Cellular microRNA networks regulate host dependency of hepatitis C virus infection

Cellular microRNAs (miRNAs) have been shown to regulate hepatitis C virus (HCV) replication, yet a systematic interrogation of the repertoire of miRNAs impacting HCV life cycle is lacking. Here we apply integrative functional genomics strategies to elucidate global HCV–miRNA interactions. Through genome-wide miRNA mimic and hairpin inhibitor phenotypic screens, and miRNA–mRNA transcriptomics analyses, we identify three proviral and nine antiviral miRNAs that interact with HCV. These miRNAs are functionally linked to particular steps of HCV life cycle and related viral host dependencies. Further mechanistic studies demonstrate that miR-25, let-7, and miR-130 families repress essential HCV co-factors, thus restricting viral infection at multiple stages. HCV subverts the antiviral actions of these miRNAs by dampening their expression in cell culture models and HCV-infected human livers. This comprehensive HCV–miRNA interaction map provides fundamental insights into HCV-mediated pathogenesis and unveils molecular pathways linking RNA biology to viral infections.

1. Line 45. The statement that miRNAs plays a role in virus-host interaction requires a reference. 2. Line 384, the inhibitors did not show the opposite effects on virus replication (supplementary results, figure10 a)? Please address it in the text. Transfected mimics for miR130a,130b and 301a led to a decrease in core staining, however their equivalent inhibitors did not increase core staining. 3. The decrease of miR130a was very low in cells ( fig. 6d) but a bit bigger in patients ( fig.  6e), do the authors have any explanation for this discrepancy? 4. Line 454. The authors should explain their speculation on page 21 that miR130a represses PPARγ at the post-transcription level to modulate LD synthesis. It is not immediately obvious that their conclusion follows from their observation. 5. Line 84. How many replicates were performed for the screen? The authors should comment about this detail in the text. 6. On lines 162-168, the authors essentially repeat their statements from the previous section. 7. Line 179. Please add reference, after your statement that host microRNAs modulate HCV infection. 8. Line 179. The authors state that "A multitude of host miRNAs that modulate HCV infection in hepatocytes have been uncovered…" Is the statement referring to the data in the manuscript, or previously published data. If the latter, then a reference is required. In addition, the section should be compressed with the previous one for clarity. 9. What do the authors mean in lines 193-194? The authors state that "these results reveal a unique feature of HCV-host interaction through reciprocal regulation…", what do the authors mean by "reciprocal regulation"? 10. Line 198, the authors state that …"biologically relevant miRNAs", what do the authors mean by "biologically relevant"? 11. Line 204, the authors state that… "productive infection", it is not clear what "productive infection means in the context of an in vitro model. 12. "Pan antiviral"? the authors describe results with 2 viruses (HCV and VSV). The statement of pan antiviral therapy may be a bit of an overreach. 13. Line 223, "…that preferentially target…' should be corrected to read "…as preferentially targeting…". 14. Line 233, please re-write. A "map" does not "elicit". 15. The text would benefit from the removal of words such as "drastically", "considerably", "barley", "interesting", etc … (for instance one line 241 and 244, and more). 16. Lines 254-256, there is an association between cellular microRNAs to antiviral effect, however causality has not been demonstrated in the manuscript. 17. Line 268, please change to "Thus, by overlapping the predicted target genes generated from both algorithms, …". 18. Please specify on line 275, at what time point the cells were transfected with miR-25 mimic? How did the authors set the 1.5x threshold? How many replicates were used in the experiment? 19. Please specify on line 279, by how much miR25 candidate targets were downregulated? 20. Line 289, Not necessarily direct. The observed phenotype could be due to an indirect effect on the target genes. 21. Line 300, are p-values for the enriched pathways FDR corrected? Ingenuity does not necessarily report FDR corrected values. 22. In line 308, "most abundantly" in hepatocytes, relative to other cell types? or relative to other let7 microRNAs? 23. Please explain in line 314, "HCV propagation" what do the authors mean by propagation? Multiple rounds of infection, or something else? 24. Lines 314-316, what is the interpretation of the downregulation of let7 in HCV infected cells (figure 5e,f)? is this mediated by virus, or a host response to infection? 25. Lines 318/340, what is the algorithm? Please specify in the text as it is hard to assess the validity of its application. 26. Line 369, "hairpin inhibitors", please add an "s". 27. Line 371, please substitute "can" to "may". 28. Lines 375-380, please review for clarity. 29. Lines 394-397, please review, the sentence is far too long. 30. Line 411, change "considerably" to "led to"? 31. Lines 421-422, the authors state that "bioinformatics tools revealed…" at what confidence interval? Were other predictions made? How many genes were in the list of predictions? As miRNA's are highly promiscuous, this can be quite relevant. 32. After line 443, please comment on whether cell viability is affected. In fact, throughout the manuscript, there is not mention on the affect of miRNA's tested on cell viability.

Reviewer #3 (Remarks to the Author):
In this manuscript Li et al. describe a comprehensive screen of oligonucleotide miRNA mimic and inhibitors to systematically identify miRNAs that can alter HCV infection. The manuscript describes a lot of work and provides a subset of higher confidence miRNAs that at least indirectly regulate HCV infection as well as partial mechanisms that may account for some of this regulation. Although there is no real complete story regarding virus life cycle or host response to HCV, this large dataset will likely be a useful reference for future studies of HCV and miRNAs. Critiques: 1. The manuscript is very long and at times the flow of prose is disjointed. Portions of the Discussion are redundant with the Results section. The paragraph in lines 178 -194 is outof-place.
2. The studies claiming to identify direct targets of miRNAs lack the important control of mutating the putative docking sites in the 3'UTR targets to show this ablate regulation.
3. The mimic studies are almost certainly performed at super-physiological levels compared to endogenous miRNA expression. No estimate for copy number of the mimics is provided. This concern is muted by the priority given to those miRNAs that score inverse in the mimic/inhibitor screens. Still, the strength of evidence that these miRNAs affect HCV infection (as opposed to off-target effects from synthetic oligonucleotides) would be improved by using DNA vectors-based miRNA expression for a few of the top-studied miRNAs.
4. Although control miR-122 is discussed in depth, consider describing the role of the top miRNAs to emerge from this screen in other viral contexts. For example, the miR-17-93 family has been shown in several previous studies to be pro-viral, perhaps by suppressing the effects of IFN & NFKappa B (PMIDs: 20643939, 24075860, 27512057).
We greatly appreciate the reviewers' insightful and constructive comments regarding our earlier version of the manuscript. We have significantly revised the manuscript per the reviewers' suggestions, and are providing a point-by-point response to the reviewers' comments (in blue) below.
Reviewer #1 (Remarks to the Author): Using genome-wide miRNA mimetic and inhibitor screens, the authors identified miRNAs that have proviral and antiviral functions in the HCV life cycle. Detailed analyses pointed to roles for identified miRs in the viral life cycle, and several putative host target mRNA were identified. This is a comprehensive analyses that, unfortunately, suffers from subtle effects of the miRs and the huge number of identified target mRNAs. The only solid effect on HCV entry and viral mRNA translation was described for let7a ( Fig. 5). Here effects on viral yield and claudin protein production are impressive. Effects of other miRs are subtle, i.e. two-fold at best (Fig. 3a).

Response:
We appreciate the reviewer for pointing out an important perspective regarding the biological relevance of the identified HCV-associated cellular miRNAs. In this study, we pursued the truly physiologically relevant miRNAs through multiple strategies. First, we conducted an effective genome-wide miRNA hairpin inhibitor screen (which masks the functions of endogenously expressed miRNAs) and identified a multitude of miRNAs that regulate HCV infection in hepatocytes. The inhibitor screen was then cross-referenced with an unbiased whole-genome miRNA mimic screen, producing 31 miRNAs that exhibited opposite phenotypes in the loss-of-function and gain-of-function assays (see Fig. 1). Next, to identify whether these 31 phenotypespecific cellular miRNAs are physiologically relevant to HCV, we elucidated miRNA expression landscapes in Huh7.5.1 cells and primary human hepatocytes (PHHs) by conducting NanoString nCounter and microarray-based transcriptome analyses (see Suppl. Tabs. 3-5). Twelve miRNAs that are abundantly expressed in hepatocytes were confirmed to modulate HCV infection in a biologically relevant manner. The study flow is illustrated in Fig. 2b. Interestingly, the combined mimic/inhibitor screen and transcriptome analyses also revealed previously unrecognized mechanism of HCV to subvert many of these biologically relevant anti-HCV miRNAs by downregulating their expression in cultured hepatocytes (see Figs. 2d, e and S5) and livers of chronic hepatitis C patients (see Figs. 4e, 5f and 6e). While some of the effects reported here may not be dramatic (2-fold), all the data are solid and highly significant. The less dramatic effects could also be due to variations in experimental reagents and conditions. Collectively, these findings unveil a unique, intrinsic feature of HCV-host interactions through reciprocal regulation between the virus and associated cellular miRNAs.

While the authors quantitate the amount of miRs in uninfected and infected cells,
what is need is the abundance of mimetics per cell. In the absence of this data, it is difficult to judge whether the observations are physiologically relevant.

Response:
We thank the reviewer for the constructive advice regarding quantification of the abundance of mimetics per cell. We agree that this measurement is instrumental in defining whether the observed phenotypes of a miRNA are physiologically relevant in hepatocytes. Indeed, the nCounter-based Nanostring analysis is reported in count per cell (see Tables S3 and S4). We apologize for not mentioning it clearly in the previous manuscript. We have stated it explicitly in the revised manuscript.
2. Many target genes were predicted for the identified miRs. Based on bias, a few were tested. While those were regulated by the miRs, effects on HCV were not rigorously tested. For example, did the employed siRNAs cause cell toxicity?

Response:
We respectively disagree with the reviewer that the selection of the tested target genes in this study are biased. We applied a combined functional, transcriptomics and bioinformatics-based algorithm that systematically identified a complete list of phenotype-specific targets for each miRNA (miR-25, let-7 and miR-130), which underwent further rigorous validation and analyses. The flow of the algorithm is elicited in Figs. 4f, S9a and S12a and described in detail in the manuscript (Pages 12~13).
Particularly, these "phenotype-specific miRNA targets" are all previously confirmed HCV host dependencies uncovered from our genome-wide siRNA functional screen (Li et al., 2009 PNAS) or other published studies. Their effects on HCV life cycle have been rigorously defined in our recent functional genomics studies (Li et al., 2014 PLOS Pathogens). In this study, we conducted multiple HCV life cycle assays and further validated the effects of these miRNA targets on various stages of HCV life cycle (see Figs. 6d-g, S9d, e, and S12d, e). The employed siRNAs did not cause appreciable cytotoxicity in these studies as well as in previous functional screens (Li et al., 2009 PNAS;Li et al., 2014 PLOS Pathogens).
3. Ago-Clip data (Luna et al.) could be used to compare predicted miR-target mRNA interactions in uninfected and infected cells. Maybe this analysis will reveal additional, relevant target mRNAs.

Response:
We appreciate the reviewer for this valuable suggestion. Indeed, we have compared the identified HCV-associated miRNA targets with the published Ago-Clip & HiTS data (Luna et al., 2015 Cell), which revealed that the majority of the functionally validated targets in our study are confirmed by the HiTS-CLIP. These include SUV420H1 for miR-25; PPIA, IQCB1, IGF2BP1 and CLDN1 for let-7a; and DDX6, NPAT, LDLR, HCCS and INTS6 for miR-130a. We have mentioned this in the "Discussion" section (see Pages 27~28 of the manuscript).
Nevertheless, individual miR-mRNA interactions cannot be reliably examined applying the Ago-CLIP database, as many other factors may affect the mRNA expression levels of predicted targets during HCV infection. The analysis performed by Luna et al. examined the predicted targets as sets of genes, and did not confirm individual genes as targets of miR-122. Given the wide variety of factors that influence the expression level of a given gene when a cell is infected, a specific analysis of individual mRNA levels would be inconclusive in evaluating the effects of these miRNAs. As mentioned above, we did use the database qualitatively, to assess whether the miR-mRNA interactions we identified are also found using the Ago-Clip method, and have described the results of this qualitative analysis of the database in the text.
4. The discussion is too long.

Response:
We thank the reviewer for pointing out this issue. The discussion section has been significantly condensed and edited for clarity.
Reviewer #2 (Remarks to the Author): In "Cellular microRNA networks regulate host dependency of hepatitis C virus infection", the authors perform a genome-wide miRNA screen for possible roles in HCV infection. The authors transfected 972 different miRNA mimics, and 970 hairpin inhibitors and affects on HCV viral infection was quantified using HCV core protein expression analysis. As a counter screen supernatants were monitored for late stage viral replication. The authors identified 276 mimics and 153 hairpin inhibitors as 'hits' and of those, 31 miRNAs had opposite phenotypes in both gain and loss of function experiments. Of these, three pro-viral miRNAs and 9 antiviral miRNAs were transcriptionally regulated during infection and their affects on HCV infection confirmed through a HCVcc-Luc infection assay. In particular, the authors identify proviral miR17-5p and antiviral miR-25 as targeting late stage HCV assembly and secretion. In addition, let7a was shown to be associated with host response to HCV infection. Finally, the miR130 family was shown to regulate HCV replication and assembly and the authors identified several possible targets for miR130.
While the manuscript provides an important catalog of miRNAs that can influence HCV infection, many of the affects reported are rather modest and the biological relevance to clinical disease remains to be validated and explored. More importantly, the text would benefit from judicious editing and careful rewrite for clarity. The authors should address the concerns below prior to publication.
Major concerns: 1. The authors compared the expression of miRs in patients and in healthy individuals in addition to their cell culture work. There is not comment or assessment as to the purity of the samples. i.e. Contamination of other tissues or various cell types could be a concern especially in individuals with clinical disease (where fibrosis and fat content can vary significantly). Normalization to tissue specific house-keeping genes would greatly alleviate such concerns and potential biases in miRNA expression.

Response:
We thank the reviewer for this insightful suggestion. We have performed two additional analyses to address this concern.
First, we analyzed miRNA expression in patients striated by Ishak score, an indication of the severity of liver fibrosis. As shown in the new figures of the revised manuscript (see Figs. S6d, S8f, and S11f), the presence or extent of liver fibrosis does not affect the expression profiles of miR-130a, let-7a, and miR-25. We hope these data would alleviates the concern that tissue/cell type "impurity" would affect the abundance of hepatic miRNAs.
As suggested by the reviewer, we also examined several hepatocyte-specific markers in the liver tissues of both healthy donors and chronic hepatitis C (CHC) patients, to address the concern that there may be fewer hepatocytes (and more immune cells) in the CHC patient biopsies. We demonstrated that the expression levels of various hepatocyte-specific markers including CYP1A2 and HNF6 are actually slightly higher in the CHC patient biopsies than the healthy donor biopsies (but not significant, see data below). As such, we conclude that the observed HCV-mediated decreased expression of these miRNAs was not due to dilution by other cell types in the liver samples.
Minor concerns: 1. Line 45. The statement that miRNAs plays a role in virus-host interaction requires a reference.

Response:
We thank the reviewer for the helpful suggestion. We have added a couple of references that specifically review virus-miRNA interactions.
2. Line 384, the inhibitors did not show the opposite effects on virus replication (supplementary results, figure10 a)? Please address it in the text. Transfected mimics for miR130a,130b and 301a led to a decrease in core staining, however their equivalent inhibitors did not increase core staining.

Response:
We agree with the reviewer that overexpressing miR-130a or miR-130b hairpin inhibitors did not considerably increase HCV core staining, comparing with the sharp inhibitory effects of their mimic counterparts (Fig. S11a). Nevertheless, the core staining data were obtained from the genome-wide screens, which are less quantitative and can generate false-negative results. Hence, we further conducted HCVcc assays by quantifying intracellular and extracellular HCV RNA levels upon miR-130a/b mimic or inhibitor transfection. We found that both miR-130a and miR-130b hairpin inhibitors, when overexpressed in hepatocytes, significantly enhanced HCV infection (see Figs. 6b and S13a).
Interestingly, inhibition of miR-130a seems to be more efficient than miR-130b inhibition in inducing productive HCV infection. We thus stated in the manuscript that "the effect of inhibitors of less abundant family members (e.g. miR-130b) was less dramatic than that of the more highly expressed family member (i.e. miR-130a). This is likely due to the continued function of the highly expressed miR-130a when the less abundant miR-130b is inhibited." In general, transfecting many miRNA hairpin inhibitors in cells exerted less dramatic effects on HCV infection than the effects of their equivalent mimics. We have attributed this difference between miRNA mimics and inhibitors to their different modes of actions, and have addressed this point in various parts of the manuscript (e.g. in the paragraph of Page 7).