Essentiality and dynamic expression of the human tRNA pool during viral infection

  1. Noa Aharon-Hefetz
  2. Michal Schwartz
  3. Orna Dahan
  4. Noam Stern-Ginossar
  5. Yitzhak Pilpel

Reviewed by

Read the full article See related articles

Discuss this preprint

Start a discussion What are Sciety discussions?

Listed in

Log in to save this article

Abstract

Human viruses depend on the translation resources of the host cell. A significant translation resource is the tRNA pool of the cell, as human viruses do not encode tRNA genes. Through tRNA sequencing, we inspected the human tRNA pool upon infection of human Cytomegalovirus (HCMV) and SARS-CoV-2. HCMV-induced alterations in tRNA expression were predominantly virus-driven, with minimal influence from the cellular immune response. Notably, specific tRNA post-transcriptional modifications appeared to modulate stability and were susceptible to HCMV manipulation. In contrast, SARS-CoV-2 infection did not significantly impact tRNA expression or modifications. We compared the codon usage of viral genes to the proliferation-differentiation codon-usage signatures of human genes. We found a marked difference between the viruses, with HCMV genes aligning with differentiation codon usage and SARS-CoV-2 genes reflecting proliferation codon usage. We further found that codon usage of structural and gene expression-related viral genes displayed high adaptation to host cell tRNA pools. Through a systematic CRISPR screen targeting human tRNA genes and modification enzymes, we identified specific tRNAs and enzymes that improve or reduce HCMV infectivity and cellular growth. These findings highlight the dynamic interplay between the tRNA pool and viral infection dynamics, shedding light on mechanisms governing host-virus interactions.

Article activity feed

  1. Review Commons

    Note: This response was posted by the corresponding author to Review Commons. The content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    Reviewer #1 (Evidence, reproducibility and clarity (Required)):

    I have mixed feelings regarding this manuscript. On the one hand, the authors did an impressive amount of work. On the other hand, the manuscript seems overly descriptive (writing should be more concise) without a clear message or hypothesis that is cohesive to all the presented evidence. Below, I will outline my concerns.

    We appreciate the comment about missing a cohesive presentation. We worked extensively to improve that in the revised manuscript.

    Reviewer #1- first part

    I am not an expert in the field of viral biology and immunology. I wonder how well the IFN treatment mimics the cellular response to infection (yet without the virus). Also, how good is ruxolitinib at blocking the IFN response ? I would appreciate it if you could explain both with one or two sentences and provide the necessary references.

    The reviewer is correct that we cannot claim that interferon treatment mimics exactly the cellular response. However, the expression of interferon-stimulated genes (ISGs) is a major arm of the antiviral response to HCMV (c.f. doi:10.3390/v10090447, doi:10.2217/fvl-2018-0189). In addition, Ruxolitinib is a potent and selective Janus kinase 1 and 2 inhibitor (doi:10.1021/ol900350k), and we have shown in the past that it very effectively reduces the expression of many ISGs (doi: 10.1038/s41590-018-0275-z). Since ISGs constitute a major part of the host response to HCMV infection, the fact that their expression leads to minor changes in the tRNA pool strongly suggests that it is mainly the virus (as opposed to the host cell) that mediates the changes seen in the tRNA pools during HCMV infection. In the revised version, these claims were amended, and relevant references were added (pages 5, lines 132-136).

    (MAJOR) Can these two treatments really allow the effects of host response and viral infection to be separated? OR in other words, are these two effects really orthogonal? In my opinion, they are NOT. Fig. 1E seems to support my opinion, as the changes seen for the "IFN" sample relative to the "uninfected" sample (referred to as "changes-A" below), are parallel to the changes seen for the "24hpi + ruxo" sample relative to the "24hpi" sample ("changes-B"). More specifically, changes-A represent the host response, as argued by the author, whereas changes-B represent the elimination of the host response (due to ruxo, conditioned on the virus-driven effect). If the virus-driven effect and the host response could really be separated, one would expect changes-A and changes-B are more or less opposite. However, they appeared to be parallel, suggesting that uninfected versus infected conditions can have totally different (even opposite) host responses. More importantly, if one cannot separate the host response from virus-driven effects, the conclusion of "tRNA changes are driven by virus, not host response" is then unfounded.

    This is an important point to clarify. Changes-A indeed represent the effect of the host antiviral response on the tRNA pool. Changes-B, however, represent a mix of two effects. 1: counteracting the effect of the host antiviral response on the tRNA pool, which we show is a minor effect, and 2: The enhanced effect of the virus, since ruxolitinib, by inhibiting the host antiviral response, enhances the viral infection. It may indeed be that both the virus and the host antiviral effects are in the same direction. However, it is clear that the antiviral effect is minor. Thus, it is likely that the second effect of ruxolitinib (i.e., allowing enhanced viral infection) is the more substantial one. Therefore, it seems as though the viral effect and the elimination of the host effect are in the same direction. This point was clarified in the revised version (page 6, lines 145-146).

    Even if we let go of this previous point and accept that these results indeed offer some support for the notion that the virus-driven effect are the main contributor to the shifts in tRNA pool, the support is at best moderate. A big gap here is "how?" I suggest the authors should at least give some insight on how virus can do that in Discussion (and mention it with one sentence in Results).

    We certainly welcome the challenge, which we now meet in the revision. In short, here, transcription regulation of tRNAs, mainly upon viral infection, is poorly studied. Unlike other herpesviruses, HCMV does not cause a host shut-off of the host transcripts. Upon HCMV infection, the tRNA transcription machinery is upregulated significantly, which probably contributes to the upregulation in pre-tRNA (doi.org/10.1016/j.semcdb.2023.01.011). However, it is still unknown what the viral factors are that promote upregulation in the tRNA transcription machinery. We now relate to this point in the results (page 6, lines 147-148) and discuss the known effects of viral infection of tRNA expression in the discussion section (page 15, lines 447-451).

    The authors compared the HCMV codon usage to the proliferation and differentiation signatures of human cells. But these two signatures are not compared with measured tRNA expression. It might shed some light on the general characteristics of tRNA pool shifts due to infection (towards a proliferation-like or differentiation-like signature). This fits in the general topic of virus-host interaction and might give more evidence for the point that HMCV is adapted to a differentiation signature (as it drives the host into that state).

    We performed the analysis suggested by the reviewer. We found that the tRNA pool of uninfected HFF cells correlated to the same extent with proliferation codon usage (r=0.29, p-value=0.029) and differentiation codon usage (r=0.26, p-value=0.05). Similar correlations to the proliferation and differentiation signature were found when analyzing the tRNA pool 72h post-infection (proliferation r=0.33, p-value=0.011, differentiation r=0.28, p-value=0.034). This result suggests no general shift in the tRNA pool towards a specific codon usage signature.

    How is the dashed box in Fig3A/B chosen?

    We determined the dashed lines based on the most prominent groups of transcripts best adapted to proliferation or differentiation codon usage signatures. Figure S3A clearly shows the two groups without viral genes. We emphasize this point in the legend of Figure S3A (page 36, lines 1157).

    The tAI values shown in Fig3C-E are extremely low (compared to other reports I am aware of). Does this mean that the adaptation of viral codon usage to human cell supply is actually very weak? This is in opposition to the major claims made in this section.

    We acknowledge that the tAI values presented here are lower than typically presented. However, this is due to how tAI was calculated rather than the potential weak adaptation between viral genes and tRNA supply. Specifically, unlike previous works that estimate tRNA availability based on tRNA gene copy number, here we calculated tAI using tRNA sequencing (in order to capture the dynamics in the tRNA pool during infection). Indeed, the value of tAI calculated by tRNA read counts is lower than tAI calculated by tRNA copy number. This is due to the skewed distribution of tRNA read counts (some tRNAs are highly expressed, and others are lowly expressed), while tRNA copy number is distributed more evenly. Thus, due to the mathematical nature of the tAI (computing geometric rather than arithmetic average of tRNA availability), the skewed distribution observed in the data results in lower tAI values. When computing tAI based on gene copy number, we get higher tAI values (0.3 on average). Nevertheless, as all tAI calculations here were done similarly, the comparisons between gene groups or genes are valid.

    I believe that the part about SARS-CoV-2 could be made more concise. It is sufficient to mention that results may differ from those obtained with HCMV in one paragraph.

    The section on SARS-Cov-2 is now made rather succinct. This virus is mainly given as a comparison to the primary virus studied in this paper - HCMV.

    Line 299 on page 11 - I do not believe codon usage between different viruses can be directly compared, let alone reaching such a conclusion. Some viruses have low CAI or tAI to humans, but they have co-evolved with humans for a long time. Furthermore, there are viruses that infect multiple hosts, but their CAI for a host with which they have long co-evolved is higher while their CAI for a host that is relatively new is lower.

    We agree with the reviewer that a direct link between co-evolution time and tAI may not always exist. Indeed, other factors might explain the observation that SARS-CoV-2 genes are less adapted than HCMV genes. These may include effective population sizes and mutation rates that vary substantially. We, therefore, removed this conclusion from the manuscript.

    (MAJOR) A more general comment is that there is a difference between tRNA expression and the abundance of translation-ready tRNA. The process of charging tRNA with amino acids may take a long time. It is the abundance of the charged-tRNA (the ternary complex of aminoacylated tRNA and EF-Tu-GTP) that is of biological importance. In this regard, the use of tRNA expression falls short.

    The reviewer raises a valid point. Indeed, our tRNA sequencing protocol measures both charged and uncharged tRNAs that constitute the cell's mature tRNA pool. Compared to previous studies that focus on the transcription process of tRNAs in viral-infection models by sequencing the pre-tRNAs, here we look at the mature tRNA pool that accounts for both transcription and post-transcription processes. Therefore, we changed the use of "tRNA expression" to "mature-tRNA levels" and "highly" or "lowly-abundant tRNAs" rather than “highly” or “lowly expressed tRNAs” in the manuscript. We note, however, that although limited in the ability to differentiate between charged and uncharged tRNAs, the tRNA sequencing protocol used here is commonly used and validated as a state-of-the-art protocol in tRNA sequencing (10.1016/j.molcel.2021.01.028, 10.1038/s41467-020-17879-x, etc.), mainly because it addresses the level of "ready-to-use" tRNA.

    Reviewer #1- second part

    (MAJOR) Prior to the actual competition assay in the first high-throughput screen (cell competition assay), the authors applied two days of antibiotic selection and two days of recovery. This could result in a serious problem of false negatives or drop outs. Specifically, an sgRNA targeting an essential gene with high efficiency would kill the cells, leaving no (or a small number of) cells in the ancestor population at the beginning of the competition process. A sgRNA's enrichment in competing populations cannot be reliably estimated in such situations. I am not certain that the FDR used in Figure 5B is sufficient to address this issue. Please clarify whether it could. Providing raw counts for competing and ancestor populations would also be helpful.

    As customary in CRISPR screens, the step of lentiviral transduction and antibiotic selection is necessary to ensure that only CRISPR-edited cells are left in the population. Indeed, essential genes like housekeeping genes are probably removed from the competing population relatively quickly, which might result in their dropouts. We could have lost some tRNA hits in the cell growth CRISPR screen (Figure 5B-C) because of their overall essentiality for cell growth. The MAGcK tool we used, the state-of-the-art in the field, filters out sgRNAs with low read counts to be able to calculate false discovery rates. Indeed, we identified 15 tRNAs that were depleted from the competing cells. We believe that our procedure minimizes the concern of dropouts. tRNA dropout in the HCMV infection CRISPR screen (Figure 6B-C) can also happen, which means our screen underestimates the essentiality of tRNAs to HCMV infection. However, this concern does not affect the significance of the hits we did find. We acknowledge this inherent difficulty in CRISPR screens and will provide the raw read counts of all samples upon full submission. We emphasize, though, that while valid, this concern applies to essentially any CRISPR screen that is commonplace in genomics these days.

    It is also highly questionable to me the nearly negligible effects of tRNA modification enzymes. This may be explained by the point above. Indeed, the dots of tRNA modification enzymes in general appear to have higher FDR (lower y values) when compared to red dots with similar enrichment levels.

    This is a valid point. We found a lack of essentiality of tRNA modification enzymes in both screens. We analyzed additional CRISPR screens and compared the effect of tRNA modification enzyme knockouts relative to the restriction and dependency factors we used in the library. The tested screens included 34 knockout CRISPR screens we downloaded from the BioGRID ORCS database that have similar parameters to our screen. Namely, they all test cell proliferation in a time-course manner, using a pooled sgRNA library and using the MAGeCK tool for data analysis. Overall, the screens use different human cell lines and diverse sgRNA libraries. Although potentially surprising, we found that the lack of essentiality of tRNA modification enzymes was also observed in the analyzed CRISPR screens (Figure S5B and on page 11, lines 322-330, and on page 18, lines 539-541). One potential reason is if some of these enzymes were "backed up" by others, which we mentioned. Another explanation is that most tRNA modification enzymes are indeed not essential for growth and for viral infection (now described in the Discussion, page 18, lines 544-545). Alternatively, dropouts can explain this result, as suggested by the reviewer. To examine the likelihood of the dropout option, we examined the average raw read count of the tRNA modification enzyme in the ancestor samples. We compared it to that of other sub-groups. We found that raw read counts of the tRNA modification enzymes are not different than other sub-groups in the CRISPR library. Thus, the dropout issue cannot explain our screens' lack of essentiality of tRNA modification enzymes.

    The screen based on IE2-GFP labeled HCMV measures a phenotype that is very difficult to interpret. Particularly, I am not sure if GFP2 and GFP3 are good controls for comparing GFP4 (GFP1 might be better). Various factors can affect GFP levels, including, but not limited to, dilution caused by a rapidly dividing host cell, unhealthy translational machinery resulting from infection or microenvironment. My point is supported by some observations in Fig6B. For example, SEC61B, a restriction factor for HCMV infection, is enriched in the GFP2 group, contrary to expectations. It is necessary for the authors to prove with firm evidence that their choice of GFP signal thresholds is appropriate.

    We acknowledge the concern. Specifically, the translation of the GFP gene itself could be affected by the tRNA manipulation done. To account for this potential concern, we tested the codon usage of the eGFP gene (which is the GFP version we used in the system) and compared it with tRNA essentiality, as determined by the cell growth CRISPR screen. We report this in the revised manuscript (page 13, lines 390-392, and added Figure S6D). We found that GFP does not tend to significantly use codons that correspond to essential or less essential tRNAs. The same lack of correlation was also found for the tRNA essentiality upon HCMV infection (not shown).

    More generally, we show that GFP intensity does correlate with viral genome copies (Figure S6A). Also, from mRNA-seq data of temporal HCMV infection (10.1016/j.celrep.2022.110653), IE2 (UL122) shows a dynamic expression- high expression pick in early infection, then a decline in expression level followed by a gradual increase.

    Altogether, we believe that the IE2-GFP level provides a good estimation for viral load.

    Regarding SEC61B, which served as a control in our screen – the referee is rightly asking why it behaves oppositely from what's expected, given that this was supposed to be a restriction factor of HCMV infection. We returned to the literature on the essentiality of this gene upon HCMV infection. In Weissman's paper (10.1038/384432a0), which was the reference for choosing control genes in our system, this gene was targeted through two different CRISPR technologies, once with CRISPR knockout and once with CRISPRi. Interestingly, only upon CRISPRi did this gene prove to be a restriction factor (i.e., improved infection upon reduction of the gene). We comment on this peculiar fact in the revised manuscript (page 13, lines 370-374). However, we note that the rest of our positive and negative controls deliver the expected results – increasing or reducing infection as expected from their role, thus lending considerable support to our experimental system. It is possible, especially in light of our screen, and since other positive and negative controls behave as expected, that the status of the SEC61B gene as a "restriction factor" of viral infection needs to be reconsidered, as we now suggest.

    I would appreciate more information regarding why restriction factors of cell growth have a high GFP2/GFP4. Intuitively, a KO of restriction factors of cell growth should result in better growth and higher GFP, thus leading to enrichment in GFP4, not GFP2.

    The reviewer raises an interesting question (although not at the heart of this work, as sgRNAs for the cell growth restriction factor mainly aim to serve as controls for the CRISPR screen). HCMV has a complex interaction with the cellular cell cycle. Specifically, it establishes a unique G1/S arrest that is both stimulatory and inhibitory since, on the one hand, it serves the virus to arrest the cell cycle, a critical step for viral genome replication. On the other hand, the virus needs many of the resources that serve cell growth. Both p53 and CDKN1A are important regulators at this stage; therefore, their interaction with the virus may indeed be complex. For example, p53 is upregulated by a viral infection. However, it is sequestered in the viral replication compartments, and its transcriptional are down-regulated, but its absence harms viral propagation (doi: 10.1128/mBio.02934-21, doi: 10.1128/jvi.72.3.2033-2039.1998, doi: 10.1128/jvi.00505-06). Therefore, it is not surprising that genes related to cell growth and cell cycle have complex effects on HCMV infection. We mention the essentiality of p53 for HCMV infection in the results (page 14, line 404).

    Line 404 "nonetheless"

    We appreciate the reviewer for noticing the typo. We corrected it.

    Reviewer #1 (Significance (Required)):

    The relation between human tRNA supply and viral translation is a topic of profound biological and biomedical importance. In this study, the authors used HCMV infection as the primary model to investigate this question. Results fall into two major parts: (i) changes in the tRNA pool during viral infection, and (ii) the impact of tRNA-related gene KO on viral infection.

    We appreciate the detailed report. We addressed the major points raised in the revised manuscript.

    Reviewer #2 (Evidence, reproducibility and clarity (Required)):

    In this study by Aharon-Hefetz et al., the researchers examined changes in tRNA pools during virus infections. The translation machinery plays a crucial role in virus replication. Consequently, host cells have developed sensors and effectors within this compartment to counteract viral mechanisms. The translation apparatus serves as a pivotal point in the virus-host conflict. Therefore, investigating alterations in the translation machinery during infections is vital for gaining a comprehensive understanding of the infection process. This study offers a thorough and high-quality analysis of data in a relevant cell culture system involving two different viruses. By conducting tRNA sequencing, the researchers studied the human tRNA pool following infections with human Cytomegalovirus (HCMV) and SARS-CoV-2. Changes in tRNA expression induced by HCMV were mainly driven by the virus infection itself, with minimal impact from the cellular immune response. Interestingly, specific tRNA post-transcriptional modifications seemed to influence stability and were subject to manipulation by HCMV. Conversely, SARS-CoV-2 did not lead to significant alterations in tRNA expression or post-transcriptional modifications. Moreover, a systematic CRISPR screen targeting human tRNA genes and modification enzymes allowed the identification of specific tRNAs and enzymes that either enhanced or reduced HCMV infectivity and cellular growth. This information enabled them to control the development of HCMV-specific tRNA modifications, highlighting the importance of these tRNA epitranscriptome modifications in virus replication. The authors concluded that the observed differences between the viruses are consistent with HCMV genes aligning with differentiation codon usage and SARS-CoV-2 genes reflecting proliferation codon usage. This observation's connection to the biology of HCMV and SARS-CoV-2 lies in the codon usage of structural and gene expression-related viral genes, showing a significant adaptation to host cell tRNA pools. Notably, these genes from both viruses demonstrated the highest adaptation to the tRNA pool of infected cells. The reason behind this phenomenon remains unclear. One hypothesis suggests that a high level of structural gene expression is necessary during activation. Testing this hypothesis could involve examining if hindering tRNA modifications affects virus morphogenesis. In summary, this study presents an interesting and innovative perspective on how viruses modify the translation machinery. The meticulous analysis sheds light on a central interaction point between viruses and their host cells.

    Reviewer #2 (Significance (Required)):

    In summary, this study presents an interesting and innovative perspective on how viruses modify the translation machinery. The meticulous analysis sheds light on a central interaction point between viruses and their host cells.

    We thank the reviewer for finding our work interesting, innovative, and well analyzed

    Reviewer #3 (Evidence, reproducibility and clarity (Required)):

    Summary

    Aharon-Hefetz et al. present the expression dynamics and modification signatures of tRNAs using DM-tRNA-seq in human foreskin fibroblasts or Calu3 cells during infections with two diverse viruses, HCMV and SARS-CoV2, respectively. They also use a newly designed tRNA-centric CRISPR library to screen the essentiality of tRNA and tRNA factors during HCMV-GFP infection. They find several tRNAs that are differentially expressed during HCMV infection, and most closely resemble the set of tRNAs shown to be used during cellular differentiation. Additionally, tRNA differential expression does not resemble that following interferon treatment, implying that virus modulation of tRNAs is unique to the general interferon response. They compare codon usage signatures during infection to their prior-defined sets of proliferation/differentiation tRNA genes. In their CRISPR screen, they find that different tRNAs can promote or restrict HCMV infection levels, as measured by the intensity of GFP fluorescence marker in their virus. Surprisingly, there were few tRNA modification factor hits that contributed to growth or infection.

    Reviewer #3- major comments

    The topic of this work is important, and the analysis performed here is assumed to be top quality, based on the previous work by the last author. The weakness with this body of work is a lack of rigor, specifically regarding validation and follow-up studies. Without these experiments, the reader lacks confidence in stated conclusions. For example: There is no validation or clue to how penetrant CRISPR is against tRNA genes. Given how duplicated some tRNA families are, it is possible that CRISPR is more effective against certain families compared to others. While this is likely an inherent caveat in all CRISPR screens, it would lend confidence in this approach to see some validation of tRNA KO by northern blot or RT-qPCR or sequencing.

    We thank the reviewer for raising this important issue. Indeed, many tRNA genes appear in multiple copies in the human genome. Yet, based on our previous work, we expect parallel editing of multiple copies using the same sgRNA. In our previous work (doi.org/10.7554/eLife.58461), we validated, based on several tRNA families, the ability of our tRNA CRISPR system to successfully target and affect tRNA expression levels. This included sequencing of the edited tRNA genes (i.e., DNA sequencing), in which we observed diverse INDEL mutations that predicted full disruption of the tRNA structure. Furthermore, we sequenced the tRNA pool of CRISPR-edited cells and found the downregulation of the targeted tRNAs to be up to 2-4-fold. This previous work provides foundations and confidence in this tRNA-CRISPR approach.

    Nevertheless, to further mitigate the reviewer's concern, we also plan to perform additional experiments in the current settings. We will choose individual tRNAs from our CRISPR screen as representatives to validate CRISPR editing. We will target each tRNA independently and test expression reduction by sequencing. We shall share the results in the full revision if granted.

    There is no validation that tRNA modification factor knockouts alter tRNA modification levels. Without this knowledge, the lack of essentiality cannot be confidently and fully interpreted. If the group does not validate whether individual tRNA modification factor knockouts alter modification profiles, then all possible explanations should be posited. For example, it is possible that 1) there could be major redundancy among tRNA modification enzymes, as the authors posit in the Discussion 2) tRNA modification enzymes are not essential for growth bc their activity/the modification they place is non-essential for growth, OR 3) the knockouts are not fully penetrant. I think this Discussion should be expanded to make caveats clearer. Perhaps referencing whether tRNA modification factors have been shown to be essential in other CRISPR screens would be helpful.

    Regarding the possible explanations for the lack of essentiality of tRNA modification enzymes, we agree with all three possibilities the reviewer raised. Reviewer #1 raised an additional option, in which tRNA modification enzymes are essential for HCMV infection and cell growth; thus, we cannot detect them in the screens because they drop out early in the process (before collecting the ancestor samples). We checked this possibility and found comparable read counts of sgRNAs targeting tRNA modification enzymes to that of other sub-libraries. This result suggests the drop-outs of sgRNA targeting are unlikely to happen on our screens.

    Furthermore, as the reviewer asked, we analyzed additional CRISPR screens and compared the effect of tRNA modification enzyme knockouts relative to the restriction and dependency factors we used in the library. The tested screens included 34 knockout CRISPR screens we downloaded from the BioGRID ORCS database that have similar parameters to our screen. Namely, they all test cell proliferation in a time-course manner, using a pooled sgRNA library and using the MAGeCK tool for data analysis. Overall, the screens use different human cell lines and diverse sgRNA libraries. Although potentially surprising, we found that the lack of essentiality of tRNA modification enzymes was also observed in the analyzed CRISPR screens (Figure S5B and on page 11, lines 322-330, and on page 18, lines 539-541).

    There is no validation that factors modulating GFP intensity in the HCMV screen actually impact virus replication. This is the point most important to this body of work. While GFP intensity does correlate to genome copies as shown by the authors, GFP read-out on a case-by-case basis could be simply due to factors required for expression/translation of GFP. Are any of the tRNA hits enriched or not represented in GFP reporter sequence? Either way, this information is informative.

    We acknowledge the concern. Specifically, the translation of the GFP gene itself could be affected by the tRNA manipulation done. To account for this potential concern, we tested the codon usage of the eGFP gene (which is the GFP version we used in the system) and compared it with tRNA essentiality, as determined by the cell growth CRISPR screen. We report this in the revised manuscript (page 13, lines 390-392, and added Figure S6D). We found that GFP does not tend to significantly use codons that correspond to essential or less essential tRNAs. The same lack of correlation was also found for the tRNA essentiality upon HCMV infection (not shown).

    Additionally, given that the hits are cross-compared ONLY to other infected (low intensitiy "GFP+") cells, and not to an uninfected population, there is no guarantee that these primarily drive HCMV infection. The top hits should be validated in HFFs, infected with HCMV, with resulting titers/viral gene expression/genome copies measured. Additionally, the reasons for not using a GFP- population as a control should be clarified.

    We agree that additional experiments on some hits may be warranted. We plan to examine for such an effect on infection using an individual gene version of the assay. In particular, we will target individually candidate tRNA genes following validation (as described previously in point 1). We will then infect the tRNA-targeted cells with HCMV and measure the effectiveness of HCMV infection using a standard titer assay.

    The reviewer also suggest comparing GFP1/2/3 to an ancestor in addition to comparing them to GFP4. Towards that we now show a GFP2 vs ancestor comparison (shown below). The results look very similar and are now added to the supplemental material of the revised manuscript (page 13, lines 385-387, Figure S6B).

    Though careful codon usage analysis for HCMV versus the human host was analyzed, it seems pertinent to analyze whether the differentially expressed tRNAs during infection correlate to either codon usage profiles. Figure 3C and S3C intend to address this point for viral gene groups; however, I would encourage the authors to expand the description of these results to make them easier to interpret, especially for those not in the tRNA field. For example, "tRNA adaptation index (tAI)" is not defined in the text, but simply referenced. For clarity, you should include a brief explanation of what this measure describes. Following, when reporting results from Figure 3, the results can then be delivered with more specific and interpretable language. These steps will ensure maximal scientific communication to the audience.

    We appreciate the reviewer's comment regarding the importance of scientific communication and making this manuscript easier to interpret, especially for readers unfamiliar with the world of tRNAs and translation efficiency. We added a description of our motivation to use tAI and the meaning of the measurement (page 9, lines 241-243). We also elaborated on the results part and made the results more interpretable (page 9, lines 245 and 249-250).

    Finally, given that changes are most visible at 72 hpi, the analysis should include expression based on this time point for comparison.

    Regarding the time point used for tAI calculation (Figure 3), we tested the tAI measured by the tRNA pool at 72hpi and got very similar results to that obtained using the tRNA pool measured at 24hpi. As 24hpi represents the pick of HCMV infection, we decided to present this analysis. In the current revised version, we also added the analysis done using the tRNA pool measured 72hpi as suggested by the reviewer (Figure S3D).

    Reviewer #3- minor comments

    I would recommend more care in terminology used for the CRISPR screen (Figures 5 and 6) to make the manuscript easier to digest. Labeling sgRNAs-containing cells as " Reduced Growth/Infection" or "Increased Growth/Infection" is not immediately easy to understand. For example, saying this sgRNA "increased growth" could refer to the knockdown increasing growth OR could mean that this sgRNA was enriched in cells with increased growth, which are opposing. It might be more clear to state to use depleted/enriched terminology in these figure labels. This also applies to the text, be sure to plainly describe the terminology and what it means each time you refer to the CRISPR results.

    This is a good point. Indeed, focusing on the significant enrichment of the sgRNAs, rather than their effect on growth or infection, is more straightforward. We changed the terminology in Figures 5C and 6C and the text in the current version.

    Is there actual evidence that the new tRNA sgRNA library is more effective than that used previously? State if so.

    We assume the referee refers to our previous paper on the smaller-scale library (doi.org/10.7554/eLife.58461). The addition here is that the library is much more comprehensive (the previous one targeted only 20 tRNAs). We point it out in the revised manuscript (page 17, lines 499-501).

    Fig 1A-C: The cutoff for "red" symbol distinction is not stringent enough. 1.05 would be red, but that is not convincingly upregulated. The cutoff should be at least FC>1.2.

    We thank the reviewer for bringing our intention to this point. In the current version, we changed the cutoff of absolute fold change higher than 1.2 in Figures 1A-C and S1A (also in legend).

    Need thorough description of tRNA bioinformatics and modification analysis (citing past work is not appropriate here-need to make accessible to your audience).

    Further thorough descriptions of tRNA bioinformatics and modification analysis are added in the revised version (page 6, lines 149-151, page 7, lines 178-183).

    Line 182- Result headings could be more informative, even with small adjustments. For example "Specific tRNA modifications are modulated in response to HCMV infection" is more clear and accurate, as there are only a few measurable changes in tRNA modification. Limitations of using sequencing techniques to analyze modifications (versus MS) should also be discussed.

    We changed that heading accordingly.

    We also mentioned the advantages and disadvantages of using sequencing to assess tRNA modification levels (page 7, lines 184-187).

    It is not immediately clear why the viral plot looks different in Fig S3B compared to Fig 3B.

    We thank the referee for spotting this. We employed different length cutoffs on the genes in each panel and have now fixed that in the revised manuscript.

    Line 254-255. This point is not immediately clear-please include more specific language detailing the logic leading to this conclusion.

    Indeed, the logic here was missing. The idea was that longer genes are associated with gene conservation, hence functionality. Thus, non-canonical HCMV genes that are both long and codon-optimized might have a function during HCMV infection. We added this explanation to the text (pages 8-9, lines 235-238).

    Line 408- "may be essential"-I would modify the language here. Especially given there is no true comparison with uninfected cells.

    We improved the language throughout the revised manuscript.

    There are a number of recent publications profiling tRNA expression in herpesviruses. These should be mentioned and discussed in the context of this work. I know some were included in the reference list, but the body of work as a whole, and how this work fits in and pushes the horizon, could be further emphasized. It is quite impressive that this is a conserved feature of herpesvirus infection. a. PMID: 36752632 b. PMID: 35110532 c. PMID: 34535641 d. PMID: 33986151 e. PMID: 33323507 f. PMID: 35458509

    We thank the reviewer for highlighting these works. We added a discussion item regarding tRNA expression in HCMV and other herpesviruses with the references (pages 15-16, lines 447-458)

    CoV2 Discussion point-The lack of tRNA expression regulation might have more to do with the length of the infection (6 hpi cov2- also didn't see much a change at 5hpi with hcmv). This should be proposed as a possibility.

    It is a possibility that due to the high stability of tRNAs, expression regulation of tRNAs will not affect the tRNA pool in short infection such as of SARS-CoV-2. We added this explanation in the discussion part, page 15, lines 441-442.

    Line 582. Misspelled schlafen in Discussion. (SLFN, not SFLN)

    The point is fixed in the revised manuscript.

    Reviewer #3 (Significance (Required)):

    General assessment: I found this paper exciting to read, given the dearth of knowledge regarding viral modulation of tRNA expression.

    We appreciate the reviewer's comment

    However, the work is highly descriptive, with a complete absence of follow-up or validation studies. At the very least, I would have hoped that the authors validated that viral titer (and not just GFP intensity) was impacted by some of the hits. The lack of confirmation and quality control overall diminishes confidence in the stated conclusions.

    However, I think the topic is timely, important, and that this manuscript offers tools to the community at large to learn more about viral manipulation or other drivers of tRNA regulation. Once follow-up/validation experiments are added to the work, as detailed below, this manuscript will be of broad importance and highly impactful.

    As mentioned above, we plan to add such validations to the fully revised manuscript.

    Advance: While there have been many studies suggesting tRNA regulation occurs during viral infection (these pubs should be referenced as mentioned above), this is an advance due to the fact that it begins to address whether tRNA expression changes functionally impact viral replication. This will be much more solid with follow-up experiments confirming that hits alter HCMV replication (rather than GFP intensity).

    Audience: This will be of broad interest to those with interest in virology and gene expression. The new sub-libraries of tRNA-related factors might be useful to be tested in other cell types and settings. Again, as the work stands, it is descriptive and hypothesis-stimulating, but the conclusions need validation and further support.

    We thank the referee for the encouraging words and the suggested analyses. We already implemented most of the suggestions in the current revised version and hope to add further experiments in a fully revised manuscript.

  2. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    Summary

    Aharon-Hefetz et al. present the expression dynamics and modification signatures of tRNAs using DM-tRNA-seq in human foreskin fibroblasts or Calu3 cells during infections with two diverse viruses, HCMV and SARS-CoV2, respectively. They also use a newly designed tRNA-centric CRISPR library to screen the essentiality of tRNA and tRNA factors during HCMV-GFP infection. They find several tRNAs that are differentially expressed during HCMV infection, and most closely resemble the set of tRNAs shown to be used during cellular differentiation. Additionally, tRNA differential expression does not resemble that following interferon treatment, implying that virus modulation of tRNAs is unique to the general interferon response. They compare codon usage signatures during infection to their prior-defined sets of proliferation/differentiation tRNA genes. In their CRISPR screen, they find that different tRNAs can promote or restrict HCMV infection levels, as measured by the intensity of GFP fluorescence marker in their virus. Surprisingly, there were few tRNA modification factor hits that contributed to growth or infection.

    Major Comments

    1. The topic of this work is important, and the analysis performed here is assumed to be top quality, based on the previous work by the last author. The weakness with this body of work is a lack of rigor, specifically regarding validation and follow-up studies. Without these experiments, the reader lacks confidence in stated conclusions. For example:
      • a. There is no validation or clue to how penetrant CRISPR is against tRNA genes. Given how duplicated some tRNA families are, it is possible that CRISPR is more effective against certain families compared to others. While this is likely an inherent caveat in all CRISPR screens, it would lend confidence in this approach to see some validation of tRNA KO by northern blot or RT-qPCR or sequencing.
      • b. There is no validation that tRNA modification factor knockouts alter tRNA modification levels. Without this knowledge, the lack of essentiality cannot be confidently and fully interpreted. If the group does not validate whether individual tRNA modification factor knockouts alter modification profiles, then all possible explanations should be posited. For example, it is possible that 1) there could be major redundancy among tRNA modification enzymes, as the authors posit in the discussion 2) tRNA modification enzymes are not essential for growth bc their activity/the modification they place is non-essential for growth, OR 3) the knockouts are not fully penetrant. I think this discussion should be expanded to make caveats clearer. Perhaps referencing whether tRNA modification factors have been shown to be essential in other CRISPR screens would be helpful.
      • c. There is no validation that factors modulating GFP intensity in the HCMV screen actually impact virus replication. This is the point most important to this body of work. While GFP intensity does correlate to genome copies as shown by the authors, GFP read-out on a case-by-case basis could be simply due to factors required for expression/translation of GFP. Are any of the tRNA hits enriched or not represented in GFP reporter sequence? Either way, this information is informative. Additionally, given that the hits are cross-compared ONLY to other infected (low intensitiy "GFP+") cells, and not to an uninfected population, there is no guarantee that these primarily drive HCMV infection. The top hits should be validated in HFFs, infected with HCMV, with resulting titers/viral gene expression/genome copies measured. Additionally, the reasons for not using a GFP- population as a control should be clarified.
    2. Though careful codon usage analysis for HCMV versus the human host was analyzed, it seems pertinent to analyze whether the differentially expressed tRNAs during infection correlate to either codon usage profiles. Figure 3C and S3C intend to address this point for viral gene groups; however, I would encourage the authors to expand the description of these results to make them easier to interpret, especially for those not in the tRNA field. For example, "tRNA adaptation index (tAI)" is not defined in the text, but simply referenced. For clarity, you should include a brief explanation of what this measure describes. Following, when reporting results from Figure 3, the results can then be delivered with more specific and interpretable language. These steps will ensure maximal scientific communication to the audience. Finally, given that changes are most visible at 72 hpi, the analysis should include expression based on this time point for comparison.

    Minor Comments

    1. I would recommend more care in terminology used for the CRISPR screen (Figures 5 and 6) to make the manuscript easier to digest. Labeling sgRNAs-containing cells as " Reduced Growth/Infection" or "Increased Growth/Infection" is not immediately easy to understand. For example, saying this sgRNA "increased growth" could refer to the knockdown increasing growth OR could mean that this sgRNA was enriched in cells with increased growth, which are opposing. It might be more clear to state to use depleted/enriched terminology in these figure labels. This also applies to the text, be sure to plainly describe the terminology and what it means each time you refer to the CRISPR results.
    2. Is there actual evidence that the new tRNA sgRNA library is more effective than that used previously? State if so.
    3. Fig 1A-C: The cutoff for "red" symbol distinction is not stringent enough. 1.05 would be red, but that is not convincingly upregulated. The cutoff should be at least FC>1.2.
    4. Need thorough description of tRNA bioinformatics and modification analysis (citing past work is not appropriate here-need to make accessible to your audience)
    5. Line 182- Result headings could be more informative, even with small adjustments. For example "Specific tRNA modifications are modulated in response to HCMV infection" is more clear and accurate, as there are only a few measurable changes in tRNA modification. Limitations of using sequencing techniques to analyze modifications (versus MS) should also be discussed.
    6. It is not immediately clear why the viral plot looks different in Fig S3B compared to Fig 3B.
    7. Line 254-255. This point is not immediately clear-please include more specific language detailing the logic leading to this conclusion.
    8. Line 408- "may be essential"-I would modify the language here. Especially given there is no true comparison with uninfected cells.
    9. There are a number of recent publications profiling tRNA expression in herpesviruses. These should be mentioned and discussed in the context of this work. I know some were included in the reference list, but the body of work as a whole, and how this work fits in and pushes the horizon, could be further emphasized. It is quite impressive that this is a conserved feature of herpesvirus infection.
      • a. PMID: 36752632
      • b. PMID: 35110532
      • c. PMID: 34535641
      • d. PMID: 33986151
      • e. PMID: 33323507
      • f. PMID: 35458509
    10. CoV2 discussion point-The lack of tRNA expression regulation might have more to do with the length of the infection (6 hpi cov2- also didn't see much a change at 5hpi with hcmv). This should be proposed as a possibility.
    11. Line 582. Misspelled schlafen in discussion. (SLFN, not SFLN)

    Significance

    General assessment: I found this paper exciting to read, given the dearth of knowledge regarding viral modulation of tRNA expression. However, the work is highly descriptive, with a complete absence of follow-up or validation studies. At the very least, I would have hoped that the authors validated that viral titer (and not just GFP intensity) was impacted by some of the hits. The lack of confirmation and quality control overall diminishes confidence in the stated conclusions. However, I think the topic is timely, important, and that this manuscript offers tools to the community at large to learn more about viral manipulation or other drivers of tRNA regulation. Once follow-up/validation experiments are added to the work, as detailed below, this manuscript will be of broad importance and highly impactful.

    Advance: While there have been many studies suggesting tRNA regulation occurs during viral infection (these pubs should be referenced as mentioned above), this is an advance due to the fact that it begins to address whether tRNA expression changes functionally impact viral replication. This will be much more solid with follow-up experiments confirming that hits alter HCMV replication (rather than GFP intensity).

    Audience: This will be of broad interest to those with interest in virology and gene expression. The new sub-libraries of tRNA-related factors might be useful to be tested in other cell types and settings. Again, as the work stands, it is descriptive and hypothesis-stimulating, but the conclusions need validation and further support.

  3. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    In this study by Aharon-Hefetz et al., the researchers examined changes in tRNA pools during virus infections. The translation machinery plays a crucial role in virus replication. Consequently, host cells have developed sensors and effectors within this compartment to counteract viral mechanisms. The translation apparatus serves as a pivotal point in the virus-host conflict. Therefore, investigating alterations in the translation machinery during infections is vital for gaining a comprehensive understanding of the infection process.

    This study offers a thorough and high-quality analysis of data in a relevant cell culture system involving two different viruses. By conducting tRNA sequencing, the researchers studied the human tRNA pool following infections with human Cytomegalovirus (HCMV) and SARS-CoV-2. Changes in tRNA expression induced by HCMV were mainly driven by the virus infection itself, with minimal impact from the cellular immune response. Interestingly, specific tRNA post-transcriptional modifications seemed to influence stability and were subject to manipulation by HCMV. Conversely, SARS-CoV-2 did not lead to significant alterations in tRNA expression or post-transcriptional modifications.

    Moreover, a systematic CRISPR screen targeting human tRNA genes and modification enzymes allowed the identification of specific tRNAs and enzymes that either enhanced or reduced HCMV infectivity and cellular growth. This information enabled them to control the development of HCMV-specific tRNA modifications, highlighting the importance of these tRNA epitranscriptome modifications in virus replication. The authors concluded that the observed differences between the viruses are consistent with HCMV genes aligning with differentiation codon usage and SARS-CoV-2 genes reflecting proliferation codon usage. This observation's connection to the biology of HCMV and SARS-CoV-2 lies in the codon usage of structural and gene expression-related viral genes, showing a significant adaptation to host cell tRNA pools. Notably, these genes from both viruses demonstrated the highest adaptation to the tRNA pool of infected cells. The reason behind this phenomenon remains unclear. One hypothesis suggests that a high level of structural gene expression is necessary during activation. Testing this hypothesis could involve examining if hindering tRNA modifications affects virus morphogenesis. In summary, this study presents an interesting and innovative perspective on how viruses modify the translation machinery. The meticulous analysis sheds light on a central interaction point between viruses and their host cells.

    Significance

    In summary, this study presents an interesting and innovative perspective on how viruses modify the translation machinery. The meticulous analysis sheds light on a central interaction point between viruses and their host cells.

  4. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    I have mixed feelings regarding this manuscript. On the one hand, the authors did an impressive amount of work. On the other hand, the manuscript seems overly descriptive (writing should be more concise) without a clear message or hypothesis that is cohesive to all the presented evidence. Below, I will outline my concerns.

    Regarding the first part.

    1. I am not an expert in the field of viral biology and immunology. I wonder how well the IFN treatment mimics the cellular response to infection (yet without the virus). Also, how good is ruxolitinib at blocking the IFN response ? I would appreciate it if you could explain both with one or two sentences and provide the necessary references.
    2. (MAJOR) Can these two treatments really allow the effects of host response and viral infection to be separated? OR in other words, are these two effects really orthogonal? In my opinion, they are NOT. Fig. 1E seems to support my opinion, as the changes seen for the "IFN" sample relative to the "uninfected" sample (referred to as "changes-A" below), are parallel to the changes seen for the "24hpi + ruxo" sample relative to the "24hpi" sample ("changes-B"). More specifically, changes-A represent the host response, as argued by the author, whereas changes-B represent the elimination of the host response (due to ruxo, conditioned on the virus-driven effect). If the virus-driven effect and the host response could really be separated, one would expect changes-A and changes-B are more or less opposite. However, they appeared to be parallel, suggesting that uninfected versus infected conditions can have totally different (even opposite) host responses. More importantly, if one cannot separate the host response from virus-driven effects, the conclusion of "tRNA changes are driven by virus, not host response" is then unfounded.3. Even if we let go of this previous point and accept that these results indeed offer some support for the notion that the virus-driven effect are the main contributor to the shifts in tRNA pool, the support is at best moderate. A big gap here is "how?" I suggest the authors should at least give some insight on how virus can do that in Discussion (and mention it with one sentence in Results).
    3. The authors compared the HCMV codon usage to the proliferation and differentiation signatures of human cells. But these two signatures are not compared with measured tRNA expression. It might shed some light on the general characteristics of tRNA pool shifts due to infection (towards a proliferation-like or differentiation-like signature). This fits in the general topic of virus-host interaction and might give more evidence for the point that HMCV is adapted to a differentiation signature (as it drives the host into that state).
    4. How is the dashed box in Fig3A/B chosen?
    5. The tAI values shown in Fig3C-E are extremely low (compared to other reports I am aware of). Does this mean that the adaptation of viral codon usage to human cell supply is actually very weak? This is in opposition to the major claims made in this section.
    6. I believe that the part about SARS-CoV-2 could be made more concise. It is sufficient to mention that results may differ from those obtained with HCMV in one paragraph.
    7. Line 299 on page 11 - I do not believe codon usage between different viruses can be directly compared, let alone reaching such a conclusion. Some viruses have low CAI or tAI to humans, but they have co-evolved with humans for a long time. Furthermore, there are viruses that infect multiple hosts, but their CAI for a host with which they have long co-evolved is higher while their CAI for a host that is relatively new is lower.
    8. (MAJOR) A more general comment is that there is a difference between tRNA expression and the abundance of translation-ready tRNA. The process of charging tRNA with amino acids may take a long time. It is the abundance of the charged-tRNA (the ternary complex of aminoacylated tRNA and EF-Tu-GTP) that is of biological importance. In this regard, the use of tRNA expression falls short.

    Regarding the second part,

    1. (MAJOR) Prior to the actual competition assay in the first high-throughput screen (cell competition assay), the authors applied two days of antibiotic selection and two days of recovery. This could result in a serious problem of false negatives or drop outs. Specifically, an sgRNA targeting an essential gene with high efficiency would kill the cells, leaving no (or a small number of) cells in the ancestor population at the beginning of the competition process. A sgRNA's enrichment in competing populations cannot be reliably estimated in such situations. I am not certain that the FDR used in Figure 5B is sufficient to address this issue. Please clarify whether it could. Providing raw counts for competing and ancestor populations would also be helpful.
    2. It is also highly questionable to me the nearly negligible effects of tRNA modification enzymes. This may be explained by the point above. Indeed, the dots of tRNA modification enzymes in general appear to have higher FDR (lower y values) when compared to red dots with similar enrichment levels.
    3. The screen based on IE2-GFP labeled HCMV measures a phenotype that is very difficult to interpret. Particularly, I am not sure if GFP2 and GFP3 are good controls for comparing GFP4 (GFP1 might be better). Various factors can affect GFP levels, including, but not limited to, dilution caused by a rapidly dividing host cell, unhealthy translational machinery resulting from infection or microenvironment. My point is supported by some observations in Fig6B. For example, SEC61B, a restriction factor for HCMV infection, is enriched in the GFP2 group, contrary to expectations. It is necessary for the authors to prove with firm evidence that their choice of GFP signal thresholds is appropriate.
    4. I would appreciate more information regarding why restriction factors of cell growth have a high GFP2/GFP4. Intuitively, a KO of restriction factors of cell growth should result in better growth and higher GFP, thus leading to enrichment in GFP4, not GFP2.
    5. Line 404 "nonetheless"

    Significance

    The relation between human tRNA supply and viral translation is a topic of profound biological and biomedical importance. In this study, the authors used HCMV infection as the primary model to investigate this question. Results fall into two major parts: (i) changes in the tRNA pool during viral infection, and (ii) the impact of tRNA-related gene KO on viral infection.

  5. Version published to 10.1101/2024.04.25.591047 on bioRxiv