated endothelial cells increased with the time of adhesion.
The adhesion of tumor cells onto the endothelium is in principle similar to that of leukocytes, but slightly differs in the molecules that mediate the adhesion process. Ligands for the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin were found to be expressed on tumor cells and to mediate tumor-endothelial cell interaction [21] . Inhibition of the v-ATPase might affect the expression of endothelial cell adhesion molecules on mRNA or protein levels. To determine the mRNA expression of ICAM-1, VCAM-1, E-selectin and Ncadherin, HUVECs were treated with archazolid for 12 h. TNFα is known to upregulate the expression of ICAM-1, VCAM-1 and E-selectin [32] and, thus, served as positive control. Quantitative real-time PCR showed that v-ATPase inhibition in HUVECs did not alter the mRNA levels of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5A) . The protein expression of these adhesion molecules on the surface of endothelial cells was analyzed by flow cytometry. Archazolid (1 nM, 24 h) did not affect the cell surface expression of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5B) .
Besides ICAM-1, VCAM-1, E-selectin and N-cadherin, also integrins are able to mediate the process of cell adhesion [33] [34] [35] . Since none of the cell adhesion molecules expressed on HUVECs were regulated upon archazolid treatment, we considered integrins as potential interaction partners. Within this protein family β1-integrins have been reported to mediate tumor cell adhesion onto quiescent endothelial cells [25] . In order to elucidate the role of β1-integrins for the archazolid-induced tumor cell adhesion, the integrin β1-subunit was blocked either on MDA-MB-231 cells, PC-3 cells or on HUVECs. (Of note, as in all experiments throughout this study, only endothelial cells were treated with archazolid.) After blocking β1-integrins on MDA-MB-231 or PC-3 cells, the archazolid-induced tumor cell adhesion was reduced almost to control level (Fig 6A and 6B , left panels), whereas blocking of β1-integrins on HUVECs had no significant effect on the increase of tumor cell adhesion by v-ATPase inhibition (Fig 6A and 6B , right panels).
Depending on their α subunit, β1-integrins have a variety of ligands including extracellular matrix (ECM) components such as collagen, fibronectin and laminin [35] . Therefore, we hypothesized that archazolid treatment of endothelial cells might lead to an upregulation of these components. MDA-MB-231 and PC-3 cells were allowed to adhere onto plastic that was coated with these ECM components. This cell adhesion assay revealed that MDA-MB-231 as well as PC-3 cells favor the interaction with the ECM component collagen, as the adhesion onto collagen is much higher than onto the uncoated plastic control (Fig 7A) . MDA-MB-231 and PC-3 cells also adhered to fibronectin-coated plastic, but to a much lesser extent compared to the collagen coating. Therefore, we focused on the interaction between these two tumor cell lines and collagen. Blocking of the integrin β1 subunit on MDA-MB-231 and PC-3 cells clearly abolished the interaction with collagen (Fig 7B) , indicating that the attachment of these tumor cells to collagen is mediated by β1-integrins.
Since collagen is the major ECM component MDA-MB-231 and PC-3 cells interact with, the next step was to prove whether v-ATPase inhibition influences the amount of collagen expressed by HUVECs as extracellular matrix. To detect collagen on the endothelial surface, archazolid-treated HUVECs were labeled with an antibody against collagen type I-IV on ice to prevent endocytosis and to ensure that the antibody does not bind to intracellular collagen. Interestingly, archazolid increased the amount of surface collagen on HUVECs by about 50% (Fig 7C) . Control stainings were performed using an antibody against the cytosolic p65 subunit of the transcription factor nuclear factor κB (NFκB) to show that intracellular proteins were not detected by this staining method (S2 Fig) .
It was reported that v-ATPase inhibition by archazolid impairs the activity of cathepsin B [28, 36] , a lysosomal enzyme that degrades extracellular matrix components including collagen [37] [38] [39] [40] [41] . As collagen is degraded by cathepsin B and the activation of cathepsin B depends on v-ATPase activity [28, [36] [37] [38] 42] , we suggested that an accumulation of collagen on the surface of endothelial cells might be a consequence of an impaired functionality of cathepsin B. Therefore, an enzyme activity assay based on the proteolysis of a fluorogenic cathepsin B substrate was performed. In archazolid-treated HUVECs and HMEC-1 the activity of cathepsin B was induce both the mRNA (1 ng/ml TNF) and the cell surface expression (10 ng/ml TNF) of ICAM-1, VCAM-1, E-selectin and Ncadherin.
https://doi.org/10.1371/journal.pone.0203053.g005 Inhibition of endothelial vATPase increases tumor cell adhesion to endothelial cells strongly decreased by approximately 50% compared to control cells at an archazolid concentration of 1 nM (Fig 8A) . In line with this result, western blot analysis showed that archazolid (1 nM) reduces the protein expression of the mature, active form of cathepsin B to less than 40% of the control in HUVECs (Fig 8B) . To proof whether the archazolid-induced tumor cell adhesion is a consequence of the decreased amount of cathepsin B, HUVECs were transfected with a plasmid coding for human cathepsin B or with the empty vector as control. After 48 h, the transfected cells were treated with 1 nM archazolid. The level of cathepsin B after transfection and treatment was assessed by western blot analysis (Fig 9A) . Overexpression of cathepsin B strongly diminished both the basal and the archazolid-induced adhesion of MDA-MB-231 cells (Fig 9B) .
Targeting the proton pump v-ATPase for cancer therapy has gained great interest since its inhibition was reported to reduce the invasiveness of cancer cells and, most importantly, also metastasis [8, 9] . Thus, intensive research related to v-ATPases was done in cancer cells, whereas there are only few studies investigating v-ATPases in endothelial cells indicating a role in migration, proliferation and possibly angiogenesis [22] [23] [24] . In the present study we used the myxobacterial natural product archazolid to investigate the consequences of v-ATPase inhibition in the endothelium on tumor-endothelial cell interactions.
For the first time, we were able to show a link between v-ATPase and the adhesion and transmigration properties of the endothelium. Inhibition of the v-ATPase in endothelial cells by archazolid significantly increased the adhesion of metastatic cancer cells and decreased the transendothelial migration of cancer cells which was attributed to augmented collagen levels on the surface on archazolid-treated endothelial cells. Of note, adhesion of the non-metastatic Jurkat cell line onto archazolid-treated endothelial cells remained unaffected. The archazolidinduced adhesion of tumor cells was independent from the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin, as their expression was not regulated by the compound. However, we found that the archazolid-induced tumor cell adhesion was mediated by β1-integrins expressed on MDA-MB-231 breast cancer and PC-3 prostate cancer cells as blocking of the integrin β1 subunit on these tumor cells reversed the pro-adhesive effect of archazolid. In adhesion experiments on plastic coated with extracellular matrix components, we could show that MDA-MB-231 and PC-3 cells clearly favored the interaction with collagen, whereas the adhesion of non-metastatic Jurkat cells was largely independent from extracellular matrix proteins (S1B Fig). The different adhesion properties of metastatic cancer cells and Jurkat cells might be a result of the distinct integrin expression pattern of each cell line. MDA-MB-231 and PC-3 cells express α2β1-and α3β1-integrins, which represent collagen receptors [43, 44] , while Jurkat cells express α4β1-integrins but lack α2β1-, α3β1-integrins [44] . α4β1integrins are receptors for VCAM-1 and fibronectin [35] and it has been shown that Jurkat cells interact with human endothelial cells that express VCAM-1 after cytokine treatment or cells transfected with VCAM-1 [45] . Our results are in line with previous studies showing that α2β1-and α3β1-integrin expressing MDA-MB-231 and PC-3 cells were able to rapidly attach to collagen in the cortical bone matrix. In contrast, Jurkat cells were not able to adhere [44] and might preferentially interact with cell adhesion molecules rather than with ECM proteins. α2β1-and α3β1-integrins can additionally act as laminin receptors [46] and at least α3β1integrins recognize fibronectin [46, 47] . Though expressing receptors for fibronectin and laminin, MDA-MB-231 and PC-3 cells adhered to fibronectin to a much lesser extent and did not adhere to laminin, probably due to lower affinities to these extracellular matrix components.
Importantly, v-ATPase inhibition by archazolid increased the surface levels of the extracellular matrix component collagen, which might explain that the increase of MDA-MB-231 and PC-3 cells onto archazolid-treated HUVECs is independent of endothelial cell adhesion molecules. By performing a live cell proteolysis assay, Cavallo-Medved et al. demonstrated ECM degradation, in particular of gelatin and collagen IV, in association with active cathepsin B in caveolae of endothelial cells during tube formation [40] . In addition, recent studies reported that v-ATPase inhibition impairs the activity of cathepsin B in cancer cells [28, 36] . Therefore, we suggested that the accumulation of collagen on the endothelial surface might be a consequence of impaired cathepsin B activity or expression in endothelial cells. In fact, we confirmed the impairment of cathepsin B activity by archazolid as the expression levels of the mature active form of this enzyme was strongly reduced. Cathepsin B is synthesized as preprocathepsin B on membrane-bound ribosomes. Following transport to the Golgi apparatus, the preprocathepsin B is glycosylated with mannose-containing oligosaccharides. The targeting of procathepsin B to lysosomes is mannose-6-phosphate receptor-dependent and its dissociation from the receptor as well as its proteolytic processing into mature cathepsin B requires acidification of the compartment [48] . In cancer cells v-ATPase inhibition by archazolid impaired the mannose-6-phosphate receptor-mediated trafficking from the trans-Golgi network to prelysosomal compartments resulting in a decrease of active lysosomal proteases like cathepsin B [28] . We assumed that the archazolid-induced decrease in cathepsin B activity and expression was based on the same mechanism. Interestingly, overexpression of cathepsin B attenuated the archazolid-induced adhesion of breast cancer cells onto endothelial cells, indicating that the adhesion negatively correlates with the expression of cathepsin B.
As cathepsin B can also degrade other extracellular matrix components such as fibronectin and laminin [38, 49] , v-ATPase inhibition could lead to an accumulation of these proteins and an increased adhesion of cells expressing fibronectin or laminin receptors. However, we did not focus on these ECM components since they were not relevant for the adhesion of MDA-MB-231 and PC-3 cells. These cells predominantly adhered to collagen, while the adhesion of Jurkat cells is mostly independent from the ECM proteins collagen, fibronectin or laminin (S1B Fig). Interestingly [50] . In hepatic cancer cells, archazolid reduces Ras/Raf/MEK/ERK signaling by altering the membrane composition and fluidity [51] . We assume that archazolid affects endothelial cells in a similar way leading to inhibition of Ras signaling and, therefore, reduced transendothelial migration of MDA-MB-231 cells.
Taken together, our study shows that archazolid reduces the activity and expression of cathepsin B in endothelial cells. As a result, the amount of collagen on the surface of endothelial cells was significantly upregulated, which finally resulted in an increased adhesion of the β1-integrin-expressing metastatic cancer cell lines MDA-MB-231 and PC-3 onto archazolidtreated endothelial cells, whereas the adhesion of non-metastatic Jurkat cells was unaffected. This study shows that the v-ATPase plays an important role in regulating the adhesion of cells expressing receptors for extracellular matrix components. Archazolid represents a promising tool to elucidate the role of v-ATPase in endothelial cells. Moreover, we for the first time linked the function of v-ATPase to the adhesion and transmigration of tumor cells onto endothelial cells as well as to the remodeling of the extracellular matrix on the surface of endothelial cells. The fact that the adhesion of metastatic tumor cells onto endothelial cells is increased while their transendothelial migration is reduced upon inhibition of endothelial v-ATPase by archazolid further supports the view of archazolid as a potential anti-metastatic compound.
|
What does angiogenesis depend on?
| 5,305 |
the proliferation, migration and differentiation of endothelial cells
| 5,195 |
1,633 |
The vacuolar-type ATPase inhibitor archazolid increases tumor cell adhesion to endothelial cells by accumulating extracellular collagen
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6133348/
SHA: f1b81916fac1ca3d50dde774df2e1bb26bf0fb39
Authors: Luong, Betty; Schwenk, Rebecca; Bräutigam, Jacqueline; Müller, Rolf; Menche, Dirk; Bischoff, Iris; Fürst, Robert
Date: 2018-09-11
DOI: 10.1371/journal.pone.0203053
License: cc-by
Abstract: The vacuolar-type H(+)-ATPase (v-ATPase) is the major proton pump that acidifies intracellular compartments of eukaryotic cells. Since the inhibition of v-ATPase resulted in anti-tumor and anti-metastatic effects in different tumor models, this enzyme has emerged as promising strategy against cancer. Here, we used the well-established v-ATPase inhibitor archazolid, a natural product first isolated from the myxobacterium Archangium gephyra, to study the consequences of v-ATPase inhibition in endothelial cells (ECs), in particular on the interaction between ECs and cancer cells, which has been neglected so far. Human endothelial cells treated with archazolid showed an increased adhesion of tumor cells, whereas the transendothelial migration of tumor cells was reduced. The adhesion process was independent from the EC adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin. Instead, the adhesion was mediated by β1-integrins expressed on tumor cells, as blocking of the integrin β1 subunit reversed this process. Tumor cells preferentially adhered to the β1-integrin ligand collagen and archazolid led to an increase in the amount of collagen on the surface of ECs. The accumulation of collagen was accompanied by a strong decrease of the expression and activity of the protease cathepsin B. Overexpression of cathepsin B in ECs prevented the capability of archazolid to increase the adhesion of tumor cells onto ECs. Our study demonstrates that the inhibition of v-ATPase by archazolid induces a pro-adhesive phenotype in endothelial cells that promotes their interaction with cancer cells, whereas the transmigration of tumor cells was reduced. These findings further support archazolid as a promising anti-metastatic compound.
Text: The vacuolar-type H + -ATPase (v-ATPase) is the major proton pump responsible for acidification of intracellular compartments in eukaryotic cells [1] . The enzyme consists of two multi-subunit complexes, the soluble V 1 transmembrane V o subcomplex required for the proton transport across membranes [1, 2] . In most cell types v-ATPases are only expressed in the endomembrane system to regulate and maintain the acidic pH of intracellular compartments such as lysosomes, endosomes, the Golgi apparatus, secretory granules and coated vesicles [3] . The function of v-ATPases is essential for cellular processes such as vesicular trafficking, receptor-mediated endocytosis and protein degradation and processing. In specialized cell types including osteoclasts and renal epithelial cells, v-ATPases can also be expressed on the plasma membrane, where they pump protons into the extracellular space [2] [3] [4] . In cancer cells v-ATPases are expressed on the plasma membrane in order to eliminate toxic cytosolic H + . Most importantly, v-ATPases contribute to the acidic tumor microenvironment, which leads to the activation of proteases, thus facilitating tumor cell migration, invasion and angiogenesis [5] [6] [7] . Since the inhibition of v-ATPase was shown to reduce the invasiveness of cancer cells and metastasis formation [8, 9] , this enzyme has emerged as a promising drug target in the recent years. Archazolid A and B are highly potent and specific inhibitors of v-ATPases [10] . They were first isolated from the myxobacterium Archangium gephyra [11] . These compounds inhibit v-ATPase at low nanomolar concentrations [10, 12] by binding to the subunit c of the V o complex. As their biological activity is comparable to the v-ATPase inhibitors bafilomycin and concanamycin [10, 11] , archazolids are natural compounds of high interest that can be used both as a tool to study the consequences of v-ATPase inhibition and as a lead for drug development. Archazolids can be either produced by fermentation [11] or by total synthesis [13, 14] .
In the field of cancer research several studies reported on interesting pharmacological effects of archazolid: It reduced the migration of different invasive tumor cells in vitro and cancer cell metastasis in vivo in a breast tumor mouse model [15] . Furthermore, archazolid activated pathways of cellular stress response and apoptosis in highly invasive tumor cells [16] . In classically activated macrophages, archazolid selectively induced the generation of tumor necrosis factor α (TNFα), which may indirectly promote tumor suppression [17] .
Up to now, the role of v-ATPases in endothelial cells has only rarely been investigated. The endothelium plays a crucial role in the pathogenesis and progression of cancer: The metastatic cascade includes local angiogenesis at the site of the primary tumor and adhesion of tumor cells at the site of metastasis [18] . Angiogenesis, the development of new blood vessels out of existing ones, depends on the proliferation, migration and differentiation of endothelial cells [19] . This process ensures the nutrient supply of the tumor and its growth [20] . Circulating cancer cells can adhere to the endothelium at distant sites. This adhesive interaction is mediated by receptors and corresponding ligands expressed on tumor and endothelial cells [18, 21] . V-ATPases have been reported to regulate intracellular pH and cell migration in microvascular endothelial cells [22, 23] . A recent study showed that the inhibition of v-ATPase by concanamycin prevented proliferation, reduced migration and impaired angiogenesis-related signaling in endothelial cells [24] . So far, there are no investigations on the role of endothelial v-ATPases for the process of tumor cell adhesion onto the endothelium. Thus, we were interested in the consequences of the inhibition of endothelial v-ATPase by archazolid on the interaction between endothelial and cancer cells. Various cell adhesion molecules on the endothelium, such as intercellular adhesion molecule 1 (ICAM-1), vascular cell adhesion protein (VCAM-1), E-selectin or N-cadherin [21] as well as integrins expressed on cancer cells have been reported to mediate cell adhesion of cancer cells onto endothelial cells [25] [26] [27] . Accordingly, we focused on these cell adhesion molecules and integrins. For the first time, our study revealed a link between the function of v-ATPases and the adhesion and transmigration properties of endothelial cells.
CellTiter-Blue Cell Viability Assay (Promega, Mannheim, Germany) was performed according to the manufacturer's protocol for determining the cell viability of cells after treatment with archazolid. This assay is based on the ability of metabolically active cells to reduce resazurin which results in fluorescent resorufin. The CellTiter-Blue Reagent was added to the cells 4 h before the endpoint of treatment. Fluorescence was measured with an Infinite F200 pro microplate reader (Tecan, Männedorf, Switzerland) at 560 nm (excitation) and 590 nm (emission).
CytoTox 96 Non-Radioactive Cytotoxicity Assay (Promega) was performed according to the manufacturer's instructions for determining the lactate dehydrogenase (LDH) release after treatment with archazolid. Lysis buffer was added to untreated cells 45 min before the end of treatment to induce the release of this enzyme. LDH is a cytosolic enzyme that is released by leaky cells. Released LDH catalyzes the enzymatic conversion of lactate to pyruvate which provides NADH for the conversion of iodonitrotetrazolium violet into a red formazan product in the presence of diaphorase. The absorbance was measured with a Varioskan Flash microplate reader (Thermo Fisher Scientific) at 490 nm.
LysoTracker Red DND-99 (Life Technologies, Thermo Fisher Scientific) is a dye to measure pH values in viable cells. HUVECs were cultured to confluence on collagen G-coated μ-slides (80826, ibidi, Martinsried, Germany) before they were treated with archazolid for 24 h. 1 μg/ ml Hoechst 33342 (Sigma-Aldrich, Munich, Germany) was used to visualize the nuclei and 50 nM LysoTracker Red DND-99 was used to visualize the acidic compartments which correspond to the lysosomes. Both dyes were incubated for 10 min at 37˚C before acquisition of single images by a Leica DMI6000 B fluorescence microscope (Leica Microsystems, Wetzlar, Germany).
HUVECs were seeded in collagen G-coated 24-well plates and grown to confluence for two days before treatment. The cells were incubated with indicated concentrations of archazolid for 24 h. Untreated MDA-MB-231 or PC-3 cells were labeled with CellTracker Green CMFDA Dye (5 μM in serum-free DMEM, 37˚C) for 30 min before 100,000 cells per well were added to HUVECs and were allowed to adhere for various time points at 37˚C. Non-adherent tumor cells were washed off three times with PBS containing Ca 2+ and Mg 2+ . Tumor cell adhesion was determined by fluorescence measurements with an Infinite F200 pro microplate reader (Tecan) at 485 nm (excitation) and 535 nm (emission).
For blocking the integrin β1 subunit on MDA-MB-231 or PC-3 cells, CellTracker Greenlabeled MDA-MB-231 or PC-3 cells were incubated with an anti-integrin β1 antibody (P5D2, ab24693, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg antibody per one million cells in 1 ml DMEM. Before adding to archazolid-treated HUVECs, MDA-MB-231 or PC-3 cells were washed once with DMEM. For blocking the integrin β1 subunit on HUVECs, the cells were incubated with the anti-integrin β1 antibody (0.1 μg/well in ECGM). HUVECs were washed once with ECGM before untreated MDA-MB-231 or PC-3 cells were added to HUVECs.
For the adhesion of MDA-MB-231 or PC-3 cells onto extracellular matrix (ECM) components 24-well plates were coated with collagen G (10 μg/ml in PBS), human plasma fibronectin (10 μg/ml PBS) or laminin-411 (10 μg/ml in Dulbecco's PBS [DPBS] containing Ca 2+ and Mg 2+ ) at 4˚C overnight. The adhesion of MDA-MB-231 and PC-3 cells onto these three most prominent ECM components was carried out as described above (10 min adhesion at 37˚C).
HUVECs were grown on a porous filter membrane (Transwell insert, polycarbonate membrane, 8 μm pores; Corning, New York, USA) for 48 h and were treated as indicated. Untreated MDA-MB-231 cells were labeled with CellTracker Green CMFDA Dye (as described in the section cell adhesion assay) and resuspended in medium 199 (PAN-Biotech) containing 0.1% BSA. HUVECs were washed twice with medium 199 containing 0.1% BSA before MDA-MB-231 cells were allowed to transmigrate through the endothelial monolayer for 24 h. Medium 199 containing 0.1% BSA was used as negative control and medium 199 containing 20% FCS was used as chemoattractant for transmigration in the lower compartment. Non-migrated cells remaining in the upper compartment were carefully removed using a cotton swab. Transmigrated cells were lysed in radioimmunoprecipitation assay (RIPA) buffer and transmigration was quantified by measuring the fluorescence signal at 485 nm (excitation) and 535 nm (emission).
HUVECs were grown to confluence on 6-well plates before they were treated with archazolid for 12 h. The cells were induced to upregulate the gene expression of cell adhesion molecules by TNFα. RNA was isolated using the RNeasy Mini Kit from Qiagen (Hilden, Germany) according to the manufacturer's protocol. On-column DNase digestion was performed to remove genomic DNA. RNA was transcribed into cDNA by Superscript II (Life Technologies, Thermo Fisher Scientific). qPCR experiments were performed using a StepOnePlus System (Applied Biosystems, Thermo Fisher Scientific) and data was analyzed by the StepOne and Ste-pOnePlus Software v2.3. Power SYBR Green PCR Master Mix (Life Technologies) and the comparative C T quantitation method (2 -ΔΔCT ) were used.
HUVECs were grown to confluence on 12-well plates before they were treated with archazolid for 24 h. Cells were treated with TNFα for 24 h to induce the expression of cell adhesion molecules. Subsequently, the cells were detached with HyClone HyQTase (GE Healthcare, Freiburg, Germany). In the case of ICAM-1 the detached cells were fixed with 4% formaldehyde (Polysciences, Hirschberg an der Bergstraße, Germany) in PBS for 10 min and washed once with PBS before incubating with the fluorescein isothiocyanate (FITC)-labeled anti-human CD54 (mouse, ICAM-1) antibody (MCA1615F, Biozol, Eching, Germany) at room temperature for 45 min. For all other proteins, the cells were not fixed and washed once with PBS before incubating with the antibodies phycoerythrin (PE)-labeled anti-human CD106 (mouse, VCAM-1), PE-labeled anti-human CD62E (mouse, E-selectin) and PE-labeled anti-human CD325 (mouse, N-cadherin) from Becton Dickinson on ice for 45 min. These antibodies were diluted in PBS containing 0.2% BSA. The surface expression of cell adhesion molecules was measured by flow cytometry (FACSVerse, Becton Dickinson, Heidelberg, Germany).
To stain the surface collagen on HUVECs, cells were incubated with an anti-human collagen antibody (rabbit, 1:40, ab36064, Abcam) on ice for 30 min. The staining procedure was performed on ice to ensure that surface proteins or antibodies are not endocytosed. The cells were washed once with PBS containing Ca 2+ and Mg 2+ before they were fixed with Roti-Histofix (Carl Roth). Alexa Fluor 488-conjugated anti-rabbit antibody (goat, 1:400, A11008, Life Technologies) was used as secondary antibody and Hoechst 33342 (1 μg/ml, Sigma-Aldrich) was used to visualize nuclei.
Confluent HUVECs in 6-well plates were treated as indicated. Cells were washed with ice-cold PBS and lysed with RIPA buffer supplemented with protease inhibitors (Complete Mini EDTA-free; Roche, Mannheim, Germany), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride, β-glycerophosphate, sodium pyrophosphate and H 2 O 2 . Protein determination was performed using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of proteins (10-20 μg) were separated by sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE; Bio-Rad Laboratories, Munich, Germany). Separated proteins were transferred onto polyvinylidene difluoride membranes by tank blotting (Bio-Rad Laboratories) for immunodetection. Membranes were blocked with 5% boltinggrade milk powder (Carl Roth) in TBS containing 0.1% Tween 20 (Sigma-Aldrich). The following antibodies were used: mouse anti-human cathepsin B antibody (IM27L, Merck) (1:500), mouse anti-β-actin-peroxidase antibody (A3854, Sigma-Aldrich) (1:100,000) and antimouse IgG horse radish peroxidase (HRP)-linked antibody (7076, Cell Signaling, Frankfurt, Germany) (1:5,000). ImageJ version 1.49m was used for densitometric analysis.
Cathepsin B activity assay was performed as described in the publication by Kubisch et al. [28] . Confluent HUVECs or HMEC-1 seeded in 6-well plates were treated as indicated. Cells were washed with PBS and lysed with the non-denaturating M-PER mammalian protein extraction reagent (78501, Thermo Fisher Scientific) supplemented with protease inhibitors (Complete Mini EDTA-free, Roche), sodium orthovanadate, sodium fluoride, phenylmethylsulphonyl fluoride. The fluorogenic cathepsin B substrate Z-Arg-Arg-7-amido-4-methylcoumarin hydrochloride (C5429, Sigma-Aldrich) was added to 30 μg of the cell lysate diluted in assay buffer supplemented with 2 mM L-cysteine (C7880, Sigma-Aldrich) and incubated for 30 min at 40˚C. Fluorescence was measured at 348 nm (excitation) and 440 nm (emission) with a microplate reader (Varioskan Flash, Thermo Fisher Scientific). The intensity of the fluorescence signal corresponded to the cathepsin B enzyme activity. For background subtraction the cathepsin B inhibitor CA-074Me (Enzo Life Sciences, Lörrach, Germany) was added to an additional reaction.
The HUVEC Nucleofector Kit (Lonza, Cologne, Germany) was used to transfect HUVECs. The transfection was performed according to the manufacturer's protocol using 2.5 μg plasmid DNA for 500,000 cells (Nucleofector 2b Device, Lonza). 48 h after transfection the cells were treated for further experiments. The addgene plasmid #11249 hCathepsin B was kindly provided by Hyeryun Choe [29] . hCathepsin B was digested with PmeI and XbaI and the linear DNA fragment not corresponding to the human CTSB gene was religated to generate the empty pcDNA3.1 (-) delta MCS plasmid that was used for control transfections. The original backbone of hCathepsin B is the pcDNA3.1 (-) from Thermo Fisher Scientific. The control vector pcDNA3.1 (-) delta MCS used for our transfections was cloned on the basis of hCathepsin B and is therefore lacking almost the whole part of the multiple cloning site of the pcDNA3.1 (-).
Statistical analyses were performed using GraphPad Prism 5.0 (San Diego, USA). One-way ANOVA followed by Tukey's post-hoc test or unpaired t-test was used for the evaluation of a minimum of three independent experiments. The numbers of independently performed experiments (n) are stated in the corresponding figure legends. p 0.05 was considered as statistically significant. Data are expressed as mean ± standard error of the mean (SEM).
Since the v-ATPase inhibitor archazolid has never been used for studies in endothelial cells, we first performed cytotoxicity assays. We treated confluent HUVECs with up to 1 nM archazolid for 24 and 48 h and observed that this treatment has neither an influence on the metabolic activity nor on the release of LDH after 24 h (Fig 1A and 1B, left panels) . The metabolic activity and the release of LDH were only slightly affected by the highest concentration of archazolid after 48 h (Fig 1A and 1B, right panels) . Consequently, the following experiments were all carried out after 24 h (or less) of archazolid treatment in order to exclude any cytotoxic effects of archazolid within our experimental settings.
Microscopic analysis revealed that also the integrity of the endothelial monolayer was not affected by archazolid, but the cells showed a slightly different morphology (Fig 2A) : Archazolid-treated cells were swollen compared to control cells, which was not unexpected, as vacuolation of the endoplasmic reticulum (ER) has been described for other cell types and is typical for v-ATPase inhibitors [11, 16, 24, 30] . This effect was obvious both in subconfluent and in confluent cells (Fig 2A) . Inhibition of v-ATPase prevents the acidification of lysosomes [1, 31] . Using the cell-permeable dye LysoTracker Red DND-99, it is possible to label the acidic lysosomes in living cells. Thus, this dye can serve as an indicator of v-ATPase inhibition. To proof that archazolid is also functionally active as a v-ATPase inhibitor in HUVECs, cells were treated with 1 nM archazolid before they were incubated with LysoTracker Red DND-99 and Hoechst 33342. As shown in Fig 2B, the red vesicular staining corresponding to acidified lysosomes in control cells disappeared completely after treatment with archazolid. In summary, archazolid treatment for 24 h was not cytotoxic to quiescent HUVECs, but inhibited the functionality of the v-ATPase.
We analyzed the adhesion of MDA-MB-231 cells onto HUVECs. Confluent HUVECs were treated with up to 1 nM archazolid for 24 h. Untreated MDA-MB-231 cells were labeled with Cell-Tracker Green CMFDA Dye. Interestingly, v-ATPase inhibition strongly increased the attachment of the metastatic breast carcinoma cell line MDA-MB-231 onto HUVECs after 10 and 120 min of adhesion (Fig 3A and 3B) . We also investigated the influence of archazolid on the transendothelial migration of MDA-MB-231 cells. HUVECs seeded in a Boyden chamber were treated with 1 nM archazolid for 24 h. CellTracker Green-labeled MDA-MB-231 cells (not treated with archazolid) were allowed to transmigrate through the endothelial monolayer for 24 h. As shown in Fig 3C, archazolid significantly decreased the transendothelial migration of MDA-MB-231 cells.
The influence of archazolid on tumor cell adhesion was not only studied in HUVECs, which represent macrovascular endothelial cells, but also in microvascular HMEC-1 cells. Moreover, besides the breast cancer cell line MDA-MB-231, also PC-3 prostate cancer cells were used as a second metastatic cancer cell line. Archazolid treatment of endothelial cells increased the attachment of MDA-MB-231 cells onto the HMEC-1 monolayer after 120 min of adhesion ( Fig 4A) and increased the attachment of PC-3 cells onto the HUVEC monolayer after 30 and 60 min of adhesion (Fig 4B) . Of note, the adhesion of non-metastatic Jurkat cells, an acute T cell leukemia cell line, remained unaffected after treatment of HUVECs with archazolid (S1A Fig). Taken together, archazolid treatment augmented the adhesive properties of both micro-and macrovascular endothelial cells for metastatic tumor cells, but not for non-metastatic ones. Of note, cancer cell adhesion onto archazolid-activated endothelial cells increased with the time of adhesion.
The adhesion of tumor cells onto the endothelium is in principle similar to that of leukocytes, but slightly differs in the molecules that mediate the adhesion process. Ligands for the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin were found to be expressed on tumor cells and to mediate tumor-endothelial cell interaction [21] . Inhibition of the v-ATPase might affect the expression of endothelial cell adhesion molecules on mRNA or protein levels. To determine the mRNA expression of ICAM-1, VCAM-1, E-selectin and Ncadherin, HUVECs were treated with archazolid for 12 h. TNFα is known to upregulate the expression of ICAM-1, VCAM-1 and E-selectin [32] and, thus, served as positive control. Quantitative real-time PCR showed that v-ATPase inhibition in HUVECs did not alter the mRNA levels of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5A) . The protein expression of these adhesion molecules on the surface of endothelial cells was analyzed by flow cytometry. Archazolid (1 nM, 24 h) did not affect the cell surface expression of ICAM-1, VCAM-1, E-selectin and N-cadherin (Fig 5B) .
Besides ICAM-1, VCAM-1, E-selectin and N-cadherin, also integrins are able to mediate the process of cell adhesion [33] [34] [35] . Since none of the cell adhesion molecules expressed on HUVECs were regulated upon archazolid treatment, we considered integrins as potential interaction partners. Within this protein family β1-integrins have been reported to mediate tumor cell adhesion onto quiescent endothelial cells [25] . In order to elucidate the role of β1-integrins for the archazolid-induced tumor cell adhesion, the integrin β1-subunit was blocked either on MDA-MB-231 cells, PC-3 cells or on HUVECs. (Of note, as in all experiments throughout this study, only endothelial cells were treated with archazolid.) After blocking β1-integrins on MDA-MB-231 or PC-3 cells, the archazolid-induced tumor cell adhesion was reduced almost to control level (Fig 6A and 6B , left panels), whereas blocking of β1-integrins on HUVECs had no significant effect on the increase of tumor cell adhesion by v-ATPase inhibition (Fig 6A and 6B , right panels).
Depending on their α subunit, β1-integrins have a variety of ligands including extracellular matrix (ECM) components such as collagen, fibronectin and laminin [35] . Therefore, we hypothesized that archazolid treatment of endothelial cells might lead to an upregulation of these components. MDA-MB-231 and PC-3 cells were allowed to adhere onto plastic that was coated with these ECM components. This cell adhesion assay revealed that MDA-MB-231 as well as PC-3 cells favor the interaction with the ECM component collagen, as the adhesion onto collagen is much higher than onto the uncoated plastic control (Fig 7A) . MDA-MB-231 and PC-3 cells also adhered to fibronectin-coated plastic, but to a much lesser extent compared to the collagen coating. Therefore, we focused on the interaction between these two tumor cell lines and collagen. Blocking of the integrin β1 subunit on MDA-MB-231 and PC-3 cells clearly abolished the interaction with collagen (Fig 7B) , indicating that the attachment of these tumor cells to collagen is mediated by β1-integrins.
Since collagen is the major ECM component MDA-MB-231 and PC-3 cells interact with, the next step was to prove whether v-ATPase inhibition influences the amount of collagen expressed by HUVECs as extracellular matrix. To detect collagen on the endothelial surface, archazolid-treated HUVECs were labeled with an antibody against collagen type I-IV on ice to prevent endocytosis and to ensure that the antibody does not bind to intracellular collagen. Interestingly, archazolid increased the amount of surface collagen on HUVECs by about 50% (Fig 7C) . Control stainings were performed using an antibody against the cytosolic p65 subunit of the transcription factor nuclear factor κB (NFκB) to show that intracellular proteins were not detected by this staining method (S2 Fig) .
It was reported that v-ATPase inhibition by archazolid impairs the activity of cathepsin B [28, 36] , a lysosomal enzyme that degrades extracellular matrix components including collagen [37] [38] [39] [40] [41] . As collagen is degraded by cathepsin B and the activation of cathepsin B depends on v-ATPase activity [28, [36] [37] [38] 42] , we suggested that an accumulation of collagen on the surface of endothelial cells might be a consequence of an impaired functionality of cathepsin B. Therefore, an enzyme activity assay based on the proteolysis of a fluorogenic cathepsin B substrate was performed. In archazolid-treated HUVECs and HMEC-1 the activity of cathepsin B was induce both the mRNA (1 ng/ml TNF) and the cell surface expression (10 ng/ml TNF) of ICAM-1, VCAM-1, E-selectin and Ncadherin.
https://doi.org/10.1371/journal.pone.0203053.g005 Inhibition of endothelial vATPase increases tumor cell adhesion to endothelial cells strongly decreased by approximately 50% compared to control cells at an archazolid concentration of 1 nM (Fig 8A) . In line with this result, western blot analysis showed that archazolid (1 nM) reduces the protein expression of the mature, active form of cathepsin B to less than 40% of the control in HUVECs (Fig 8B) . To proof whether the archazolid-induced tumor cell adhesion is a consequence of the decreased amount of cathepsin B, HUVECs were transfected with a plasmid coding for human cathepsin B or with the empty vector as control. After 48 h, the transfected cells were treated with 1 nM archazolid. The level of cathepsin B after transfection and treatment was assessed by western blot analysis (Fig 9A) . Overexpression of cathepsin B strongly diminished both the basal and the archazolid-induced adhesion of MDA-MB-231 cells (Fig 9B) .
Targeting the proton pump v-ATPase for cancer therapy has gained great interest since its inhibition was reported to reduce the invasiveness of cancer cells and, most importantly, also metastasis [8, 9] . Thus, intensive research related to v-ATPases was done in cancer cells, whereas there are only few studies investigating v-ATPases in endothelial cells indicating a role in migration, proliferation and possibly angiogenesis [22] [23] [24] . In the present study we used the myxobacterial natural product archazolid to investigate the consequences of v-ATPase inhibition in the endothelium on tumor-endothelial cell interactions.
For the first time, we were able to show a link between v-ATPase and the adhesion and transmigration properties of the endothelium. Inhibition of the v-ATPase in endothelial cells by archazolid significantly increased the adhesion of metastatic cancer cells and decreased the transendothelial migration of cancer cells which was attributed to augmented collagen levels on the surface on archazolid-treated endothelial cells. Of note, adhesion of the non-metastatic Jurkat cell line onto archazolid-treated endothelial cells remained unaffected. The archazolidinduced adhesion of tumor cells was independent from the endothelial cell adhesion molecules ICAM-1, VCAM-1, E-selectin and N-cadherin, as their expression was not regulated by the compound. However, we found that the archazolid-induced tumor cell adhesion was mediated by β1-integrins expressed on MDA-MB-231 breast cancer and PC-3 prostate cancer cells as blocking of the integrin β1 subunit on these tumor cells reversed the pro-adhesive effect of archazolid. In adhesion experiments on plastic coated with extracellular matrix components, we could show that MDA-MB-231 and PC-3 cells clearly favored the interaction with collagen, whereas the adhesion of non-metastatic Jurkat cells was largely independent from extracellular matrix proteins (S1B Fig). The different adhesion properties of metastatic cancer cells and Jurkat cells might be a result of the distinct integrin expression pattern of each cell line. MDA-MB-231 and PC-3 cells express α2β1-and α3β1-integrins, which represent collagen receptors [43, 44] , while Jurkat cells express α4β1-integrins but lack α2β1-, α3β1-integrins [44] . α4β1integrins are receptors for VCAM-1 and fibronectin [35] and it has been shown that Jurkat cells interact with human endothelial cells that express VCAM-1 after cytokine treatment or cells transfected with VCAM-1 [45] . Our results are in line with previous studies showing that α2β1-and α3β1-integrin expressing MDA-MB-231 and PC-3 cells were able to rapidly attach to collagen in the cortical bone matrix. In contrast, Jurkat cells were not able to adhere [44] and might preferentially interact with cell adhesion molecules rather than with ECM proteins. α2β1-and α3β1-integrins can additionally act as laminin receptors [46] and at least α3β1integrins recognize fibronectin [46, 47] . Though expressing receptors for fibronectin and laminin, MDA-MB-231 and PC-3 cells adhered to fibronectin to a much lesser extent and did not adhere to laminin, probably due to lower affinities to these extracellular matrix components.
Importantly, v-ATPase inhibition by archazolid increased the surface levels of the extracellular matrix component collagen, which might explain that the increase of MDA-MB-231 and PC-3 cells onto archazolid-treated HUVECs is independent of endothelial cell adhesion molecules. By performing a live cell proteolysis assay, Cavallo-Medved et al. demonstrated ECM degradation, in particular of gelatin and collagen IV, in association with active cathepsin B in caveolae of endothelial cells during tube formation [40] . In addition, recent studies reported that v-ATPase inhibition impairs the activity of cathepsin B in cancer cells [28, 36] . Therefore, we suggested that the accumulation of collagen on the endothelial surface might be a consequence of impaired cathepsin B activity or expression in endothelial cells. In fact, we confirmed the impairment of cathepsin B activity by archazolid as the expression levels of the mature active form of this enzyme was strongly reduced. Cathepsin B is synthesized as preprocathepsin B on membrane-bound ribosomes. Following transport to the Golgi apparatus, the preprocathepsin B is glycosylated with mannose-containing oligosaccharides. The targeting of procathepsin B to lysosomes is mannose-6-phosphate receptor-dependent and its dissociation from the receptor as well as its proteolytic processing into mature cathepsin B requires acidification of the compartment [48] . In cancer cells v-ATPase inhibition by archazolid impaired the mannose-6-phosphate receptor-mediated trafficking from the trans-Golgi network to prelysosomal compartments resulting in a decrease of active lysosomal proteases like cathepsin B [28] . We assumed that the archazolid-induced decrease in cathepsin B activity and expression was based on the same mechanism. Interestingly, overexpression of cathepsin B attenuated the archazolid-induced adhesion of breast cancer cells onto endothelial cells, indicating that the adhesion negatively correlates with the expression of cathepsin B.
As cathepsin B can also degrade other extracellular matrix components such as fibronectin and laminin [38, 49] , v-ATPase inhibition could lead to an accumulation of these proteins and an increased adhesion of cells expressing fibronectin or laminin receptors. However, we did not focus on these ECM components since they were not relevant for the adhesion of MDA-MB-231 and PC-3 cells. These cells predominantly adhered to collagen, while the adhesion of Jurkat cells is mostly independent from the ECM proteins collagen, fibronectin or laminin (S1B Fig). Interestingly [50] . In hepatic cancer cells, archazolid reduces Ras/Raf/MEK/ERK signaling by altering the membrane composition and fluidity [51] . We assume that archazolid affects endothelial cells in a similar way leading to inhibition of Ras signaling and, therefore, reduced transendothelial migration of MDA-MB-231 cells.
Taken together, our study shows that archazolid reduces the activity and expression of cathepsin B in endothelial cells. As a result, the amount of collagen on the surface of endothelial cells was significantly upregulated, which finally resulted in an increased adhesion of the β1-integrin-expressing metastatic cancer cell lines MDA-MB-231 and PC-3 onto archazolidtreated endothelial cells, whereas the adhesion of non-metastatic Jurkat cells was unaffected. This study shows that the v-ATPase plays an important role in regulating the adhesion of cells expressing receptors for extracellular matrix components. Archazolid represents a promising tool to elucidate the role of v-ATPase in endothelial cells. Moreover, we for the first time linked the function of v-ATPase to the adhesion and transmigration of tumor cells onto endothelial cells as well as to the remodeling of the extracellular matrix on the surface of endothelial cells. The fact that the adhesion of metastatic tumor cells onto endothelial cells is increased while their transendothelial migration is reduced upon inhibition of endothelial v-ATPase by archazolid further supports the view of archazolid as a potential anti-metastatic compound.
|
How were untreated MDA-MB-231 cells labeled?
| 5,306 |
CellTracker Green CMFDA Dye
| 10,518 |
1,634 |
Gastroenteritis and respiratory infection outbreaks in French nursing homes from 2007 to 2018: Morbidity and all-cause lethality according to the individual characteristics of residents
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6759171/
SHA: f1d456ea268266ff3c21317c4190e4fcb49b5e4f
Authors: Gaspard, Philippe; Mosnier, Anne; Simon, Loic; Ali-Brandmeyer, Olivia; Rabaud, Christian; Larocca, Sabrina; Heck, Béatrice; Aho-Glélé, Serge; Pothier, Pierre; Ambert-Balay, Katia
Date: 2019-09-24
DOI: 10.1371/journal.pone.0222321
License: cc-by
Abstract: BACKGROUND: Gastroenteritis (GE) and respiratory tract infection (RTI) outbreaks are a significant issue in nursing homes. This study aimed to describe GE and RTI outbreaks with infection and all-cause lethality rates according to the individual characteristics of nursing home residents. METHODS: Clinical and virological surveillance were conducted (2007 to 2018). Virus stratifications for the analysis were: outbreaks with positive norovirus or influenza identifications (respectively NoV+ or Flu+), episodes with no NoV or influenza identification or testing (respectively NoV- or Flu-). Associations between individual variables (sex, age, length of stay (LOS), autonomy status) and infection and lethality rates were tested with univariate and Mantel-Haenszel (MH) methods. RESULTS: 61 GE outbreaks and 76 RTI oubreaks (total 137 outbreaks) were recorded involving respectively 4309 and 5862 residents. In univariate analysis, higher infection rates and age were associated in NoV+, NoV-, and Flu+ contexts, and lower infection rates were associated with longer stays (NoV+ and NoV-). In MH stratified analysis (virus, sex (female/male)) adjusted for LOS (<4 or ≥4 years), the odds of being infected remained significant among older residents (≥86 years): NoV+/male (Odds ratio (OR(MH)): 1.64, 95% confidence interval (CI): 1.16–2.30) and Flu+/female and male (respectively OR(MH): 1.50, CI: 1.27–1.79 and 1.73, CI: 1.28–2.33). In univariate analysis, lower autonomy status (NoV+, Flu+ and Flu-) and increased age (Flu+) were associated with higher lethality. In MH adjusted analysis, significant OR(age) adjusted for autonomy was: Flu+/ ≥86 years compared with <86 years, 1.97 (1.19–3.25) and OR(autonomy) adjusted for age for the more autonomous group (compared with the less autonomous group) was: Flu+, 0.41 (0.24–0.69); Flu-, 0.42 (0.20, 0.90). CONCLUSION: The residents of nursing homes are increasingly elderly and dependent. The specific infection and lethality risks according to these two factors indicate that surveillance and infection control measures are essential and of high priority.
Text: Introduction Gastroenteritis (GE) and respiratory tract infection (RTI) outbreaks represent a significant burden of illness in nursing homes. Viruses cause the majority of these outbreaks, and noroviruses and influenza viruses are the most common pathogens [1, 2] .
Previous studies have suggested that viral respiratory infections and norovirus outbreaks are a common cause of hospitalization or death, particularly among elderly individuals [3] [4] [5] .
The impact of outbreaks has been described in terms of both frequency and epidemiology, but little is known about infection rates and all-cause lethality in GE and RTI nursing home outbreaks in relation to the individual characteristics of the residents [6] . The residents of these institutions are increasingly elderly and dependent, and the impact of this trend on the seasonal outbreak burden requires in-depth investigation. The results of studies focused on this issue could yield valuable information for nursing homes, allowing them to adapt their infection control strategies, in particular for improved assessment of infection risk.
Our objective was to describe GE and RTI infection and all-cause lethality rates according to the individual characteristics of nursing home residents (sex, age, length of stay, autonomy status), and to identify specific susceptibility patterns related to these types of viral outbreaks in these facilities.
The present study explored outbreaks in 14 sites (28 units with geriatric nursing home activities for a total of 1121 beds) caring for dependent people in southern Alsace (an area in northeastern France). Data were collected between September 2007 and August 2018 [7, 8] .
Each site was geographically independent and autonomous for social and care management. Units were located within the larger sites and were defined as a place having a dedicated team at one location.
During outbreaks at one site, only the residents in the units with confirmed cases were included. Outbreak inclusion depended on institutional alert to the hygiene team. Surveillance was done in each unit independently, and the members of staff had to inform a physician or charge nurse when two or more potential related cases of pneumonia or GE were observed within four days and when three or more cases were observed for other RTI. Units also had to inform the hygiene team when these threshold values were exceeded. For influenza, the first suspected case led to a local alert and the hygiene team was contacted. A practitioner from the hygiene team collected the information and evaluated the clinical signs, the virology information and the epidemiological context with the physician in the affected unit. The detected cluster was only put under surveillance if the hygiene team considered that there was a potential outbreak phenomenon. The duration of 4 days was in relation with the national protocol with alert to the authorities when 5 cases occurred within 4 days [9, 10] . On a local level and in addition to the clusters reported to the authorities, clusters with at least 3 cases within a period of seven days in one unit could be recorded if they were reported to the hygiene team. Because several outbreaks could potentially occur in the same unit during the surveillance period, a resident could be included repeatedly in different clusters. As a result, the observed patterns reflected the characteristics of an institutional population with longitudinal and pluriannual exposures.
Personal information and clinical information was collected by a practitioner from the hygiene team directly from the residents' health care records. Personal information was collected for all those present the first day of the outbreak. The collected information included: month and year of birth, sex, date of arrival at the nursing home and autonomy status. The autonomy status of residents in French nursing homes is assessed using the AGGIR scale (Autonomy Gerontology Groups Iso-Resources), which is the legal instrument for evaluating dependency in the elderly and whose primary purpose is the allocation of means and resources [11] .
With the AGGIR scale, autonomy is classified into 6 Iso-Resource Groups (GIR): GIR 1 (bedridden or armchair-bound persons, mental functions seriously altered and requiring continuous presence), GIR 2 (bedridden or armchair-bound persons, mental functions not totally altered and requiring assistance in most activities of daily living, or mental functions altered with preserved ability to get around), GIR 3 (preserved mental autonomy with partially preserved motor autonomy and assistance several times a day for physical autonomy), GIR 4 (moves around the home and sometimes assistance for washing, dressing, physical activities or eating), GIR 5 (only occasional assistance for washing, meal preparation, and housework), GIR 6 (autonomy for essential tasks of daily living). In outbreaks where the influenza virus was identified, influenza vaccination status and oseltamivir prescriptions were recorded as well. A file is transmitted in the Supporting Information with all previous data (S1 Data).
GE was defined as the sudden onset of vomiting and/or diarrhea over a 24 h period: (i) diarrhea �3 episodes, (ii) and/or vomiting �3 episodes, (iii) or diarrhea or vomiting <3 episodes with two or more other symptoms (diarrhea, vomiting, stomach ache, abdominal cramps, nausea, fever, mucus in stools) [1] .
RTI presentation in older adults may be atypical, like for other acute illnesses in this age group [12] . We used the recommended definitions for RTI surveillance in geriatric units, divided in 3 subcategories: (i) common cold syndromes or pharyngitis (at least two of the following criteria: runny nose or sneezing, stuffy nose (i.e. congestion), sore throat or hoarseness or difficulty swallowing, dry cough, swollen or tender glands in the neck (cervical lymphadenopathy)), (ii) influenza-like illness (both the following criteria must be met: fever AND at least three other symptoms (chills, new headache or eye pain, myalgia or body aches, malaise or loss of appetite, sore throat, new or increased dry cough)) and (iii) lower respiratory tract infection (both of the following criteria must be met: at least two respiratory signs or symptoms (new or increased cough, new/increased sputum production, O 2 saturation <94% or reduced >3% from baseline, abnormal lung examination (new or changed), pleuritic chest pain, respiratory rate �25 breaths/min AND one or more constitutional signs/symptoms (fever, leukocytosis, confusion, acute functional decline)) [13] [14] [15] [16] . Infection corresponding to one of these three subcategories was included in this study and classified as RTI.
For both infection types, the practitioner from the hygiene team obtained clinical information from the patient's health care records, and members of the health care team were consulted if necessary to complete any missing information. At the end of the episode (within seven days after the last identified case), case inclusion as exposed and not infected (ENI) or exposed and infected (EI) was determined with a resident physician.
In order to study the lethality, the presence of each infected resident was evaluated once at least 56 days after the last case of each outbreak. Each resident was followed up retrospectively during eighth 7-day interval (I n, n = 1 to 8, total 56 days, between the date of onset of symptoms and the fifty-sixth day of the studied period) with three different possibilities: present (alive and officially residing in the institution), lost to follow-up (alive at the date of departure but no longer residing in the institution (return home, transfer to another institution)) or death (death recorded in the health care record). The dates of death and lost to follow-up were recorded.
Testing for the virus was not systematic and was decided by the physicians in each institution in the presence of clinical signs.
For GE, stool samples were sent to the National Reference Centre for Gastroenteritis Viruses in Dijon for laboratory testing, as previously described [8] . For RTI surveillance, rapid tests were used to identify the influenza virus. The rapid immunoassay diagnosis tests used for influenza detection were: Clearview1 Exact Influenza A and B (Inverness Medical, Cologne, Germany) from 2007 to 2014 and InfluenzaTop1 (Alldiag, Strasbourg, France) from 2014 to 2018. Given the low sensitivity of influenza rapid tests, they were no longer used once the control measures had been implemented and the influenza outbreak was under control. Samples were also occasionally sent to hospital laboratories or to the National Reference Centre for Influenza Viruses to detect viruses with real-time RT-PCR [17] . Most testing targeted the norovirus (NoV) and influenza virus, but other tests were occasionally performed by the National Reference Centre for Influenza Viruses (rhinovirus, respiratory syncytial virus, human metapneumovirus, parainfluenza 1, 2, 3 and 4, and coronavirus) and the National Reference Centre for Enteric Viruses (rotavirus, astrovirus, and adenovirus).
Because testing for the viruses was variable (from one institution/physician to another, not used in some outbreaks, types of virus sought) and considering the poor sensitivity of the rapid influenza tests, these two sources of data were used to define the epidemiological context of confirmed outbreaks. Consequently, individual cases were included consistently in all episodes according to clinical signs and medical evaluation. When virus testing was negative, the clinical signs were recorded and medical evaluation was used as previously to classify the included residents as infected or not infected.
The epidemiological context of each outbreak was defined according to whether the virus had been identified or not. One or more positive samples led to the qualification of a NoV (NoV+) or flu (Flu+) context. The other episodes were qualified as flu or NoV outbreaks with no specific identification or testing (NoV-and Flu-).
Flu and NoV contexts did not eliminate other potential enteric or respiratory pathogens.
Sex, age, length of stay (LOS, in years) and autonomy status were described for all exposed residents. Influenza vaccination and oseltamivir administration rates were calculated for confirmed influenza outbreaks. Dichotomous or categorical variables were expressed as percentages. In univariate analysis, the categories were specific in order to obtain a precise description of the age and LOS variables. Class intervals were 5 years for age and one year for LOS. The residents classified as GIR 4 to 6 (sometimes, occasional and no assistance) were grouped together because they were few.
For the multi-level analysis with 2x2 tables, a median value was used to define the two-level age categories. For LOS, assessing the longest stays was necessary to identify the effect of longer exposure in a nursing home. Consequently, a four-year cutoff was chosen to create the two categories. For autonomy, the two most dependent categories (GIR � 2) were grouped together and compared with the more autonomous categories (GIR � 3).
The outbreak epidemiological contexts were used with the four categories: NoV+, Flu+, NoV-and flu-. Other GE or RTI viruses were occasionally identified, but the number of results was too limited to develop separate analyses. However, all the results are available in the tables about the virus investigations along with NoV and influenza identifications.
For the different categories, infection rate (EI/Exposed Residents (ER), in percentage) was calculated according to sex, age group, LOS and autonomy status. To investigate all-cause lethality and define the appropriate period for the 56-day monitoring (D 1 to 56 ), the all-cause lethality rate per 7-day interval (LR n /I n, n = 1 to 8 ) was calculated: (number of deaths during interval I n /(EI alive the first day of n th studied interval minus lost to follow-up EI during the interval I n ) � 100).
Seeing as successive clusters could occur within the same site, potentially influencing allcause lethality, the serial interval in days (SI d ) was calculated. The SI d was the time period between the onset of symptoms of the last case in initial outbreak (N) and the onset of symptoms of the first case in the following outbreak (N+1). Investigations were performed when SI d was shorter or equal to the length of the previous D 1 to 56 and the following parameters were evaluated for these specific situations: number of episodes, residents infected in both outbreaks, and death among the identified individuals.
Finally, according to the death rate and the impact of successive outbreaks, the number of 7-day intervals (N.I n ) to take into account was defined, and the all-cause lethality rate was analyzed during these periods (I 1 to N th .I n or D 1 to 7 � N ).
All-cause lethality rates were calculated with the following formula: (number of deaths from D 1 to 7 � N /(EI number at D 1 minus lost to follow-up among EI during the period D 1 to 7 � N ) � 100). Estimation of the turnover rate per 7-day interval among the infected residents was calculated on the base of the LOS (median in years) with the following formula: (proportion of discharged residents: 50.0% in the case of the median)/[(median LOS � 365)/7)]. The average rate of residents discharged per 7-day period was calculated: [(number of lost to follow up during the period D 1 to 7 � N /number of exposed and infected residents at D 1 )/N 7-day interval] � 100.
As some residents were included in several outbreaks during the surveillance, the observations were not completely independent; non-parametric tests were used as a result. In univariate analysis, Chi-square or Fisher exact tests (expected number of frequencies fewer than 5) were used to compare infection and lethality rates according to the studied parameters and the odds ratio was calculated by median-unbiased estimation. The Kruskal-Wallis test was used to compare median values. Confidence intervals for medians were calculated with bootstrap methods.
Covariate adjusted analyses were performed with two tables (2x2). The respective impact of each individual factor was tested with Mantel-Haenszel chi-squared tests. The equality of the stratum odds ratios was tested with the Woolf test of homogeneity. Finally, for each virus context, multiple tables (2x2) were generated and tested with confounding variables, effect modifiers or covariables. Statistical analyses were done using R for Mas OS X version R 3.4.1 software with RStudio version 1.0.153. A file is transmitted in the Supporting Information with all R codes and the packages used (S1 R Codes). Differences were considered significant at p � 0.05.
The French Data Protection Authority approved data collection and analysis (DE-2013-074) and the local ethics committee (Espace Local de Réflexion Ethique, Centre Hospitalier de Rouffach) approved the study protocol (ERLE-32). According to the French law for biomedical research and human experimentation, individual written consent was not required from the patients or their relatives for data collection. Each year, the referring local practitioner of the study coordinated with the doctors working in the nursing home. At the beginning of the surveillance period, information regarding participation in the study was displayed in the family vising area, including a document about their right to access and rectify personal data. After collection, data were rendered anonymous. No specific authorization was needed to retrospectively analyze anonymous data collected during routine care in the context of routine surveillance.
A total of 137 outbreaks were recorded in the 14 sites. RTI outbreaks were more frequent than GE outbreaks (76 outbreaks and 5862 exposed residents vs. 61 outbreaks and 4309 exposed residents, respectively). Overall, 7643 of the exposed residents were women and 2528 were men. The median age was 86.7 years old (interquartile range: 81.1-91.0 years).
Virus investigations (respectively 389 samples for RTI and 143 for GE with all the detailed results in S1-S4 Tables) confirmed a considerable number of norovirus-related GE outbreaks (34/61) and influenza-related RTI outbreaks (46/76). For GE outbreaks, 2524 residents were in a NoV+ context versus 1785 in a NoV-context, and for RTI outbreaks, 3479 residents were in a Flu+ context versus 2383 in a Flu-context.
For GE surveillance in the NoV+ context, there were 1093 EI residents versus 1431 ENI residents, whereas in the NoV-context, there were 583 EI residents versus 1202 ENI residents. Therefore, the infection rate was higher in the NoV+ context (43.3%) than in the NoV-context (32.7%, (odds ratio (OR): 0.63, 95% confidence interval (CI): 0.56-0.72), p < 0.001).
For RTI surveillance, the rates of infection were similar with and without confirmed influenza: 31.5% (N = 1095 EI residents /3479 exposed residents) vs. 30.5% (N = 728 EI residents/ 2383 exposed residents, OR: 0.96, CI: 0.85-1.07, p = 0.47). Moreover, infection rate in the NoV + context was higher than the three other contexts: NoV-(OR: 0.63, CI: 0.56-0.72), Flu+ (OR: 0.60, CI:0.54-0.67) and Flu-(OR: 0.58, CI: 0.51-0.65).
In univariate analysis, certain individual characteristics were associated with significant variations in the infection rate (S5 Table) . The infection rate increased with age (except in the Flucontext) and, decreased with LOS during GE outbreaks. The covariate adjusted analysis revealed specific significant effect modification according to sex (NoV+) and LOS (NoV-) (S6 Table) . In analyses stratified according to virus and sex, age adjusted for LOS remained significant for Flu+ and NoV+ outbreaks (males). In NoV-context, the effect modification of LOS remained significant (Table 1) . Finally, when autonomy was included and adjusted for age (virus, sex, LOS stratification), the less autonomous residents (female/LOS<4 years/age<86/ GIR 1-2) were affected more severely by Flu+ outbreaks with specific effect modification according to age (S7 Table) . The study of lethality rates in infected residents over the 56 days after onset indicated that there were significant variations for RTI but no change for GE (Table 2 ). Significant differences appeared after 28 days in the context of Flu+ outbreaks and other RTI outbreaks.
The analysis of successive or simultaneous clusters in the same institutions was performed when the time period between the onset of symptoms of the last case in outbreak N and the onset of symptoms of the first case in outbreak N+1 was �56 days (S8 Table) . 44 of the 137 outbreaks (32.12%) were identified, and 194 of the 3499 exposed and infected residents contracted multiple infections. The percentage of exposed and infected residents implicated in more than one virus stratification was 11.09% ((194 � 2)/3499). Moreover, two deceased residents were included in the NoV-Na and Flu lethality analyses because death occurred within 56 days for both infections. The analysis of virus stratification of the 44 outbreaks showed the absence of successive clusters for the same category. The same analysis for the first four 7-day intervals (Days 1 to 28) showed the respective values: 26 outbreaks (18.98%), 117 residents ((6.69% ((117 � 2)/3499)), one dead resident.
Finally, according to the higher lethality impact during the first four 7-day intervals and to limit the impact of successive clusters in the same site, all cause lethality rates were studied according to individual parameters for the four 7 days intervals with the respective number of According to the surveillance type (GE or RTI), the lethality rates differed significantly: 1.6% versus 3.4% (respectively NoV+ and NoV-contexts, OR: 2.24, CI: 1.16-4.39, p = 0.02) and 8.3% versus 5.6% (respectively Flu+ and Flu-, OR: 0.67, CI: 0.45-0.97, p = 0.04).
In univariate analysis (S9 Table) , low autonomy status in the NoV+, Flu+ and Flu-contexts was most significantly associated with increased all-cause lethality, and age was associated with higher lethality in the Flu+ context. In the adjusted analysis, no significant statistical differences were identified in GE outbreaks. For RTI episodes, the adjusted analysis showed that autonomy had a significant impact when adjusted for sex, age or LOS (Flu+ and Flu-NA) and that age had a significant impact when adjusted for sex, autonomy or LOS (Flu+) (S10 Table) .
In Table 3 , the specific effects of age or autonomy were tested. Significant OR age adjusted for autonomy were: Flu+/age �86 years (compared with the <86 group), 1.97 (1.19-3.25). OR autonomy adjusted for age were for GIR 3-6 (compared with GIR 1-2): Flu+, 0.41 (0.24-0.69); Flu-, 0.42 (0.20, 0.90).
Finally, despite the low number of residents and deaths per category, and consequently the limited robustness of the results, autonomy adjusted for age with stratification according to virus, sex and LOS showed that the effects were higher among subgroups of less autonomous residents (female or male/LOS<4 years/GIR 1-2) in Flu+ outbreaks, and there was also higher mortality in the small subgroup of autonomous men with LOS � 4 years (higher mortality) (S11 Table) .
In the Flu+ context, data regarding vaccination status and oseltamivir prescriptions were available but not used in this study.
In the present study, surveillance data obtained during GE and RTI outbreaks in nursing homes were used to construct stratified analyses and to identify specific infection and all-cause lethality rates according to the residents' individual characteristics.
The infection rates observed here were similar to those found in previous studies of NoV and Influenza outbreaks (odds of being infected during a Flu+ outbreak were around 40% less than during a NoV+ outbreak). Reported infection rates were close to 30.0% in influenza outbreaks and 40.0% in NoV outbreaks [18] [19] [20] .
Older age appeared to increase the likelihood of GE and influenza infection, with increasing rates among older residents. Age is a well-known factor for influenza and norovirus severity in the elderly and in nursing homes [21, 22] . For NoV, the highest incidence estimates (5-year age strata) was found in the �85 year-category (approximately 800 men and for 1,400 women per 100,000 inhabitants). In our study, univariate analysis (NoV+) showed that the odds of being infected were 1.5 to 1.6 times higher if a resident was older than 85. Moreover, an adjusted analysis of GE outbreaks highlighted different effects among subgroups of residents according to sex and LOS. Indeed, multiple and/or repeated exposure to GE viruses while institutionalized may lead to susceptibility or possible increased immunity in some residents [23] . For the sex variable, two factors could explain the effect: a possible selection bias with men reporting mild infections less than women (particularly in the <86 years subgroup) or that male susceptibility was different (age, LOS, immunity,. . .). A German study from 2013 also reported a greater impact in women [21] . Moreover, when age analysis was stratified by sex and LOS, no Epidemic impacts in nursing homes significant impact was observed in women; the only significant differences were fewer infections in men in the <86-subgroup (except in the NoV-with LOS <4 years).
For the RTI outbreaks, sex and LOS variables did not have a significant effect. In residents older than 86, the odds of being infected in Flu+ context were 1.5 times higher for women and 1.7 for men. In univariate analysis, contrary to the other virus contexts where odds ratios were rarely above 2, residents over 95 years old had increased odds of infection of � 2.8 compared with the 70-year-old category, and for the 100 year-old group the odds were approximately 3.8. In the Flu+ context, autonomy adjusted for age (virus, sex and LOS stratification) revealed a possible increase in infection rates among less autonomous residents. A previous study found that when elderly residents were exposed to the A(H3N2) virus, there were higher rates of infection and reinfection, and more significant effects on the institution than with other influenza types/subtypes. In the community, the relative illness ratio (RIR) in the [22, 24] .
The incidence of influenza infection and the associated risks were well described by age group, but the specific impact according to age was not studied. The results of this work highlighted the specific age distribution of influenza illnesses among the nursing home residents and the more significant impact among the older residents. This specific susceptibility could be a critical factor in the institutional exposure and dissemination of influenza and could partly explain the high infection impact in the elderly institutional population.
In this work, autonomy status was not the main factor associated with infection (no significant impact in GE and in Flu-contexts). However, in Flu+ outbreaks, a high level of dependency was associated with a higher risk of falling ill. This observation implies that staff could play a role in the spread of infection (highly dependent and less mobile residents are less likely to contaminate themselves) or that the more active residents may be less fragile and/or have a greater involvement in the recommended infection control measures. Finally, improving compliance with personal hygiene measures both for nursing staff and residents might be expected to have a beneficial effect on infection rates. Previous studies identified higher NoV infection rates in highly dependent individuals, but the results were not adjusted for age and LOS to take into account the potential correlation with the autonomy status [20] .
Lethality is difficult to assess in nursing homes because death is frequent. Our GE and RTI episodes occurred during the winter seasons, and there are possible interactions between outbreaks and increased mortality at this time of the year [25] . The all-cause lethality rate of the infected residents in our study reflected global mortality including GE and RTI outbreaks and the global epidemiological context. Not surprisingly, a higher all-cause lethality rate was observed in the influenza contexts, as reported in previous studies [25] [26] [27] . Age and autonomy had similar effects in the different contexts, but in GE and to a lesser degree in Flu-outbreaks, the relatively small number of deaths could have limited the power of the statistical tests. In nursing homes, residents are generally discharged due to death. The number of residents lost to follow up was low (0.14% in the first 28 days), so the 7-day interval turnover rate calculated on the base of the median LOS provides a good indication of the average case fatality rate. The lethality rate for NoV+ outbreaks was similar to the estimated 7-day interval turnover rate (1.6%) indicating that this context had a limited impact on the death rate.
The all-cause lethality rate was most affected by age and autonomy. Both individual characteristics were significant in the Flu+ outbreaks, and autonomy adjusted for age was significant in the Flu-episodes. The influence of age on mortality in a context of influenza has already been described: a very high mortality rate (831/100,000 inhabitants) was reported in persons 90 years of age and older compared with those aged 65-69 years (23/100,000 inhabitants) [28] . In our univariate analysis, the higher risk was observed in the �90 group whose risk of death was at least 2.6 higher than the <70 group. When adjusted for autonomy, the impact of age was not significant in more autonomous residents in the Flu+ context and not at all in the Flucontext. The opposite analysis (autonomy adjusted for age) showed higher global impact in the less autonomous group (Flu-) or only in the �86 age group (Flu+). Age and autonomy are a reflection of resident's level of frailty. Clinical frailty scores were not used in this study, but in a previous study of patients with critical illness, they were associated with greater mortality, regardless of age [29] . This suggests that in addition to age, autonomy can be a valuable indicator for the assessment of outbreak impact in outbreak surveillance. Other studies have suggested that age and certain comorbidities are independent risk factors for the influenza mortality rate or that mortality increase according to the number of risk factors [28, 30] . Comorbidities and underlying diseases of various severities could reflect overall frailty and consequently the risk of death. In nursing homes, information about autonomy and age are easier to collect and interpret than data on comorbidities. These various approaches should be evaluated and compared in the goal of optimizing risk assessment among nursing home residents.
The present work has two main limitations. First, the virus information was incomplete (limited identification, mainly influenza rapid tests for the RTI). Consequently, some episodes in the levels with no available identification may also have been associated with influenza or norovirus, and multiple contaminations could have been underestimated or not taken into account. Moreover, vaccination and oseltamivir prescriptions were recorded but not included because the influenza genotype was not determined and identification was limited.
Secondly, the deaths of uninfected residents were not recorded in this protocol even though such data would have provided valuable information about the global epidemiological context.
In conclusion, specific susceptibility patterns were observed among exposed residents. In this cohort of nursing homes, infection rates varied according to virus, sex, length of stay and age, and there were major differences in lethality depending on virus, age and autonomy score. The collected data were easy to record and could be used to improve the characterization of seasonal outbreaks in nursing homes, whose residents are particularly vulnerable. Finally, as the average age and dependency level of residents continues to increase, subsequently increasing the risk of infection and death, health care staff will have to be increasingly vigilant during seasonal outbreaks and targeted interventions should be implemented.
Supporting information S1 Data. (CSV) S1 R Codes. (R) S1 Table. (XLSX) S10 Table. (XLSX) S11 Table. (XLSX)
|
What was the goal of the study?
| 5,307 |
describe GE and RTI outbreaks with infection and all-cause lethality rates according to the individual characteristics of nursing home residents
| 697 |
1,634 |
Gastroenteritis and respiratory infection outbreaks in French nursing homes from 2007 to 2018: Morbidity and all-cause lethality according to the individual characteristics of residents
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6759171/
SHA: f1d456ea268266ff3c21317c4190e4fcb49b5e4f
Authors: Gaspard, Philippe; Mosnier, Anne; Simon, Loic; Ali-Brandmeyer, Olivia; Rabaud, Christian; Larocca, Sabrina; Heck, Béatrice; Aho-Glélé, Serge; Pothier, Pierre; Ambert-Balay, Katia
Date: 2019-09-24
DOI: 10.1371/journal.pone.0222321
License: cc-by
Abstract: BACKGROUND: Gastroenteritis (GE) and respiratory tract infection (RTI) outbreaks are a significant issue in nursing homes. This study aimed to describe GE and RTI outbreaks with infection and all-cause lethality rates according to the individual characteristics of nursing home residents. METHODS: Clinical and virological surveillance were conducted (2007 to 2018). Virus stratifications for the analysis were: outbreaks with positive norovirus or influenza identifications (respectively NoV+ or Flu+), episodes with no NoV or influenza identification or testing (respectively NoV- or Flu-). Associations between individual variables (sex, age, length of stay (LOS), autonomy status) and infection and lethality rates were tested with univariate and Mantel-Haenszel (MH) methods. RESULTS: 61 GE outbreaks and 76 RTI oubreaks (total 137 outbreaks) were recorded involving respectively 4309 and 5862 residents. In univariate analysis, higher infection rates and age were associated in NoV+, NoV-, and Flu+ contexts, and lower infection rates were associated with longer stays (NoV+ and NoV-). In MH stratified analysis (virus, sex (female/male)) adjusted for LOS (<4 or ≥4 years), the odds of being infected remained significant among older residents (≥86 years): NoV+/male (Odds ratio (OR(MH)): 1.64, 95% confidence interval (CI): 1.16–2.30) and Flu+/female and male (respectively OR(MH): 1.50, CI: 1.27–1.79 and 1.73, CI: 1.28–2.33). In univariate analysis, lower autonomy status (NoV+, Flu+ and Flu-) and increased age (Flu+) were associated with higher lethality. In MH adjusted analysis, significant OR(age) adjusted for autonomy was: Flu+/ ≥86 years compared with <86 years, 1.97 (1.19–3.25) and OR(autonomy) adjusted for age for the more autonomous group (compared with the less autonomous group) was: Flu+, 0.41 (0.24–0.69); Flu-, 0.42 (0.20, 0.90). CONCLUSION: The residents of nursing homes are increasingly elderly and dependent. The specific infection and lethality risks according to these two factors indicate that surveillance and infection control measures are essential and of high priority.
Text: Introduction Gastroenteritis (GE) and respiratory tract infection (RTI) outbreaks represent a significant burden of illness in nursing homes. Viruses cause the majority of these outbreaks, and noroviruses and influenza viruses are the most common pathogens [1, 2] .
Previous studies have suggested that viral respiratory infections and norovirus outbreaks are a common cause of hospitalization or death, particularly among elderly individuals [3] [4] [5] .
The impact of outbreaks has been described in terms of both frequency and epidemiology, but little is known about infection rates and all-cause lethality in GE and RTI nursing home outbreaks in relation to the individual characteristics of the residents [6] . The residents of these institutions are increasingly elderly and dependent, and the impact of this trend on the seasonal outbreak burden requires in-depth investigation. The results of studies focused on this issue could yield valuable information for nursing homes, allowing them to adapt their infection control strategies, in particular for improved assessment of infection risk.
Our objective was to describe GE and RTI infection and all-cause lethality rates according to the individual characteristics of nursing home residents (sex, age, length of stay, autonomy status), and to identify specific susceptibility patterns related to these types of viral outbreaks in these facilities.
The present study explored outbreaks in 14 sites (28 units with geriatric nursing home activities for a total of 1121 beds) caring for dependent people in southern Alsace (an area in northeastern France). Data were collected between September 2007 and August 2018 [7, 8] .
Each site was geographically independent and autonomous for social and care management. Units were located within the larger sites and were defined as a place having a dedicated team at one location.
During outbreaks at one site, only the residents in the units with confirmed cases were included. Outbreak inclusion depended on institutional alert to the hygiene team. Surveillance was done in each unit independently, and the members of staff had to inform a physician or charge nurse when two or more potential related cases of pneumonia or GE were observed within four days and when three or more cases were observed for other RTI. Units also had to inform the hygiene team when these threshold values were exceeded. For influenza, the first suspected case led to a local alert and the hygiene team was contacted. A practitioner from the hygiene team collected the information and evaluated the clinical signs, the virology information and the epidemiological context with the physician in the affected unit. The detected cluster was only put under surveillance if the hygiene team considered that there was a potential outbreak phenomenon. The duration of 4 days was in relation with the national protocol with alert to the authorities when 5 cases occurred within 4 days [9, 10] . On a local level and in addition to the clusters reported to the authorities, clusters with at least 3 cases within a period of seven days in one unit could be recorded if they were reported to the hygiene team. Because several outbreaks could potentially occur in the same unit during the surveillance period, a resident could be included repeatedly in different clusters. As a result, the observed patterns reflected the characteristics of an institutional population with longitudinal and pluriannual exposures.
Personal information and clinical information was collected by a practitioner from the hygiene team directly from the residents' health care records. Personal information was collected for all those present the first day of the outbreak. The collected information included: month and year of birth, sex, date of arrival at the nursing home and autonomy status. The autonomy status of residents in French nursing homes is assessed using the AGGIR scale (Autonomy Gerontology Groups Iso-Resources), which is the legal instrument for evaluating dependency in the elderly and whose primary purpose is the allocation of means and resources [11] .
With the AGGIR scale, autonomy is classified into 6 Iso-Resource Groups (GIR): GIR 1 (bedridden or armchair-bound persons, mental functions seriously altered and requiring continuous presence), GIR 2 (bedridden or armchair-bound persons, mental functions not totally altered and requiring assistance in most activities of daily living, or mental functions altered with preserved ability to get around), GIR 3 (preserved mental autonomy with partially preserved motor autonomy and assistance several times a day for physical autonomy), GIR 4 (moves around the home and sometimes assistance for washing, dressing, physical activities or eating), GIR 5 (only occasional assistance for washing, meal preparation, and housework), GIR 6 (autonomy for essential tasks of daily living). In outbreaks where the influenza virus was identified, influenza vaccination status and oseltamivir prescriptions were recorded as well. A file is transmitted in the Supporting Information with all previous data (S1 Data).
GE was defined as the sudden onset of vomiting and/or diarrhea over a 24 h period: (i) diarrhea �3 episodes, (ii) and/or vomiting �3 episodes, (iii) or diarrhea or vomiting <3 episodes with two or more other symptoms (diarrhea, vomiting, stomach ache, abdominal cramps, nausea, fever, mucus in stools) [1] .
RTI presentation in older adults may be atypical, like for other acute illnesses in this age group [12] . We used the recommended definitions for RTI surveillance in geriatric units, divided in 3 subcategories: (i) common cold syndromes or pharyngitis (at least two of the following criteria: runny nose or sneezing, stuffy nose (i.e. congestion), sore throat or hoarseness or difficulty swallowing, dry cough, swollen or tender glands in the neck (cervical lymphadenopathy)), (ii) influenza-like illness (both the following criteria must be met: fever AND at least three other symptoms (chills, new headache or eye pain, myalgia or body aches, malaise or loss of appetite, sore throat, new or increased dry cough)) and (iii) lower respiratory tract infection (both of the following criteria must be met: at least two respiratory signs or symptoms (new or increased cough, new/increased sputum production, O 2 saturation <94% or reduced >3% from baseline, abnormal lung examination (new or changed), pleuritic chest pain, respiratory rate �25 breaths/min AND one or more constitutional signs/symptoms (fever, leukocytosis, confusion, acute functional decline)) [13] [14] [15] [16] . Infection corresponding to one of these three subcategories was included in this study and classified as RTI.
For both infection types, the practitioner from the hygiene team obtained clinical information from the patient's health care records, and members of the health care team were consulted if necessary to complete any missing information. At the end of the episode (within seven days after the last identified case), case inclusion as exposed and not infected (ENI) or exposed and infected (EI) was determined with a resident physician.
In order to study the lethality, the presence of each infected resident was evaluated once at least 56 days after the last case of each outbreak. Each resident was followed up retrospectively during eighth 7-day interval (I n, n = 1 to 8, total 56 days, between the date of onset of symptoms and the fifty-sixth day of the studied period) with three different possibilities: present (alive and officially residing in the institution), lost to follow-up (alive at the date of departure but no longer residing in the institution (return home, transfer to another institution)) or death (death recorded in the health care record). The dates of death and lost to follow-up were recorded.
Testing for the virus was not systematic and was decided by the physicians in each institution in the presence of clinical signs.
For GE, stool samples were sent to the National Reference Centre for Gastroenteritis Viruses in Dijon for laboratory testing, as previously described [8] . For RTI surveillance, rapid tests were used to identify the influenza virus. The rapid immunoassay diagnosis tests used for influenza detection were: Clearview1 Exact Influenza A and B (Inverness Medical, Cologne, Germany) from 2007 to 2014 and InfluenzaTop1 (Alldiag, Strasbourg, France) from 2014 to 2018. Given the low sensitivity of influenza rapid tests, they were no longer used once the control measures had been implemented and the influenza outbreak was under control. Samples were also occasionally sent to hospital laboratories or to the National Reference Centre for Influenza Viruses to detect viruses with real-time RT-PCR [17] . Most testing targeted the norovirus (NoV) and influenza virus, but other tests were occasionally performed by the National Reference Centre for Influenza Viruses (rhinovirus, respiratory syncytial virus, human metapneumovirus, parainfluenza 1, 2, 3 and 4, and coronavirus) and the National Reference Centre for Enteric Viruses (rotavirus, astrovirus, and adenovirus).
Because testing for the viruses was variable (from one institution/physician to another, not used in some outbreaks, types of virus sought) and considering the poor sensitivity of the rapid influenza tests, these two sources of data were used to define the epidemiological context of confirmed outbreaks. Consequently, individual cases were included consistently in all episodes according to clinical signs and medical evaluation. When virus testing was negative, the clinical signs were recorded and medical evaluation was used as previously to classify the included residents as infected or not infected.
The epidemiological context of each outbreak was defined according to whether the virus had been identified or not. One or more positive samples led to the qualification of a NoV (NoV+) or flu (Flu+) context. The other episodes were qualified as flu or NoV outbreaks with no specific identification or testing (NoV-and Flu-).
Flu and NoV contexts did not eliminate other potential enteric or respiratory pathogens.
Sex, age, length of stay (LOS, in years) and autonomy status were described for all exposed residents. Influenza vaccination and oseltamivir administration rates were calculated for confirmed influenza outbreaks. Dichotomous or categorical variables were expressed as percentages. In univariate analysis, the categories were specific in order to obtain a precise description of the age and LOS variables. Class intervals were 5 years for age and one year for LOS. The residents classified as GIR 4 to 6 (sometimes, occasional and no assistance) were grouped together because they were few.
For the multi-level analysis with 2x2 tables, a median value was used to define the two-level age categories. For LOS, assessing the longest stays was necessary to identify the effect of longer exposure in a nursing home. Consequently, a four-year cutoff was chosen to create the two categories. For autonomy, the two most dependent categories (GIR � 2) were grouped together and compared with the more autonomous categories (GIR � 3).
The outbreak epidemiological contexts were used with the four categories: NoV+, Flu+, NoV-and flu-. Other GE or RTI viruses were occasionally identified, but the number of results was too limited to develop separate analyses. However, all the results are available in the tables about the virus investigations along with NoV and influenza identifications.
For the different categories, infection rate (EI/Exposed Residents (ER), in percentage) was calculated according to sex, age group, LOS and autonomy status. To investigate all-cause lethality and define the appropriate period for the 56-day monitoring (D 1 to 56 ), the all-cause lethality rate per 7-day interval (LR n /I n, n = 1 to 8 ) was calculated: (number of deaths during interval I n /(EI alive the first day of n th studied interval minus lost to follow-up EI during the interval I n ) � 100).
Seeing as successive clusters could occur within the same site, potentially influencing allcause lethality, the serial interval in days (SI d ) was calculated. The SI d was the time period between the onset of symptoms of the last case in initial outbreak (N) and the onset of symptoms of the first case in the following outbreak (N+1). Investigations were performed when SI d was shorter or equal to the length of the previous D 1 to 56 and the following parameters were evaluated for these specific situations: number of episodes, residents infected in both outbreaks, and death among the identified individuals.
Finally, according to the death rate and the impact of successive outbreaks, the number of 7-day intervals (N.I n ) to take into account was defined, and the all-cause lethality rate was analyzed during these periods (I 1 to N th .I n or D 1 to 7 � N ).
All-cause lethality rates were calculated with the following formula: (number of deaths from D 1 to 7 � N /(EI number at D 1 minus lost to follow-up among EI during the period D 1 to 7 � N ) � 100). Estimation of the turnover rate per 7-day interval among the infected residents was calculated on the base of the LOS (median in years) with the following formula: (proportion of discharged residents: 50.0% in the case of the median)/[(median LOS � 365)/7)]. The average rate of residents discharged per 7-day period was calculated: [(number of lost to follow up during the period D 1 to 7 � N /number of exposed and infected residents at D 1 )/N 7-day interval] � 100.
As some residents were included in several outbreaks during the surveillance, the observations were not completely independent; non-parametric tests were used as a result. In univariate analysis, Chi-square or Fisher exact tests (expected number of frequencies fewer than 5) were used to compare infection and lethality rates according to the studied parameters and the odds ratio was calculated by median-unbiased estimation. The Kruskal-Wallis test was used to compare median values. Confidence intervals for medians were calculated with bootstrap methods.
Covariate adjusted analyses were performed with two tables (2x2). The respective impact of each individual factor was tested with Mantel-Haenszel chi-squared tests. The equality of the stratum odds ratios was tested with the Woolf test of homogeneity. Finally, for each virus context, multiple tables (2x2) were generated and tested with confounding variables, effect modifiers or covariables. Statistical analyses were done using R for Mas OS X version R 3.4.1 software with RStudio version 1.0.153. A file is transmitted in the Supporting Information with all R codes and the packages used (S1 R Codes). Differences were considered significant at p � 0.05.
The French Data Protection Authority approved data collection and analysis (DE-2013-074) and the local ethics committee (Espace Local de Réflexion Ethique, Centre Hospitalier de Rouffach) approved the study protocol (ERLE-32). According to the French law for biomedical research and human experimentation, individual written consent was not required from the patients or their relatives for data collection. Each year, the referring local practitioner of the study coordinated with the doctors working in the nursing home. At the beginning of the surveillance period, information regarding participation in the study was displayed in the family vising area, including a document about their right to access and rectify personal data. After collection, data were rendered anonymous. No specific authorization was needed to retrospectively analyze anonymous data collected during routine care in the context of routine surveillance.
A total of 137 outbreaks were recorded in the 14 sites. RTI outbreaks were more frequent than GE outbreaks (76 outbreaks and 5862 exposed residents vs. 61 outbreaks and 4309 exposed residents, respectively). Overall, 7643 of the exposed residents were women and 2528 were men. The median age was 86.7 years old (interquartile range: 81.1-91.0 years).
Virus investigations (respectively 389 samples for RTI and 143 for GE with all the detailed results in S1-S4 Tables) confirmed a considerable number of norovirus-related GE outbreaks (34/61) and influenza-related RTI outbreaks (46/76). For GE outbreaks, 2524 residents were in a NoV+ context versus 1785 in a NoV-context, and for RTI outbreaks, 3479 residents were in a Flu+ context versus 2383 in a Flu-context.
For GE surveillance in the NoV+ context, there were 1093 EI residents versus 1431 ENI residents, whereas in the NoV-context, there were 583 EI residents versus 1202 ENI residents. Therefore, the infection rate was higher in the NoV+ context (43.3%) than in the NoV-context (32.7%, (odds ratio (OR): 0.63, 95% confidence interval (CI): 0.56-0.72), p < 0.001).
For RTI surveillance, the rates of infection were similar with and without confirmed influenza: 31.5% (N = 1095 EI residents /3479 exposed residents) vs. 30.5% (N = 728 EI residents/ 2383 exposed residents, OR: 0.96, CI: 0.85-1.07, p = 0.47). Moreover, infection rate in the NoV + context was higher than the three other contexts: NoV-(OR: 0.63, CI: 0.56-0.72), Flu+ (OR: 0.60, CI:0.54-0.67) and Flu-(OR: 0.58, CI: 0.51-0.65).
In univariate analysis, certain individual characteristics were associated with significant variations in the infection rate (S5 Table) . The infection rate increased with age (except in the Flucontext) and, decreased with LOS during GE outbreaks. The covariate adjusted analysis revealed specific significant effect modification according to sex (NoV+) and LOS (NoV-) (S6 Table) . In analyses stratified according to virus and sex, age adjusted for LOS remained significant for Flu+ and NoV+ outbreaks (males). In NoV-context, the effect modification of LOS remained significant (Table 1) . Finally, when autonomy was included and adjusted for age (virus, sex, LOS stratification), the less autonomous residents (female/LOS<4 years/age<86/ GIR 1-2) were affected more severely by Flu+ outbreaks with specific effect modification according to age (S7 Table) . The study of lethality rates in infected residents over the 56 days after onset indicated that there were significant variations for RTI but no change for GE (Table 2 ). Significant differences appeared after 28 days in the context of Flu+ outbreaks and other RTI outbreaks.
The analysis of successive or simultaneous clusters in the same institutions was performed when the time period between the onset of symptoms of the last case in outbreak N and the onset of symptoms of the first case in outbreak N+1 was �56 days (S8 Table) . 44 of the 137 outbreaks (32.12%) were identified, and 194 of the 3499 exposed and infected residents contracted multiple infections. The percentage of exposed and infected residents implicated in more than one virus stratification was 11.09% ((194 � 2)/3499). Moreover, two deceased residents were included in the NoV-Na and Flu lethality analyses because death occurred within 56 days for both infections. The analysis of virus stratification of the 44 outbreaks showed the absence of successive clusters for the same category. The same analysis for the first four 7-day intervals (Days 1 to 28) showed the respective values: 26 outbreaks (18.98%), 117 residents ((6.69% ((117 � 2)/3499)), one dead resident.
Finally, according to the higher lethality impact during the first four 7-day intervals and to limit the impact of successive clusters in the same site, all cause lethality rates were studied according to individual parameters for the four 7 days intervals with the respective number of According to the surveillance type (GE or RTI), the lethality rates differed significantly: 1.6% versus 3.4% (respectively NoV+ and NoV-contexts, OR: 2.24, CI: 1.16-4.39, p = 0.02) and 8.3% versus 5.6% (respectively Flu+ and Flu-, OR: 0.67, CI: 0.45-0.97, p = 0.04).
In univariate analysis (S9 Table) , low autonomy status in the NoV+, Flu+ and Flu-contexts was most significantly associated with increased all-cause lethality, and age was associated with higher lethality in the Flu+ context. In the adjusted analysis, no significant statistical differences were identified in GE outbreaks. For RTI episodes, the adjusted analysis showed that autonomy had a significant impact when adjusted for sex, age or LOS (Flu+ and Flu-NA) and that age had a significant impact when adjusted for sex, autonomy or LOS (Flu+) (S10 Table) .
In Table 3 , the specific effects of age or autonomy were tested. Significant OR age adjusted for autonomy were: Flu+/age �86 years (compared with the <86 group), 1.97 (1.19-3.25). OR autonomy adjusted for age were for GIR 3-6 (compared with GIR 1-2): Flu+, 0.41 (0.24-0.69); Flu-, 0.42 (0.20, 0.90).
Finally, despite the low number of residents and deaths per category, and consequently the limited robustness of the results, autonomy adjusted for age with stratification according to virus, sex and LOS showed that the effects were higher among subgroups of less autonomous residents (female or male/LOS<4 years/GIR 1-2) in Flu+ outbreaks, and there was also higher mortality in the small subgroup of autonomous men with LOS � 4 years (higher mortality) (S11 Table) .
In the Flu+ context, data regarding vaccination status and oseltamivir prescriptions were available but not used in this study.
In the present study, surveillance data obtained during GE and RTI outbreaks in nursing homes were used to construct stratified analyses and to identify specific infection and all-cause lethality rates according to the residents' individual characteristics.
The infection rates observed here were similar to those found in previous studies of NoV and Influenza outbreaks (odds of being infected during a Flu+ outbreak were around 40% less than during a NoV+ outbreak). Reported infection rates were close to 30.0% in influenza outbreaks and 40.0% in NoV outbreaks [18] [19] [20] .
Older age appeared to increase the likelihood of GE and influenza infection, with increasing rates among older residents. Age is a well-known factor for influenza and norovirus severity in the elderly and in nursing homes [21, 22] . For NoV, the highest incidence estimates (5-year age strata) was found in the �85 year-category (approximately 800 men and for 1,400 women per 100,000 inhabitants). In our study, univariate analysis (NoV+) showed that the odds of being infected were 1.5 to 1.6 times higher if a resident was older than 85. Moreover, an adjusted analysis of GE outbreaks highlighted different effects among subgroups of residents according to sex and LOS. Indeed, multiple and/or repeated exposure to GE viruses while institutionalized may lead to susceptibility or possible increased immunity in some residents [23] . For the sex variable, two factors could explain the effect: a possible selection bias with men reporting mild infections less than women (particularly in the <86 years subgroup) or that male susceptibility was different (age, LOS, immunity,. . .). A German study from 2013 also reported a greater impact in women [21] . Moreover, when age analysis was stratified by sex and LOS, no Epidemic impacts in nursing homes significant impact was observed in women; the only significant differences were fewer infections in men in the <86-subgroup (except in the NoV-with LOS <4 years).
For the RTI outbreaks, sex and LOS variables did not have a significant effect. In residents older than 86, the odds of being infected in Flu+ context were 1.5 times higher for women and 1.7 for men. In univariate analysis, contrary to the other virus contexts where odds ratios were rarely above 2, residents over 95 years old had increased odds of infection of � 2.8 compared with the 70-year-old category, and for the 100 year-old group the odds were approximately 3.8. In the Flu+ context, autonomy adjusted for age (virus, sex and LOS stratification) revealed a possible increase in infection rates among less autonomous residents. A previous study found that when elderly residents were exposed to the A(H3N2) virus, there were higher rates of infection and reinfection, and more significant effects on the institution than with other influenza types/subtypes. In the community, the relative illness ratio (RIR) in the [22, 24] .
The incidence of influenza infection and the associated risks were well described by age group, but the specific impact according to age was not studied. The results of this work highlighted the specific age distribution of influenza illnesses among the nursing home residents and the more significant impact among the older residents. This specific susceptibility could be a critical factor in the institutional exposure and dissemination of influenza and could partly explain the high infection impact in the elderly institutional population.
In this work, autonomy status was not the main factor associated with infection (no significant impact in GE and in Flu-contexts). However, in Flu+ outbreaks, a high level of dependency was associated with a higher risk of falling ill. This observation implies that staff could play a role in the spread of infection (highly dependent and less mobile residents are less likely to contaminate themselves) or that the more active residents may be less fragile and/or have a greater involvement in the recommended infection control measures. Finally, improving compliance with personal hygiene measures both for nursing staff and residents might be expected to have a beneficial effect on infection rates. Previous studies identified higher NoV infection rates in highly dependent individuals, but the results were not adjusted for age and LOS to take into account the potential correlation with the autonomy status [20] .
Lethality is difficult to assess in nursing homes because death is frequent. Our GE and RTI episodes occurred during the winter seasons, and there are possible interactions between outbreaks and increased mortality at this time of the year [25] . The all-cause lethality rate of the infected residents in our study reflected global mortality including GE and RTI outbreaks and the global epidemiological context. Not surprisingly, a higher all-cause lethality rate was observed in the influenza contexts, as reported in previous studies [25] [26] [27] . Age and autonomy had similar effects in the different contexts, but in GE and to a lesser degree in Flu-outbreaks, the relatively small number of deaths could have limited the power of the statistical tests. In nursing homes, residents are generally discharged due to death. The number of residents lost to follow up was low (0.14% in the first 28 days), so the 7-day interval turnover rate calculated on the base of the median LOS provides a good indication of the average case fatality rate. The lethality rate for NoV+ outbreaks was similar to the estimated 7-day interval turnover rate (1.6%) indicating that this context had a limited impact on the death rate.
The all-cause lethality rate was most affected by age and autonomy. Both individual characteristics were significant in the Flu+ outbreaks, and autonomy adjusted for age was significant in the Flu-episodes. The influence of age on mortality in a context of influenza has already been described: a very high mortality rate (831/100,000 inhabitants) was reported in persons 90 years of age and older compared with those aged 65-69 years (23/100,000 inhabitants) [28] . In our univariate analysis, the higher risk was observed in the �90 group whose risk of death was at least 2.6 higher than the <70 group. When adjusted for autonomy, the impact of age was not significant in more autonomous residents in the Flu+ context and not at all in the Flucontext. The opposite analysis (autonomy adjusted for age) showed higher global impact in the less autonomous group (Flu-) or only in the �86 age group (Flu+). Age and autonomy are a reflection of resident's level of frailty. Clinical frailty scores were not used in this study, but in a previous study of patients with critical illness, they were associated with greater mortality, regardless of age [29] . This suggests that in addition to age, autonomy can be a valuable indicator for the assessment of outbreak impact in outbreak surveillance. Other studies have suggested that age and certain comorbidities are independent risk factors for the influenza mortality rate or that mortality increase according to the number of risk factors [28, 30] . Comorbidities and underlying diseases of various severities could reflect overall frailty and consequently the risk of death. In nursing homes, information about autonomy and age are easier to collect and interpret than data on comorbidities. These various approaches should be evaluated and compared in the goal of optimizing risk assessment among nursing home residents.
The present work has two main limitations. First, the virus information was incomplete (limited identification, mainly influenza rapid tests for the RTI). Consequently, some episodes in the levels with no available identification may also have been associated with influenza or norovirus, and multiple contaminations could have been underestimated or not taken into account. Moreover, vaccination and oseltamivir prescriptions were recorded but not included because the influenza genotype was not determined and identification was limited.
Secondly, the deaths of uninfected residents were not recorded in this protocol even though such data would have provided valuable information about the global epidemiological context.
In conclusion, specific susceptibility patterns were observed among exposed residents. In this cohort of nursing homes, infection rates varied according to virus, sex, length of stay and age, and there were major differences in lethality depending on virus, age and autonomy score. The collected data were easy to record and could be used to improve the characterization of seasonal outbreaks in nursing homes, whose residents are particularly vulnerable. Finally, as the average age and dependency level of residents continues to increase, subsequently increasing the risk of infection and death, health care staff will have to be increasingly vigilant during seasonal outbreaks and targeted interventions should be implemented.
Supporting information S1 Data. (CSV) S1 R Codes. (R) S1 Table. (XLSX) S10 Table. (XLSX) S11 Table. (XLSX)
|
What was the reported infection rate for influenza?
| 5,308 |
30.0%
| 24,623 |
1,634 |
Gastroenteritis and respiratory infection outbreaks in French nursing homes from 2007 to 2018: Morbidity and all-cause lethality according to the individual characteristics of residents
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6759171/
SHA: f1d456ea268266ff3c21317c4190e4fcb49b5e4f
Authors: Gaspard, Philippe; Mosnier, Anne; Simon, Loic; Ali-Brandmeyer, Olivia; Rabaud, Christian; Larocca, Sabrina; Heck, Béatrice; Aho-Glélé, Serge; Pothier, Pierre; Ambert-Balay, Katia
Date: 2019-09-24
DOI: 10.1371/journal.pone.0222321
License: cc-by
Abstract: BACKGROUND: Gastroenteritis (GE) and respiratory tract infection (RTI) outbreaks are a significant issue in nursing homes. This study aimed to describe GE and RTI outbreaks with infection and all-cause lethality rates according to the individual characteristics of nursing home residents. METHODS: Clinical and virological surveillance were conducted (2007 to 2018). Virus stratifications for the analysis were: outbreaks with positive norovirus or influenza identifications (respectively NoV+ or Flu+), episodes with no NoV or influenza identification or testing (respectively NoV- or Flu-). Associations between individual variables (sex, age, length of stay (LOS), autonomy status) and infection and lethality rates were tested with univariate and Mantel-Haenszel (MH) methods. RESULTS: 61 GE outbreaks and 76 RTI oubreaks (total 137 outbreaks) were recorded involving respectively 4309 and 5862 residents. In univariate analysis, higher infection rates and age were associated in NoV+, NoV-, and Flu+ contexts, and lower infection rates were associated with longer stays (NoV+ and NoV-). In MH stratified analysis (virus, sex (female/male)) adjusted for LOS (<4 or ≥4 years), the odds of being infected remained significant among older residents (≥86 years): NoV+/male (Odds ratio (OR(MH)): 1.64, 95% confidence interval (CI): 1.16–2.30) and Flu+/female and male (respectively OR(MH): 1.50, CI: 1.27–1.79 and 1.73, CI: 1.28–2.33). In univariate analysis, lower autonomy status (NoV+, Flu+ and Flu-) and increased age (Flu+) were associated with higher lethality. In MH adjusted analysis, significant OR(age) adjusted for autonomy was: Flu+/ ≥86 years compared with <86 years, 1.97 (1.19–3.25) and OR(autonomy) adjusted for age for the more autonomous group (compared with the less autonomous group) was: Flu+, 0.41 (0.24–0.69); Flu-, 0.42 (0.20, 0.90). CONCLUSION: The residents of nursing homes are increasingly elderly and dependent. The specific infection and lethality risks according to these two factors indicate that surveillance and infection control measures are essential and of high priority.
Text: Introduction Gastroenteritis (GE) and respiratory tract infection (RTI) outbreaks represent a significant burden of illness in nursing homes. Viruses cause the majority of these outbreaks, and noroviruses and influenza viruses are the most common pathogens [1, 2] .
Previous studies have suggested that viral respiratory infections and norovirus outbreaks are a common cause of hospitalization or death, particularly among elderly individuals [3] [4] [5] .
The impact of outbreaks has been described in terms of both frequency and epidemiology, but little is known about infection rates and all-cause lethality in GE and RTI nursing home outbreaks in relation to the individual characteristics of the residents [6] . The residents of these institutions are increasingly elderly and dependent, and the impact of this trend on the seasonal outbreak burden requires in-depth investigation. The results of studies focused on this issue could yield valuable information for nursing homes, allowing them to adapt their infection control strategies, in particular for improved assessment of infection risk.
Our objective was to describe GE and RTI infection and all-cause lethality rates according to the individual characteristics of nursing home residents (sex, age, length of stay, autonomy status), and to identify specific susceptibility patterns related to these types of viral outbreaks in these facilities.
The present study explored outbreaks in 14 sites (28 units with geriatric nursing home activities for a total of 1121 beds) caring for dependent people in southern Alsace (an area in northeastern France). Data were collected between September 2007 and August 2018 [7, 8] .
Each site was geographically independent and autonomous for social and care management. Units were located within the larger sites and were defined as a place having a dedicated team at one location.
During outbreaks at one site, only the residents in the units with confirmed cases were included. Outbreak inclusion depended on institutional alert to the hygiene team. Surveillance was done in each unit independently, and the members of staff had to inform a physician or charge nurse when two or more potential related cases of pneumonia or GE were observed within four days and when three or more cases were observed for other RTI. Units also had to inform the hygiene team when these threshold values were exceeded. For influenza, the first suspected case led to a local alert and the hygiene team was contacted. A practitioner from the hygiene team collected the information and evaluated the clinical signs, the virology information and the epidemiological context with the physician in the affected unit. The detected cluster was only put under surveillance if the hygiene team considered that there was a potential outbreak phenomenon. The duration of 4 days was in relation with the national protocol with alert to the authorities when 5 cases occurred within 4 days [9, 10] . On a local level and in addition to the clusters reported to the authorities, clusters with at least 3 cases within a period of seven days in one unit could be recorded if they were reported to the hygiene team. Because several outbreaks could potentially occur in the same unit during the surveillance period, a resident could be included repeatedly in different clusters. As a result, the observed patterns reflected the characteristics of an institutional population with longitudinal and pluriannual exposures.
Personal information and clinical information was collected by a practitioner from the hygiene team directly from the residents' health care records. Personal information was collected for all those present the first day of the outbreak. The collected information included: month and year of birth, sex, date of arrival at the nursing home and autonomy status. The autonomy status of residents in French nursing homes is assessed using the AGGIR scale (Autonomy Gerontology Groups Iso-Resources), which is the legal instrument for evaluating dependency in the elderly and whose primary purpose is the allocation of means and resources [11] .
With the AGGIR scale, autonomy is classified into 6 Iso-Resource Groups (GIR): GIR 1 (bedridden or armchair-bound persons, mental functions seriously altered and requiring continuous presence), GIR 2 (bedridden or armchair-bound persons, mental functions not totally altered and requiring assistance in most activities of daily living, or mental functions altered with preserved ability to get around), GIR 3 (preserved mental autonomy with partially preserved motor autonomy and assistance several times a day for physical autonomy), GIR 4 (moves around the home and sometimes assistance for washing, dressing, physical activities or eating), GIR 5 (only occasional assistance for washing, meal preparation, and housework), GIR 6 (autonomy for essential tasks of daily living). In outbreaks where the influenza virus was identified, influenza vaccination status and oseltamivir prescriptions were recorded as well. A file is transmitted in the Supporting Information with all previous data (S1 Data).
GE was defined as the sudden onset of vomiting and/or diarrhea over a 24 h period: (i) diarrhea �3 episodes, (ii) and/or vomiting �3 episodes, (iii) or diarrhea or vomiting <3 episodes with two or more other symptoms (diarrhea, vomiting, stomach ache, abdominal cramps, nausea, fever, mucus in stools) [1] .
RTI presentation in older adults may be atypical, like for other acute illnesses in this age group [12] . We used the recommended definitions for RTI surveillance in geriatric units, divided in 3 subcategories: (i) common cold syndromes or pharyngitis (at least two of the following criteria: runny nose or sneezing, stuffy nose (i.e. congestion), sore throat or hoarseness or difficulty swallowing, dry cough, swollen or tender glands in the neck (cervical lymphadenopathy)), (ii) influenza-like illness (both the following criteria must be met: fever AND at least three other symptoms (chills, new headache or eye pain, myalgia or body aches, malaise or loss of appetite, sore throat, new or increased dry cough)) and (iii) lower respiratory tract infection (both of the following criteria must be met: at least two respiratory signs or symptoms (new or increased cough, new/increased sputum production, O 2 saturation <94% or reduced >3% from baseline, abnormal lung examination (new or changed), pleuritic chest pain, respiratory rate �25 breaths/min AND one or more constitutional signs/symptoms (fever, leukocytosis, confusion, acute functional decline)) [13] [14] [15] [16] . Infection corresponding to one of these three subcategories was included in this study and classified as RTI.
For both infection types, the practitioner from the hygiene team obtained clinical information from the patient's health care records, and members of the health care team were consulted if necessary to complete any missing information. At the end of the episode (within seven days after the last identified case), case inclusion as exposed and not infected (ENI) or exposed and infected (EI) was determined with a resident physician.
In order to study the lethality, the presence of each infected resident was evaluated once at least 56 days after the last case of each outbreak. Each resident was followed up retrospectively during eighth 7-day interval (I n, n = 1 to 8, total 56 days, between the date of onset of symptoms and the fifty-sixth day of the studied period) with three different possibilities: present (alive and officially residing in the institution), lost to follow-up (alive at the date of departure but no longer residing in the institution (return home, transfer to another institution)) or death (death recorded in the health care record). The dates of death and lost to follow-up were recorded.
Testing for the virus was not systematic and was decided by the physicians in each institution in the presence of clinical signs.
For GE, stool samples were sent to the National Reference Centre for Gastroenteritis Viruses in Dijon for laboratory testing, as previously described [8] . For RTI surveillance, rapid tests were used to identify the influenza virus. The rapid immunoassay diagnosis tests used for influenza detection were: Clearview1 Exact Influenza A and B (Inverness Medical, Cologne, Germany) from 2007 to 2014 and InfluenzaTop1 (Alldiag, Strasbourg, France) from 2014 to 2018. Given the low sensitivity of influenza rapid tests, they were no longer used once the control measures had been implemented and the influenza outbreak was under control. Samples were also occasionally sent to hospital laboratories or to the National Reference Centre for Influenza Viruses to detect viruses with real-time RT-PCR [17] . Most testing targeted the norovirus (NoV) and influenza virus, but other tests were occasionally performed by the National Reference Centre for Influenza Viruses (rhinovirus, respiratory syncytial virus, human metapneumovirus, parainfluenza 1, 2, 3 and 4, and coronavirus) and the National Reference Centre for Enteric Viruses (rotavirus, astrovirus, and adenovirus).
Because testing for the viruses was variable (from one institution/physician to another, not used in some outbreaks, types of virus sought) and considering the poor sensitivity of the rapid influenza tests, these two sources of data were used to define the epidemiological context of confirmed outbreaks. Consequently, individual cases were included consistently in all episodes according to clinical signs and medical evaluation. When virus testing was negative, the clinical signs were recorded and medical evaluation was used as previously to classify the included residents as infected or not infected.
The epidemiological context of each outbreak was defined according to whether the virus had been identified or not. One or more positive samples led to the qualification of a NoV (NoV+) or flu (Flu+) context. The other episodes were qualified as flu or NoV outbreaks with no specific identification or testing (NoV-and Flu-).
Flu and NoV contexts did not eliminate other potential enteric or respiratory pathogens.
Sex, age, length of stay (LOS, in years) and autonomy status were described for all exposed residents. Influenza vaccination and oseltamivir administration rates were calculated for confirmed influenza outbreaks. Dichotomous or categorical variables were expressed as percentages. In univariate analysis, the categories were specific in order to obtain a precise description of the age and LOS variables. Class intervals were 5 years for age and one year for LOS. The residents classified as GIR 4 to 6 (sometimes, occasional and no assistance) were grouped together because they were few.
For the multi-level analysis with 2x2 tables, a median value was used to define the two-level age categories. For LOS, assessing the longest stays was necessary to identify the effect of longer exposure in a nursing home. Consequently, a four-year cutoff was chosen to create the two categories. For autonomy, the two most dependent categories (GIR � 2) were grouped together and compared with the more autonomous categories (GIR � 3).
The outbreak epidemiological contexts were used with the four categories: NoV+, Flu+, NoV-and flu-. Other GE or RTI viruses were occasionally identified, but the number of results was too limited to develop separate analyses. However, all the results are available in the tables about the virus investigations along with NoV and influenza identifications.
For the different categories, infection rate (EI/Exposed Residents (ER), in percentage) was calculated according to sex, age group, LOS and autonomy status. To investigate all-cause lethality and define the appropriate period for the 56-day monitoring (D 1 to 56 ), the all-cause lethality rate per 7-day interval (LR n /I n, n = 1 to 8 ) was calculated: (number of deaths during interval I n /(EI alive the first day of n th studied interval minus lost to follow-up EI during the interval I n ) � 100).
Seeing as successive clusters could occur within the same site, potentially influencing allcause lethality, the serial interval in days (SI d ) was calculated. The SI d was the time period between the onset of symptoms of the last case in initial outbreak (N) and the onset of symptoms of the first case in the following outbreak (N+1). Investigations were performed when SI d was shorter or equal to the length of the previous D 1 to 56 and the following parameters were evaluated for these specific situations: number of episodes, residents infected in both outbreaks, and death among the identified individuals.
Finally, according to the death rate and the impact of successive outbreaks, the number of 7-day intervals (N.I n ) to take into account was defined, and the all-cause lethality rate was analyzed during these periods (I 1 to N th .I n or D 1 to 7 � N ).
All-cause lethality rates were calculated with the following formula: (number of deaths from D 1 to 7 � N /(EI number at D 1 minus lost to follow-up among EI during the period D 1 to 7 � N ) � 100). Estimation of the turnover rate per 7-day interval among the infected residents was calculated on the base of the LOS (median in years) with the following formula: (proportion of discharged residents: 50.0% in the case of the median)/[(median LOS � 365)/7)]. The average rate of residents discharged per 7-day period was calculated: [(number of lost to follow up during the period D 1 to 7 � N /number of exposed and infected residents at D 1 )/N 7-day interval] � 100.
As some residents were included in several outbreaks during the surveillance, the observations were not completely independent; non-parametric tests were used as a result. In univariate analysis, Chi-square or Fisher exact tests (expected number of frequencies fewer than 5) were used to compare infection and lethality rates according to the studied parameters and the odds ratio was calculated by median-unbiased estimation. The Kruskal-Wallis test was used to compare median values. Confidence intervals for medians were calculated with bootstrap methods.
Covariate adjusted analyses were performed with two tables (2x2). The respective impact of each individual factor was tested with Mantel-Haenszel chi-squared tests. The equality of the stratum odds ratios was tested with the Woolf test of homogeneity. Finally, for each virus context, multiple tables (2x2) were generated and tested with confounding variables, effect modifiers or covariables. Statistical analyses were done using R for Mas OS X version R 3.4.1 software with RStudio version 1.0.153. A file is transmitted in the Supporting Information with all R codes and the packages used (S1 R Codes). Differences were considered significant at p � 0.05.
The French Data Protection Authority approved data collection and analysis (DE-2013-074) and the local ethics committee (Espace Local de Réflexion Ethique, Centre Hospitalier de Rouffach) approved the study protocol (ERLE-32). According to the French law for biomedical research and human experimentation, individual written consent was not required from the patients or their relatives for data collection. Each year, the referring local practitioner of the study coordinated with the doctors working in the nursing home. At the beginning of the surveillance period, information regarding participation in the study was displayed in the family vising area, including a document about their right to access and rectify personal data. After collection, data were rendered anonymous. No specific authorization was needed to retrospectively analyze anonymous data collected during routine care in the context of routine surveillance.
A total of 137 outbreaks were recorded in the 14 sites. RTI outbreaks were more frequent than GE outbreaks (76 outbreaks and 5862 exposed residents vs. 61 outbreaks and 4309 exposed residents, respectively). Overall, 7643 of the exposed residents were women and 2528 were men. The median age was 86.7 years old (interquartile range: 81.1-91.0 years).
Virus investigations (respectively 389 samples for RTI and 143 for GE with all the detailed results in S1-S4 Tables) confirmed a considerable number of norovirus-related GE outbreaks (34/61) and influenza-related RTI outbreaks (46/76). For GE outbreaks, 2524 residents were in a NoV+ context versus 1785 in a NoV-context, and for RTI outbreaks, 3479 residents were in a Flu+ context versus 2383 in a Flu-context.
For GE surveillance in the NoV+ context, there were 1093 EI residents versus 1431 ENI residents, whereas in the NoV-context, there were 583 EI residents versus 1202 ENI residents. Therefore, the infection rate was higher in the NoV+ context (43.3%) than in the NoV-context (32.7%, (odds ratio (OR): 0.63, 95% confidence interval (CI): 0.56-0.72), p < 0.001).
For RTI surveillance, the rates of infection were similar with and without confirmed influenza: 31.5% (N = 1095 EI residents /3479 exposed residents) vs. 30.5% (N = 728 EI residents/ 2383 exposed residents, OR: 0.96, CI: 0.85-1.07, p = 0.47). Moreover, infection rate in the NoV + context was higher than the three other contexts: NoV-(OR: 0.63, CI: 0.56-0.72), Flu+ (OR: 0.60, CI:0.54-0.67) and Flu-(OR: 0.58, CI: 0.51-0.65).
In univariate analysis, certain individual characteristics were associated with significant variations in the infection rate (S5 Table) . The infection rate increased with age (except in the Flucontext) and, decreased with LOS during GE outbreaks. The covariate adjusted analysis revealed specific significant effect modification according to sex (NoV+) and LOS (NoV-) (S6 Table) . In analyses stratified according to virus and sex, age adjusted for LOS remained significant for Flu+ and NoV+ outbreaks (males). In NoV-context, the effect modification of LOS remained significant (Table 1) . Finally, when autonomy was included and adjusted for age (virus, sex, LOS stratification), the less autonomous residents (female/LOS<4 years/age<86/ GIR 1-2) were affected more severely by Flu+ outbreaks with specific effect modification according to age (S7 Table) . The study of lethality rates in infected residents over the 56 days after onset indicated that there were significant variations for RTI but no change for GE (Table 2 ). Significant differences appeared after 28 days in the context of Flu+ outbreaks and other RTI outbreaks.
The analysis of successive or simultaneous clusters in the same institutions was performed when the time period between the onset of symptoms of the last case in outbreak N and the onset of symptoms of the first case in outbreak N+1 was �56 days (S8 Table) . 44 of the 137 outbreaks (32.12%) were identified, and 194 of the 3499 exposed and infected residents contracted multiple infections. The percentage of exposed and infected residents implicated in more than one virus stratification was 11.09% ((194 � 2)/3499). Moreover, two deceased residents were included in the NoV-Na and Flu lethality analyses because death occurred within 56 days for both infections. The analysis of virus stratification of the 44 outbreaks showed the absence of successive clusters for the same category. The same analysis for the first four 7-day intervals (Days 1 to 28) showed the respective values: 26 outbreaks (18.98%), 117 residents ((6.69% ((117 � 2)/3499)), one dead resident.
Finally, according to the higher lethality impact during the first four 7-day intervals and to limit the impact of successive clusters in the same site, all cause lethality rates were studied according to individual parameters for the four 7 days intervals with the respective number of According to the surveillance type (GE or RTI), the lethality rates differed significantly: 1.6% versus 3.4% (respectively NoV+ and NoV-contexts, OR: 2.24, CI: 1.16-4.39, p = 0.02) and 8.3% versus 5.6% (respectively Flu+ and Flu-, OR: 0.67, CI: 0.45-0.97, p = 0.04).
In univariate analysis (S9 Table) , low autonomy status in the NoV+, Flu+ and Flu-contexts was most significantly associated with increased all-cause lethality, and age was associated with higher lethality in the Flu+ context. In the adjusted analysis, no significant statistical differences were identified in GE outbreaks. For RTI episodes, the adjusted analysis showed that autonomy had a significant impact when adjusted for sex, age or LOS (Flu+ and Flu-NA) and that age had a significant impact when adjusted for sex, autonomy or LOS (Flu+) (S10 Table) .
In Table 3 , the specific effects of age or autonomy were tested. Significant OR age adjusted for autonomy were: Flu+/age �86 years (compared with the <86 group), 1.97 (1.19-3.25). OR autonomy adjusted for age were for GIR 3-6 (compared with GIR 1-2): Flu+, 0.41 (0.24-0.69); Flu-, 0.42 (0.20, 0.90).
Finally, despite the low number of residents and deaths per category, and consequently the limited robustness of the results, autonomy adjusted for age with stratification according to virus, sex and LOS showed that the effects were higher among subgroups of less autonomous residents (female or male/LOS<4 years/GIR 1-2) in Flu+ outbreaks, and there was also higher mortality in the small subgroup of autonomous men with LOS � 4 years (higher mortality) (S11 Table) .
In the Flu+ context, data regarding vaccination status and oseltamivir prescriptions were available but not used in this study.
In the present study, surveillance data obtained during GE and RTI outbreaks in nursing homes were used to construct stratified analyses and to identify specific infection and all-cause lethality rates according to the residents' individual characteristics.
The infection rates observed here were similar to those found in previous studies of NoV and Influenza outbreaks (odds of being infected during a Flu+ outbreak were around 40% less than during a NoV+ outbreak). Reported infection rates were close to 30.0% in influenza outbreaks and 40.0% in NoV outbreaks [18] [19] [20] .
Older age appeared to increase the likelihood of GE and influenza infection, with increasing rates among older residents. Age is a well-known factor for influenza and norovirus severity in the elderly and in nursing homes [21, 22] . For NoV, the highest incidence estimates (5-year age strata) was found in the �85 year-category (approximately 800 men and for 1,400 women per 100,000 inhabitants). In our study, univariate analysis (NoV+) showed that the odds of being infected were 1.5 to 1.6 times higher if a resident was older than 85. Moreover, an adjusted analysis of GE outbreaks highlighted different effects among subgroups of residents according to sex and LOS. Indeed, multiple and/or repeated exposure to GE viruses while institutionalized may lead to susceptibility or possible increased immunity in some residents [23] . For the sex variable, two factors could explain the effect: a possible selection bias with men reporting mild infections less than women (particularly in the <86 years subgroup) or that male susceptibility was different (age, LOS, immunity,. . .). A German study from 2013 also reported a greater impact in women [21] . Moreover, when age analysis was stratified by sex and LOS, no Epidemic impacts in nursing homes significant impact was observed in women; the only significant differences were fewer infections in men in the <86-subgroup (except in the NoV-with LOS <4 years).
For the RTI outbreaks, sex and LOS variables did not have a significant effect. In residents older than 86, the odds of being infected in Flu+ context were 1.5 times higher for women and 1.7 for men. In univariate analysis, contrary to the other virus contexts where odds ratios were rarely above 2, residents over 95 years old had increased odds of infection of � 2.8 compared with the 70-year-old category, and for the 100 year-old group the odds were approximately 3.8. In the Flu+ context, autonomy adjusted for age (virus, sex and LOS stratification) revealed a possible increase in infection rates among less autonomous residents. A previous study found that when elderly residents were exposed to the A(H3N2) virus, there were higher rates of infection and reinfection, and more significant effects on the institution than with other influenza types/subtypes. In the community, the relative illness ratio (RIR) in the [22, 24] .
The incidence of influenza infection and the associated risks were well described by age group, but the specific impact according to age was not studied. The results of this work highlighted the specific age distribution of influenza illnesses among the nursing home residents and the more significant impact among the older residents. This specific susceptibility could be a critical factor in the institutional exposure and dissemination of influenza and could partly explain the high infection impact in the elderly institutional population.
In this work, autonomy status was not the main factor associated with infection (no significant impact in GE and in Flu-contexts). However, in Flu+ outbreaks, a high level of dependency was associated with a higher risk of falling ill. This observation implies that staff could play a role in the spread of infection (highly dependent and less mobile residents are less likely to contaminate themselves) or that the more active residents may be less fragile and/or have a greater involvement in the recommended infection control measures. Finally, improving compliance with personal hygiene measures both for nursing staff and residents might be expected to have a beneficial effect on infection rates. Previous studies identified higher NoV infection rates in highly dependent individuals, but the results were not adjusted for age and LOS to take into account the potential correlation with the autonomy status [20] .
Lethality is difficult to assess in nursing homes because death is frequent. Our GE and RTI episodes occurred during the winter seasons, and there are possible interactions between outbreaks and increased mortality at this time of the year [25] . The all-cause lethality rate of the infected residents in our study reflected global mortality including GE and RTI outbreaks and the global epidemiological context. Not surprisingly, a higher all-cause lethality rate was observed in the influenza contexts, as reported in previous studies [25] [26] [27] . Age and autonomy had similar effects in the different contexts, but in GE and to a lesser degree in Flu-outbreaks, the relatively small number of deaths could have limited the power of the statistical tests. In nursing homes, residents are generally discharged due to death. The number of residents lost to follow up was low (0.14% in the first 28 days), so the 7-day interval turnover rate calculated on the base of the median LOS provides a good indication of the average case fatality rate. The lethality rate for NoV+ outbreaks was similar to the estimated 7-day interval turnover rate (1.6%) indicating that this context had a limited impact on the death rate.
The all-cause lethality rate was most affected by age and autonomy. Both individual characteristics were significant in the Flu+ outbreaks, and autonomy adjusted for age was significant in the Flu-episodes. The influence of age on mortality in a context of influenza has already been described: a very high mortality rate (831/100,000 inhabitants) was reported in persons 90 years of age and older compared with those aged 65-69 years (23/100,000 inhabitants) [28] . In our univariate analysis, the higher risk was observed in the �90 group whose risk of death was at least 2.6 higher than the <70 group. When adjusted for autonomy, the impact of age was not significant in more autonomous residents in the Flu+ context and not at all in the Flucontext. The opposite analysis (autonomy adjusted for age) showed higher global impact in the less autonomous group (Flu-) or only in the �86 age group (Flu+). Age and autonomy are a reflection of resident's level of frailty. Clinical frailty scores were not used in this study, but in a previous study of patients with critical illness, they were associated with greater mortality, regardless of age [29] . This suggests that in addition to age, autonomy can be a valuable indicator for the assessment of outbreak impact in outbreak surveillance. Other studies have suggested that age and certain comorbidities are independent risk factors for the influenza mortality rate or that mortality increase according to the number of risk factors [28, 30] . Comorbidities and underlying diseases of various severities could reflect overall frailty and consequently the risk of death. In nursing homes, information about autonomy and age are easier to collect and interpret than data on comorbidities. These various approaches should be evaluated and compared in the goal of optimizing risk assessment among nursing home residents.
The present work has two main limitations. First, the virus information was incomplete (limited identification, mainly influenza rapid tests for the RTI). Consequently, some episodes in the levels with no available identification may also have been associated with influenza or norovirus, and multiple contaminations could have been underestimated or not taken into account. Moreover, vaccination and oseltamivir prescriptions were recorded but not included because the influenza genotype was not determined and identification was limited.
Secondly, the deaths of uninfected residents were not recorded in this protocol even though such data would have provided valuable information about the global epidemiological context.
In conclusion, specific susceptibility patterns were observed among exposed residents. In this cohort of nursing homes, infection rates varied according to virus, sex, length of stay and age, and there were major differences in lethality depending on virus, age and autonomy score. The collected data were easy to record and could be used to improve the characterization of seasonal outbreaks in nursing homes, whose residents are particularly vulnerable. Finally, as the average age and dependency level of residents continues to increase, subsequently increasing the risk of infection and death, health care staff will have to be increasingly vigilant during seasonal outbreaks and targeted interventions should be implemented.
Supporting information S1 Data. (CSV) S1 R Codes. (R) S1 Table. (XLSX) S10 Table. (XLSX) S11 Table. (XLSX)
|
How many times more likely was an infection found in patients over 85 years old?
| 5,309 |
1.5 to 1.6 times
| 25,180 |
1,621 |
Vesicular stomatitis virus with the rabies virus glycoprotein directs retrograde transsynaptic transport among neurons in vivo
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3566411/
SHA: ee48061797d29eeef5a9e606841bf8ab04b1d75b
Authors: Beier, Kevin T.; Saunders, Arpiar B.; Oldenburg, Ian A.; Sabatini, Bernardo L.; Cepko, Constance L.
Date: 2013-02-07
DOI: 10.3389/fncir.2013.00011
License: cc-by
Abstract: Defining the connections among neurons is critical to our understanding of the structure and function of the nervous system. Recombinant viruses engineered to transmit across synapses provide a powerful approach for the dissection of neuronal circuitry in vivo. We recently demonstrated that recombinant vesicular stomatitis virus (VSV) can be endowed with anterograde or retrograde transsynaptic tracing ability by providing the virus with different glycoproteins. Here we extend the characterization of the transmission and gene expression of recombinant VSV (rVSV) with the rabies virus glycoprotein (RABV-G), and provide examples of its activity relative to the anterograde transsynaptic tracer form of rVSV. rVSV with RABV-G was found to drive strong expression of transgenes and to spread rapidly from neuron to neuron in only a retrograde manner. Depending upon how the RABV-G was delivered, VSV served as a polysynaptic or monosynaptic tracer, or was able to define projections through axonal uptake and retrograde transport. In animals co-infected with rVSV in its anterograde form, rVSV with RABV-G could be used to begin to characterize the similarities and differences in connections to different areas. rVSV with RABV-G provides a flexible, rapid, and versatile tracing tool that complements the previously described VSV-based anterograde transsynaptic tracer.
Text: Mapping neuronal connectivity in the central nervous system (CNS) of even simple organisms is a difficult task. Recombinant viruses engineered to trace synaptic connections and express transgenes promise to enable higher-throughput mapping of connections among neurons than other methods, e.g., serial reconstruction from electron micrographs (Bock et al., 2011; Briggman et al., 2011) . The Pseudorabies (PRV) and Rabies viruses (RABV) have been the best characterized and most utilized circuit tracing viruses to date (Ugolini et al., 1989; Kelly and Strick, 2000) . RABV was recently modified by Wickersham and colleagues such that it can travel across only one synapse, allowing for a straightforward definition of monosynaptic connections (Wickersham et al., 2007b) . This strategy permitted the first unambiguous identification of retrogradely connected cells from an initially infected cell ("starter cell"), without the need for electrophysiology. Moreover, the starter cell could be defined through the expression of a specific viral receptor that limited the initial infection.
Recently, we created an anterograde monosynaptic virus that complements the previously available retrograde viral tracers (Beier et al., 2011) . Vesicular stomatitis virus (VSV), a virus related to RABV, with its own glycoprotein (G) gene (VSV-G), or with a G from the unrelated lymphocytic choriomeningitis virus (LCMV-G), spreads in the anterograde direction across synapses. VSV can be used as a polysynaptic tracer that spreads across many synapses, owing to the fact that the normal, replicationcompetent form of the virus does not cause serious diseases in humans (Brandly and Hanson, 1957; Johnson et al., 1966; Brody et al., 1967) . Whether the virus is a monosynaptic or polysynaptic tracer is determined by the method of delivery of the G gene ( Figure 1A) . Advantages of VSV are that it is well-characterized, is relatively simple in comparison to PRV, and it rapidly grows to high titer in tissue culture cells. It is also being developed as a vaccine vector, often using a G of another virus as the immunogen, as well as being developed as a cytocidal agent that will target tumor cells in humans (Balachandran and Barber, 2000; Stojdl et al., 2000 Stojdl et al., , 2003 .
Previous studies of the anatomical patterns of transmission, as well as physiological recordings, have shown that the transmission of VSV and RABV among neurons is via synapses (Kelly and Strick, 2000; Wickersham et al., 2007b; Beier et al., 2011) . In addition, it has been shown that RABV, as well as lentiviruses with RABV-G in their envelope, travel retrogradely from an injection site (Mazarakis et al., 2001; Wickersham et al., 2007a) . We hypothesized that providing a recombinant VSV (rVSV) with the RABV-G would create a retrograde polysynaptic transsynaptic tracer without the biosafety concerns inherent to RABV. Our initial characterization of rVSV with RABV-G showed that indeed FIGURE 1 | Synaptic tracing strategies using VSV. (A) Schematic illustrating the strategies for polysynaptic or monosynaptic retrograde or anterograde transsynaptic transmission of rVSV encoding GFP. The initially infected cell is indicated by an asterisk. VSV encoding a glycoprotein (G) within its genome can spread polysynaptically. The direction of the spread depends on the identity of the glycoprotein. Infected neurons are shown in green. In some cases, the initially infected starter cell can be defined by the expression of an avian receptor, TVA (tagged with a red fluorescent protein). The TVA-expressing neurons can then be specifically infected by rVSV G with the EnvA/RABV-G (A/RG) glycoprotein (Wickersham et al., 2007b) on the virion surface [rVSV G(A/RG)]. These starter cells are then yellow, due to viral GFP and mCherry from TVA-mCherry expression. For monosynaptic tracing, the G protein is expressed in trans in the TVA-expressing cell, and thus complements rVSV G to allow transmission in a specific direction. (B) Genomic diagrams of rVSV vectors. All VSVs contain four essential proteins: N, P, M, and L. Some viruses encode a G gene in their genome, which allows them to spread polysynaptically. rVSV vectors typically encode a transgene in the first position, while others carry an additional transgene in the G position. (C) Morphological characterization of rVSV-infected neurons in several locations within the mouse brain. (i,ii) Caudate-putamen (CP) neurons at 4 dpi from an injection of the CP with rVSV(VSV-G) viruses encoding (i) CFP or (ii) Korange. (iii) Labeled neurons of the CA1 region of the hippocampus are shown at 5 dpi following injection into the hippocampus of rVSV(VSV-G) encoding Venus. (iv,v) Cortical pyramidal neurons are shown following injection into the CP of rVSV(RABV-G) expressing (iv) GFP at 24 hpi, or (v) mCherry at 48 hpi. Inset in (iv) is a high magnification of the neuron in panel (iv), highlighting labeling of dendritic spines. (vi) Multiple viruses can be co-injected into the same animal. Here, individual rVSV G(VSV-G) viruses encoding CFP, GFP , Venus, Korange, and mCherry were used to infect the cortex. Scale bars = 50 µm.
www.frontiersin.org February 2013 | Volume 7 | Article 11 | 2 it could be taken up as a retrograde tracer (Beier et al., 2011) .
To determine if it could transmit among neurons following its replication in neurons, and to further analyze the transmission patterns of both the monosynaptic and polysynaptic forms of rVSV with RABV-G, we made injections into several CNS and peripheral locations. In addition, we performed co-infections of rVSV with RABV-G and the anterograde form of rVSV in order to exploit the differences in the directionality of transmission of these two viruses in mapping circuits.
Schematics of viruses created and used throughout this study are shown in Figure 1 . We created rVSV vector plasmids carrying different transgenes in either the first or fifth genomic positions ( Figure 1B) . After rescuing each virus, we tested the ability of each to express transgenes in different brain regions through intracranial injections ( Figure 1C ). All rVSV vectors drove robust fluorophore expression 1 or 2 days post-infection (hpi) ( Figure 1C ) (van den Pol et al., 2009) . In fact, by 12 hpi, labeling was sufficiently bright to image fine morphological details, such as dendritic spines ( Figure 1C ,iv).
To characterize the physiological properties of cells infected with rVSV, we tested a replication-competent rVSV encoding GFP, with RABV-G in the genome in place of VSV-G [hereafter designated rVSV(RABV-G)]. van den Pol et al. reported that hippocampal neurons infected with replication-incompetent (G-deleted or " G") rVSV were physiologically healthy at 12-14 hpi, but were less so by 1 day post-infection (dpi) (van den Pol et al., 2009) . Given the known toxicity of both VSV and RABV-G (Coulon et al., 1982) , we tested the physiology of cortical pyramidal neurons in the motor cortex (M1) infected with rVSV(RABV-G). Between 12 and 18 hpi, the membrane capacitance, input resistance, resting membrane potential, and current-to-action potential firing relationship were indistinguishable between infected and uninfected neurons (Figure 2) . However, by 2 dpi, electrophysiological properties were so abnormal in the infected cortical pyramidal cells that physiological measurements could not be made.
The speed and strength of the expression of transgenes encoded by VSV depends upon the gene's genomic position (van den Pol et al., 2009; Beier et al., 2011) . Genes in the first position are expressed the most highly, with a decrease in the level of expression in positions more 3 within the viral plus strand. When GFP was inserted into the first position of VSV, GFP fluorescence was first detectable at approximately 1 hpi in cultured cells (van den Pol et al., 2009) . In order to quantify the relative expression of a fluorescent protein in the first genomic position in neurons, rat hippocampal slices were infected with a replication-incompetent rVSV that expresses mCherry (rVSV G, Figures 1A,B) . This was a G virus which had the RABV-G supplied in trans during the preparation of the virus stock [referred to as rVSV G(RABV-G)].
Average fluorescence intensity of the infected cells was measured every hour over the course of 18 h. By 4 hpi at 37 • C, red fluorescence was clearly visible, and reached maximal levels by approximately 14 hpi (N = 3, Figure 3 ). Similar results were obtained with a virus encoding GFP in the first genomic position rather than mCherry (i.e., Figure 1B ) (N = 3).
We previously demonstrated that rVSV(RABV-G) could be taken up retrogradely by neurons (Beier et al., 2011) , but these experiments did not distinguish between direct axonal uptake of the initial inoculum vs. retrograde transsynaptic transmission following viral replication. To distinguish between these two mechanisms and to extend the previous analyses, we conducted further experiments in the mammalian visual system (Figures 4A-G) . As visual cortex area 1 (V1) does not receive direct projections from retinal ganglion cells (RGCs), but rather receives secondary input from RGCs via the lateral geniculate nucleus (LGN), infection of RGCs from injection of V1 would demonstrate retrograde transmission from cells which supported at least one round of viral replication. Following a V1 injection with rVSV(RABV-G), GFP-positive RGCs were observed in the retina by 3 dpi (N = 3; Figure 4G ). Importantly, viral labeling in the brain was restricted to primary and secondary projection areas, even at 7 dpi. These included the LGN ( Figure 4D ) and the hypothalamus (Figure 4E) , two areas known to project directly to V1 (Kandel, 2000) . Selective labeling was observed in other areas, such as cortical areas surrounding V1 (Figure 4C) , which project directly to V1, and also in the superior colliculus (SC) stratum griseum centrale, which projects to the LGN ( Figure 4F) . Labeling was also observed in the nucleus basalis, which projects to the cortex, as well as many components of the basal ganglia circuit, which provide input to the thalamus [such as the caudate-putamen (CP), globus pallidus (GP), and the subthalamic nucleus (STn)]. The amygdala, which projects to the hypothalamus, was also labeled. Consistent with a lack of widespread viral transmission, animals did not exhibit signs of disease at 7 dpi.
These data show that rVSV(RABV-G) can spread in a retrograde direction from the injection site, but do not address whether the virus can spread exclusively in the retrograde direction. Directional transsynaptic specificity can only be definitively addressed using a unidirectional circuit. We therefore turned to the primary motor cortex (M1) to CP connection, in which neurons project from the cortex to the CP, but not in the other direction ( Figure 4H ) (Beier et al., 2011) . Injections of rVSV(RABV-G) into M1 should not label neurons in the CP if the virus can only label cells across synapses in the retrograde direction. Indeed, at 2 dpi, areas directly projecting to the injection site, including the contralateral cortex, were labeled ( Figure 4I ). Only axons from cortical cells were observed in the CP, with no GFP-labeled cell bodies present in the CP (Figure 4J) , consistent with lack of anterograde transsynaptic spread. By 3 dpi, a small number of medium spiny neurons (MSNs) in the CP were observed, likely via secondary spread from initially infected thalamic or GP neurons (data not shown).
A particular advantage of retrograde viral tracers is the ability to label CNS neurons projecting to peripheral sites. This has been a powerful application of both RABV and PRV (Ugolini et al., 1989; Standish et al., 1994) . To test if rVSV(RABV-G) could also perform this function, we examined the innervation of the dura surface by neurons of the trigeminal ganglion, a neuronal circuit thought to be involved in migraine headaches (Penfield and McNaughton, 1940; Mayberg et al., 1984) . These neurons have axons, but not canonical dendrites, and send projections into the spinal cord and brainstem. Therefore, the only way trigeminal neurons could become labeled from viral application to the dura is through retrograde uptake of the virus. We applied rVSV(RABV-G) to the intact dura mater and analyzed the dura, trigeminal ganglion, and CNS for labeling ( Figure 4K ). At the earliest time point examined, 3 dpi, we observed axons traveling along the dura, but little other evidence of infection ( Figure 4L) . No labeled neuronal cell bodies on the dura were observed, consistent with the lack of neurons on this surface. In contrast, we did find labeled cell bodies in the trigeminal ganglion ( Figure 4M ). No infection was seen in the CNS, even at 4 dpi, consistent with the lack of inputs from the brain into the trigeminal ganglion (N = 4 animals).
To further characterize patterns and kinetics of viral transmission and directional specificity of transsynaptic spread, injections of rVSV(RABV-G) were made into the CP ( Figure 5A ). In order to determine which cells were labeled by direct uptake of virus in the inoculum, a separate set of animals were injected into the CP with the replication-incompetent rVSV G(RABV-G) (N = 3 animals, analyzed 3 dpi). Cells labeled by rVSV G(RABV-G) were observed in the CP, GP, substantia nigra (SN), thalamus, and layers 3 and 5 of the cortex, consistent with infection at the axon terminal and retrograde labeling of cell bodies of neurons known to project directly to the CP ( Figure 5C ) (Albin et al., 1995) . Areas labeled by CP injection are indicated in Figure 5B .
The patterns of spread for the replication-competent rVSV(RABV-G) were characterized over the course of 1-5 dpi ( Figures 5D-H) . During this interval, progressively more cells in infected regions were labeled by rVSV(RABV-G), including within the CP, nucleus basalis, cortex, and GP (listed in Figure 5B ). In addition, more cortical cells were labeled in clusters near cortical pyramidal neurons, both ipsilateral and contralateral to the injected side, including neurogliaform cells (data not shown). These data are in contrast to those observed following infection with an anterograde transsynaptic tracing virus, such as rVSV with its own G gene, rVSV(VSV-G) (Figure 5B) . At 3 dpi following rVSV(VSV-G) injection into the CP, the cerebral cortex was not labeled, but regions receiving projections from the CP, such as the STn, GP, and SN, were labeled (Beier et al., 2011) .
In order to investigate other areas for evidence of cell-to-cell retrograde transsynaptic spread, the nucleus basalis was examined following infection of the CP with replication-competent rVSV(RABV-G). The nucleus basalis was labeled by 2 dpi (Figures 5E-H) , consistent with at least a single transsynaptic jump, as this area does not directly project to the CP. The virus appeared to travel transsynaptically at the rate of roughly 1 synapse per day, as evidenced by the lack of labeled neurogliaform cells in the cortex, and lack of neurons in the nucleus basalis at 1 dpi, and label appearing in these cell types/areas at 2 dpi, as previously observed (Beier et al., 2011) . Labeling remained well-restricted to the expected corticostriatal circuits at 5 dpi, suggesting that viral spread becomes less efficient after crossing one or two connections, consistent with injections into V1 (Figure 4) . While glial cells can be infected and were observed near the injection site (van den Pol et al., 2002; Chauhan et al., 2010) , infected glial cells away from the injection site generally were not observed.
One advantage of having both anterograde and retrograde forms of the same virus is that they can be used in parallel, or in tandem, to trace circuitry to and from a single or multiple sites of injection, with each virus having similar kinetics of spread and gene expression. In fact, if different fluorophores are used in different viruses, e.g., rVSV(VSV-G) and rVSV(RABV-G), then the viruses can be co-injected into the same site and their transmission can be traced independently ( Figure 6A ). This is most straightforward if there are no cells at the injection site that are initially infected by both viruses. Co-infected cells can be easily detected, as they would express both fluorescent proteins shortly after injection.
In order to determine whether two viruses would allow simultaneous anterograde and retrograde transsynaptic tracing from a single injection site, a rVSV(VSV-G) expressing Venus and a rVSV(RABV-G) expressing mCherry were injected individually (Figures 6B-D) or co-injected (Figures 6E-G) into the motor cortex, and brains were examined 3 dpi. The pattern of labeling from the co-injected brains was equivalent to the patterns observed when each virus was injected individually: rVSV(VSV-G) was observed to infect neurons in the cortex, CP, and downstream nuclei, whereas the rVSV(RABV-G) was not observed to infect neurons in the CP, but rather in the thalamus and nucleus basalis (N = 4). The initial co-infection rate is dependent upon The presence or absence of labeling is indicated by (+) and (−), respectively. The extent of labeling is indicated by the number of (+). Some animals were infected with G viruses to determine which areas were labeled by direct uptake of the virions, rather than by replication and transmission. These were sacrificed at 3 dpi. (C) Parasaggital section of a brain infected with VSV[greek delta]G(RABV-G). The injection site is marked by a red arrow. Several areas that project directly to the CP were labeled due to direct uptake of the virions, including the cortex, thalamus, and GP (arrowheads), 3 dpi. the dose of the initial inocula. When injecting 3 × 10 3 focus forming units (ffu) rVSV(VSV-G) and 3 × 10 4 ffu rVSV(RABV-G), no co-infection was observed at the injection site. Thus, co-infection of the same brain region, without co-infection of the same cells,
does not alter the spreading behavior of either rVSV(VSV-G) or rVSV(RABV-G). One example of how this dual retrograde and anterograde transsynaptic tracing system can be used is to determine if three Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 8 distinct regions are connected and the directionality of any connections. For example, the anterograde transsynaptic virus can be injected into one region, the retrograde into another, and a third region can then be examined for evidence of labeling by either or both viruses (e.g., Figure 6H ). To test this possibility, rVSV(VSV-G) was injected into the motor cortex, rVSV(RABV-G) was injected into the substantia nigra pars reticulara (SNr), and animals were sacrificed at 3 dpi. We observed that cells were singly labeled, either with Venus [rVSV(VSV-G)] or with mCherry [rVSV(RABV-G)], and were located largely in different regions of the CP (Figures 6I,J) (N = 3) . These results suggest that the anterograde connections from the cells infected with rVSV(VSV-G) in the M1 were with CP MSNs that did not project to the region of the SNr injected with rVSV(RABV-G) (N = 3 animals).
In addition to polysynaptic tracing, VSV can be modified to trace circuits monosynaptically (Beier et al., 2011) . With RABV, this was achieved in vivo by first infecting with an adeno-associated virus (AAV) expressing TVA, a receptor for an avian retrovirus, and RABV-G (Wall et al., 2010) . This was followed 3 weeks later by infection with a G RABV with an EnvA/RABV-G chimeric glycoprotein on the virion surface (Wickersham et al., 2007b) , which allowed infection specifically of the cells expressing TVA. A similar strategy was used to test rVSV's ability to monosynaptically trace retrogradely connected neurons in vivo. Inputs to choline acetyltransferase (ChAT)-expressing neurons in the striatum were used for this test. These neurons primarily receive input from the cortex and the thalamus (Thomas et al., 2000; Bloomfield et al., 2007) (Figure 7A) . In order to mark this population, we crossed ChAT-Cre mice to Ai9 mice, which express tdTomato in cells with a Cre expression history (Madisen et al., 2010) . Six-week-old mice from this cross were injected in the CP with two AAV vectors: one expressing a Cre-conditional ("floxed") TVA-mCherry fusion protein, and another expressing a floxed RABV-G. Two weeks later, the mice were injected in the same coordinates with rVSV G with the EnvA/RABV-G chimeric glycoprotein on the virion surface [rVSV G(A/RG)] (Beier et al., 2011) . Cells successfully infected with these two AAV vectors could host infection by a rVSV and should be able to produce rVSV virions with RABV-G on the surface. Such starter cells should also express tdTomato and GFP. If rVSV were to be produced, and if it were to transmit across the synapse retrogradely, cortical and thalamic neurons should be labeled by GFP. Mice injected with these AAV and rVSV viruses were sacrificed 5 days after rVSV infection, and brains analyzed for fluorescence. As expected for starter cells, some neurons in the CP expressed both tdTomato and GFP (Figure 7B) . Outside of the CP, small numbers of GFP+ neurons that were not mCherry+ were observed in the cortex (Figures 7C,D) and thalamus (Figure 7E) , consistent with retrograde spread. Control animals not expressing Cre, or not injected with AAV encoding RABV-G, did not label cells in the cortex or thalamus (N = 3 for both controls and experimental condition).
Here, we report on the use of rVSV as a retrograde transsynaptic tracer for CNS circuitry. VSV can be modified to encode the RABV-G protein in the viral genome, allowing the virus to replicate and transmit across multiple synaptically connected cells, i.e., as a polysynaptic tracer. Alternatively, if the virus has the G gene deleted from its genome and RABV-G is provided in trans, it behaves as a monosynaptic tracer (Beier et al., 2011) . Although it has been known for many years that RABV travels retrogradely among neurons (Astic et al., 1993; Ugolini, 1995; Kelly and Strick, 2003) , and pseudotyping lentiviruses with RABV-G is sufficient for axonal transport (Mazarakis et al., 2001) , the retrograde transmission specificity among neurons had not been clearly shown to be a property of the G protein itself, as it might have been due to other viral proteins in addition to, or instead of, the viral G protein. Since native VSV does not have these retrograde transsynaptic properties (van den Pol et al., 2002; Beier et al., 2011) , and the only alteration to the VSV genome was the substitution of the VSV G gene with the G gene of RABV, it is clear that the RABV glycoprotein is responsible for retrograde direction of viral transmission across synapses, at least in the case of rVSV.
The early onset of gene expression from VSV relative to RABV (one hour vs. multiple hours) makes it beneficial in experimental paradigms in which the experiment needs to be done within a narrow window of time, such as tissue slices and explants. In addition, more than one transgene can be encoded in the viral genome without the need of a 2A or IRES element. The use of the first position of the genome enhances the expression level of the transgene inserted at that location, since VSV (and RABV) express genes in a transcriptional gradient; therefore, the first gene is the most highly transcribed (Knipe, 2007) . This leads to rational predictions of expression levels so that one can choose the position of insertion of a transgene, or transgenes, according to this gradient and the desired level of expression. The size of the viral capsid is apparently not rigid, allowing for the inclusion of genomes that are substantially larger than the native genome, unlike the rigid capacity for some other viral vectors, such as AAV Yan et al., 2000) .
The fact that VSV can be made to spread anterogradely (Beier et al., 2011) or retrogradely across synapses with the change of a single gene affords several advantages over viral tracers that heretofore have not shown such flexibility in the directionality of tracing. In addition to the obvious application of tracing anterograde connections, combinations can be made to exploit the different forms of the virus. One example that employs the simultaneous infection with an anterograde and retrograde form of VSV is demonstrated in Figure 6 . This experiment was designed to address whether the anterograde projections from the cortex to the CP would label the same brain regions as were labeled by a retrograde virus injected into the SN. Although a block of superinfection by the virus may preclude infection of the same cell with multiple rVSVs, adjacent cells could still become labeled by different viruses (Whitaker-Dowling et al., 1983) . The observed results could be due to a preferential labeling by the anterograde transsynaptic virus of indirect pathway MSNs in this experiment, which then synapse onto the GP, thereby reflecting a viral bias. Alternatively, it could indicate that the cortical neurons in the injected region largely do not label the MSNs that project to the area of the SN injected with the retrograde virus. One further possibility is that too little virus was used to observe co-labeling of a given region. However, given the density of infection (i.e., Figures 6I,J) , the latter possibility seems unlikely. Additionally, the spread of the polysynaptic rVSV(RABV-G) appears to attenuate with increasing numbers of synapses crossed, permitting an analysis of more restricted viral spread. This is quite fortuitous, as if spread were to continue, it would lead to widespread infection and lethality. In addition, reconstruction of connectivity would be more difficult. This reduced efficiency appears to also hold for the monosynaptic form of VSV complemented with RABV-G, as the efficiency of transmission appeared lower than the comparable experiment with RABV (Watabe-Uchida et al., 2012) . This is likely due to viral attenuation when VSV-G is replaced with RABV-G.
We were attracted to the use of VSV as a viral tracer due to its long track record as a safe, replication-competent laboratory agent. Laboratory workers using VSV have not contracted any diseases, and natural VSV infections among human populations in Central America and the southwestern United States (Rodríguez, 2002) occur without evident pathology (Johnson et al., 1966; Brody et al., 1967) . VSV was thus an attractive candidate for its use as a polysynaptic tracer for CNS studies, which requires an ability to replicate through multiple transmission cycles. Both replicationcompetent and incompetent forms of VSV are in use under Biosafety Level 2 containment. Replication-competent RABV is Biosafety Level 3, due to the fact that infection with replicationcompetent RABV is almost always fatal to humans and in mice when infected intracerebrally (Smith, 1981; Knipe, 2007) . Differences in pathogenicity between VSV and RABV are likely due to the ability of RABV to evade the innate immune system, particularly interferon (Hangartner et al., 2006; Junt et al., 2007; Lyles and Rupprecht, 2007; Rieder and Conzelmann, 2009; Iannacone et al., 2010) . VSV infection efficiently triggers an interferon response, and it has not evolved a method of escape from this response, unlike RABV (Brzózka et al., 2006) . In fact, VSV is being pursued as a vaccine for other viruses, including RABV (Lichty et al., 2004; Publicover et al., 2004; Kapadia et al., 2005; Schwartz et al., 2007; Iyer et al., 2009; Geisbert and Feldmann, 2011) . VSV does not typically spread beyond the initially infected site in the periphery (Kramer et al., 1983; Vogel and Fertsch, 1987) . This likely is the cause of the minor or absent symptoms in humans and animals infected in nature. Polysynaptic VSV vectors are thus predicted to be much safer than polysynaptic RABV vectors. We have tested this prediction by injecting a series of mice in the footpads and hind leg muscles with rVSV(RABV-G), with the result that no injected animals showed any evidence of morbidity or mortality (Beier, Goz et al., in preparation) .
While safer for laboratory workers than RABV, the main drawback to using VSV is its rapid cellular toxicity (van den Pol et al., 2009; Beier et al., 2011) . Toxicity is due to suppression of cellular transcription and a block in the export of cellular RNAs from the nucleus to the cytoplasm (Black and Lyles, 1992; Her et al., 1997; Ahmed and Lyles, 1998; Petersen et al., 2000; von Kobbe et al., 2000) , as well as inhibition of the translation of cellular mRNAs (Francoeur et al., 1987; Jayakar et al., 2000; Kopecky et al., Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 10 2001). VSV is much quicker to enact its gene expression program than is RABV, such that cells suffer the toxic effects more quickly than after RABV infection. One aspect of VSV that can be exploited in the future to ameliorate the speed of toxicity is the use of VSV mutants and variants. One such mutant is the M51R, which permitted us to conduct physiological analyses of pre-and post-synaptic cells (Beier et al., 2011) . We are in the process of examining the transmission properties of this mutant in vivo, as well as the effects of other mutations or viral variants on prolonging the health of neurons after infection.
rVSV vectors can be used to study the connectivity of neuronal circuitry. In addition to combinations of replication-competent forms of VSV, the replication-incompetent, monosynaptic forms of the virus can be easily combined, without the need to change viruses (Beier et al., 2011) . This allows a straightforward way to study both the projections into, and out from, a genetically defined cell population. This can be done with the same viral genome, with the only change needed being the glycoprotein, for the selection of the direction of transmission. This flexibility of VSV makes it a powerful, multi-application vector for studying connectivity in the CNS.
All rVSV clones were cloned from the rVSV G backbone (Chandran et al., 2005) . mCherry, Kusabira orange, Venus, and CFP were cloned into the first (GFP) position using XhoI and MscI sites, and VSV-G (a gift from Richard Mulligan, Harvard Medical School, Boston, MA) and RABV-G (a gift from Ed Callaway, Salk Institute, San Diego, CA) were cloned into the fifth (G) position using the MluI and NotI restriction sites. Genes for fluorescent proteins were obtained from Clontech. Viruses were rescued as previously described (Whelan et al., 1995) . At 95% confluency, eight 10 cm plates of BSR cells were infected at an MOI of 0.01. Viral supernatants were collected at 24-h time intervals and ultracentrifuged at 21,000 RPM using a SW28 rotor and resuspended in 0.2% of the original volume. For titering, concentrated viral stocks were applied in a dilution series to 100% confluent BSR cells and plates were examined at 12 hpi. Viral stocks were stored at −80 • C.
For G viruses, 293T cells were transfected with PEI (Ehrhardt et al., 2006) at 70% confluency on 10 cm dishes with 5 µg of pCAG-RABV-G. Twenty-four hours post-infection, the cells were infected at an MOI of 0.01 with rVSV G expressing either GFP or mCherry. Viral supernatants were collected for the subsequent 4 days at 24 h intervals.
Virus preparations are now available from the Salk GT3 viral core (http://vectorcore.salk.edu/). All plasmids are available from Addgene (http://www.addgene.org/).
AAV-FLEx-RABV-G and AAV-FLEx-TVA-mCherry plasmids originated from the Lab of Naoshige Uchida (Watabe-Uchida et al., 2012) , and virus stocks were generous gifts from Brad Lowell, Harvard Medical School.
ChAT-Cre (B6;129S6-Chat tm1(cre)Lowl /J) and Ai9 (B6.Cg-Gt(ROSA)26Sor<tm9(CAG-tdTomato)Hze>/J) mice were obtained from the Jackson Laboratory (Madisen et al., 2010) .
Eight-week-old CD-1 mice were injected using pulled capillary microdispensers (Drummond Scientific, Cat. No: 5-000-2005) , using coordinates from The Mouse Brain in Stereotaxic Coordinates (Franklin and Paxinos, 1997) . Injection coordinates (in mm) used were: For multi-color analysis (Figures 1C,D) , 3 × 10 9 ffu/mL rVSV was injected into various regions. For CP injections, 100 nL of rVSV(RABV-G) or rVSV(VSV-G) at 3 × 10 7 ffu/mL was injected at a rate of 100 nL/min. For the replication-incompetent viruses, 100 nL of 1 × 10 7 ffu/mL rVSV G(RABV-G) or rVSV G (VSV-G) was injected. In the motor cortex, 100 nL of 1 × 10 7 ffu/mL rVSV(RABV-G) was injected, and mice harvested 2 dpi. For V1 injections, 100 nL of 3 × 10 10 ffu/mL rVSV(RABV-G) was injected, and mice were examined 3 or 7 dpi.
For infections of the dura mater, 1 µL of 3 × 10 10 ffu/mL rVSV(RABV-G) was applied to the surface of the dura. The virus was allowed to absorb, and the surface was subsequently covered in bone wax, and the wound sutured.
For co-injections of virus into the same animal, 100 nL of a combination of 3 × 10 7 ffu/mL rVSV(VSV-G) and 3 × 10 8 ffu/mL rVSV(RABV-G) were co-injected into the motor cortex, and brains examined 3 dpi. For injections of the viruses into different regions, 100 nL of 3 × 10 7 ffu/mL rVSV(VSV-G) was injected into M1, and 100 nL of 3 × 10 8 ffu/mL rVSV(RABV-G) into the SNr, and brains examined 3 dpi. A lower titer of rVSV(VSV-G) was used, as rVSV(RABV-G) is attenuated.
All mouse work was conducted in biosafety containment level 2 conditions and was approved by the Longwood Medical Area Institutional Animal Care and Use Committee.
Recordings were made from cortical pyramidal neurons in slices taken from postnatal day 12-18 mice, inoculated in the CP 12-18 h prior with rVSV(RABV-G). Coronal slices (300 µm thick) were cut in ice-cold external solution containing (in mM): 110 choline, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 7 MgCl 2 , 0.5 CaCl 2 , 25 glucose, 11.6 Na-ascorbate, and 3.1 Na-pyruvate, bubbled with 95% O 2 and 5% CO 2 . Slices were then transferred to artificial cerebrospinal fluid (ACSF) containing (in mM): 127 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 1 MgCl 2 , 2 CaCl 2 , and 25 glucose, bubbled with 95% O 2 and 5% CO 2 . After an incubation period Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 11 of 30-40 min at 34 • C, slices were stored at room temperature. All experiments were conducted at room temperature (25 • C). In all experiments, 50 µM picrotoxin, 10 µM 2,3-Dioxo-6-nitro-1, 2, 3, 4 -tetrahydrobenzo [f]quinoxaline -7-sulfonamide (NBQX), and 10 µM 3-((R)-2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) were present in the ACSF to block GABAA/C, AMPA, and NMDA receptor-mediated transmission, respectively. All chemicals were from Sigma or Tocris.
Whole-cell recordings were obtained from infected and uninfected deep layer cortical pyramidal neurons identified with video-IR/DIC and GFP fluorescence was detected using epifluorescence illumination. With the deep layers of the cortex, 2-photon laser scanning microscopy (2PLSM) was used to confirm the cell types based on morphology. Deep layer pyramidal neurons had large cell bodies, classic pyramidal shape and dendritic spines. Glass electrodes (2-4 M ) were filled with internal solution containing (in mM): 135 KMeSO 4 , 5 KCl, 5 HEPES, 4 MgATP, 0.3 NaGTP, 10 Na 2 HPO 4 , 1 EGTA, and 0.01 Alexa Fluor-594 (to image neuronal morphology) adjusted to pH 7.4 with KOH. Current and voltage recordings were made at room temperature using a AxoPatch 200B or a Multiclamp 700B amplifier. Data was filtered at 5 kHz and digitized at 10 kHz.
Imaging and physiology data were acquired and analyzed as described previously (Carter and Sabatini, 2004) . Resting membrane potential was determined by the average of three 5-s sweeps with no injected current. Passive properties of the cell, membrane (Rm) and series resistance (Rs) and capacitance (Cm), were measured while clamping cells at −65 mV and applying voltage steps from −55 to −75 mV. The current-firing relationship was determined in current clamp with 1-s periods of injected current from 100 to 500 pA.
The time course of viral gene expression experiments were carried out in organotypic hippocampal slice cultures prepared from postnatal day 5-7 Sprague-Dawley rats as described previously (Stoppini et al., 1991) . Slices were infected after 7 days in vitro, and images were acquired on a two-photon microscope.
|
What virus are used by the most successful neuronal circuit tracing methods?
| 1,935 |
Pseudorabies (PRV) and Rabies viruses (RABV)
| 2,185 |
1,621 |
Vesicular stomatitis virus with the rabies virus glycoprotein directs retrograde transsynaptic transport among neurons in vivo
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3566411/
SHA: ee48061797d29eeef5a9e606841bf8ab04b1d75b
Authors: Beier, Kevin T.; Saunders, Arpiar B.; Oldenburg, Ian A.; Sabatini, Bernardo L.; Cepko, Constance L.
Date: 2013-02-07
DOI: 10.3389/fncir.2013.00011
License: cc-by
Abstract: Defining the connections among neurons is critical to our understanding of the structure and function of the nervous system. Recombinant viruses engineered to transmit across synapses provide a powerful approach for the dissection of neuronal circuitry in vivo. We recently demonstrated that recombinant vesicular stomatitis virus (VSV) can be endowed with anterograde or retrograde transsynaptic tracing ability by providing the virus with different glycoproteins. Here we extend the characterization of the transmission and gene expression of recombinant VSV (rVSV) with the rabies virus glycoprotein (RABV-G), and provide examples of its activity relative to the anterograde transsynaptic tracer form of rVSV. rVSV with RABV-G was found to drive strong expression of transgenes and to spread rapidly from neuron to neuron in only a retrograde manner. Depending upon how the RABV-G was delivered, VSV served as a polysynaptic or monosynaptic tracer, or was able to define projections through axonal uptake and retrograde transport. In animals co-infected with rVSV in its anterograde form, rVSV with RABV-G could be used to begin to characterize the similarities and differences in connections to different areas. rVSV with RABV-G provides a flexible, rapid, and versatile tracing tool that complements the previously described VSV-based anterograde transsynaptic tracer.
Text: Mapping neuronal connectivity in the central nervous system (CNS) of even simple organisms is a difficult task. Recombinant viruses engineered to trace synaptic connections and express transgenes promise to enable higher-throughput mapping of connections among neurons than other methods, e.g., serial reconstruction from electron micrographs (Bock et al., 2011; Briggman et al., 2011) . The Pseudorabies (PRV) and Rabies viruses (RABV) have been the best characterized and most utilized circuit tracing viruses to date (Ugolini et al., 1989; Kelly and Strick, 2000) . RABV was recently modified by Wickersham and colleagues such that it can travel across only one synapse, allowing for a straightforward definition of monosynaptic connections (Wickersham et al., 2007b) . This strategy permitted the first unambiguous identification of retrogradely connected cells from an initially infected cell ("starter cell"), without the need for electrophysiology. Moreover, the starter cell could be defined through the expression of a specific viral receptor that limited the initial infection.
Recently, we created an anterograde monosynaptic virus that complements the previously available retrograde viral tracers (Beier et al., 2011) . Vesicular stomatitis virus (VSV), a virus related to RABV, with its own glycoprotein (G) gene (VSV-G), or with a G from the unrelated lymphocytic choriomeningitis virus (LCMV-G), spreads in the anterograde direction across synapses. VSV can be used as a polysynaptic tracer that spreads across many synapses, owing to the fact that the normal, replicationcompetent form of the virus does not cause serious diseases in humans (Brandly and Hanson, 1957; Johnson et al., 1966; Brody et al., 1967) . Whether the virus is a monosynaptic or polysynaptic tracer is determined by the method of delivery of the G gene ( Figure 1A) . Advantages of VSV are that it is well-characterized, is relatively simple in comparison to PRV, and it rapidly grows to high titer in tissue culture cells. It is also being developed as a vaccine vector, often using a G of another virus as the immunogen, as well as being developed as a cytocidal agent that will target tumor cells in humans (Balachandran and Barber, 2000; Stojdl et al., 2000 Stojdl et al., , 2003 .
Previous studies of the anatomical patterns of transmission, as well as physiological recordings, have shown that the transmission of VSV and RABV among neurons is via synapses (Kelly and Strick, 2000; Wickersham et al., 2007b; Beier et al., 2011) . In addition, it has been shown that RABV, as well as lentiviruses with RABV-G in their envelope, travel retrogradely from an injection site (Mazarakis et al., 2001; Wickersham et al., 2007a) . We hypothesized that providing a recombinant VSV (rVSV) with the RABV-G would create a retrograde polysynaptic transsynaptic tracer without the biosafety concerns inherent to RABV. Our initial characterization of rVSV with RABV-G showed that indeed FIGURE 1 | Synaptic tracing strategies using VSV. (A) Schematic illustrating the strategies for polysynaptic or monosynaptic retrograde or anterograde transsynaptic transmission of rVSV encoding GFP. The initially infected cell is indicated by an asterisk. VSV encoding a glycoprotein (G) within its genome can spread polysynaptically. The direction of the spread depends on the identity of the glycoprotein. Infected neurons are shown in green. In some cases, the initially infected starter cell can be defined by the expression of an avian receptor, TVA (tagged with a red fluorescent protein). The TVA-expressing neurons can then be specifically infected by rVSV G with the EnvA/RABV-G (A/RG) glycoprotein (Wickersham et al., 2007b) on the virion surface [rVSV G(A/RG)]. These starter cells are then yellow, due to viral GFP and mCherry from TVA-mCherry expression. For monosynaptic tracing, the G protein is expressed in trans in the TVA-expressing cell, and thus complements rVSV G to allow transmission in a specific direction. (B) Genomic diagrams of rVSV vectors. All VSVs contain four essential proteins: N, P, M, and L. Some viruses encode a G gene in their genome, which allows them to spread polysynaptically. rVSV vectors typically encode a transgene in the first position, while others carry an additional transgene in the G position. (C) Morphological characterization of rVSV-infected neurons in several locations within the mouse brain. (i,ii) Caudate-putamen (CP) neurons at 4 dpi from an injection of the CP with rVSV(VSV-G) viruses encoding (i) CFP or (ii) Korange. (iii) Labeled neurons of the CA1 region of the hippocampus are shown at 5 dpi following injection into the hippocampus of rVSV(VSV-G) encoding Venus. (iv,v) Cortical pyramidal neurons are shown following injection into the CP of rVSV(RABV-G) expressing (iv) GFP at 24 hpi, or (v) mCherry at 48 hpi. Inset in (iv) is a high magnification of the neuron in panel (iv), highlighting labeling of dendritic spines. (vi) Multiple viruses can be co-injected into the same animal. Here, individual rVSV G(VSV-G) viruses encoding CFP, GFP , Venus, Korange, and mCherry were used to infect the cortex. Scale bars = 50 µm.
www.frontiersin.org February 2013 | Volume 7 | Article 11 | 2 it could be taken up as a retrograde tracer (Beier et al., 2011) .
To determine if it could transmit among neurons following its replication in neurons, and to further analyze the transmission patterns of both the monosynaptic and polysynaptic forms of rVSV with RABV-G, we made injections into several CNS and peripheral locations. In addition, we performed co-infections of rVSV with RABV-G and the anterograde form of rVSV in order to exploit the differences in the directionality of transmission of these two viruses in mapping circuits.
Schematics of viruses created and used throughout this study are shown in Figure 1 . We created rVSV vector plasmids carrying different transgenes in either the first or fifth genomic positions ( Figure 1B) . After rescuing each virus, we tested the ability of each to express transgenes in different brain regions through intracranial injections ( Figure 1C ). All rVSV vectors drove robust fluorophore expression 1 or 2 days post-infection (hpi) ( Figure 1C ) (van den Pol et al., 2009) . In fact, by 12 hpi, labeling was sufficiently bright to image fine morphological details, such as dendritic spines ( Figure 1C ,iv).
To characterize the physiological properties of cells infected with rVSV, we tested a replication-competent rVSV encoding GFP, with RABV-G in the genome in place of VSV-G [hereafter designated rVSV(RABV-G)]. van den Pol et al. reported that hippocampal neurons infected with replication-incompetent (G-deleted or " G") rVSV were physiologically healthy at 12-14 hpi, but were less so by 1 day post-infection (dpi) (van den Pol et al., 2009) . Given the known toxicity of both VSV and RABV-G (Coulon et al., 1982) , we tested the physiology of cortical pyramidal neurons in the motor cortex (M1) infected with rVSV(RABV-G). Between 12 and 18 hpi, the membrane capacitance, input resistance, resting membrane potential, and current-to-action potential firing relationship were indistinguishable between infected and uninfected neurons (Figure 2) . However, by 2 dpi, electrophysiological properties were so abnormal in the infected cortical pyramidal cells that physiological measurements could not be made.
The speed and strength of the expression of transgenes encoded by VSV depends upon the gene's genomic position (van den Pol et al., 2009; Beier et al., 2011) . Genes in the first position are expressed the most highly, with a decrease in the level of expression in positions more 3 within the viral plus strand. When GFP was inserted into the first position of VSV, GFP fluorescence was first detectable at approximately 1 hpi in cultured cells (van den Pol et al., 2009) . In order to quantify the relative expression of a fluorescent protein in the first genomic position in neurons, rat hippocampal slices were infected with a replication-incompetent rVSV that expresses mCherry (rVSV G, Figures 1A,B) . This was a G virus which had the RABV-G supplied in trans during the preparation of the virus stock [referred to as rVSV G(RABV-G)].
Average fluorescence intensity of the infected cells was measured every hour over the course of 18 h. By 4 hpi at 37 • C, red fluorescence was clearly visible, and reached maximal levels by approximately 14 hpi (N = 3, Figure 3 ). Similar results were obtained with a virus encoding GFP in the first genomic position rather than mCherry (i.e., Figure 1B ) (N = 3).
We previously demonstrated that rVSV(RABV-G) could be taken up retrogradely by neurons (Beier et al., 2011) , but these experiments did not distinguish between direct axonal uptake of the initial inoculum vs. retrograde transsynaptic transmission following viral replication. To distinguish between these two mechanisms and to extend the previous analyses, we conducted further experiments in the mammalian visual system (Figures 4A-G) . As visual cortex area 1 (V1) does not receive direct projections from retinal ganglion cells (RGCs), but rather receives secondary input from RGCs via the lateral geniculate nucleus (LGN), infection of RGCs from injection of V1 would demonstrate retrograde transmission from cells which supported at least one round of viral replication. Following a V1 injection with rVSV(RABV-G), GFP-positive RGCs were observed in the retina by 3 dpi (N = 3; Figure 4G ). Importantly, viral labeling in the brain was restricted to primary and secondary projection areas, even at 7 dpi. These included the LGN ( Figure 4D ) and the hypothalamus (Figure 4E) , two areas known to project directly to V1 (Kandel, 2000) . Selective labeling was observed in other areas, such as cortical areas surrounding V1 (Figure 4C) , which project directly to V1, and also in the superior colliculus (SC) stratum griseum centrale, which projects to the LGN ( Figure 4F) . Labeling was also observed in the nucleus basalis, which projects to the cortex, as well as many components of the basal ganglia circuit, which provide input to the thalamus [such as the caudate-putamen (CP), globus pallidus (GP), and the subthalamic nucleus (STn)]. The amygdala, which projects to the hypothalamus, was also labeled. Consistent with a lack of widespread viral transmission, animals did not exhibit signs of disease at 7 dpi.
These data show that rVSV(RABV-G) can spread in a retrograde direction from the injection site, but do not address whether the virus can spread exclusively in the retrograde direction. Directional transsynaptic specificity can only be definitively addressed using a unidirectional circuit. We therefore turned to the primary motor cortex (M1) to CP connection, in which neurons project from the cortex to the CP, but not in the other direction ( Figure 4H ) (Beier et al., 2011) . Injections of rVSV(RABV-G) into M1 should not label neurons in the CP if the virus can only label cells across synapses in the retrograde direction. Indeed, at 2 dpi, areas directly projecting to the injection site, including the contralateral cortex, were labeled ( Figure 4I ). Only axons from cortical cells were observed in the CP, with no GFP-labeled cell bodies present in the CP (Figure 4J) , consistent with lack of anterograde transsynaptic spread. By 3 dpi, a small number of medium spiny neurons (MSNs) in the CP were observed, likely via secondary spread from initially infected thalamic or GP neurons (data not shown).
A particular advantage of retrograde viral tracers is the ability to label CNS neurons projecting to peripheral sites. This has been a powerful application of both RABV and PRV (Ugolini et al., 1989; Standish et al., 1994) . To test if rVSV(RABV-G) could also perform this function, we examined the innervation of the dura surface by neurons of the trigeminal ganglion, a neuronal circuit thought to be involved in migraine headaches (Penfield and McNaughton, 1940; Mayberg et al., 1984) . These neurons have axons, but not canonical dendrites, and send projections into the spinal cord and brainstem. Therefore, the only way trigeminal neurons could become labeled from viral application to the dura is through retrograde uptake of the virus. We applied rVSV(RABV-G) to the intact dura mater and analyzed the dura, trigeminal ganglion, and CNS for labeling ( Figure 4K ). At the earliest time point examined, 3 dpi, we observed axons traveling along the dura, but little other evidence of infection ( Figure 4L) . No labeled neuronal cell bodies on the dura were observed, consistent with the lack of neurons on this surface. In contrast, we did find labeled cell bodies in the trigeminal ganglion ( Figure 4M ). No infection was seen in the CNS, even at 4 dpi, consistent with the lack of inputs from the brain into the trigeminal ganglion (N = 4 animals).
To further characterize patterns and kinetics of viral transmission and directional specificity of transsynaptic spread, injections of rVSV(RABV-G) were made into the CP ( Figure 5A ). In order to determine which cells were labeled by direct uptake of virus in the inoculum, a separate set of animals were injected into the CP with the replication-incompetent rVSV G(RABV-G) (N = 3 animals, analyzed 3 dpi). Cells labeled by rVSV G(RABV-G) were observed in the CP, GP, substantia nigra (SN), thalamus, and layers 3 and 5 of the cortex, consistent with infection at the axon terminal and retrograde labeling of cell bodies of neurons known to project directly to the CP ( Figure 5C ) (Albin et al., 1995) . Areas labeled by CP injection are indicated in Figure 5B .
The patterns of spread for the replication-competent rVSV(RABV-G) were characterized over the course of 1-5 dpi ( Figures 5D-H) . During this interval, progressively more cells in infected regions were labeled by rVSV(RABV-G), including within the CP, nucleus basalis, cortex, and GP (listed in Figure 5B ). In addition, more cortical cells were labeled in clusters near cortical pyramidal neurons, both ipsilateral and contralateral to the injected side, including neurogliaform cells (data not shown). These data are in contrast to those observed following infection with an anterograde transsynaptic tracing virus, such as rVSV with its own G gene, rVSV(VSV-G) (Figure 5B) . At 3 dpi following rVSV(VSV-G) injection into the CP, the cerebral cortex was not labeled, but regions receiving projections from the CP, such as the STn, GP, and SN, were labeled (Beier et al., 2011) .
In order to investigate other areas for evidence of cell-to-cell retrograde transsynaptic spread, the nucleus basalis was examined following infection of the CP with replication-competent rVSV(RABV-G). The nucleus basalis was labeled by 2 dpi (Figures 5E-H) , consistent with at least a single transsynaptic jump, as this area does not directly project to the CP. The virus appeared to travel transsynaptically at the rate of roughly 1 synapse per day, as evidenced by the lack of labeled neurogliaform cells in the cortex, and lack of neurons in the nucleus basalis at 1 dpi, and label appearing in these cell types/areas at 2 dpi, as previously observed (Beier et al., 2011) . Labeling remained well-restricted to the expected corticostriatal circuits at 5 dpi, suggesting that viral spread becomes less efficient after crossing one or two connections, consistent with injections into V1 (Figure 4) . While glial cells can be infected and were observed near the injection site (van den Pol et al., 2002; Chauhan et al., 2010) , infected glial cells away from the injection site generally were not observed.
One advantage of having both anterograde and retrograde forms of the same virus is that they can be used in parallel, or in tandem, to trace circuitry to and from a single or multiple sites of injection, with each virus having similar kinetics of spread and gene expression. In fact, if different fluorophores are used in different viruses, e.g., rVSV(VSV-G) and rVSV(RABV-G), then the viruses can be co-injected into the same site and their transmission can be traced independently ( Figure 6A ). This is most straightforward if there are no cells at the injection site that are initially infected by both viruses. Co-infected cells can be easily detected, as they would express both fluorescent proteins shortly after injection.
In order to determine whether two viruses would allow simultaneous anterograde and retrograde transsynaptic tracing from a single injection site, a rVSV(VSV-G) expressing Venus and a rVSV(RABV-G) expressing mCherry were injected individually (Figures 6B-D) or co-injected (Figures 6E-G) into the motor cortex, and brains were examined 3 dpi. The pattern of labeling from the co-injected brains was equivalent to the patterns observed when each virus was injected individually: rVSV(VSV-G) was observed to infect neurons in the cortex, CP, and downstream nuclei, whereas the rVSV(RABV-G) was not observed to infect neurons in the CP, but rather in the thalamus and nucleus basalis (N = 4). The initial co-infection rate is dependent upon The presence or absence of labeling is indicated by (+) and (−), respectively. The extent of labeling is indicated by the number of (+). Some animals were infected with G viruses to determine which areas were labeled by direct uptake of the virions, rather than by replication and transmission. These were sacrificed at 3 dpi. (C) Parasaggital section of a brain infected with VSV[greek delta]G(RABV-G). The injection site is marked by a red arrow. Several areas that project directly to the CP were labeled due to direct uptake of the virions, including the cortex, thalamus, and GP (arrowheads), 3 dpi. the dose of the initial inocula. When injecting 3 × 10 3 focus forming units (ffu) rVSV(VSV-G) and 3 × 10 4 ffu rVSV(RABV-G), no co-infection was observed at the injection site. Thus, co-infection of the same brain region, without co-infection of the same cells,
does not alter the spreading behavior of either rVSV(VSV-G) or rVSV(RABV-G). One example of how this dual retrograde and anterograde transsynaptic tracing system can be used is to determine if three Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 8 distinct regions are connected and the directionality of any connections. For example, the anterograde transsynaptic virus can be injected into one region, the retrograde into another, and a third region can then be examined for evidence of labeling by either or both viruses (e.g., Figure 6H ). To test this possibility, rVSV(VSV-G) was injected into the motor cortex, rVSV(RABV-G) was injected into the substantia nigra pars reticulara (SNr), and animals were sacrificed at 3 dpi. We observed that cells were singly labeled, either with Venus [rVSV(VSV-G)] or with mCherry [rVSV(RABV-G)], and were located largely in different regions of the CP (Figures 6I,J) (N = 3) . These results suggest that the anterograde connections from the cells infected with rVSV(VSV-G) in the M1 were with CP MSNs that did not project to the region of the SNr injected with rVSV(RABV-G) (N = 3 animals).
In addition to polysynaptic tracing, VSV can be modified to trace circuits monosynaptically (Beier et al., 2011) . With RABV, this was achieved in vivo by first infecting with an adeno-associated virus (AAV) expressing TVA, a receptor for an avian retrovirus, and RABV-G (Wall et al., 2010) . This was followed 3 weeks later by infection with a G RABV with an EnvA/RABV-G chimeric glycoprotein on the virion surface (Wickersham et al., 2007b) , which allowed infection specifically of the cells expressing TVA. A similar strategy was used to test rVSV's ability to monosynaptically trace retrogradely connected neurons in vivo. Inputs to choline acetyltransferase (ChAT)-expressing neurons in the striatum were used for this test. These neurons primarily receive input from the cortex and the thalamus (Thomas et al., 2000; Bloomfield et al., 2007) (Figure 7A) . In order to mark this population, we crossed ChAT-Cre mice to Ai9 mice, which express tdTomato in cells with a Cre expression history (Madisen et al., 2010) . Six-week-old mice from this cross were injected in the CP with two AAV vectors: one expressing a Cre-conditional ("floxed") TVA-mCherry fusion protein, and another expressing a floxed RABV-G. Two weeks later, the mice were injected in the same coordinates with rVSV G with the EnvA/RABV-G chimeric glycoprotein on the virion surface [rVSV G(A/RG)] (Beier et al., 2011) . Cells successfully infected with these two AAV vectors could host infection by a rVSV and should be able to produce rVSV virions with RABV-G on the surface. Such starter cells should also express tdTomato and GFP. If rVSV were to be produced, and if it were to transmit across the synapse retrogradely, cortical and thalamic neurons should be labeled by GFP. Mice injected with these AAV and rVSV viruses were sacrificed 5 days after rVSV infection, and brains analyzed for fluorescence. As expected for starter cells, some neurons in the CP expressed both tdTomato and GFP (Figure 7B) . Outside of the CP, small numbers of GFP+ neurons that were not mCherry+ were observed in the cortex (Figures 7C,D) and thalamus (Figure 7E) , consistent with retrograde spread. Control animals not expressing Cre, or not injected with AAV encoding RABV-G, did not label cells in the cortex or thalamus (N = 3 for both controls and experimental condition).
Here, we report on the use of rVSV as a retrograde transsynaptic tracer for CNS circuitry. VSV can be modified to encode the RABV-G protein in the viral genome, allowing the virus to replicate and transmit across multiple synaptically connected cells, i.e., as a polysynaptic tracer. Alternatively, if the virus has the G gene deleted from its genome and RABV-G is provided in trans, it behaves as a monosynaptic tracer (Beier et al., 2011) . Although it has been known for many years that RABV travels retrogradely among neurons (Astic et al., 1993; Ugolini, 1995; Kelly and Strick, 2003) , and pseudotyping lentiviruses with RABV-G is sufficient for axonal transport (Mazarakis et al., 2001) , the retrograde transmission specificity among neurons had not been clearly shown to be a property of the G protein itself, as it might have been due to other viral proteins in addition to, or instead of, the viral G protein. Since native VSV does not have these retrograde transsynaptic properties (van den Pol et al., 2002; Beier et al., 2011) , and the only alteration to the VSV genome was the substitution of the VSV G gene with the G gene of RABV, it is clear that the RABV glycoprotein is responsible for retrograde direction of viral transmission across synapses, at least in the case of rVSV.
The early onset of gene expression from VSV relative to RABV (one hour vs. multiple hours) makes it beneficial in experimental paradigms in which the experiment needs to be done within a narrow window of time, such as tissue slices and explants. In addition, more than one transgene can be encoded in the viral genome without the need of a 2A or IRES element. The use of the first position of the genome enhances the expression level of the transgene inserted at that location, since VSV (and RABV) express genes in a transcriptional gradient; therefore, the first gene is the most highly transcribed (Knipe, 2007) . This leads to rational predictions of expression levels so that one can choose the position of insertion of a transgene, or transgenes, according to this gradient and the desired level of expression. The size of the viral capsid is apparently not rigid, allowing for the inclusion of genomes that are substantially larger than the native genome, unlike the rigid capacity for some other viral vectors, such as AAV Yan et al., 2000) .
The fact that VSV can be made to spread anterogradely (Beier et al., 2011) or retrogradely across synapses with the change of a single gene affords several advantages over viral tracers that heretofore have not shown such flexibility in the directionality of tracing. In addition to the obvious application of tracing anterograde connections, combinations can be made to exploit the different forms of the virus. One example that employs the simultaneous infection with an anterograde and retrograde form of VSV is demonstrated in Figure 6 . This experiment was designed to address whether the anterograde projections from the cortex to the CP would label the same brain regions as were labeled by a retrograde virus injected into the SN. Although a block of superinfection by the virus may preclude infection of the same cell with multiple rVSVs, adjacent cells could still become labeled by different viruses (Whitaker-Dowling et al., 1983) . The observed results could be due to a preferential labeling by the anterograde transsynaptic virus of indirect pathway MSNs in this experiment, which then synapse onto the GP, thereby reflecting a viral bias. Alternatively, it could indicate that the cortical neurons in the injected region largely do not label the MSNs that project to the area of the SN injected with the retrograde virus. One further possibility is that too little virus was used to observe co-labeling of a given region. However, given the density of infection (i.e., Figures 6I,J) , the latter possibility seems unlikely. Additionally, the spread of the polysynaptic rVSV(RABV-G) appears to attenuate with increasing numbers of synapses crossed, permitting an analysis of more restricted viral spread. This is quite fortuitous, as if spread were to continue, it would lead to widespread infection and lethality. In addition, reconstruction of connectivity would be more difficult. This reduced efficiency appears to also hold for the monosynaptic form of VSV complemented with RABV-G, as the efficiency of transmission appeared lower than the comparable experiment with RABV (Watabe-Uchida et al., 2012) . This is likely due to viral attenuation when VSV-G is replaced with RABV-G.
We were attracted to the use of VSV as a viral tracer due to its long track record as a safe, replication-competent laboratory agent. Laboratory workers using VSV have not contracted any diseases, and natural VSV infections among human populations in Central America and the southwestern United States (Rodríguez, 2002) occur without evident pathology (Johnson et al., 1966; Brody et al., 1967) . VSV was thus an attractive candidate for its use as a polysynaptic tracer for CNS studies, which requires an ability to replicate through multiple transmission cycles. Both replicationcompetent and incompetent forms of VSV are in use under Biosafety Level 2 containment. Replication-competent RABV is Biosafety Level 3, due to the fact that infection with replicationcompetent RABV is almost always fatal to humans and in mice when infected intracerebrally (Smith, 1981; Knipe, 2007) . Differences in pathogenicity between VSV and RABV are likely due to the ability of RABV to evade the innate immune system, particularly interferon (Hangartner et al., 2006; Junt et al., 2007; Lyles and Rupprecht, 2007; Rieder and Conzelmann, 2009; Iannacone et al., 2010) . VSV infection efficiently triggers an interferon response, and it has not evolved a method of escape from this response, unlike RABV (Brzózka et al., 2006) . In fact, VSV is being pursued as a vaccine for other viruses, including RABV (Lichty et al., 2004; Publicover et al., 2004; Kapadia et al., 2005; Schwartz et al., 2007; Iyer et al., 2009; Geisbert and Feldmann, 2011) . VSV does not typically spread beyond the initially infected site in the periphery (Kramer et al., 1983; Vogel and Fertsch, 1987) . This likely is the cause of the minor or absent symptoms in humans and animals infected in nature. Polysynaptic VSV vectors are thus predicted to be much safer than polysynaptic RABV vectors. We have tested this prediction by injecting a series of mice in the footpads and hind leg muscles with rVSV(RABV-G), with the result that no injected animals showed any evidence of morbidity or mortality (Beier, Goz et al., in preparation) .
While safer for laboratory workers than RABV, the main drawback to using VSV is its rapid cellular toxicity (van den Pol et al., 2009; Beier et al., 2011) . Toxicity is due to suppression of cellular transcription and a block in the export of cellular RNAs from the nucleus to the cytoplasm (Black and Lyles, 1992; Her et al., 1997; Ahmed and Lyles, 1998; Petersen et al., 2000; von Kobbe et al., 2000) , as well as inhibition of the translation of cellular mRNAs (Francoeur et al., 1987; Jayakar et al., 2000; Kopecky et al., Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 10 2001). VSV is much quicker to enact its gene expression program than is RABV, such that cells suffer the toxic effects more quickly than after RABV infection. One aspect of VSV that can be exploited in the future to ameliorate the speed of toxicity is the use of VSV mutants and variants. One such mutant is the M51R, which permitted us to conduct physiological analyses of pre-and post-synaptic cells (Beier et al., 2011) . We are in the process of examining the transmission properties of this mutant in vivo, as well as the effects of other mutations or viral variants on prolonging the health of neurons after infection.
rVSV vectors can be used to study the connectivity of neuronal circuitry. In addition to combinations of replication-competent forms of VSV, the replication-incompetent, monosynaptic forms of the virus can be easily combined, without the need to change viruses (Beier et al., 2011) . This allows a straightforward way to study both the projections into, and out from, a genetically defined cell population. This can be done with the same viral genome, with the only change needed being the glycoprotein, for the selection of the direction of transmission. This flexibility of VSV makes it a powerful, multi-application vector for studying connectivity in the CNS.
All rVSV clones were cloned from the rVSV G backbone (Chandran et al., 2005) . mCherry, Kusabira orange, Venus, and CFP were cloned into the first (GFP) position using XhoI and MscI sites, and VSV-G (a gift from Richard Mulligan, Harvard Medical School, Boston, MA) and RABV-G (a gift from Ed Callaway, Salk Institute, San Diego, CA) were cloned into the fifth (G) position using the MluI and NotI restriction sites. Genes for fluorescent proteins were obtained from Clontech. Viruses were rescued as previously described (Whelan et al., 1995) . At 95% confluency, eight 10 cm plates of BSR cells were infected at an MOI of 0.01. Viral supernatants were collected at 24-h time intervals and ultracentrifuged at 21,000 RPM using a SW28 rotor and resuspended in 0.2% of the original volume. For titering, concentrated viral stocks were applied in a dilution series to 100% confluent BSR cells and plates were examined at 12 hpi. Viral stocks were stored at −80 • C.
For G viruses, 293T cells were transfected with PEI (Ehrhardt et al., 2006) at 70% confluency on 10 cm dishes with 5 µg of pCAG-RABV-G. Twenty-four hours post-infection, the cells were infected at an MOI of 0.01 with rVSV G expressing either GFP or mCherry. Viral supernatants were collected for the subsequent 4 days at 24 h intervals.
Virus preparations are now available from the Salk GT3 viral core (http://vectorcore.salk.edu/). All plasmids are available from Addgene (http://www.addgene.org/).
AAV-FLEx-RABV-G and AAV-FLEx-TVA-mCherry plasmids originated from the Lab of Naoshige Uchida (Watabe-Uchida et al., 2012) , and virus stocks were generous gifts from Brad Lowell, Harvard Medical School.
ChAT-Cre (B6;129S6-Chat tm1(cre)Lowl /J) and Ai9 (B6.Cg-Gt(ROSA)26Sor<tm9(CAG-tdTomato)Hze>/J) mice were obtained from the Jackson Laboratory (Madisen et al., 2010) .
Eight-week-old CD-1 mice were injected using pulled capillary microdispensers (Drummond Scientific, Cat. No: 5-000-2005) , using coordinates from The Mouse Brain in Stereotaxic Coordinates (Franklin and Paxinos, 1997) . Injection coordinates (in mm) used were: For multi-color analysis (Figures 1C,D) , 3 × 10 9 ffu/mL rVSV was injected into various regions. For CP injections, 100 nL of rVSV(RABV-G) or rVSV(VSV-G) at 3 × 10 7 ffu/mL was injected at a rate of 100 nL/min. For the replication-incompetent viruses, 100 nL of 1 × 10 7 ffu/mL rVSV G(RABV-G) or rVSV G (VSV-G) was injected. In the motor cortex, 100 nL of 1 × 10 7 ffu/mL rVSV(RABV-G) was injected, and mice harvested 2 dpi. For V1 injections, 100 nL of 3 × 10 10 ffu/mL rVSV(RABV-G) was injected, and mice were examined 3 or 7 dpi.
For infections of the dura mater, 1 µL of 3 × 10 10 ffu/mL rVSV(RABV-G) was applied to the surface of the dura. The virus was allowed to absorb, and the surface was subsequently covered in bone wax, and the wound sutured.
For co-injections of virus into the same animal, 100 nL of a combination of 3 × 10 7 ffu/mL rVSV(VSV-G) and 3 × 10 8 ffu/mL rVSV(RABV-G) were co-injected into the motor cortex, and brains examined 3 dpi. For injections of the viruses into different regions, 100 nL of 3 × 10 7 ffu/mL rVSV(VSV-G) was injected into M1, and 100 nL of 3 × 10 8 ffu/mL rVSV(RABV-G) into the SNr, and brains examined 3 dpi. A lower titer of rVSV(VSV-G) was used, as rVSV(RABV-G) is attenuated.
All mouse work was conducted in biosafety containment level 2 conditions and was approved by the Longwood Medical Area Institutional Animal Care and Use Committee.
Recordings were made from cortical pyramidal neurons in slices taken from postnatal day 12-18 mice, inoculated in the CP 12-18 h prior with rVSV(RABV-G). Coronal slices (300 µm thick) were cut in ice-cold external solution containing (in mM): 110 choline, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 7 MgCl 2 , 0.5 CaCl 2 , 25 glucose, 11.6 Na-ascorbate, and 3.1 Na-pyruvate, bubbled with 95% O 2 and 5% CO 2 . Slices were then transferred to artificial cerebrospinal fluid (ACSF) containing (in mM): 127 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 1 MgCl 2 , 2 CaCl 2 , and 25 glucose, bubbled with 95% O 2 and 5% CO 2 . After an incubation period Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 11 of 30-40 min at 34 • C, slices were stored at room temperature. All experiments were conducted at room temperature (25 • C). In all experiments, 50 µM picrotoxin, 10 µM 2,3-Dioxo-6-nitro-1, 2, 3, 4 -tetrahydrobenzo [f]quinoxaline -7-sulfonamide (NBQX), and 10 µM 3-((R)-2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) were present in the ACSF to block GABAA/C, AMPA, and NMDA receptor-mediated transmission, respectively. All chemicals were from Sigma or Tocris.
Whole-cell recordings were obtained from infected and uninfected deep layer cortical pyramidal neurons identified with video-IR/DIC and GFP fluorescence was detected using epifluorescence illumination. With the deep layers of the cortex, 2-photon laser scanning microscopy (2PLSM) was used to confirm the cell types based on morphology. Deep layer pyramidal neurons had large cell bodies, classic pyramidal shape and dendritic spines. Glass electrodes (2-4 M ) were filled with internal solution containing (in mM): 135 KMeSO 4 , 5 KCl, 5 HEPES, 4 MgATP, 0.3 NaGTP, 10 Na 2 HPO 4 , 1 EGTA, and 0.01 Alexa Fluor-594 (to image neuronal morphology) adjusted to pH 7.4 with KOH. Current and voltage recordings were made at room temperature using a AxoPatch 200B or a Multiclamp 700B amplifier. Data was filtered at 5 kHz and digitized at 10 kHz.
Imaging and physiology data were acquired and analyzed as described previously (Carter and Sabatini, 2004) . Resting membrane potential was determined by the average of three 5-s sweeps with no injected current. Passive properties of the cell, membrane (Rm) and series resistance (Rs) and capacitance (Cm), were measured while clamping cells at −65 mV and applying voltage steps from −55 to −75 mV. The current-firing relationship was determined in current clamp with 1-s periods of injected current from 100 to 500 pA.
The time course of viral gene expression experiments were carried out in organotypic hippocampal slice cultures prepared from postnatal day 5-7 Sprague-Dawley rats as described previously (Stoppini et al., 1991) . Slices were infected after 7 days in vitro, and images were acquired on a two-photon microscope.
|
In what direction does the Vesicular stomatitis virus spread through the nervous system?
| 1,936 |
anterograde
| 3,221 |
1,621 |
Vesicular stomatitis virus with the rabies virus glycoprotein directs retrograde transsynaptic transport among neurons in vivo
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3566411/
SHA: ee48061797d29eeef5a9e606841bf8ab04b1d75b
Authors: Beier, Kevin T.; Saunders, Arpiar B.; Oldenburg, Ian A.; Sabatini, Bernardo L.; Cepko, Constance L.
Date: 2013-02-07
DOI: 10.3389/fncir.2013.00011
License: cc-by
Abstract: Defining the connections among neurons is critical to our understanding of the structure and function of the nervous system. Recombinant viruses engineered to transmit across synapses provide a powerful approach for the dissection of neuronal circuitry in vivo. We recently demonstrated that recombinant vesicular stomatitis virus (VSV) can be endowed with anterograde or retrograde transsynaptic tracing ability by providing the virus with different glycoproteins. Here we extend the characterization of the transmission and gene expression of recombinant VSV (rVSV) with the rabies virus glycoprotein (RABV-G), and provide examples of its activity relative to the anterograde transsynaptic tracer form of rVSV. rVSV with RABV-G was found to drive strong expression of transgenes and to spread rapidly from neuron to neuron in only a retrograde manner. Depending upon how the RABV-G was delivered, VSV served as a polysynaptic or monosynaptic tracer, or was able to define projections through axonal uptake and retrograde transport. In animals co-infected with rVSV in its anterograde form, rVSV with RABV-G could be used to begin to characterize the similarities and differences in connections to different areas. rVSV with RABV-G provides a flexible, rapid, and versatile tracing tool that complements the previously described VSV-based anterograde transsynaptic tracer.
Text: Mapping neuronal connectivity in the central nervous system (CNS) of even simple organisms is a difficult task. Recombinant viruses engineered to trace synaptic connections and express transgenes promise to enable higher-throughput mapping of connections among neurons than other methods, e.g., serial reconstruction from electron micrographs (Bock et al., 2011; Briggman et al., 2011) . The Pseudorabies (PRV) and Rabies viruses (RABV) have been the best characterized and most utilized circuit tracing viruses to date (Ugolini et al., 1989; Kelly and Strick, 2000) . RABV was recently modified by Wickersham and colleagues such that it can travel across only one synapse, allowing for a straightforward definition of monosynaptic connections (Wickersham et al., 2007b) . This strategy permitted the first unambiguous identification of retrogradely connected cells from an initially infected cell ("starter cell"), without the need for electrophysiology. Moreover, the starter cell could be defined through the expression of a specific viral receptor that limited the initial infection.
Recently, we created an anterograde monosynaptic virus that complements the previously available retrograde viral tracers (Beier et al., 2011) . Vesicular stomatitis virus (VSV), a virus related to RABV, with its own glycoprotein (G) gene (VSV-G), or with a G from the unrelated lymphocytic choriomeningitis virus (LCMV-G), spreads in the anterograde direction across synapses. VSV can be used as a polysynaptic tracer that spreads across many synapses, owing to the fact that the normal, replicationcompetent form of the virus does not cause serious diseases in humans (Brandly and Hanson, 1957; Johnson et al., 1966; Brody et al., 1967) . Whether the virus is a monosynaptic or polysynaptic tracer is determined by the method of delivery of the G gene ( Figure 1A) . Advantages of VSV are that it is well-characterized, is relatively simple in comparison to PRV, and it rapidly grows to high titer in tissue culture cells. It is also being developed as a vaccine vector, often using a G of another virus as the immunogen, as well as being developed as a cytocidal agent that will target tumor cells in humans (Balachandran and Barber, 2000; Stojdl et al., 2000 Stojdl et al., , 2003 .
Previous studies of the anatomical patterns of transmission, as well as physiological recordings, have shown that the transmission of VSV and RABV among neurons is via synapses (Kelly and Strick, 2000; Wickersham et al., 2007b; Beier et al., 2011) . In addition, it has been shown that RABV, as well as lentiviruses with RABV-G in their envelope, travel retrogradely from an injection site (Mazarakis et al., 2001; Wickersham et al., 2007a) . We hypothesized that providing a recombinant VSV (rVSV) with the RABV-G would create a retrograde polysynaptic transsynaptic tracer without the biosafety concerns inherent to RABV. Our initial characterization of rVSV with RABV-G showed that indeed FIGURE 1 | Synaptic tracing strategies using VSV. (A) Schematic illustrating the strategies for polysynaptic or monosynaptic retrograde or anterograde transsynaptic transmission of rVSV encoding GFP. The initially infected cell is indicated by an asterisk. VSV encoding a glycoprotein (G) within its genome can spread polysynaptically. The direction of the spread depends on the identity of the glycoprotein. Infected neurons are shown in green. In some cases, the initially infected starter cell can be defined by the expression of an avian receptor, TVA (tagged with a red fluorescent protein). The TVA-expressing neurons can then be specifically infected by rVSV G with the EnvA/RABV-G (A/RG) glycoprotein (Wickersham et al., 2007b) on the virion surface [rVSV G(A/RG)]. These starter cells are then yellow, due to viral GFP and mCherry from TVA-mCherry expression. For monosynaptic tracing, the G protein is expressed in trans in the TVA-expressing cell, and thus complements rVSV G to allow transmission in a specific direction. (B) Genomic diagrams of rVSV vectors. All VSVs contain four essential proteins: N, P, M, and L. Some viruses encode a G gene in their genome, which allows them to spread polysynaptically. rVSV vectors typically encode a transgene in the first position, while others carry an additional transgene in the G position. (C) Morphological characterization of rVSV-infected neurons in several locations within the mouse brain. (i,ii) Caudate-putamen (CP) neurons at 4 dpi from an injection of the CP with rVSV(VSV-G) viruses encoding (i) CFP or (ii) Korange. (iii) Labeled neurons of the CA1 region of the hippocampus are shown at 5 dpi following injection into the hippocampus of rVSV(VSV-G) encoding Venus. (iv,v) Cortical pyramidal neurons are shown following injection into the CP of rVSV(RABV-G) expressing (iv) GFP at 24 hpi, or (v) mCherry at 48 hpi. Inset in (iv) is a high magnification of the neuron in panel (iv), highlighting labeling of dendritic spines. (vi) Multiple viruses can be co-injected into the same animal. Here, individual rVSV G(VSV-G) viruses encoding CFP, GFP , Venus, Korange, and mCherry were used to infect the cortex. Scale bars = 50 µm.
www.frontiersin.org February 2013 | Volume 7 | Article 11 | 2 it could be taken up as a retrograde tracer (Beier et al., 2011) .
To determine if it could transmit among neurons following its replication in neurons, and to further analyze the transmission patterns of both the monosynaptic and polysynaptic forms of rVSV with RABV-G, we made injections into several CNS and peripheral locations. In addition, we performed co-infections of rVSV with RABV-G and the anterograde form of rVSV in order to exploit the differences in the directionality of transmission of these two viruses in mapping circuits.
Schematics of viruses created and used throughout this study are shown in Figure 1 . We created rVSV vector plasmids carrying different transgenes in either the first or fifth genomic positions ( Figure 1B) . After rescuing each virus, we tested the ability of each to express transgenes in different brain regions through intracranial injections ( Figure 1C ). All rVSV vectors drove robust fluorophore expression 1 or 2 days post-infection (hpi) ( Figure 1C ) (van den Pol et al., 2009) . In fact, by 12 hpi, labeling was sufficiently bright to image fine morphological details, such as dendritic spines ( Figure 1C ,iv).
To characterize the physiological properties of cells infected with rVSV, we tested a replication-competent rVSV encoding GFP, with RABV-G in the genome in place of VSV-G [hereafter designated rVSV(RABV-G)]. van den Pol et al. reported that hippocampal neurons infected with replication-incompetent (G-deleted or " G") rVSV were physiologically healthy at 12-14 hpi, but were less so by 1 day post-infection (dpi) (van den Pol et al., 2009) . Given the known toxicity of both VSV and RABV-G (Coulon et al., 1982) , we tested the physiology of cortical pyramidal neurons in the motor cortex (M1) infected with rVSV(RABV-G). Between 12 and 18 hpi, the membrane capacitance, input resistance, resting membrane potential, and current-to-action potential firing relationship were indistinguishable between infected and uninfected neurons (Figure 2) . However, by 2 dpi, electrophysiological properties were so abnormal in the infected cortical pyramidal cells that physiological measurements could not be made.
The speed and strength of the expression of transgenes encoded by VSV depends upon the gene's genomic position (van den Pol et al., 2009; Beier et al., 2011) . Genes in the first position are expressed the most highly, with a decrease in the level of expression in positions more 3 within the viral plus strand. When GFP was inserted into the first position of VSV, GFP fluorescence was first detectable at approximately 1 hpi in cultured cells (van den Pol et al., 2009) . In order to quantify the relative expression of a fluorescent protein in the first genomic position in neurons, rat hippocampal slices were infected with a replication-incompetent rVSV that expresses mCherry (rVSV G, Figures 1A,B) . This was a G virus which had the RABV-G supplied in trans during the preparation of the virus stock [referred to as rVSV G(RABV-G)].
Average fluorescence intensity of the infected cells was measured every hour over the course of 18 h. By 4 hpi at 37 • C, red fluorescence was clearly visible, and reached maximal levels by approximately 14 hpi (N = 3, Figure 3 ). Similar results were obtained with a virus encoding GFP in the first genomic position rather than mCherry (i.e., Figure 1B ) (N = 3).
We previously demonstrated that rVSV(RABV-G) could be taken up retrogradely by neurons (Beier et al., 2011) , but these experiments did not distinguish between direct axonal uptake of the initial inoculum vs. retrograde transsynaptic transmission following viral replication. To distinguish between these two mechanisms and to extend the previous analyses, we conducted further experiments in the mammalian visual system (Figures 4A-G) . As visual cortex area 1 (V1) does not receive direct projections from retinal ganglion cells (RGCs), but rather receives secondary input from RGCs via the lateral geniculate nucleus (LGN), infection of RGCs from injection of V1 would demonstrate retrograde transmission from cells which supported at least one round of viral replication. Following a V1 injection with rVSV(RABV-G), GFP-positive RGCs were observed in the retina by 3 dpi (N = 3; Figure 4G ). Importantly, viral labeling in the brain was restricted to primary and secondary projection areas, even at 7 dpi. These included the LGN ( Figure 4D ) and the hypothalamus (Figure 4E) , two areas known to project directly to V1 (Kandel, 2000) . Selective labeling was observed in other areas, such as cortical areas surrounding V1 (Figure 4C) , which project directly to V1, and also in the superior colliculus (SC) stratum griseum centrale, which projects to the LGN ( Figure 4F) . Labeling was also observed in the nucleus basalis, which projects to the cortex, as well as many components of the basal ganglia circuit, which provide input to the thalamus [such as the caudate-putamen (CP), globus pallidus (GP), and the subthalamic nucleus (STn)]. The amygdala, which projects to the hypothalamus, was also labeled. Consistent with a lack of widespread viral transmission, animals did not exhibit signs of disease at 7 dpi.
These data show that rVSV(RABV-G) can spread in a retrograde direction from the injection site, but do not address whether the virus can spread exclusively in the retrograde direction. Directional transsynaptic specificity can only be definitively addressed using a unidirectional circuit. We therefore turned to the primary motor cortex (M1) to CP connection, in which neurons project from the cortex to the CP, but not in the other direction ( Figure 4H ) (Beier et al., 2011) . Injections of rVSV(RABV-G) into M1 should not label neurons in the CP if the virus can only label cells across synapses in the retrograde direction. Indeed, at 2 dpi, areas directly projecting to the injection site, including the contralateral cortex, were labeled ( Figure 4I ). Only axons from cortical cells were observed in the CP, with no GFP-labeled cell bodies present in the CP (Figure 4J) , consistent with lack of anterograde transsynaptic spread. By 3 dpi, a small number of medium spiny neurons (MSNs) in the CP were observed, likely via secondary spread from initially infected thalamic or GP neurons (data not shown).
A particular advantage of retrograde viral tracers is the ability to label CNS neurons projecting to peripheral sites. This has been a powerful application of both RABV and PRV (Ugolini et al., 1989; Standish et al., 1994) . To test if rVSV(RABV-G) could also perform this function, we examined the innervation of the dura surface by neurons of the trigeminal ganglion, a neuronal circuit thought to be involved in migraine headaches (Penfield and McNaughton, 1940; Mayberg et al., 1984) . These neurons have axons, but not canonical dendrites, and send projections into the spinal cord and brainstem. Therefore, the only way trigeminal neurons could become labeled from viral application to the dura is through retrograde uptake of the virus. We applied rVSV(RABV-G) to the intact dura mater and analyzed the dura, trigeminal ganglion, and CNS for labeling ( Figure 4K ). At the earliest time point examined, 3 dpi, we observed axons traveling along the dura, but little other evidence of infection ( Figure 4L) . No labeled neuronal cell bodies on the dura were observed, consistent with the lack of neurons on this surface. In contrast, we did find labeled cell bodies in the trigeminal ganglion ( Figure 4M ). No infection was seen in the CNS, even at 4 dpi, consistent with the lack of inputs from the brain into the trigeminal ganglion (N = 4 animals).
To further characterize patterns and kinetics of viral transmission and directional specificity of transsynaptic spread, injections of rVSV(RABV-G) were made into the CP ( Figure 5A ). In order to determine which cells were labeled by direct uptake of virus in the inoculum, a separate set of animals were injected into the CP with the replication-incompetent rVSV G(RABV-G) (N = 3 animals, analyzed 3 dpi). Cells labeled by rVSV G(RABV-G) were observed in the CP, GP, substantia nigra (SN), thalamus, and layers 3 and 5 of the cortex, consistent with infection at the axon terminal and retrograde labeling of cell bodies of neurons known to project directly to the CP ( Figure 5C ) (Albin et al., 1995) . Areas labeled by CP injection are indicated in Figure 5B .
The patterns of spread for the replication-competent rVSV(RABV-G) were characterized over the course of 1-5 dpi ( Figures 5D-H) . During this interval, progressively more cells in infected regions were labeled by rVSV(RABV-G), including within the CP, nucleus basalis, cortex, and GP (listed in Figure 5B ). In addition, more cortical cells were labeled in clusters near cortical pyramidal neurons, both ipsilateral and contralateral to the injected side, including neurogliaform cells (data not shown). These data are in contrast to those observed following infection with an anterograde transsynaptic tracing virus, such as rVSV with its own G gene, rVSV(VSV-G) (Figure 5B) . At 3 dpi following rVSV(VSV-G) injection into the CP, the cerebral cortex was not labeled, but regions receiving projections from the CP, such as the STn, GP, and SN, were labeled (Beier et al., 2011) .
In order to investigate other areas for evidence of cell-to-cell retrograde transsynaptic spread, the nucleus basalis was examined following infection of the CP with replication-competent rVSV(RABV-G). The nucleus basalis was labeled by 2 dpi (Figures 5E-H) , consistent with at least a single transsynaptic jump, as this area does not directly project to the CP. The virus appeared to travel transsynaptically at the rate of roughly 1 synapse per day, as evidenced by the lack of labeled neurogliaform cells in the cortex, and lack of neurons in the nucleus basalis at 1 dpi, and label appearing in these cell types/areas at 2 dpi, as previously observed (Beier et al., 2011) . Labeling remained well-restricted to the expected corticostriatal circuits at 5 dpi, suggesting that viral spread becomes less efficient after crossing one or two connections, consistent with injections into V1 (Figure 4) . While glial cells can be infected and were observed near the injection site (van den Pol et al., 2002; Chauhan et al., 2010) , infected glial cells away from the injection site generally were not observed.
One advantage of having both anterograde and retrograde forms of the same virus is that they can be used in parallel, or in tandem, to trace circuitry to and from a single or multiple sites of injection, with each virus having similar kinetics of spread and gene expression. In fact, if different fluorophores are used in different viruses, e.g., rVSV(VSV-G) and rVSV(RABV-G), then the viruses can be co-injected into the same site and their transmission can be traced independently ( Figure 6A ). This is most straightforward if there are no cells at the injection site that are initially infected by both viruses. Co-infected cells can be easily detected, as they would express both fluorescent proteins shortly after injection.
In order to determine whether two viruses would allow simultaneous anterograde and retrograde transsynaptic tracing from a single injection site, a rVSV(VSV-G) expressing Venus and a rVSV(RABV-G) expressing mCherry were injected individually (Figures 6B-D) or co-injected (Figures 6E-G) into the motor cortex, and brains were examined 3 dpi. The pattern of labeling from the co-injected brains was equivalent to the patterns observed when each virus was injected individually: rVSV(VSV-G) was observed to infect neurons in the cortex, CP, and downstream nuclei, whereas the rVSV(RABV-G) was not observed to infect neurons in the CP, but rather in the thalamus and nucleus basalis (N = 4). The initial co-infection rate is dependent upon The presence or absence of labeling is indicated by (+) and (−), respectively. The extent of labeling is indicated by the number of (+). Some animals were infected with G viruses to determine which areas were labeled by direct uptake of the virions, rather than by replication and transmission. These were sacrificed at 3 dpi. (C) Parasaggital section of a brain infected with VSV[greek delta]G(RABV-G). The injection site is marked by a red arrow. Several areas that project directly to the CP were labeled due to direct uptake of the virions, including the cortex, thalamus, and GP (arrowheads), 3 dpi. the dose of the initial inocula. When injecting 3 × 10 3 focus forming units (ffu) rVSV(VSV-G) and 3 × 10 4 ffu rVSV(RABV-G), no co-infection was observed at the injection site. Thus, co-infection of the same brain region, without co-infection of the same cells,
does not alter the spreading behavior of either rVSV(VSV-G) or rVSV(RABV-G). One example of how this dual retrograde and anterograde transsynaptic tracing system can be used is to determine if three Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 8 distinct regions are connected and the directionality of any connections. For example, the anterograde transsynaptic virus can be injected into one region, the retrograde into another, and a third region can then be examined for evidence of labeling by either or both viruses (e.g., Figure 6H ). To test this possibility, rVSV(VSV-G) was injected into the motor cortex, rVSV(RABV-G) was injected into the substantia nigra pars reticulara (SNr), and animals were sacrificed at 3 dpi. We observed that cells were singly labeled, either with Venus [rVSV(VSV-G)] or with mCherry [rVSV(RABV-G)], and were located largely in different regions of the CP (Figures 6I,J) (N = 3) . These results suggest that the anterograde connections from the cells infected with rVSV(VSV-G) in the M1 were with CP MSNs that did not project to the region of the SNr injected with rVSV(RABV-G) (N = 3 animals).
In addition to polysynaptic tracing, VSV can be modified to trace circuits monosynaptically (Beier et al., 2011) . With RABV, this was achieved in vivo by first infecting with an adeno-associated virus (AAV) expressing TVA, a receptor for an avian retrovirus, and RABV-G (Wall et al., 2010) . This was followed 3 weeks later by infection with a G RABV with an EnvA/RABV-G chimeric glycoprotein on the virion surface (Wickersham et al., 2007b) , which allowed infection specifically of the cells expressing TVA. A similar strategy was used to test rVSV's ability to monosynaptically trace retrogradely connected neurons in vivo. Inputs to choline acetyltransferase (ChAT)-expressing neurons in the striatum were used for this test. These neurons primarily receive input from the cortex and the thalamus (Thomas et al., 2000; Bloomfield et al., 2007) (Figure 7A) . In order to mark this population, we crossed ChAT-Cre mice to Ai9 mice, which express tdTomato in cells with a Cre expression history (Madisen et al., 2010) . Six-week-old mice from this cross were injected in the CP with two AAV vectors: one expressing a Cre-conditional ("floxed") TVA-mCherry fusion protein, and another expressing a floxed RABV-G. Two weeks later, the mice were injected in the same coordinates with rVSV G with the EnvA/RABV-G chimeric glycoprotein on the virion surface [rVSV G(A/RG)] (Beier et al., 2011) . Cells successfully infected with these two AAV vectors could host infection by a rVSV and should be able to produce rVSV virions with RABV-G on the surface. Such starter cells should also express tdTomato and GFP. If rVSV were to be produced, and if it were to transmit across the synapse retrogradely, cortical and thalamic neurons should be labeled by GFP. Mice injected with these AAV and rVSV viruses were sacrificed 5 days after rVSV infection, and brains analyzed for fluorescence. As expected for starter cells, some neurons in the CP expressed both tdTomato and GFP (Figure 7B) . Outside of the CP, small numbers of GFP+ neurons that were not mCherry+ were observed in the cortex (Figures 7C,D) and thalamus (Figure 7E) , consistent with retrograde spread. Control animals not expressing Cre, or not injected with AAV encoding RABV-G, did not label cells in the cortex or thalamus (N = 3 for both controls and experimental condition).
Here, we report on the use of rVSV as a retrograde transsynaptic tracer for CNS circuitry. VSV can be modified to encode the RABV-G protein in the viral genome, allowing the virus to replicate and transmit across multiple synaptically connected cells, i.e., as a polysynaptic tracer. Alternatively, if the virus has the G gene deleted from its genome and RABV-G is provided in trans, it behaves as a monosynaptic tracer (Beier et al., 2011) . Although it has been known for many years that RABV travels retrogradely among neurons (Astic et al., 1993; Ugolini, 1995; Kelly and Strick, 2003) , and pseudotyping lentiviruses with RABV-G is sufficient for axonal transport (Mazarakis et al., 2001) , the retrograde transmission specificity among neurons had not been clearly shown to be a property of the G protein itself, as it might have been due to other viral proteins in addition to, or instead of, the viral G protein. Since native VSV does not have these retrograde transsynaptic properties (van den Pol et al., 2002; Beier et al., 2011) , and the only alteration to the VSV genome was the substitution of the VSV G gene with the G gene of RABV, it is clear that the RABV glycoprotein is responsible for retrograde direction of viral transmission across synapses, at least in the case of rVSV.
The early onset of gene expression from VSV relative to RABV (one hour vs. multiple hours) makes it beneficial in experimental paradigms in which the experiment needs to be done within a narrow window of time, such as tissue slices and explants. In addition, more than one transgene can be encoded in the viral genome without the need of a 2A or IRES element. The use of the first position of the genome enhances the expression level of the transgene inserted at that location, since VSV (and RABV) express genes in a transcriptional gradient; therefore, the first gene is the most highly transcribed (Knipe, 2007) . This leads to rational predictions of expression levels so that one can choose the position of insertion of a transgene, or transgenes, according to this gradient and the desired level of expression. The size of the viral capsid is apparently not rigid, allowing for the inclusion of genomes that are substantially larger than the native genome, unlike the rigid capacity for some other viral vectors, such as AAV Yan et al., 2000) .
The fact that VSV can be made to spread anterogradely (Beier et al., 2011) or retrogradely across synapses with the change of a single gene affords several advantages over viral tracers that heretofore have not shown such flexibility in the directionality of tracing. In addition to the obvious application of tracing anterograde connections, combinations can be made to exploit the different forms of the virus. One example that employs the simultaneous infection with an anterograde and retrograde form of VSV is demonstrated in Figure 6 . This experiment was designed to address whether the anterograde projections from the cortex to the CP would label the same brain regions as were labeled by a retrograde virus injected into the SN. Although a block of superinfection by the virus may preclude infection of the same cell with multiple rVSVs, adjacent cells could still become labeled by different viruses (Whitaker-Dowling et al., 1983) . The observed results could be due to a preferential labeling by the anterograde transsynaptic virus of indirect pathway MSNs in this experiment, which then synapse onto the GP, thereby reflecting a viral bias. Alternatively, it could indicate that the cortical neurons in the injected region largely do not label the MSNs that project to the area of the SN injected with the retrograde virus. One further possibility is that too little virus was used to observe co-labeling of a given region. However, given the density of infection (i.e., Figures 6I,J) , the latter possibility seems unlikely. Additionally, the spread of the polysynaptic rVSV(RABV-G) appears to attenuate with increasing numbers of synapses crossed, permitting an analysis of more restricted viral spread. This is quite fortuitous, as if spread were to continue, it would lead to widespread infection and lethality. In addition, reconstruction of connectivity would be more difficult. This reduced efficiency appears to also hold for the monosynaptic form of VSV complemented with RABV-G, as the efficiency of transmission appeared lower than the comparable experiment with RABV (Watabe-Uchida et al., 2012) . This is likely due to viral attenuation when VSV-G is replaced with RABV-G.
We were attracted to the use of VSV as a viral tracer due to its long track record as a safe, replication-competent laboratory agent. Laboratory workers using VSV have not contracted any diseases, and natural VSV infections among human populations in Central America and the southwestern United States (Rodríguez, 2002) occur without evident pathology (Johnson et al., 1966; Brody et al., 1967) . VSV was thus an attractive candidate for its use as a polysynaptic tracer for CNS studies, which requires an ability to replicate through multiple transmission cycles. Both replicationcompetent and incompetent forms of VSV are in use under Biosafety Level 2 containment. Replication-competent RABV is Biosafety Level 3, due to the fact that infection with replicationcompetent RABV is almost always fatal to humans and in mice when infected intracerebrally (Smith, 1981; Knipe, 2007) . Differences in pathogenicity between VSV and RABV are likely due to the ability of RABV to evade the innate immune system, particularly interferon (Hangartner et al., 2006; Junt et al., 2007; Lyles and Rupprecht, 2007; Rieder and Conzelmann, 2009; Iannacone et al., 2010) . VSV infection efficiently triggers an interferon response, and it has not evolved a method of escape from this response, unlike RABV (Brzózka et al., 2006) . In fact, VSV is being pursued as a vaccine for other viruses, including RABV (Lichty et al., 2004; Publicover et al., 2004; Kapadia et al., 2005; Schwartz et al., 2007; Iyer et al., 2009; Geisbert and Feldmann, 2011) . VSV does not typically spread beyond the initially infected site in the periphery (Kramer et al., 1983; Vogel and Fertsch, 1987) . This likely is the cause of the minor or absent symptoms in humans and animals infected in nature. Polysynaptic VSV vectors are thus predicted to be much safer than polysynaptic RABV vectors. We have tested this prediction by injecting a series of mice in the footpads and hind leg muscles with rVSV(RABV-G), with the result that no injected animals showed any evidence of morbidity or mortality (Beier, Goz et al., in preparation) .
While safer for laboratory workers than RABV, the main drawback to using VSV is its rapid cellular toxicity (van den Pol et al., 2009; Beier et al., 2011) . Toxicity is due to suppression of cellular transcription and a block in the export of cellular RNAs from the nucleus to the cytoplasm (Black and Lyles, 1992; Her et al., 1997; Ahmed and Lyles, 1998; Petersen et al., 2000; von Kobbe et al., 2000) , as well as inhibition of the translation of cellular mRNAs (Francoeur et al., 1987; Jayakar et al., 2000; Kopecky et al., Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 10 2001). VSV is much quicker to enact its gene expression program than is RABV, such that cells suffer the toxic effects more quickly than after RABV infection. One aspect of VSV that can be exploited in the future to ameliorate the speed of toxicity is the use of VSV mutants and variants. One such mutant is the M51R, which permitted us to conduct physiological analyses of pre-and post-synaptic cells (Beier et al., 2011) . We are in the process of examining the transmission properties of this mutant in vivo, as well as the effects of other mutations or viral variants on prolonging the health of neurons after infection.
rVSV vectors can be used to study the connectivity of neuronal circuitry. In addition to combinations of replication-competent forms of VSV, the replication-incompetent, monosynaptic forms of the virus can be easily combined, without the need to change viruses (Beier et al., 2011) . This allows a straightforward way to study both the projections into, and out from, a genetically defined cell population. This can be done with the same viral genome, with the only change needed being the glycoprotein, for the selection of the direction of transmission. This flexibility of VSV makes it a powerful, multi-application vector for studying connectivity in the CNS.
All rVSV clones were cloned from the rVSV G backbone (Chandran et al., 2005) . mCherry, Kusabira orange, Venus, and CFP were cloned into the first (GFP) position using XhoI and MscI sites, and VSV-G (a gift from Richard Mulligan, Harvard Medical School, Boston, MA) and RABV-G (a gift from Ed Callaway, Salk Institute, San Diego, CA) were cloned into the fifth (G) position using the MluI and NotI restriction sites. Genes for fluorescent proteins were obtained from Clontech. Viruses were rescued as previously described (Whelan et al., 1995) . At 95% confluency, eight 10 cm plates of BSR cells were infected at an MOI of 0.01. Viral supernatants were collected at 24-h time intervals and ultracentrifuged at 21,000 RPM using a SW28 rotor and resuspended in 0.2% of the original volume. For titering, concentrated viral stocks were applied in a dilution series to 100% confluent BSR cells and plates were examined at 12 hpi. Viral stocks were stored at −80 • C.
For G viruses, 293T cells were transfected with PEI (Ehrhardt et al., 2006) at 70% confluency on 10 cm dishes with 5 µg of pCAG-RABV-G. Twenty-four hours post-infection, the cells were infected at an MOI of 0.01 with rVSV G expressing either GFP or mCherry. Viral supernatants were collected for the subsequent 4 days at 24 h intervals.
Virus preparations are now available from the Salk GT3 viral core (http://vectorcore.salk.edu/). All plasmids are available from Addgene (http://www.addgene.org/).
AAV-FLEx-RABV-G and AAV-FLEx-TVA-mCherry plasmids originated from the Lab of Naoshige Uchida (Watabe-Uchida et al., 2012) , and virus stocks were generous gifts from Brad Lowell, Harvard Medical School.
ChAT-Cre (B6;129S6-Chat tm1(cre)Lowl /J) and Ai9 (B6.Cg-Gt(ROSA)26Sor<tm9(CAG-tdTomato)Hze>/J) mice were obtained from the Jackson Laboratory (Madisen et al., 2010) .
Eight-week-old CD-1 mice were injected using pulled capillary microdispensers (Drummond Scientific, Cat. No: 5-000-2005) , using coordinates from The Mouse Brain in Stereotaxic Coordinates (Franklin and Paxinos, 1997) . Injection coordinates (in mm) used were: For multi-color analysis (Figures 1C,D) , 3 × 10 9 ffu/mL rVSV was injected into various regions. For CP injections, 100 nL of rVSV(RABV-G) or rVSV(VSV-G) at 3 × 10 7 ffu/mL was injected at a rate of 100 nL/min. For the replication-incompetent viruses, 100 nL of 1 × 10 7 ffu/mL rVSV G(RABV-G) or rVSV G (VSV-G) was injected. In the motor cortex, 100 nL of 1 × 10 7 ffu/mL rVSV(RABV-G) was injected, and mice harvested 2 dpi. For V1 injections, 100 nL of 3 × 10 10 ffu/mL rVSV(RABV-G) was injected, and mice were examined 3 or 7 dpi.
For infections of the dura mater, 1 µL of 3 × 10 10 ffu/mL rVSV(RABV-G) was applied to the surface of the dura. The virus was allowed to absorb, and the surface was subsequently covered in bone wax, and the wound sutured.
For co-injections of virus into the same animal, 100 nL of a combination of 3 × 10 7 ffu/mL rVSV(VSV-G) and 3 × 10 8 ffu/mL rVSV(RABV-G) were co-injected into the motor cortex, and brains examined 3 dpi. For injections of the viruses into different regions, 100 nL of 3 × 10 7 ffu/mL rVSV(VSV-G) was injected into M1, and 100 nL of 3 × 10 8 ffu/mL rVSV(RABV-G) into the SNr, and brains examined 3 dpi. A lower titer of rVSV(VSV-G) was used, as rVSV(RABV-G) is attenuated.
All mouse work was conducted in biosafety containment level 2 conditions and was approved by the Longwood Medical Area Institutional Animal Care and Use Committee.
Recordings were made from cortical pyramidal neurons in slices taken from postnatal day 12-18 mice, inoculated in the CP 12-18 h prior with rVSV(RABV-G). Coronal slices (300 µm thick) were cut in ice-cold external solution containing (in mM): 110 choline, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 7 MgCl 2 , 0.5 CaCl 2 , 25 glucose, 11.6 Na-ascorbate, and 3.1 Na-pyruvate, bubbled with 95% O 2 and 5% CO 2 . Slices were then transferred to artificial cerebrospinal fluid (ACSF) containing (in mM): 127 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 1 MgCl 2 , 2 CaCl 2 , and 25 glucose, bubbled with 95% O 2 and 5% CO 2 . After an incubation period Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 11 of 30-40 min at 34 • C, slices were stored at room temperature. All experiments were conducted at room temperature (25 • C). In all experiments, 50 µM picrotoxin, 10 µM 2,3-Dioxo-6-nitro-1, 2, 3, 4 -tetrahydrobenzo [f]quinoxaline -7-sulfonamide (NBQX), and 10 µM 3-((R)-2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) were present in the ACSF to block GABAA/C, AMPA, and NMDA receptor-mediated transmission, respectively. All chemicals were from Sigma or Tocris.
Whole-cell recordings were obtained from infected and uninfected deep layer cortical pyramidal neurons identified with video-IR/DIC and GFP fluorescence was detected using epifluorescence illumination. With the deep layers of the cortex, 2-photon laser scanning microscopy (2PLSM) was used to confirm the cell types based on morphology. Deep layer pyramidal neurons had large cell bodies, classic pyramidal shape and dendritic spines. Glass electrodes (2-4 M ) were filled with internal solution containing (in mM): 135 KMeSO 4 , 5 KCl, 5 HEPES, 4 MgATP, 0.3 NaGTP, 10 Na 2 HPO 4 , 1 EGTA, and 0.01 Alexa Fluor-594 (to image neuronal morphology) adjusted to pH 7.4 with KOH. Current and voltage recordings were made at room temperature using a AxoPatch 200B or a Multiclamp 700B amplifier. Data was filtered at 5 kHz and digitized at 10 kHz.
Imaging and physiology data were acquired and analyzed as described previously (Carter and Sabatini, 2004) . Resting membrane potential was determined by the average of three 5-s sweeps with no injected current. Passive properties of the cell, membrane (Rm) and series resistance (Rs) and capacitance (Cm), were measured while clamping cells at −65 mV and applying voltage steps from −55 to −75 mV. The current-firing relationship was determined in current clamp with 1-s periods of injected current from 100 to 500 pA.
The time course of viral gene expression experiments were carried out in organotypic hippocampal slice cultures prepared from postnatal day 5-7 Sprague-Dawley rats as described previously (Stoppini et al., 1991) . Slices were infected after 7 days in vitro, and images were acquired on a two-photon microscope.
|
What determines the whether the spread of Vesicular stomatitis virus is monosynaptic or polysynaptic?
| 1,937 |
method of delivery of the G gene
| 3,603 |
1,621 |
Vesicular stomatitis virus with the rabies virus glycoprotein directs retrograde transsynaptic transport among neurons in vivo
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3566411/
SHA: ee48061797d29eeef5a9e606841bf8ab04b1d75b
Authors: Beier, Kevin T.; Saunders, Arpiar B.; Oldenburg, Ian A.; Sabatini, Bernardo L.; Cepko, Constance L.
Date: 2013-02-07
DOI: 10.3389/fncir.2013.00011
License: cc-by
Abstract: Defining the connections among neurons is critical to our understanding of the structure and function of the nervous system. Recombinant viruses engineered to transmit across synapses provide a powerful approach for the dissection of neuronal circuitry in vivo. We recently demonstrated that recombinant vesicular stomatitis virus (VSV) can be endowed with anterograde or retrograde transsynaptic tracing ability by providing the virus with different glycoproteins. Here we extend the characterization of the transmission and gene expression of recombinant VSV (rVSV) with the rabies virus glycoprotein (RABV-G), and provide examples of its activity relative to the anterograde transsynaptic tracer form of rVSV. rVSV with RABV-G was found to drive strong expression of transgenes and to spread rapidly from neuron to neuron in only a retrograde manner. Depending upon how the RABV-G was delivered, VSV served as a polysynaptic or monosynaptic tracer, or was able to define projections through axonal uptake and retrograde transport. In animals co-infected with rVSV in its anterograde form, rVSV with RABV-G could be used to begin to characterize the similarities and differences in connections to different areas. rVSV with RABV-G provides a flexible, rapid, and versatile tracing tool that complements the previously described VSV-based anterograde transsynaptic tracer.
Text: Mapping neuronal connectivity in the central nervous system (CNS) of even simple organisms is a difficult task. Recombinant viruses engineered to trace synaptic connections and express transgenes promise to enable higher-throughput mapping of connections among neurons than other methods, e.g., serial reconstruction from electron micrographs (Bock et al., 2011; Briggman et al., 2011) . The Pseudorabies (PRV) and Rabies viruses (RABV) have been the best characterized and most utilized circuit tracing viruses to date (Ugolini et al., 1989; Kelly and Strick, 2000) . RABV was recently modified by Wickersham and colleagues such that it can travel across only one synapse, allowing for a straightforward definition of monosynaptic connections (Wickersham et al., 2007b) . This strategy permitted the first unambiguous identification of retrogradely connected cells from an initially infected cell ("starter cell"), without the need for electrophysiology. Moreover, the starter cell could be defined through the expression of a specific viral receptor that limited the initial infection.
Recently, we created an anterograde monosynaptic virus that complements the previously available retrograde viral tracers (Beier et al., 2011) . Vesicular stomatitis virus (VSV), a virus related to RABV, with its own glycoprotein (G) gene (VSV-G), or with a G from the unrelated lymphocytic choriomeningitis virus (LCMV-G), spreads in the anterograde direction across synapses. VSV can be used as a polysynaptic tracer that spreads across many synapses, owing to the fact that the normal, replicationcompetent form of the virus does not cause serious diseases in humans (Brandly and Hanson, 1957; Johnson et al., 1966; Brody et al., 1967) . Whether the virus is a monosynaptic or polysynaptic tracer is determined by the method of delivery of the G gene ( Figure 1A) . Advantages of VSV are that it is well-characterized, is relatively simple in comparison to PRV, and it rapidly grows to high titer in tissue culture cells. It is also being developed as a vaccine vector, often using a G of another virus as the immunogen, as well as being developed as a cytocidal agent that will target tumor cells in humans (Balachandran and Barber, 2000; Stojdl et al., 2000 Stojdl et al., , 2003 .
Previous studies of the anatomical patterns of transmission, as well as physiological recordings, have shown that the transmission of VSV and RABV among neurons is via synapses (Kelly and Strick, 2000; Wickersham et al., 2007b; Beier et al., 2011) . In addition, it has been shown that RABV, as well as lentiviruses with RABV-G in their envelope, travel retrogradely from an injection site (Mazarakis et al., 2001; Wickersham et al., 2007a) . We hypothesized that providing a recombinant VSV (rVSV) with the RABV-G would create a retrograde polysynaptic transsynaptic tracer without the biosafety concerns inherent to RABV. Our initial characterization of rVSV with RABV-G showed that indeed FIGURE 1 | Synaptic tracing strategies using VSV. (A) Schematic illustrating the strategies for polysynaptic or monosynaptic retrograde or anterograde transsynaptic transmission of rVSV encoding GFP. The initially infected cell is indicated by an asterisk. VSV encoding a glycoprotein (G) within its genome can spread polysynaptically. The direction of the spread depends on the identity of the glycoprotein. Infected neurons are shown in green. In some cases, the initially infected starter cell can be defined by the expression of an avian receptor, TVA (tagged with a red fluorescent protein). The TVA-expressing neurons can then be specifically infected by rVSV G with the EnvA/RABV-G (A/RG) glycoprotein (Wickersham et al., 2007b) on the virion surface [rVSV G(A/RG)]. These starter cells are then yellow, due to viral GFP and mCherry from TVA-mCherry expression. For monosynaptic tracing, the G protein is expressed in trans in the TVA-expressing cell, and thus complements rVSV G to allow transmission in a specific direction. (B) Genomic diagrams of rVSV vectors. All VSVs contain four essential proteins: N, P, M, and L. Some viruses encode a G gene in their genome, which allows them to spread polysynaptically. rVSV vectors typically encode a transgene in the first position, while others carry an additional transgene in the G position. (C) Morphological characterization of rVSV-infected neurons in several locations within the mouse brain. (i,ii) Caudate-putamen (CP) neurons at 4 dpi from an injection of the CP with rVSV(VSV-G) viruses encoding (i) CFP or (ii) Korange. (iii) Labeled neurons of the CA1 region of the hippocampus are shown at 5 dpi following injection into the hippocampus of rVSV(VSV-G) encoding Venus. (iv,v) Cortical pyramidal neurons are shown following injection into the CP of rVSV(RABV-G) expressing (iv) GFP at 24 hpi, or (v) mCherry at 48 hpi. Inset in (iv) is a high magnification of the neuron in panel (iv), highlighting labeling of dendritic spines. (vi) Multiple viruses can be co-injected into the same animal. Here, individual rVSV G(VSV-G) viruses encoding CFP, GFP , Venus, Korange, and mCherry were used to infect the cortex. Scale bars = 50 µm.
www.frontiersin.org February 2013 | Volume 7 | Article 11 | 2 it could be taken up as a retrograde tracer (Beier et al., 2011) .
To determine if it could transmit among neurons following its replication in neurons, and to further analyze the transmission patterns of both the monosynaptic and polysynaptic forms of rVSV with RABV-G, we made injections into several CNS and peripheral locations. In addition, we performed co-infections of rVSV with RABV-G and the anterograde form of rVSV in order to exploit the differences in the directionality of transmission of these two viruses in mapping circuits.
Schematics of viruses created and used throughout this study are shown in Figure 1 . We created rVSV vector plasmids carrying different transgenes in either the first or fifth genomic positions ( Figure 1B) . After rescuing each virus, we tested the ability of each to express transgenes in different brain regions through intracranial injections ( Figure 1C ). All rVSV vectors drove robust fluorophore expression 1 or 2 days post-infection (hpi) ( Figure 1C ) (van den Pol et al., 2009) . In fact, by 12 hpi, labeling was sufficiently bright to image fine morphological details, such as dendritic spines ( Figure 1C ,iv).
To characterize the physiological properties of cells infected with rVSV, we tested a replication-competent rVSV encoding GFP, with RABV-G in the genome in place of VSV-G [hereafter designated rVSV(RABV-G)]. van den Pol et al. reported that hippocampal neurons infected with replication-incompetent (G-deleted or " G") rVSV were physiologically healthy at 12-14 hpi, but were less so by 1 day post-infection (dpi) (van den Pol et al., 2009) . Given the known toxicity of both VSV and RABV-G (Coulon et al., 1982) , we tested the physiology of cortical pyramidal neurons in the motor cortex (M1) infected with rVSV(RABV-G). Between 12 and 18 hpi, the membrane capacitance, input resistance, resting membrane potential, and current-to-action potential firing relationship were indistinguishable between infected and uninfected neurons (Figure 2) . However, by 2 dpi, electrophysiological properties were so abnormal in the infected cortical pyramidal cells that physiological measurements could not be made.
The speed and strength of the expression of transgenes encoded by VSV depends upon the gene's genomic position (van den Pol et al., 2009; Beier et al., 2011) . Genes in the first position are expressed the most highly, with a decrease in the level of expression in positions more 3 within the viral plus strand. When GFP was inserted into the first position of VSV, GFP fluorescence was first detectable at approximately 1 hpi in cultured cells (van den Pol et al., 2009) . In order to quantify the relative expression of a fluorescent protein in the first genomic position in neurons, rat hippocampal slices were infected with a replication-incompetent rVSV that expresses mCherry (rVSV G, Figures 1A,B) . This was a G virus which had the RABV-G supplied in trans during the preparation of the virus stock [referred to as rVSV G(RABV-G)].
Average fluorescence intensity of the infected cells was measured every hour over the course of 18 h. By 4 hpi at 37 • C, red fluorescence was clearly visible, and reached maximal levels by approximately 14 hpi (N = 3, Figure 3 ). Similar results were obtained with a virus encoding GFP in the first genomic position rather than mCherry (i.e., Figure 1B ) (N = 3).
We previously demonstrated that rVSV(RABV-G) could be taken up retrogradely by neurons (Beier et al., 2011) , but these experiments did not distinguish between direct axonal uptake of the initial inoculum vs. retrograde transsynaptic transmission following viral replication. To distinguish between these two mechanisms and to extend the previous analyses, we conducted further experiments in the mammalian visual system (Figures 4A-G) . As visual cortex area 1 (V1) does not receive direct projections from retinal ganglion cells (RGCs), but rather receives secondary input from RGCs via the lateral geniculate nucleus (LGN), infection of RGCs from injection of V1 would demonstrate retrograde transmission from cells which supported at least one round of viral replication. Following a V1 injection with rVSV(RABV-G), GFP-positive RGCs were observed in the retina by 3 dpi (N = 3; Figure 4G ). Importantly, viral labeling in the brain was restricted to primary and secondary projection areas, even at 7 dpi. These included the LGN ( Figure 4D ) and the hypothalamus (Figure 4E) , two areas known to project directly to V1 (Kandel, 2000) . Selective labeling was observed in other areas, such as cortical areas surrounding V1 (Figure 4C) , which project directly to V1, and also in the superior colliculus (SC) stratum griseum centrale, which projects to the LGN ( Figure 4F) . Labeling was also observed in the nucleus basalis, which projects to the cortex, as well as many components of the basal ganglia circuit, which provide input to the thalamus [such as the caudate-putamen (CP), globus pallidus (GP), and the subthalamic nucleus (STn)]. The amygdala, which projects to the hypothalamus, was also labeled. Consistent with a lack of widespread viral transmission, animals did not exhibit signs of disease at 7 dpi.
These data show that rVSV(RABV-G) can spread in a retrograde direction from the injection site, but do not address whether the virus can spread exclusively in the retrograde direction. Directional transsynaptic specificity can only be definitively addressed using a unidirectional circuit. We therefore turned to the primary motor cortex (M1) to CP connection, in which neurons project from the cortex to the CP, but not in the other direction ( Figure 4H ) (Beier et al., 2011) . Injections of rVSV(RABV-G) into M1 should not label neurons in the CP if the virus can only label cells across synapses in the retrograde direction. Indeed, at 2 dpi, areas directly projecting to the injection site, including the contralateral cortex, were labeled ( Figure 4I ). Only axons from cortical cells were observed in the CP, with no GFP-labeled cell bodies present in the CP (Figure 4J) , consistent with lack of anterograde transsynaptic spread. By 3 dpi, a small number of medium spiny neurons (MSNs) in the CP were observed, likely via secondary spread from initially infected thalamic or GP neurons (data not shown).
A particular advantage of retrograde viral tracers is the ability to label CNS neurons projecting to peripheral sites. This has been a powerful application of both RABV and PRV (Ugolini et al., 1989; Standish et al., 1994) . To test if rVSV(RABV-G) could also perform this function, we examined the innervation of the dura surface by neurons of the trigeminal ganglion, a neuronal circuit thought to be involved in migraine headaches (Penfield and McNaughton, 1940; Mayberg et al., 1984) . These neurons have axons, but not canonical dendrites, and send projections into the spinal cord and brainstem. Therefore, the only way trigeminal neurons could become labeled from viral application to the dura is through retrograde uptake of the virus. We applied rVSV(RABV-G) to the intact dura mater and analyzed the dura, trigeminal ganglion, and CNS for labeling ( Figure 4K ). At the earliest time point examined, 3 dpi, we observed axons traveling along the dura, but little other evidence of infection ( Figure 4L) . No labeled neuronal cell bodies on the dura were observed, consistent with the lack of neurons on this surface. In contrast, we did find labeled cell bodies in the trigeminal ganglion ( Figure 4M ). No infection was seen in the CNS, even at 4 dpi, consistent with the lack of inputs from the brain into the trigeminal ganglion (N = 4 animals).
To further characterize patterns and kinetics of viral transmission and directional specificity of transsynaptic spread, injections of rVSV(RABV-G) were made into the CP ( Figure 5A ). In order to determine which cells were labeled by direct uptake of virus in the inoculum, a separate set of animals were injected into the CP with the replication-incompetent rVSV G(RABV-G) (N = 3 animals, analyzed 3 dpi). Cells labeled by rVSV G(RABV-G) were observed in the CP, GP, substantia nigra (SN), thalamus, and layers 3 and 5 of the cortex, consistent with infection at the axon terminal and retrograde labeling of cell bodies of neurons known to project directly to the CP ( Figure 5C ) (Albin et al., 1995) . Areas labeled by CP injection are indicated in Figure 5B .
The patterns of spread for the replication-competent rVSV(RABV-G) were characterized over the course of 1-5 dpi ( Figures 5D-H) . During this interval, progressively more cells in infected regions were labeled by rVSV(RABV-G), including within the CP, nucleus basalis, cortex, and GP (listed in Figure 5B ). In addition, more cortical cells were labeled in clusters near cortical pyramidal neurons, both ipsilateral and contralateral to the injected side, including neurogliaform cells (data not shown). These data are in contrast to those observed following infection with an anterograde transsynaptic tracing virus, such as rVSV with its own G gene, rVSV(VSV-G) (Figure 5B) . At 3 dpi following rVSV(VSV-G) injection into the CP, the cerebral cortex was not labeled, but regions receiving projections from the CP, such as the STn, GP, and SN, were labeled (Beier et al., 2011) .
In order to investigate other areas for evidence of cell-to-cell retrograde transsynaptic spread, the nucleus basalis was examined following infection of the CP with replication-competent rVSV(RABV-G). The nucleus basalis was labeled by 2 dpi (Figures 5E-H) , consistent with at least a single transsynaptic jump, as this area does not directly project to the CP. The virus appeared to travel transsynaptically at the rate of roughly 1 synapse per day, as evidenced by the lack of labeled neurogliaform cells in the cortex, and lack of neurons in the nucleus basalis at 1 dpi, and label appearing in these cell types/areas at 2 dpi, as previously observed (Beier et al., 2011) . Labeling remained well-restricted to the expected corticostriatal circuits at 5 dpi, suggesting that viral spread becomes less efficient after crossing one or two connections, consistent with injections into V1 (Figure 4) . While glial cells can be infected and were observed near the injection site (van den Pol et al., 2002; Chauhan et al., 2010) , infected glial cells away from the injection site generally were not observed.
One advantage of having both anterograde and retrograde forms of the same virus is that they can be used in parallel, or in tandem, to trace circuitry to and from a single or multiple sites of injection, with each virus having similar kinetics of spread and gene expression. In fact, if different fluorophores are used in different viruses, e.g., rVSV(VSV-G) and rVSV(RABV-G), then the viruses can be co-injected into the same site and their transmission can be traced independently ( Figure 6A ). This is most straightforward if there are no cells at the injection site that are initially infected by both viruses. Co-infected cells can be easily detected, as they would express both fluorescent proteins shortly after injection.
In order to determine whether two viruses would allow simultaneous anterograde and retrograde transsynaptic tracing from a single injection site, a rVSV(VSV-G) expressing Venus and a rVSV(RABV-G) expressing mCherry were injected individually (Figures 6B-D) or co-injected (Figures 6E-G) into the motor cortex, and brains were examined 3 dpi. The pattern of labeling from the co-injected brains was equivalent to the patterns observed when each virus was injected individually: rVSV(VSV-G) was observed to infect neurons in the cortex, CP, and downstream nuclei, whereas the rVSV(RABV-G) was not observed to infect neurons in the CP, but rather in the thalamus and nucleus basalis (N = 4). The initial co-infection rate is dependent upon The presence or absence of labeling is indicated by (+) and (−), respectively. The extent of labeling is indicated by the number of (+). Some animals were infected with G viruses to determine which areas were labeled by direct uptake of the virions, rather than by replication and transmission. These were sacrificed at 3 dpi. (C) Parasaggital section of a brain infected with VSV[greek delta]G(RABV-G). The injection site is marked by a red arrow. Several areas that project directly to the CP were labeled due to direct uptake of the virions, including the cortex, thalamus, and GP (arrowheads), 3 dpi. the dose of the initial inocula. When injecting 3 × 10 3 focus forming units (ffu) rVSV(VSV-G) and 3 × 10 4 ffu rVSV(RABV-G), no co-infection was observed at the injection site. Thus, co-infection of the same brain region, without co-infection of the same cells,
does not alter the spreading behavior of either rVSV(VSV-G) or rVSV(RABV-G). One example of how this dual retrograde and anterograde transsynaptic tracing system can be used is to determine if three Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 8 distinct regions are connected and the directionality of any connections. For example, the anterograde transsynaptic virus can be injected into one region, the retrograde into another, and a third region can then be examined for evidence of labeling by either or both viruses (e.g., Figure 6H ). To test this possibility, rVSV(VSV-G) was injected into the motor cortex, rVSV(RABV-G) was injected into the substantia nigra pars reticulara (SNr), and animals were sacrificed at 3 dpi. We observed that cells were singly labeled, either with Venus [rVSV(VSV-G)] or with mCherry [rVSV(RABV-G)], and were located largely in different regions of the CP (Figures 6I,J) (N = 3) . These results suggest that the anterograde connections from the cells infected with rVSV(VSV-G) in the M1 were with CP MSNs that did not project to the region of the SNr injected with rVSV(RABV-G) (N = 3 animals).
In addition to polysynaptic tracing, VSV can be modified to trace circuits monosynaptically (Beier et al., 2011) . With RABV, this was achieved in vivo by first infecting with an adeno-associated virus (AAV) expressing TVA, a receptor for an avian retrovirus, and RABV-G (Wall et al., 2010) . This was followed 3 weeks later by infection with a G RABV with an EnvA/RABV-G chimeric glycoprotein on the virion surface (Wickersham et al., 2007b) , which allowed infection specifically of the cells expressing TVA. A similar strategy was used to test rVSV's ability to monosynaptically trace retrogradely connected neurons in vivo. Inputs to choline acetyltransferase (ChAT)-expressing neurons in the striatum were used for this test. These neurons primarily receive input from the cortex and the thalamus (Thomas et al., 2000; Bloomfield et al., 2007) (Figure 7A) . In order to mark this population, we crossed ChAT-Cre mice to Ai9 mice, which express tdTomato in cells with a Cre expression history (Madisen et al., 2010) . Six-week-old mice from this cross were injected in the CP with two AAV vectors: one expressing a Cre-conditional ("floxed") TVA-mCherry fusion protein, and another expressing a floxed RABV-G. Two weeks later, the mice were injected in the same coordinates with rVSV G with the EnvA/RABV-G chimeric glycoprotein on the virion surface [rVSV G(A/RG)] (Beier et al., 2011) . Cells successfully infected with these two AAV vectors could host infection by a rVSV and should be able to produce rVSV virions with RABV-G on the surface. Such starter cells should also express tdTomato and GFP. If rVSV were to be produced, and if it were to transmit across the synapse retrogradely, cortical and thalamic neurons should be labeled by GFP. Mice injected with these AAV and rVSV viruses were sacrificed 5 days after rVSV infection, and brains analyzed for fluorescence. As expected for starter cells, some neurons in the CP expressed both tdTomato and GFP (Figure 7B) . Outside of the CP, small numbers of GFP+ neurons that were not mCherry+ were observed in the cortex (Figures 7C,D) and thalamus (Figure 7E) , consistent with retrograde spread. Control animals not expressing Cre, or not injected with AAV encoding RABV-G, did not label cells in the cortex or thalamus (N = 3 for both controls and experimental condition).
Here, we report on the use of rVSV as a retrograde transsynaptic tracer for CNS circuitry. VSV can be modified to encode the RABV-G protein in the viral genome, allowing the virus to replicate and transmit across multiple synaptically connected cells, i.e., as a polysynaptic tracer. Alternatively, if the virus has the G gene deleted from its genome and RABV-G is provided in trans, it behaves as a monosynaptic tracer (Beier et al., 2011) . Although it has been known for many years that RABV travels retrogradely among neurons (Astic et al., 1993; Ugolini, 1995; Kelly and Strick, 2003) , and pseudotyping lentiviruses with RABV-G is sufficient for axonal transport (Mazarakis et al., 2001) , the retrograde transmission specificity among neurons had not been clearly shown to be a property of the G protein itself, as it might have been due to other viral proteins in addition to, or instead of, the viral G protein. Since native VSV does not have these retrograde transsynaptic properties (van den Pol et al., 2002; Beier et al., 2011) , and the only alteration to the VSV genome was the substitution of the VSV G gene with the G gene of RABV, it is clear that the RABV glycoprotein is responsible for retrograde direction of viral transmission across synapses, at least in the case of rVSV.
The early onset of gene expression from VSV relative to RABV (one hour vs. multiple hours) makes it beneficial in experimental paradigms in which the experiment needs to be done within a narrow window of time, such as tissue slices and explants. In addition, more than one transgene can be encoded in the viral genome without the need of a 2A or IRES element. The use of the first position of the genome enhances the expression level of the transgene inserted at that location, since VSV (and RABV) express genes in a transcriptional gradient; therefore, the first gene is the most highly transcribed (Knipe, 2007) . This leads to rational predictions of expression levels so that one can choose the position of insertion of a transgene, or transgenes, according to this gradient and the desired level of expression. The size of the viral capsid is apparently not rigid, allowing for the inclusion of genomes that are substantially larger than the native genome, unlike the rigid capacity for some other viral vectors, such as AAV Yan et al., 2000) .
The fact that VSV can be made to spread anterogradely (Beier et al., 2011) or retrogradely across synapses with the change of a single gene affords several advantages over viral tracers that heretofore have not shown such flexibility in the directionality of tracing. In addition to the obvious application of tracing anterograde connections, combinations can be made to exploit the different forms of the virus. One example that employs the simultaneous infection with an anterograde and retrograde form of VSV is demonstrated in Figure 6 . This experiment was designed to address whether the anterograde projections from the cortex to the CP would label the same brain regions as were labeled by a retrograde virus injected into the SN. Although a block of superinfection by the virus may preclude infection of the same cell with multiple rVSVs, adjacent cells could still become labeled by different viruses (Whitaker-Dowling et al., 1983) . The observed results could be due to a preferential labeling by the anterograde transsynaptic virus of indirect pathway MSNs in this experiment, which then synapse onto the GP, thereby reflecting a viral bias. Alternatively, it could indicate that the cortical neurons in the injected region largely do not label the MSNs that project to the area of the SN injected with the retrograde virus. One further possibility is that too little virus was used to observe co-labeling of a given region. However, given the density of infection (i.e., Figures 6I,J) , the latter possibility seems unlikely. Additionally, the spread of the polysynaptic rVSV(RABV-G) appears to attenuate with increasing numbers of synapses crossed, permitting an analysis of more restricted viral spread. This is quite fortuitous, as if spread were to continue, it would lead to widespread infection and lethality. In addition, reconstruction of connectivity would be more difficult. This reduced efficiency appears to also hold for the monosynaptic form of VSV complemented with RABV-G, as the efficiency of transmission appeared lower than the comparable experiment with RABV (Watabe-Uchida et al., 2012) . This is likely due to viral attenuation when VSV-G is replaced with RABV-G.
We were attracted to the use of VSV as a viral tracer due to its long track record as a safe, replication-competent laboratory agent. Laboratory workers using VSV have not contracted any diseases, and natural VSV infections among human populations in Central America and the southwestern United States (Rodríguez, 2002) occur without evident pathology (Johnson et al., 1966; Brody et al., 1967) . VSV was thus an attractive candidate for its use as a polysynaptic tracer for CNS studies, which requires an ability to replicate through multiple transmission cycles. Both replicationcompetent and incompetent forms of VSV are in use under Biosafety Level 2 containment. Replication-competent RABV is Biosafety Level 3, due to the fact that infection with replicationcompetent RABV is almost always fatal to humans and in mice when infected intracerebrally (Smith, 1981; Knipe, 2007) . Differences in pathogenicity between VSV and RABV are likely due to the ability of RABV to evade the innate immune system, particularly interferon (Hangartner et al., 2006; Junt et al., 2007; Lyles and Rupprecht, 2007; Rieder and Conzelmann, 2009; Iannacone et al., 2010) . VSV infection efficiently triggers an interferon response, and it has not evolved a method of escape from this response, unlike RABV (Brzózka et al., 2006) . In fact, VSV is being pursued as a vaccine for other viruses, including RABV (Lichty et al., 2004; Publicover et al., 2004; Kapadia et al., 2005; Schwartz et al., 2007; Iyer et al., 2009; Geisbert and Feldmann, 2011) . VSV does not typically spread beyond the initially infected site in the periphery (Kramer et al., 1983; Vogel and Fertsch, 1987) . This likely is the cause of the minor or absent symptoms in humans and animals infected in nature. Polysynaptic VSV vectors are thus predicted to be much safer than polysynaptic RABV vectors. We have tested this prediction by injecting a series of mice in the footpads and hind leg muscles with rVSV(RABV-G), with the result that no injected animals showed any evidence of morbidity or mortality (Beier, Goz et al., in preparation) .
While safer for laboratory workers than RABV, the main drawback to using VSV is its rapid cellular toxicity (van den Pol et al., 2009; Beier et al., 2011) . Toxicity is due to suppression of cellular transcription and a block in the export of cellular RNAs from the nucleus to the cytoplasm (Black and Lyles, 1992; Her et al., 1997; Ahmed and Lyles, 1998; Petersen et al., 2000; von Kobbe et al., 2000) , as well as inhibition of the translation of cellular mRNAs (Francoeur et al., 1987; Jayakar et al., 2000; Kopecky et al., Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 10 2001). VSV is much quicker to enact its gene expression program than is RABV, such that cells suffer the toxic effects more quickly than after RABV infection. One aspect of VSV that can be exploited in the future to ameliorate the speed of toxicity is the use of VSV mutants and variants. One such mutant is the M51R, which permitted us to conduct physiological analyses of pre-and post-synaptic cells (Beier et al., 2011) . We are in the process of examining the transmission properties of this mutant in vivo, as well as the effects of other mutations or viral variants on prolonging the health of neurons after infection.
rVSV vectors can be used to study the connectivity of neuronal circuitry. In addition to combinations of replication-competent forms of VSV, the replication-incompetent, monosynaptic forms of the virus can be easily combined, without the need to change viruses (Beier et al., 2011) . This allows a straightforward way to study both the projections into, and out from, a genetically defined cell population. This can be done with the same viral genome, with the only change needed being the glycoprotein, for the selection of the direction of transmission. This flexibility of VSV makes it a powerful, multi-application vector for studying connectivity in the CNS.
All rVSV clones were cloned from the rVSV G backbone (Chandran et al., 2005) . mCherry, Kusabira orange, Venus, and CFP were cloned into the first (GFP) position using XhoI and MscI sites, and VSV-G (a gift from Richard Mulligan, Harvard Medical School, Boston, MA) and RABV-G (a gift from Ed Callaway, Salk Institute, San Diego, CA) were cloned into the fifth (G) position using the MluI and NotI restriction sites. Genes for fluorescent proteins were obtained from Clontech. Viruses were rescued as previously described (Whelan et al., 1995) . At 95% confluency, eight 10 cm plates of BSR cells were infected at an MOI of 0.01. Viral supernatants were collected at 24-h time intervals and ultracentrifuged at 21,000 RPM using a SW28 rotor and resuspended in 0.2% of the original volume. For titering, concentrated viral stocks were applied in a dilution series to 100% confluent BSR cells and plates were examined at 12 hpi. Viral stocks were stored at −80 • C.
For G viruses, 293T cells were transfected with PEI (Ehrhardt et al., 2006) at 70% confluency on 10 cm dishes with 5 µg of pCAG-RABV-G. Twenty-four hours post-infection, the cells were infected at an MOI of 0.01 with rVSV G expressing either GFP or mCherry. Viral supernatants were collected for the subsequent 4 days at 24 h intervals.
Virus preparations are now available from the Salk GT3 viral core (http://vectorcore.salk.edu/). All plasmids are available from Addgene (http://www.addgene.org/).
AAV-FLEx-RABV-G and AAV-FLEx-TVA-mCherry plasmids originated from the Lab of Naoshige Uchida (Watabe-Uchida et al., 2012) , and virus stocks were generous gifts from Brad Lowell, Harvard Medical School.
ChAT-Cre (B6;129S6-Chat tm1(cre)Lowl /J) and Ai9 (B6.Cg-Gt(ROSA)26Sor<tm9(CAG-tdTomato)Hze>/J) mice were obtained from the Jackson Laboratory (Madisen et al., 2010) .
Eight-week-old CD-1 mice were injected using pulled capillary microdispensers (Drummond Scientific, Cat. No: 5-000-2005) , using coordinates from The Mouse Brain in Stereotaxic Coordinates (Franklin and Paxinos, 1997) . Injection coordinates (in mm) used were: For multi-color analysis (Figures 1C,D) , 3 × 10 9 ffu/mL rVSV was injected into various regions. For CP injections, 100 nL of rVSV(RABV-G) or rVSV(VSV-G) at 3 × 10 7 ffu/mL was injected at a rate of 100 nL/min. For the replication-incompetent viruses, 100 nL of 1 × 10 7 ffu/mL rVSV G(RABV-G) or rVSV G (VSV-G) was injected. In the motor cortex, 100 nL of 1 × 10 7 ffu/mL rVSV(RABV-G) was injected, and mice harvested 2 dpi. For V1 injections, 100 nL of 3 × 10 10 ffu/mL rVSV(RABV-G) was injected, and mice were examined 3 or 7 dpi.
For infections of the dura mater, 1 µL of 3 × 10 10 ffu/mL rVSV(RABV-G) was applied to the surface of the dura. The virus was allowed to absorb, and the surface was subsequently covered in bone wax, and the wound sutured.
For co-injections of virus into the same animal, 100 nL of a combination of 3 × 10 7 ffu/mL rVSV(VSV-G) and 3 × 10 8 ffu/mL rVSV(RABV-G) were co-injected into the motor cortex, and brains examined 3 dpi. For injections of the viruses into different regions, 100 nL of 3 × 10 7 ffu/mL rVSV(VSV-G) was injected into M1, and 100 nL of 3 × 10 8 ffu/mL rVSV(RABV-G) into the SNr, and brains examined 3 dpi. A lower titer of rVSV(VSV-G) was used, as rVSV(RABV-G) is attenuated.
All mouse work was conducted in biosafety containment level 2 conditions and was approved by the Longwood Medical Area Institutional Animal Care and Use Committee.
Recordings were made from cortical pyramidal neurons in slices taken from postnatal day 12-18 mice, inoculated in the CP 12-18 h prior with rVSV(RABV-G). Coronal slices (300 µm thick) were cut in ice-cold external solution containing (in mM): 110 choline, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 7 MgCl 2 , 0.5 CaCl 2 , 25 glucose, 11.6 Na-ascorbate, and 3.1 Na-pyruvate, bubbled with 95% O 2 and 5% CO 2 . Slices were then transferred to artificial cerebrospinal fluid (ACSF) containing (in mM): 127 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 2.5 KCl, 1 MgCl 2 , 2 CaCl 2 , and 25 glucose, bubbled with 95% O 2 and 5% CO 2 . After an incubation period Frontiers in Neural Circuits www.frontiersin.org February 2013 | Volume 7 | Article 11 | 11 of 30-40 min at 34 • C, slices were stored at room temperature. All experiments were conducted at room temperature (25 • C). In all experiments, 50 µM picrotoxin, 10 µM 2,3-Dioxo-6-nitro-1, 2, 3, 4 -tetrahydrobenzo [f]quinoxaline -7-sulfonamide (NBQX), and 10 µM 3-((R)-2-Carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP) were present in the ACSF to block GABAA/C, AMPA, and NMDA receptor-mediated transmission, respectively. All chemicals were from Sigma or Tocris.
Whole-cell recordings were obtained from infected and uninfected deep layer cortical pyramidal neurons identified with video-IR/DIC and GFP fluorescence was detected using epifluorescence illumination. With the deep layers of the cortex, 2-photon laser scanning microscopy (2PLSM) was used to confirm the cell types based on morphology. Deep layer pyramidal neurons had large cell bodies, classic pyramidal shape and dendritic spines. Glass electrodes (2-4 M ) were filled with internal solution containing (in mM): 135 KMeSO 4 , 5 KCl, 5 HEPES, 4 MgATP, 0.3 NaGTP, 10 Na 2 HPO 4 , 1 EGTA, and 0.01 Alexa Fluor-594 (to image neuronal morphology) adjusted to pH 7.4 with KOH. Current and voltage recordings were made at room temperature using a AxoPatch 200B or a Multiclamp 700B amplifier. Data was filtered at 5 kHz and digitized at 10 kHz.
Imaging and physiology data were acquired and analyzed as described previously (Carter and Sabatini, 2004) . Resting membrane potential was determined by the average of three 5-s sweeps with no injected current. Passive properties of the cell, membrane (Rm) and series resistance (Rs) and capacitance (Cm), were measured while clamping cells at −65 mV and applying voltage steps from −55 to −75 mV. The current-firing relationship was determined in current clamp with 1-s periods of injected current from 100 to 500 pA.
The time course of viral gene expression experiments were carried out in organotypic hippocampal slice cultures prepared from postnatal day 5-7 Sprague-Dawley rats as described previously (Stoppini et al., 1991) . Slices were infected after 7 days in vitro, and images were acquired on a two-photon microscope.
|
What types of viruses can be used to study the connectivity of neuronal circuitry?
| 1,938 |
rVSV vectors
| 31,209 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What is Carrageenan?
| 2,141 |
a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections
| 587 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What is a potential therapeutic benefit of carageenan?
| 2,142 |
carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7)
| 1,184 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What is the optimal window for initiating treatment with carageenan and Zanamivir?
| 2,143 |
72 hours post infection
| 1,719 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What was the mortality rate of influenza a virus subtype h7n9 (avian or bird flu)?
| 2,144 |
more than 35%
| 2,124 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
How many human cases were there of influenza a virus subtype h7n9?
| 2,145 |
more than 400 human cases
| 2,048 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
How did most patients contract influenza a virus subtype h7n9?
| 2,146 |
Most patients with A(H7N9) infections had contact with poultry or visited live animal markets
| 2,139 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
Can influenza a virus subtype h7n9 be transmit human to human?
| 2,147 |
some sporadic cases seemed to be a result of human to human transmissions
| 2,243 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What kind of disease is caused by influenza?
| 2,148 |
influenza is an acute respiratory disease
| 2,637 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
How many severe cases are there for annual influenza epidemics?
| 2,149 |
Worldwide, annual epidemics result in about three to five million cases of severe illness
| 2,751 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
How many deaths occur annually as a result of annual influenza epidemics?
| 2,150 |
250,000 to 500,000
| 2,853 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
Is coinfection common in influenza infection?
| 2,151 |
influenza virus infections are often accompanied by other viral pathogens
| 3,095 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What percentage of people infected with influenza have a viral coinfection?
| 2,153 |
Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] .
| 3,292 |
1,629 |
The Intranasal Application of Zanamivir and Carrageenan Is Synergistically Active against Influenza A Virus in the Murine Model
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4459876/
SHA: f0b1fa4036434b57c8307d43c39a4193f7e8053a
Authors: Morokutti-Kurz, Martina; König-Schuster, Marielle; Koller, Christiane; Graf, Christine; Graf, Philipp; Kirchoff, Norman; Reutterer, Benjamin; Seifert, Jan-Marcus; Unger, Hermann; Grassauer, Andreas; Prieschl-Grassauer, Eva; Nakowitsch, Sabine
Date: 2015-06-08
DOI: 10.1371/journal.pone.0128794
License: cc-by
Abstract: BACKGROUND: Carrageenan is a clinically proven and marketed compound for the treatment of viral upper respiratory tract infections. As infections caused by influenza virus are often accompanied by infections with other respiratory viruses the combination of a specific anti-influenza compound with the broadly active antiviral polymer has huge potential for the treatment of respiratory infections. Thus, the combination of the specific anti-influenza drug Zanamivir together with carrageenan in a formulation suitable for intranasal application was evaluated in-vitro and in-vivo. PRINCIPAL FINDINGS: We show in-vitro that carrageenan and Zanamivir act synergistically against several influenza A virus strains (H1N1(09)pdm, H3N2, H5N1, H7N7). Moreover, we demonstrate in a lethal influenza model with a low pathogenic H7N7 virus (HA closely related to the avian influenza A(H7N9) virus) and a H1N1(09)pdm influenza virus in C57BL/6 mice that the combined use of both compounds significantly increases survival of infected animals in comparison with both mono-therapies or placebo. Remarkably, this benefit is maintained even when the treatment starts up to 72 hours post infection. CONCLUSION: A nasal spray containing carrageenan and Zanamivir should therefore be tested for prevention and treatment of uncomplicated influenza in clinical trials.
Text: The periodic appearance of new influenza variants poses a worldwide pandemic threat. Since the emergence of the new A(H7N9) virus, more than 400 human cases were reported to the WHO with a mortality rate of more than 35%. Most patients with A(H7N9) infections had contact with poultry or visited live animal markets. However, some sporadic cases seemed to be a result of human to human transmissions [1, 2] . In contrast to pandemic viruses which fulminantly enter the human population and cause high mortality rates, seasonal influenza viruses generally cause uncomplicated and transient infections in humans, with virus replication localized to the upper respiratory tract [3, 4] . However, in its fully developed form influenza is an acute respiratory disease resulting in hospitalizations and deaths mainly among high-risk groups. Worldwide, annual epidemics result in about three to five million cases of severe illness, and about 250,000 to 500,000 deaths [5] . For this reason WHO [6] and CDC [7] recommend antiviral treatment for any patient with suspected influenza who is at risk for influenza complications without previous laboratory confirmation.
It is known that influenza virus infections are often accompanied by other viral pathogens [8] . Depending on the detection method (qRT-PCR or immunofluorescence) different ratios of co-infections have been found. Analysis by qRT-PCR revealed that 54.5-83.3% of influenza A or B positive patients were found to have at least one concomitant respiratory viral infection [9] [10] [11] [12] . The detection frequency with immunofluorescence was found to be even higher (90-100%) [13, 14] . Potential concomitant viral pathogens of influenza virus infections include human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus [14, 15] .
As a result of the multiple infections, a specific anti-influenza mono-therapy treats the influenza virus infection only, but not the infection with the concomitant viral pathogen. Hence, the therapy often fails to sufficiently resolve symptoms. This is also reflected by the fact that neuraminidase inhibitors (NI) are highly efficacious in animal models investigating influenza mono-infections [16, 17] but show lower efficacy against influenza symptoms in clinical trials in adults with natural infections [18] . Therefore, there is a high medical need for a broadly acting antiviral therapy in combination with a specific anti-influenza therapy for treatment of patients suffering from upper respiratory tract symptoms. Ideally, the substances present in the combination complement each other by different modes of action, leading to a treatment that provides full protection against a broad range of different respiratory viruses as well as different influenza strains with a low probability to induce escape mutations.
One approach for a broad antiviral therapy is the creation of a protective physical barrier in the nasal cavity using carrageenan. Carrageenan is a high molecular weight sulfated polymer derived from red seaweed (Rhodophyceae) that has been extensively used in food, cosmetic and pharmaceutical industry and is generally recognized as safe by the FDA (GRAS) (reviewed in [19] ). Three main forms of carrageenans are commercially used: kappa, iota and lambda. They differ from each other in the degree of sulfation, solubility and gelling properties [20] . The antiviral mechanism of carrageenan is based on the interference with viral attachment; as a consequence, viral entry is inhibited [21, 22] . Its antiviral activity is dependent on the type of polymer as well as the virus and the host cells [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] and has been reviewed in [33] [34] [35] . We published that iota-carrageenan is a potent inhibitor of hRV [36] and influenza A [37] replication and demonstrated the antiviral efficacy of iota-carrageenan against common cold viruses by intranasal application in several randomized, double-blind, parallel group, placebo-controlled clinical trials [38] [39] [40] . The pooled analysis of two studies conducted in 153 children and 203 adults revealed that patients infected with any respiratory virus, who were intranasally treated with iota-carrageenan showed a 1.9 day faster recovery from common cold symptoms than placebo treated patients in the intention-to-treat population [41, 42] . The anti-influenza activity was shown by subgroup analysis of 49 influenza infected patients who benefited from a 3.3 days faster recovery from symptoms. The use of carrageenan nasal spray was associated with a significant reduction of the influenza viral load in nasal fluids and a significant increase in the number of virus free patients within the treatment period of 7 days. In good accordance Prieschl-Grassauer are co-founders of Marinomed Biotechnologie GmbH. Marinomed Biotechnologie GmbH had a role in study design, data collection and analysis, decision to publish, preparation of the manuscript and is financing the processing charge of the manuscript. with the literature [9] [10] [11] [12] [13] [14] we observed that the majority of influenza virus infected patients suffered from a concomitant respiratory viral infection (66%) as determined by real-time PCR. Carrageenan containing nasal sprays are already marketed for the treatment of respiratory viral infections under different brand names in 18 countries.
At present the only available effective drugs for treatment and post exposure prevention of influenza are the NI (Oseltamivir and Zanamivir worldwide; Peramivir in Japan and South Korea). Since the large-scale use of M2 blockers for prophylaxis and treatment in humans [43] and farming [44] , the currently circulating influenza viruses already lack sensitivity to this drug group [45] .
We have already shown an additive therapeutic effect of a combination therapy with intranasally applied iota-carrageenan and orally administered Oseltamivir in lethally H1N1 A/PR/ 8/34 infected mice and a treatment start 48 hours post infection (hpi) [37] .
Due to these very promising results we further developed the concept of combining carrageenan with an NI therapy. In contrast to Oseltamivir, which needs to be activated by metabolic conversion, Zanamivir is directly applied as active drug and can also be administered intranasally [46] [47] [48] [49] [50] [51] [52] . The potential of an intranasal administration of Zanamivir was investigated by GlaxoSmithKline. In seven clinical challenge trials 66 volunteers were infected with influenza B/Yamagata/16/88 and 213 with influenza A/Texas/36/91 (H1N1). 156 of these participants got intranasally applied Zanamivir at different doses (daily dose levels from 6.4 mg to 96 mg) for prophylaxis or therapy [46, 47, 53, 54] . These challenge trials showed that treatment starting before and up to 36 hours post virus inoculation was associated with prevention of laboratory confirmed influenza and febrile illness as well as a reduction in viral titers, duration of shedding and symptoms. In total, safety data from 1092 patients after intranasal application of Zanamivir were published and no evidence for Zanamivir induced adverse events or increased frequencies of local nasal intolerance in comparison to placebo groups was found [46, 49, 52] .
Taken together, the combination of a carrageenan nasal spray that provides broad antiviral activity against upper respiratory infections-including influenza-with Zanamivir, a specific anti-influenza drug, meets the existing medical need to treat multiple viral infections. In the present work we investigate the therapeutic effect of a combination of carrageenan and Zanamivir in-vitro and in an animal model.
Kappa-carrageenan and iota-carrageenan were purchased from FMC Biopolymers (Philadelphia, PA). The identity, purity (>95%) of carrageenan subtypes and the molecular weight (>100,000) was confirmed by NMR analysis as described elsewhere [55] and the presence of lambda-carrageenan was below the detection limit of 3%. The dry polymer powders were dissolved in aqua bidest (Fresenius Kabi, Austria) to a final concentration of 2.4 mg/ml iota-and 0.8 mg/ml kappa-carrageenan. This 2x stock solution was sterile filtered through a 0.22 μm filter (PAA, Switzerland) and stored at room temperature until use. For further testing the stock solution was diluted to a mixture containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan (hereinafter referred to as "carrageenan").
Zanamivir was purchased as powder (Haosun Pharma, China) and the identity and purity was confirmed by NMR analysis. Zanamivir was either dissolved in carrageenan or placebo solutions, followed by sterile filtration through a 0.22 μm filter (Sarstedt, Germany). For in-vivo studies all Zanamivir containing solutions were freshly prepared.
Madin-Darby canine kidney (MDCK) cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultivated in a 37°C incubator (Sanyo, Japan; CO 2 : 5%, relative humidity: >95%). MDCK cells were grown in Dulbecco's minimal essential (DMEM) high glucose medium (PAA, Austria) supplemented with 10% fetal bovine serum (FBS; PAA, Austria; heat inactivated).
Influenza virus A/Hansa Hamburg/01/09 (H1N1(09)pdm) was kindly provided by Peter Staeheli Department of Virology, University of Freiburg, Germany and previously described in [56] ; A/Teal/Germany/Wv632/05 (H5N1) previously published in [57] (accession numbers CY061882-9) and A/Turkey/Germany/R11/01 (H7N7) (taxonomy ID 278191, accession number AEZ68716) were supplied by courtesy of Martin Beer, Institute of Diagnostic Virology, Friedrich-Loeffler-Institute, Riems, Germany; A/Aichi/2/68 (H3N2) was purchased from the ATCC. All influenza viruses were propagated in MDCK cells at 37°C and 5% CO 2 in influenza medium [Opti-Pro serum free medium (Gibco, Austria) supplemented with 4 mM L-glutamine (PAA, Austria), 1% antibiotic-antimycotic mix (PAA, Austria) and 5 μg/ml trypsin (Sigma Aldrich, Austria)].
To determine the 50% inhibitory concentration (IC 50 ) and the combination effect of carrageenan and Zanamivir, a semi-liquid plaque assay was developed. Into 96 well tissue culture plates 1.7x10 4 MDCK cells/well were seeded and infected at 90% confluence (24-28 hours later). Serial dilutions of carrageenan and Zanamivir were prepared in assay medium (influenza medium without trypsin). For infection, viruses were diluted to an MOI of 0.003 (H1N1(09)pdm and H3N2 Aichi), 0.015 (H5N1) or 0.004 (H7N7), respectively, in assay medium and incubated at room temperature (RT) for 10 min with the serial dilutions of carrageenan and/or Zanamivir, respectively. For evaluation of the combination effect of carrageenan and Zanamivir, viruses were diluted in assay medium containing constant concentrations of either carrageenan or Zanamivir. The other substance was serially diluted and used for virus incubation. Cells were infected in 6 replicates/compound dilution, respectively, and incubated at RT for 45 min before inoculum removal. Cells were further incubated with the respective concentration of the investigated substances present in the overlay [influenza medium with 2.25% Carboxymethylcellulose (CMC, Fluka, Austria)] for 30-42 hours at 37°C. Evolving plaques were evaluated after methanol/acetone cell fixation by immune staining with antibodies either directed against the influenza A nucleoprotein (AbD Serotec, Germany) (for H1N1(09)pdm, H5N1 and H7N7) or the hemagglutinin (AbD Serotec, Germany) (for H3N2). Analysis was done with a HRP labeled detection antibody (Thermo Scientific, Germany) using TMB (Biolegend, Germany) as substrate and a microplate reader at 450 nm. The reduction of detected signal represents a reduction in the number and size of plaques and indicates suppression of viral replication during infection and cultivation.
After the immunostaining cells were stained with 0.005% crystal violet solution to assess the condition of the cell layer and the toxicity of the compounds. IC 50 values and standard deviations were calculated for a sigmoidal dose response model using XLfit Excel add-in version 5.3.1.3.
All animal experiments were carried out according to the guidelines of the "European Convention for the Protection of Vertebrate Animals used for Experimental and other Scientific Purposes" and the Austrian law for animal experiments. All animal experiments were approved by the Veterinary University of Vienna institutional ethics committee and performed under the Austrian Federal Ministry of Science and Research experimental animal license numbers BMWF-68.205/0262-II/3b/2011 and BMWF-68.205/0142-II/3b2012. C57BL/6 mice were purchased from Janvier Labs, France and maintained under standard laboratory conditions in the animal facilities of the Veterinary University of Vienna. For euthanasia and anesthesia asphyxiation through CO 2 was used and all efforts were made to minimize suffering.
For infection experiments, 3-5 weeks old female mice were intranasally inoculated with 50 μl influenza virus solution (25 μl/nostril) containing 2.27x10 3 or 1.65x10 3 plaque-forming unit of H1N1(09)pdm or H7N7, respectively. Subsequently, treatment started 24, 48 or 72 hpi, as indicated for the different experiments. Treatment was performed intranasally either with 50 μl therapeutic solution or placebo twice per day for 5 days. As therapy either carrageenan (containing 1.2 mg/ml iota-carrageenan and 0.4 mg/ml kappa-carrageenan to provide a daily dose of 12 mg/kg body weight (BW)), Zanamivir (containing either 130 μg/ml or 390 μg/ml Zanamivir, to provide a daily dose of 1 or 3 mg/kg BW, respectively) or a combination of carrageenan and Zanamivir were used. Carrageenan and Zanamivir are used at non-toxic concentrations as shown by [58] and [59] . Mice were monitored twice daily for 15 days for survival and weight loss. Mortality also includes mice that were sacrificed for ethical considerations when they had lost more than 25% of their initial body weight. We confirm the viral infection in these animals by necropsy and scoring of the lung inflammation.
As the mechanisms underlying the antiviral activity of NI and carrageenans are fundamentally distinct, they are likely to exhibit different activities towards the individual influenza virus strains. As a result, in combination they could complement each other to provide protection against a broader spectrum of influenza virus strains than the individual compounds.
To test this hypothesis, we investigated the sensitivity of various influenza virus strains to Zanamivir and carrageenan in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells [60, 61] . Using this method, we determined the IC 50 of Zanamivir and carrageenan against influenza A viruses of human and animal origin, namely H1N1(09)pdm (A/Hansa Hamburg/01/09), H3N2 (A/Aichi/2/68), low pathogenic (LP) H5N1 (A/Teal/Germany/ Wv632/05) and LP H7N7 (A/Turkey/Germany/R11/01) ( Table 1) . Both substances were nontoxic at the highest tested concentration (400 μM Zanamivir and 533 μg/ml carrageenan), neither was their combination. Furthermore, CMC in the overlay did not show any virus inhibitory effect (data not shown). Inhibition of viral replication of all tested influenza strains was achieved with both substances. However, the IC 50 values varied widely depending on the influenza virus strain. The IC 50 values of Zanamivir ranged between 0.18 μM for H5N1 and 22.97 μM for H7N7 and that of carrageenan from 0.39 μg/ml to 118.48 μg/ml for H1N1(09)pdm and H7N7, respectively (see Table 1 ). These results demonstrate that carrageenan and Zanamivir target individual influenza strains to different extents so that they may complement each other to provide broader anti-influenza activity.
The type of compound interaction was characterized by employing isobolograms (Fig 1) . As described in [62] , isobolograms graphically compare the doses of two compounds needed to reach 50% inhibition to the predicted doses calculated based on a model of drug additivity. A curve linearity of~1 is expected for an additive compound interaction whereas a curve progression <1 argue for synergistic and >1 for an antagonistic compound interaction.
Two virus strains were selected for those experiments, one being the most sensitive to carrageenan (H1N1(09)pdm) and one being the least sensitive (H7N7). In both cases the isobolograms show a synergistic interaction of carrageenan and Zanamivir (Fig 1) . Thus, it was shown that Zanamivir and carrageenan target individual influenza viruses with different efficiencies, most probably due to their different antiviral strategies. As a result, the combination provides synergistic activity with higher protection against a broader spectrum of influenza virus strains than the individual compounds.
In the influenza animal model, C57Bl/6 mice are challenged with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). Infection and treatment (twice a day for 5 days) are done intranasally without anesthesia. We investigated whether the combination of Zanamivir and carrageenan is more efficacious in reducing mortality than the corresponding mono-therapies.
First, we determined the minimal effective dose of a Zanamivir mono-therapy that significantly improved survival time of H1N1 and H7N7 infected mice. For the H7N7 lethal infection the minimal effective dose of Zanamivir as mono-therapy ranged between 1 and 3 mg/kg BW/ day (data not shown). Next, we compared the antiviral activity of carrageenan (12 mg/kg BW/ day) and Zanamivir (1 and 3 mg/kg BW/day) mono-therapies with the respective combination versus placebo treatment. Survival rates of mice with treatment starting 24 hpi are shown in Fig 2A. All placebo treated mice died between day 7 and 9 and also in all mono-therapy groups 100% lethality was observed until day 15. In contrast, the combination therapies led to 50% and 90% survival, depending on the Zanamivir concentration. Statistical analysis showed that the Zanamivir mono-therapy 1 mg/kg BW/day did not show a significant benefit (p = 0.1810), whereas the mono-therapy with 3 mg/kg BW/day significantly increased the survival rate compared with placebo treated mice (p = 0.0016). Both Zanamivir concentrations experienced significant benefit in survival by the combination with carrageenan (p<0.0001). Similarly, the combination therapies resulted in remarkably increased survival (p = 0.0421 for 1 mg and p<0.0001 for 3 mg/kg BW/day) when compared to the carrageenan mono-therapy. No statistically significant difference was observed between the combination containing 3 mg/kg BW/day Zanamivir and that containing 1 mg/kg BW/day (p = 0.0525). However, a trend for an increased survival rate with the higher Zanamivir concentration was evident. Therefore, for further investigation the combination therapy containing 3 mg/kg BW/day Zanamivir was evaluated in lethally H7N7 infected mice.
Next, the therapeutic potential of the combination with a delayed therapy start 48 or 72 hpi versus placebo treatment was explored. The survival rates of mice are shown in Fig 2B. All placebo treated mice died until day 10 and also in the group with the treatment start 72 hpi 100% lethality was found. In contrast, the combination therapy starting 48 hpi provided a statistically significant enhanced survival rate in comparison to placebo-treated mice (p = 0.0010).
In summary, the combination of two effective, established mono-therapies resulted in a significantly enhanced survival in lethally H7N7 infected mice. Additionally, the combination therapy was highly efficient in comparison to placebo treatment even after a treatment onset up to 48 hpi.
Intranasal therapy with carrageenan and Zanamivir starting 72 hpi significantly protects lethally influenza H1N1(09)pdm infected mice Next, the minimal effective dose of Zanamivir used as mono-therapy was evaluated in a lethal H1N1(09)pdm mouse model, following the same scheme as described in the H7N7 experiments. The lowest effective dose of Zanamivir after a treatment start 24 hpi was 1 mg/kg BW/ day and its combination with carrageenan was highly effective (data not shown). In the following experiment the therapeutic potential of the combination with a therapy start 48 or 72 hpi was investigated in comparison with the respective placebo treatment.
As shown in Fig 3, the survival rates of mice treated with the combination therapy were highly significantly increased in comparison to the placebo group (p<0.0001). There was no difference in survival between the two therapy starting points, 48 or 72 hpi, which both resulted
We investigated the antiviral effect of a combination of carrageenan with the NI Zanamivir in cell culture studies and in mouse influenza infection models. We have previously shown that a combined therapy of iota-carrageenan with the NI Oseltamivir led to significantly enhanced survival in mice infected with H1N1 PR/8/34 in comparison with the respective mono-therapies [37] . However, Oseltamivir is an orally administered prodrug, which has to be converted into its active form by metabolic processing. Therefore, a further development of a combination nasal spray was not possible with Oseltamivir. Instead Zanamivir-a NI that is applied as active drug-was chosen for the development of a compound combination.
During the evaluation process we found that the binding efficiency of different carrageenan subtypes on different influenza strains varies. The combined use of iota-and kappa-carrageenan for the treatment of lethally influenza infected C57Bl/6 mice revealed a better therapeutic effect than the use of iota-carrageenan alone (S1 Fig). Thus, to provide a broader spectrum of activity against different influenza virus strains, a mixture of iota-and kappa-carrageenan (designated as carrageenan) was used for further evaluation.
For investigation of the effect of a compound combination of carrageenan and Zanamivir, we examined their inhibition efficiency, individually and in combination, against influenza viruses in an adapted plaque reduction assay with semi-liquid overlay in MDCK cells. The combination showed a synergistic inhibition of virus replication in in-vitro assays with all tested influenza viruses (Fig 1) . This indicates that the physical interaction of the polymer with the virus does not disturb the inhibition of the neuraminidase by Zanamivir. This was confirmed in in-vitro tests examining a potential influence of the polymer on the neuraminidase inhibiting activity of Zanamivir (data not shown). Hence, the observed synergistic effect is based on the combination of two distinct underlying mechanisms. As a result, in the proposed combination both mechanisms would complement each other to provide more efficient protection against a broader spectrum of influenza virus strains than the individual compounds.
The synergistic effect was also shown in lethal mice models (Fig 2 and Fig 3) . The pathogenicity of influenza viruses in mice varies and is dependent on the strain and its adaptation to the host. Depending on virus dose and strain, influenza viruses can induce lethal infections in certain mouse strains usually within two weeks [37, 63] . In our model, C57Bl/6 mice are challenged intranasally with a lethal dose of the respective virus and treated with different regimens in comparison to a vehicle control (placebo). In such a model, early virus replication takes place in the upper respiratory tract. From there, virus spreads to the lung and causes lethal pneumonia. The effect of the treatment on mortality is assessed in comparison to placebotreated control mice. Of all in-vitro tested influenza strains the H1N1(09)pdm and the LP H7N7 are particularly interesting for two reasons. First, they are highly relevant pathogens, as placebo or with the mono-therapies consisting of carrageenan (12 mg/kg BW/day) or Zanamivir (1 and 3 mg/ kg BW/day) or a combination thereof. Treatment started 24 hpi and continued for 5 days. (B) Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and intranasally treated twice per day either with placebo or a combination of carrageenan with Zanamivir (3 mg/kg BW/day). Treatment started either 48 hpi or 72 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated uninfected control mice showed 100% survival in both experiments (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. both are involved in recent influenza outbreaks. The H1N1(09)pdm is associated with more than 18,400 deaths in the season 2009/2010 while the LP H7N7 carries an HA closely related to that of the avian influenza H7N9 virus which has caused more than 175 deaths until October 2014 [64] . Second, they are of special interest for the carrageenan/Zanamivir combination approach. They have shown to differ in in-vitro susceptibility to carrageenan, Zanamivir (Table 1 ) and the combination thereof (Fig 1) . While H1N1(09)pdm was highly sensitive to inhibition by both substances alone, H7N7 required much higher concentrations of carrageenan and Zanamivir, respectively, to achieve similar inhibition efficiencies. Therefore, both virus strains were chosen to further explore the efficiency of the combination therapy in a mouse model.
We established lethal mouse models with both viruses that resulted in 6.8 and 8.5 mean survival days for LP H7N7 and H1N1(09)pdm, respectively. These results are in good accordance to similar already published lethal influenza models [65] [66] [67] . In our models the lowest effective dose for Zanamivir at a treatment start 24 hpi was found to be between 1 to 3 mg/kg BW/day for both viruses. This concentration range is relatively high in comparison to other published studies. However, these studies were done under anesthesia with different viruses and a prophylactic therapy start [65, 66] . The fact that a higher dose of NI is needed for an effective treatment when the therapy starts 24 hpi is already known for Oseltamivir [68] . Nonetheless, also data with much higher effective concentrations (10 mg/kg BW/day [69] ) and with similar concentrations of Zanamivir (2.5 mg/kg BW/day [67] ) were published as well.
We found that the combination of carrageenan with 3 mg/kg BW/day Zanamivir used for treatment of H7N7 infected mice resulted in significantly enhanced survival of mice in comparison to both mono-therapies (Fig 2) . The significantly enhanced survival compared to the placebo treated group was also found after a delayed treatment start 48 hpi. Furthermore, in the H1N1(09)pdm model the combination of carrageenan with 1 mg/kg BW/day Zanamivir showed statistically significant enhanced survival in comparison to placebo treatment even after a treatment start 72 hpi. This is a remarkable finding since NIs are normally not effective when applied 72 hpi.
The finding supports the development of the Zanamivir and carrageenan combination approach. As the intranasal treatment regime is incapable to effectively treat virus infections of the lung, the primary target of such a product is the prophylaxis and therapy of uncomplicated influenza. Since the majority of influenza infections causes uncomplicated illnesses and practically all cases of influenza start with an infection of the nasal cavity or the upper respiratory tract, the therapeutic potential is huge. However, clinical studies are required to elucidate and demonstrate the potential of the proposed combination therapy.
Combination of antiviral strategies has led to impressive achievements in the combat against other viral disease like HIV. In particular the problem of antiviral resistance could be addressed with this strategy. In the last decade concerns have been raised about the increased emergence of Oseltamivir resistant influenza viruses. The augmented appearance of viruses carrying the mutation H275Y in the neuraminidase of H1N1(09)pdm viruses that confers resistance to Oseltamivir left Zanamivir as only treatment option for symptomatic patients infected with an Oseltamivir resistant influenza strain [70] . In contrast to Oseltamivir, resistance to Zanamivir is less frequent. To date, Zanamivir resistant influenza has been detected only once, in an immunocompromised patient [71, 72] . However, lessons should be learned from previous anti-influenza interventions which resulted in occurrence of resistance against currently approved drugs [73] . Therefore, concerns are comprehensible that an increased Zanamivir use may also lead to the rapid emergence of resistances [74] . To overcome this threat, a combination of antivirals which inhibits virus replication by distinct mechanisms is a valid strategy. We checked for the possibility of generating double compound escape mutant viruses while passaging viruses in the presence of increasing concentrations of compound combinations. After 10 passages in MDCK cells no resistance to the compound combination for any tested influenza virus could be found (data not shown). However, this finding does not guarantee that emergence of Zanamivir escape mutants can be completely halted.
In summary, we demonstrated that the anti-influenza mechanisms of both single compounds complement each other. The combination provides synergistically better protection against a broader spectrum of influenza viruses than the individual compounds.
A nasal spray containing carrageenan together with Zanamivir provides an easy to apply treatment of upper respiratory tract infections in patients under suspicion to be influenza infected. Patients would benefit from the fast and efficient treatment of uncomplicated influenza in the upper respiratory tract. Due to the faster influenza virus clearance from the upper respiratory tract and the independent antiviral mechanism of carrageenan and Zanamivir the likelihood to develop escape mutations against Zanamivir will be reduced. Both individual compounds are able to reduce severity and/or duration of the influenza illness and a combination is expected to work similarly. Additionally, due to the broad antiviral effectiveness of carrageenan, patients will receive in parallel a treatment of concomitant viral infections. Therefore, patients will benefit from a decreased probability to develop complications. In consideration of the complications known to accompany an influenza virus illness this combinational therapy meets an urgent medical need.
A second scope of this combination is the protection against newly emerging pandemic viruses during the time until identification of the virus followed by manufacturing and distribution of vaccines [43] . Even if, due to new reverse genetic techniques, less time for production of vaccines is needed, it still takes months before large quantities of vaccine are available [75] . During this time the human population should be protected to decelerate viral spread. At the moment the only available opportunities for personal protection are hygiene measures and the use of Tamiflu (brand name of Oseltamivir).
Novel protection and treatment options for influenza are desperately needed. Based on our encouraging results in mice we suggest testing a nasal spray containing carrageenan in combination with the neuraminidase inhibitor Zanamivir in clinical trials for prevention or treatment of uncomplicated influenza infections.
Supporting Information S1 Fig. Therapeutic efficacy of iota-carrageenan solely or together with kappa-carrageenan in influenza H7N7 lethal infected mice. Mice (n = 20 per group) were lethally intranasally infected without anesthesia on day 0 and accordingly intranasally treated twice per day either with placebo or with iota-carrageenan or with a mixture of iota-and kappa-carrageenan. Treatment started 24 hpi and continued for 5 days. On the y-axis the survival of mice [%] and on the x-axis the time post infection [days] is given. Placebo treated, uninfected control mice showed 100% survival (data not shown). Statistical analyses were conducted using log rank test and are shown beneath the graphs. Values of p<0.05 were considered statistically significant; non-significance (n.s.) was obtained with p-values >0.05. (TIFF)
|
What are common concamitant infections during the course of influenza infection?
| 2,154 |
human rhinovirus (hRV), respiratory syncytial virus, adenovirus, human coronavirus, human metapneumovirus and parainfluenza virus
| 3,641 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.