The short isoform of the host antiviral protein ZAP acts as an inhibitor of SARS-CoV-2 programmed ribosomal frameshifting

Programmed ribosomal frameshifting (PRF) is a fundamental gene expression event in many viruses, including SARS-CoV-2. It allows production of essential viral, structural and replicative enzymes that are encoded in an alternative reading frame. Despite the importance of PRF for the viral life cycle, it is still largely unknown how and to what extent cellular factors alter mechanical properties of frameshift elements and thereby impact virulence. This prompted us to comprehensively dissect the interplay between the SARS-CoV-2 frameshift element and the host proteome. We reveal that the short isoform of the zinc-finger antiviral protein (ZAP-S) is a direct regulator of PRF in SARS-CoV-2 infected cells. ZAP-S overexpression strongly impairs frameshifting and inhibits viral replication. Using in vitro ensemble and single-molecule techniques, we further demonstrate that ZAP-S directly interacts with the SARS-CoV-2 RNA and interferes with the folding of the frameshift RNA element. Together, these data identify ZAP-S as a host-encoded inhibitor of SARS-CoV-2 frameshifting and expand our understanding of RNA-based gene regulation.

, The authors may discuss how ZAP protein overexpression affects the mRNA levels of the report genes as well as viral RNA expression levels. The reason is that ZAP is known to be involved in modulating RNA stability (page 7, line 176). The mRNA levels may affect ribosome density and thus frameshifting efficiency as shown before. In addition, repression of translation initiation by ZAP-S (as stated in page 17, line 443) may also affect frameshifting efficiency. Figure 3B, It seems that ZAP-S inhibits the production of zero-frame protein expression. Again, the authors may check if ZAP-S affects mRNA integrity here. Figure 4, In the single-molecule experiments, a control RNA, e.g., a stem-loop or a different pseudoknot may be added. It seems that ZAP-s affects the folding of the stem-loop within the SARS-CoV-2 pseudoknot structure. Thus, it is very likely that a non-specific molecular crowding effect causes the slowing down of the folding of stem-loops within the pseudoknot. Such potential effect was discussed in ref: JACS 2018, 140, 8172-8184. The statement in page 17, line 435 about "structural plasticity" is inconsistent with the model proposed in page 18, second paragraph. In the "structural plasticity" model, dynamics of multiple secondary structures are proposed to be important in enhancing ribosomal frameshifting. In fact, a recent Science paper (DOI: 10.1126/science.abf3546 ) clearly shows that specific mRNA pseudoknotribosome interaction is critical for stimulating ribosomal frameshifting. In addition, the frameshifting assay data reported in this Science paper suggest that a small molecule did not directly affect frameshifting, although the small molecule was previously reported to be a frameshifting inhibitor and affect single-molecule mechanical folding of pseudoknots.
Reviewer #3 (Remarks to the Author): The authors identify the short isoform of ZAP as a binder to SARS-CoV2 frame shift element and show that the interaction influences RNA folding. Furthermore, they show that overexpression of ZAP results in changing frameshift frequency and ZAP associates with the ribosome. Major Figure 2D: Why has the RNA levels only be determined for ZAP and not by isoform specific primers? In principle, the amount of information on mRNA changes upon infection should not be just 3 transcripts, but the authors should conduct an RNA-Seq experiment. It is just simple enough nowadays and can be even outsourced. It would give a much better picture on the results of infection of their cell line and also be useable to check for upregulated proteins among all their interactors. The ANOVA tests need to be followed by e.g. a Tukey's Posthoc test to actually judge which are the differentially regulated conditions. Applies to several occasions throughout the manuscript. Figure 2C: How is controlled that the amount of overexpressed protein is correlated to the effect size?
The authors place such an emphasis on PRF efficiency to select their candidates that this needs to be controlled. Reading through the manuscript, it is not clear to my why ZAP-S should have a stronger effect than ZAP-L, but one explanation might be actual expression levels in this assay here. It would have been a much stronger study if ZAP-L would have been taken along and for a publication at this level, I would like to see the data also for ZAP-L. For me, it looks like a little overselling of the short ZAP isoform, while there is no proof that the long ZAP isoform should not work either. The authors should acknowledge and discuss this better in the manuscript. (Especially the comment that "ZAP-S is specific" I don't see justified with the presented data.) I also would remind the authors to not overinterpret that only CoV and ZAP worked in their assay. While they tested a few structures, they can't completely rule out that there are more. It is a kind of lame argument on my side, but I don't think that the human ZAP-S isoform just evolved to recognize CoV and CoV2. For Figure 3, the authors should also include data from IGF2BP1 as they seem to suggest that it has a functional relevance and not just IGF2BP3. Why weren't they actually tested in Figure 2C, given they are very similar, but show different effects? Figure 4: For this assay (E and F) IGF2B3 should be used as a control to rule out just general effects of protein in the lysate. The authors already expressed IGF2BP3, thus it is not a large effort to include a more appropriate control. Why not endogenous studies as a ZC3HAV1 antibody (Fig. S2D) was available? This limits in my view the value of the study. There could be overexpression artefacts. A orthogonal knockdown/knock-out analysis is also not available. The association with ribosomes has only been done biochemically and looks like a case of sticky behavior. Can this be validated with microscopy or alternatively better suited assay? For sure another RNA-binding protein as negative control is missing in Figure 5b and 5c experiments (actin is not enough). Overall: I think that the inclusion of RNA-Seq, adding appropriate controls to experiments in Figure 3, Figure 4 and Figure 5 plus the data on ZAP-L is crucial and should be a prerequisite to accepting this study.
Minor I would suggest to rename the heading: "Revealing the short isoform of the host antiviral protein ZAP as an …" to avoid the misguiding information that ZAP-S is a full protein name. Abstract: Also in the text, I would add: "reveal that the short isoform of zinc-finger antiviral protein (ZAP-S) is" (afterwards the use of ZAP-S is justified as introduced as an isoform definition here). Abstract: It is unclear to me why it is a "de novo host encoded factor" and not just a "host encoded factor". In Figure 1B, the SSB writing is not clear enough visible. The single-molecule argument might be misleading (see e.g ZAP interacts with translating ribosome, line 2) as not a single molecule ZAP was studied, but a single RNA molecule. Figure S4C: The data is not really interpretable and look more similar than the examples in S4A and S4B. Can this be better represented? I would actually recommend to move Fig. S4D or at least the 0 part to the main figure to not have a cherry-picked representation there, but some global analysis from all measurements.

Point by Point Responses
We thank the reviewers for their time and efforts and for their thoughtful and constructive comments that allowed us to strengthen our manuscript.
The major revisions to the manuscript are as follows: . We have included new in vivo frameshifting reporter assay data on additional human coronaviruses, namely HKU and OC as well as different SARS CoV mutant RNA variants. In order to strengthen the reason for not selecting ZAP L, we provide in vivo data of ZAP L effect on SARS CoV frameshift RNA variants. . We have performed additional single molecule optical tweezers experiments and ensemble microscale thermophoresis analysis with these RNA variants in the presence and absence of ZAP S and could illustrate that the Stem and regions of the putative pseudoknot are the main ZAP S interaction sites. Furthermore, we have carried out structural probing experiments by DMS MaPseq, which suggest that reactivities of these stem regions change in the presence of ZAP S. . A control RNA binding protein IMP was tested in ensemble microscale thermophoresis and single molecule optical tweezers assays to monitor the effect of non specific interactions, and technical artefacts on RNA. To make the point clearer that ZAP S interferes with the energetics of the refolding of the RNA variants with different physical properties, we carried out work calculations during unfolding and refolding. We also improved the overall representation of this complex data. . In order to keep the storyline focused on the in vitro characterization of ZAP S function on frameshifting, we de emphasized the data showing ZAP:ribosome interactions, removed ZAP co immunoprecipitations and present the polysome profiling data in a different context to illustrate that ZAP S interactions with the viral RNA occur during translation in vivo and in vitro.
Point by point replies to individual reviewer comments is below. The original reviews and specific points are reproduced in blue for convenience. Remarks to the Author Zimmer et al. describe the discovery and characterization of cellular proteins that bind the programmed ribosomal frameshift PRF signal derived from the SARS CoV virus. Starting with an RNA affinity pulldown experiment, they identify a number of host proteins previously known to interact with RNA, including some that have previously been shown to be important for viral translation and activation of the innate immune response. This was followed up with assays to monitor the effects of these proteins on SARS CoV mediated PRF, further whittling down the number of candidates. After another round of selection based on changes in expression in virus infected cells and PRF inhibitory activity, they focused on ZAP S, which is known to target interferon mRNAs for degradation. Surprisingly, ZAP S appeared to specifically inhibit PRF mediated by SARS CoV and the nearly identical SARS CoV elements: it did not affect PRF promoted by two other coronaviruses MERS CoV and Bat CoV , nor of PRF signals of other origin. Follow up experiments showed that ZAP S can inhibit SARS CoV PRF in vitro, that it directly interacts with the PRF motif, and a series of biophysical experiments were used to characterize its binding affinity and ability to prevent re folding of the SARS CoV PRF stimulating mRNA pseudoknot. Additionally, pulldown experiments show that it interacts with the ribosome and identified additional interacting proteins. A final series of experiments demonstrated that ZAP S can promote an approximately fold inhibition of SARS CoV replication in Huh cells. In general, this is an outstanding piece of research.
Major Comments.
. This work has great breadth but not such great depth. It is not clear whether it is about ZAP S or about frameshifting. In particular, the final section detailing what pulls down with ZAP S and its association with the ribosome seem to be peripheral and distracting to what from the title should be a story about this cellular anti frameshifting story. There seems to be two different narratives competing here. The authors should consider separating them into two distinct papers.
We have now rephrased the section describing the association of ZAP S with the ribosome and combined the polysome profiling data with in vitro translation experiments in Fig. . We hope that the implemented changes re focus the narrative of the manuscript on the anti frameshifting activity of ZAP S. To this end, we also removed the co immunoprecipitation data from the manuscript. Nevertheless, several proteins that play a role in PRF have previously been shown to interact with ribosomes e.g., A, SFL . Since we see a similar association of ZAP S with the ribosome in vitro and in naïve cells, we find it important to include these data here, although we agree with the reviewer that a detailed analysis of the functional relevance of ZAP S interactions with the ribosome is beyond the scope of the current manuscript.
. Lines and Fig. B: The binding of ZAP S to a mutant of the bulged A was not reported. Why is this mutant mentioned here? Indeed, the only mutants that were assayed were those that fully disrupted the structure. This is an example of lack of depth. The authors should assay a few well selected ZAP S mutants, e.g. pick from a few that are described in the recent Science paper from the Ban group, in order to obtain a clearer picture of the structural requirements for ZAP S binding to the SARS CoV pseudoknot. These mutants should also be assayed using the optical tweezers seen next comment , and microscale thermophoresis.
We mentioned the bulged A mainly to emphasize that Stem loop SL is extremely important for frameshifting and that even a single mutation within this stimulatory RNA element can interfere with frameshifting efficiencies. Following the suggestions of the reviewer, we now expanded our work with additional mutants and performed rigorous analysis with microscale thermophoresis, DMS MaPseq and optical tweezers.
We took advantage of the published structure of the SARS CoV pseudoknot and generated truncation mutants of individual stem loops SL , SL and SL , which helped us to assess the structural requirements for ZAP S function. While the deletion of both SL and SL completely abolishes ZAP S binding, deletion of the SL alone showed only a moderate effect on the interaction KD nM . Upon deletion of the ' distal region, which would likely prevent formation of SL , we observed a lower binding affinity delta SL , KD nM . These data suggested two putative ZAP S binding sites, involving the stem and stem regions and we confirmed these observations further by EMSAs. KD calculations derived from two binding events nM, calculated from the formula KD KD KD are within the range of our experimentally determined KD nM . We employed DMS MaPseq to probe ZAP S binding, which, in accordance with the results discussed above, also suggested altered accessibility in SL and SL regions upon ZAP S binding please refer to text change lines and Suppl. Fig. . We also carried out optical tweezer experiments and analysis with the mutant RNAs in the absence and presence of ZAP S. As discussed in detail under point below, we found that unfolding of all RNA variants was not altered but the refolding of wild type was impeded by ZAP S. Furthermore, while we could still observe this effect of ZAP S on the refolding of delta SL and delta SL mutants, deletion of both SL and SL delta SL completely abolished this effect of ZAP S. Please refer to ms text for details, lines , .
We carefully considered the frameshift site mutations reported by the Ban and Atkins labs. However, because we cannot fully predict the effect of these deletions on the pseudoknot fold, we felt that truncation mutants of individual pseudoknot stem loops would be the better approach to study the binding interactions of ZAP.
Altogether, these new data further define the putative ZAP S binding sites and strengthen our original conclusion that ZAP S binds specifically to the SARS CoV pseudoknot structure.
. Fig. E. The unfolding and refolding profiles of the pseudoknot that are shown here do not compare well with the profiles generated by the Woodside lab. They have a quick and dirty look to them. Likewise, Figs. S A and B. In particular, the two step unwinding profile is completely absent in the wild type and is barely apparent in the refolding profile. These data are unconvincing. This is important because it is the most convincing data for specific binding of ZAP S to this element and is the basis for the model shown in Fig. of ZAP S altering the pseudoknot structure. It would be nice to see cleaner data and or an orthogonal approach.
As discussed in point above, we have now included additional data using orthogonal approaches DMS seq and MST analyses and using new RNA variants with sequential deletions within the putative stem regions of the predicted pseudoknot delta SL , SL and SL .
Further, to strengthen our single molecule optical tweezers experiments: We now include additional data on the wild type pseudoknot WT PK and the additional mutant RNA variants discussed above. The total number of curves analyzed in the study has now reached Supp. Table . We include DMS MaPSeq as an alternative approach to probe ZAP S binding and its effects on the overall RNA fold. As an additional control, we also conducted WT PK RNA experiments in the presence of an RNA binding protein IMP , which was identified in our proteomic screen as a weak binder. While we observe a slight effect on refolding curves with IMP , probably due to non specific interactions, the effect of ZAP S on PK RNA is clearly distinct from what is seen with IMP and all other controls that we employed Together with other in vitro ensemble measurements of RNA binding by MST and DMS MaPseq analysis our data strongly support our initial hypothesis that an interaction with ZAP S triggers an alteration in PK RNA refolding refer to ms, lines ; .
Regarding potential differences between our data and the results reported by the Woodside lab Neupane et al. , PMID: , we would like to point out that there are differences in the construct design between both studies. In particular, the Woodside group used a longer ' spacer derived from the native SARS CoV genome. Therefore, our results cannot be directly compared. In our analysis, of the unfolding events were marked by a single unfolding event. In a minority of the curves, we were able to see the two step unfolding that was reported by the Woodside lab now included in Fig  below . Regardless of these differences, the PK like conformer with the contour length change value of .
. nm was in a very good agreement with the values reported by Neupane et al. PMID: . . nm . The authors also proposed an additional conformer, which was not prominent in our data. We speculate this state could form due to the presence of ' spacer sequence. In that regard, Schlick et al. recently reported that the ' region preceding Stem loop of the SARS CoV frameshifting PK can form an alternative structure by interacting with the ' part of stem loop PMID: . We have now included this discussion in the manuscript text line .

Rebuttal Fig
Optical tweezers data from Neupane et al., , PMID: top and the predicted structure formed by ' spacer bottom .
. The specificity of ZAP S for the SARS CoV family of PRF signals is surprising given that prior to , humans had never been exposed to these viruses. Evolution does not tend to be anticipatory . Therefore, it is logical to hypothesize that this activity of ZAP S evolved in response to another viral PRF signal. There are four other, well established human coronaviruses which cause approximately of common colds: E alpha coronavirus ; NL alpha coronavirus ; OC beta coronavirus ; and HKU beta coronavirus . A reasonable hypothesis is that this function of ZAP S may have evolved in response to one or a few of them. This should be tested. This is an interesting point. Indeed, we were also surprised that ZAP S did not show any effect on other PRF sites we tested. As suggested by the reviewer, we have now included PRF elements from other coronaviruses namely HKU and OC . We observed a slight, albeit not statistically significant reduction in frameshift levels mediated by ZAP S.
While coronaviral PRF elements are relatively conserved Rebuttal Fig. , there are notable differences in their sequence and hence their structure. For example, SARS CoV and SARS CoV have a deletion of three nucleotides between SL and SL , which might result in a different fold of the pseudoknot, as proposed by Plant et al., PMID: . In addition, SARS CoV and SARS CoV are the only coronaviruses that share a high sequence similarity in SL , which, according to our biophysical interaction studies and DMS Seq results, is one of the potential ZAP S binding sites.
Future studies in our lab are aimed at understanding the structural basis of the ZAP S mediated reduction of frameshift efficiency specifically in SARS CoV and . At this point, however, we can only speculate about the evolutionary origins of this effect. We agree with the reviewer that it is highly unlikely that this function of ZAP evolved in response to coronaviruses. ZAP is a multi functional host protein, with several other antiviral roles, perhaps the regulation of SARS CoV frameshifting is a serendipitous side effect.

Rebuttal Fig
Alignment of PRF elements of several coronaviruses. Asterisks denote perfect conservation between all sequences shown.
Minor comments: . The term for the first time is considered to be bad form in scientific writing. Please try not to use it. Changed.
. Lines : the statement that even a change in frameshifting inhibited SARS CoV viral propagation and reduced infectivity is incorrect. In the cited reference, the mutant with the smallest change in frameshifting that was tested with regard to virus propagation conferred an reduction, not .
We thank the reviewer for spotting this oversight. Corrected.
. A fold reduction in viral titers is not considered to be big by virology standards. orders of magnitude or greater are considered to be the threshold for a strong effect on virus replication.
While we agree that the effect of ZAP overexpression on viral titers fold might be considered modest, it is nevertheless reproducible and statistically significant. It is important to note that ZAP is only one of many other interferon stimulated genes ISGs that are induced upon infection as part of the host defense system, and therefore its overall contribution might be small, but functionally relevant. In order to strengthen our conclusions that ZAP S impact viral replication, we have now also carried out an immunofluorescence assay to detect the viral N protein, which is an early marker of SARS CoV infection. ZAP S decreases the expression of N, further indicating that virus replication is impaired.
. Line : the data show a significant amount of residual frameshifting, indicating that the roadblock effect is lessened, not abrogated. Changed.

Reviewer
Remarks to the Author The manuscript by Zimmer et al. reports the discovery of host proteins including ZAP S that bind to SARS CoV viral RNA programmed ribosomal frameshifting stimulating RNA structure through pull down assays. In this work, the authors focused on characterizing ZAP S by cell culture reporter gene assays, viral replication and cell free translation assay and ensemble binding studies. The data suggest that ZAP S interacts with frameshifting stimulating viral RNA pseudoknot in SARS CoV infected cells, and inhibits frameshifting and viral replication. In addition, the authors utilized single molecule optical tweezers techniques to study how affects RNA pseudoknot folding. The authors need to make the below clarifications and corrections before the paper may be published.
Page , line , A related ref Biochemistry , , provides insights into how dsRNA binders may stimulate ribosomal frameshifting.
We thank the reviewer for this comment. We added a brief discussion on this topic and included the reference lines .
Page , line , Fig. D may not be referred to here. Corrected.
Page , line , the exact sequences and structures of the mutants of the RNAs should be shown in SI. Fig. A, B and the exact sequences of the RNA variants are provided in Supplementary Table . Page , line , The authors may have a discussion about why ZAP mRNA transcription is activated.

Schematics of the RNA variants used in this study have been included in
ZAP S is an interferon induced gene that is part of the innate immune response. The transcriptional regulation of ZAP has now been discussed in page of the revised manuscript.
Figure C, The authors may discuss how ZAP protein overexpression affects the mRNA levels of the report genes as well as viral RNA expression levels. The reason is that ZAP is known to be involved in modulating RNA stability page , line . The mRNA levels may affect ribosome density and thus frameshifting efficiency as shown before. In addition, repression of translation initiation by ZAP S as stated in page , line may also affect frameshifting efficiency.
In the original manuscript, we had demonstrated that ZAP S overexpression does not affect the stability of the reporter mRNA in cells Supplementary Fig. B . To further test whether ZAP S influences translation or ribosome density in cells, we compared polysome profiles in control and ZAP S overexpressing cells but observed no difference Rebuttal. Again, we observed very similar polysome profiles. Combined, these data indicate that the effect of ZAP S on frameshifting is not due to degradation of the reporter mRNA, a global decrease in the rate of translation or ribosome density Supplementary Fig. B, please refer to text change in lines .
As pointed out by the reviewer, it has been reported that ZAP S might bind to ' UTR of mRNAs and recruit RNA surveillance factors to regulate viral and some host RNAs. How and whether ZAP S affects global translation or RNA surveillance during coronavirus infections is an interesting question, which we are actively pursuing in collaboration with other groups, but this analysis goes beyond the scope of the current manuscript, which is focused on the effects of ZAP S on coronaviral frameshifting. In that context, we did not observe an effect on RNA stability of reporter constructs in our experimental set up.

Rebuttal Fig Polysome profiles of naïve HEK cells vs HEK cells overexpressing ZAP S.
Figure B, It seems that ZAP S inhibits the production of zero frame protein expression. Again, the authors may check if ZAP S affects mRNA integrity here.
As mentioned above, ZAP S overexpression did not affect the abundance of the reporter mRNA in cells Supplementary Fig. B . In addition, we did not observe a difference in the expression levels of GFP frame and mCherry frame control constructs in the presence and absence of ZAP S.
The decrease in the frame controls seen in vitro is likely an artifact of the in vitro translation systems when high concentrations of ZAP S are used. Other studies have previously observed a similar effect upon increasing concentrations of trans acting proteins A and SFL in in vitro reporter assays Napthine et al , ; PMID: , PMID: Rebuttal Fig. . To account for this effect, frameshifting efficiencies are calculated by dividing the intensity of the frame product by the sum of the intensity of the and frame product. Therefore, the calculated frameshifting efficiency is normalized to the frame band and therefore independent of any effect on the mRNA integrity please refer to Methods . Figure , In the single molecule experiments, a control RNA, e.g., a stem loop or a different pseudoknot may be added. It seems that ZAP s affects the folding of the stem loop within the SARS CoV pseudoknot structure. Thus, it is very likely that a non specific molecular crowding effect causes the slowing down of the folding of stem loops within the pseudoknot. Such potential effect was discussed in ref: JACS , , .

Rebuttal Fig
We followed the reviewer's suggestion and have now included additional RNA mutants of the putative SARS CoV PK to dissect the interactions between ZAP S and the SARS CoV RNA. As an example of an additional stem loop control, we employed the Δ SL SL mutant, which is predicted to fold into SL about nt . ZAP S did not substantially affect refolding of this stem loop alone. We also employed a PK variant, referred to as 'compensatory mut. , which is predicted to form a stack of Gs in the stems. Notably, the unfolding of RNA occurred at higher forces suggesting a stabilization of the RNA fold. This variant still interacts with ZAP S with similar affinity as the wt PK, but we did not observe a drastic change in refolding which suggest either the structure or the local thermodynamic stability of the RNA is different. An alternative could also be that the faster kinetics of RNA refolding in this pseudoknot masks the effect of ZAP S please also see lines .
To rule out the possibility that the energetics of refolding are altered simply due to molecular crowding, we conducted measurements of the PK RNA in the presence of IMP , another unrelated RBP. IMP was identified as a weak binder in our proteomic screen and MST confirmed its lower affinity compared to ZAP KD around nM, Supplementary Fig. H, also ms text lines . In this case, we observed a very mild effect on RNA refolding, which was likely due to non specific steric interactions or molecular crowding. Similar effects were also observed with the RNA variants ΔSL SL in the presence of ZAP S, although ZAP S clearly did not bind to that RNA variant or affect frameshifting Supplementary Fig. F, ms lines .
To quantify these effects, we now calculate the work during refolding in the presence of ZAP S Fig. K . The normalized work values of the PK and mutant RNAs in the presence of control protein IMP were very similar to the values obtained for the ΔSL mutant and the compensatory mutant in presence of ZAP S Fig. K . Based on these data, we conclude that the effect of ZAP S on PK RNA refolding is distinct and stronger than observed for the various control conditions, hence molecular crowding or non specific steric contributions are unlikely to account for the observed effect.
The statement in page , line about structural plasticity is inconsistent with the model proposed in page , second paragraph. In the structural plasticity model, dynamics of multiple secondary structures are proposed to be important in enhancing ribosomal frameshifting. In fact, a recent Science paper DOI: . science.abf clearly shows that specific mRNA pseudoknot ribosome interaction is critical for stimulating ribosomal frameshifting. In addition, the frameshifting assay data reported in this Science paper suggest that a small molecule did not directly affect frameshifting, although the small molecule was previously reported to be a frameshifting inhibitor and affect single molecule mechanical folding of pseudoknots.
We revised our discussion of the model accordingly and further discuss the possible effects of trans acting factors page . Indeed, not every protein that we identified in our proteomic screen as an interactor with the frameshift PK also altered frameshifting levels these were previously presented in SI now added in Fig.  C . In addition, even if we detect binding of ZAP S to some of the RNA variants tested, we did not necessarily see a difference in the refolding of the RNA comp. mut., Fig G . Therefore, binding to the RNA alone is not sufficient for altering RNA dynamics or frameshifting. We are currently pursuing structural studies to address how the structural dynamics of the RNA and trans acting factors modulate translation, but resolving this issue is clearly beyond the scope of the current work.

Reviewer
Remarks to the Author The authors identify the short isoform of ZAP as a binder to SARS CoV frame shift element and show that the interaction influences RNA folding. Furthermore, they show that overexpression of ZAP results in changing frameshift frequency and ZAP associates with the ribosome. Figure D: Why has the RNA levels only be determined for ZAP and not by isoform specific primers? In principle, the amount of information on mRNA changes upon infection should not be just transcripts, but the authors should conduct an RNA Seq experiment. It is just simple enough nowadays and can be even outsourced. It would give a much better picture on the results of infection of their cell line and also be useable to check for upregulated proteins among all their interactors.
To address this point, we took advantage of published RNAseq data and compared gene expression levels in infected Huh . . , Calu cells and patient samples Blanco Melo et al., ; Sun et al., as referenced in the ms . These data allowed us to better illustrate differences in expression levels of the core interactors identified in our screen. ZAP was still the most enriched transcript in the RNAseq data sets, which aligned well with our qRT PCR analysis Supplementary Fig. E and F please refer to lines .
We also looked at the Blanco Melo data to quantify upregulation of ZAP S and ZAP L isoforms upon infection by analyzing reads mapping to the splice junction. We found that both ZAP S and ZAP L are upregulated upon infection, . fold and . fold respectively. Mock: ZAPL fragments and ZAPS fragments Infected: ZAPL fragments and ZAPS fragments However, due to the low coverage at the splice junction we consider this data qualitative rather than quantitative. We include these data in the rebuttal for the reviewer's interest.
The ANOVA tests need to be followed by e.g. a Tukey's Posthoc test to actually judge which are the differentially regulated conditions. Applies to several occasions throughout the manuscript.
We apologize for not mentioning that the ANOVA tests were followed by a Brown Forsythe test to ensure equal variance among the samples. Also, a Dunnett s multiple comparisons test was employed in order to identify the differentially regulated conditions compared to the control constructs using GraphPad Prism version . . . This is now clarified in the manuscript text.
Figure C: How is controlled that the amount of overexpressed protein is correlated to the effect size? The authors place such an emphasis on PRF efficiency to select their candidates that this needs to be controlled.
We acknowledge that there could be differences in their expression levels that may change the observed effect. To account for that, in the dual fluorescence assay, the putative trans regulators of frameshifting are expressed as ECFP fusion proteins. Thus, we can apply a gating technique to analyze the frameshifting ratio in cells that are positive for ECFP, and hence express the putative trans acting regulators ms lines . This gating strategy is illustrated in Supplementary Fig. A. We now also present the mean ECFP values for a selection of proteins interrogated in this study Rebuttal Fig. . Based on the mean ECFP values expression levels of both ZAP isoforms are actually lower than what we observe for most other candidates. Thus, the data we show in the manuscript likely underestimates the effect of ZAP S. Nevertheless, we can clearly show expression levels do not differ significantly between ZAP S and ZAP L, which in turn also confirms that the difference in effect of ZAP L and ZAP S is not due to their expression levels.

Rebuttal Fig
Mean ECFP intensity for various proteins interrogated in the dual fluorescence assay as described in Fig. of the manuscript. Reading through the manuscript, it is not clear to my why ZAP S should have a stronger effect than ZAP L, but one explanation might be actual expression levels in this assay here. It would have been a much stronger study if ZAP L would have been taken along and for a publication at this level, I would like to see the data also for ZAP L. For me, it looks like a little overselling of the short ZAP isoform, while there is no proof that the long ZAP isoform should not work either. The authors should acknowledge and discuss this better in the manuscript. Especially the comment that ZAP S is specific I don't see justified with the presented data.
We have now analyzed the mean ECFP intensity of the ZAP L and ZAP S fusion proteins see Rebuttal Fig. and demonstrate that the expression levels do not differ significantly between the two isoforms. Originally, we selected the shorter isoform of ZAP for further investigation based on the cellular reporter assays, where a stronger decrease in relative frameshift levels were observed with ZAP S compared to ZAP L see Rebuttal . Previously, it has also been reported that ZAP S and ZAP L have different subcellular localization. While ZAP S is mainly localized in the cytoplasm, ZAP L can be prenylated, triggering its membrane associated. This differential localization was proposed to have a direct influence on the roles these proteins play during infection Charron et al., , DOI: . pnas. ; Vyas et al., DOI: . ncomms ; Schwerk et al., , DOI: . s . Despite our efforts to quantify endogenous levels of both isoforms during infection using available RNA seq datasets see point above , we were unable to obtain reliable results due to low coverage of the splice junction.

Rebuttal Fig
Effect of ZAP L on FE of SARS CoV , HKU , OC as well as mutants of the SARS CoV PRF site compare to ms Fig. E . I also would remind the authors to not overinterpret that only CoV and ZAP worked in their assay. While they tested a few structures, they can't completely rule out that there are more. It is a kind of lame argument on my side, but I don't think that the human ZAP S isoform just evolved to recognize CoV and CoV .
Reviewer raised a similar point pt. . As suggested by this reviewer, we tested additional human coronaviruses namely HKU and OC , but we did not observe a statistically significant reduction in frameshift efficiency Fig. E . However, we agree that ZAP is very unlikely to have evolved to specifically recognize the frameshift motifs of SARS CoV and , given that these are very new human pathogens. We have therefore toned down the manuscript text accordingly.
For Figure , the authors should also include data from IGF BP as they seem to suggest that it has a functional relevance and not just IGF BP . Why weren't they actually tested in Figure C, given they are very similar, but show different effects?
First, we would like to clarify that we had not intended to suggest that IGF BP IMP or IGF BP IMP have functional relevance. We initially included IMP as a 'negative' control, because the protein is only very marginally enriched in our interactome screen. Following the suggestions of the reviewer, we have now included both IMP and IMP in our dual fluorescence reporter assay Fig. C  We also used IMP in our in vitro optical tweezer analysis, but only as a control to corroborate the specificity of this assay. We are not entirely certain what experiment this comment refers to. Co immunoprecipitation was performed in infected Calu lysate removed in the updated ms and in these experiments, we analyzed endogenous ZAP. However, as an interferon induced gene, basal cellular ZAP S levels in absence of viral infection are insufficient to be reliably detected by western blotting in a polysome gradient. Hence, ZAP S had to be overexpressed in those experiments. However, we conducted ribosome pelleting in naïve Calu lysates to show the interactions of endogenous ZAP with ribosomes Fig. F . Nonetheless, to re focus the manuscript on the in vitro characterization of ZAP S function on frameshifting, as requested by reviewer , we de emphasized the interactions of ZAP S with the ribosome and removed the ZAP co immunoprecipitation data from the manuscript ms lines .
A orthogonal knock down knock out analysis is also not available.
We assume that this comment refers to the SARS CoV infection experiments. To address this point, we were able to take advantage of the recently published data by the Kim lab, which shows that depletion of ZAP in their study referred to as ZC HAV leads to an increase in viral titers Lee et al., , DOI: .
j.molcel. . . Rebuttal Fig. . We now discuss these findings, which are well in line with our observations, and citation is added in the manuscript text ms text lines .

Rebuttal Fig
Depiction of antiviral effect of ZAP upon knockdown as shown in Lee et al., .
The association with ribosomes has only been done biochemically and looks like a case of sticky behavior. Can this be validated with microscopy or alternatively better suited assay? For sure another RNA binding protein as negative control is missing in Figure b and c experiments actin is not enough .
The initial aim of the ribosome association experiments was to monitor whether ZAP S interacts with the SARS CoV RNA during active translation. For that purpose, we would argue that polysome profiling is a suitable and well accepted assay in the translation field. In addition, ribosome pelleting is generally employed to monitor if a specific factor associates with the ribosomes.
We have now included ribosome pelleting in naïve Calu cells in the main figure to show that endogenous ZAP S also associates with the ribosomes Fig. F . As an additional control, we also tested ribosome binding of a random RNA binding protein SFPQ. As shown in Rebuttal Fig. , compared to the input, SFPQ protein is not significantly enriched in the ribosome pellet, while endogenous ZAP is highly enriched.
Based on comments of reviewer , who found our initial presentation of the ZAP ribosome interactions distracting, we restructured of the manuscript and are presenting these data now in Fig. . We also de emphasized the observation and functional relevance of the ZAP ribosome interactions please also see point above . While we agree that it would be interesting to investigate whether this association occurs in the cell, single RNA localization and single ribosome tracking experiments would be challenging and would require substantial optimizations in cell culture. Therefore, we think that such an in depth analysis is better suited for follow up work, in line with the suggestions of reviewer . Ribosome pelleting of ZAP in naïve Calu cells highlighting that not every RNA binding protein is enriched in the ribosome pellet ZAP vs. SFPQ .
Overall: I think that the inclusion of RNA Seq, adding appropriate controls to experiments in Figure  We included additional data and controls in the revised manuscript. Briefly, we analyzed available RNA seq datasets from infected and uninfected cells as well as patient samples Blanco Melo et al. , DOI: .
j.cell. . . , Sun et al. , DOI: . j.cell. . . confirming that ZAP is one of the most highly upregulated transcripts upon infection among the proteins we investigated.
Four new RNA variants were added in ensemble and single molecule experiments to further pinpoint the interactions of ZAP S with the RNA element. Furthermore, we include additional negative controls IMP as an RNA binding protein and the SUMO tag to our in vitro translation assays and observed no effect on frameshifting levels Fig. C, Supplementary Table . We have also carried out structure probing using DMS MaPseq, which allowed us to monitor potential ZAP S mediated alterations in the RNA structure Supplementary Fig . ms text lines .
These additional control experiments confirmed the interaction of ZAP S with the pseudoknot and the effects of this interaction on RNA dynamics.
Minor I would suggest to rename the heading: Revealing the short isoform of the host antiviral protein ZAP as an … to avoid the misguiding information that ZAP S is a full protein name. Abstract: Also in the text, I would add: reveal that the short isoform of zinc finger antiviral protein ZAP S is afterwards the use of ZAP S is justified as introduced as an isoform definition here . Abstract: It is unclear to me why it is a de novo host encoded factor and not just a host encoded factor .
Wording is changed to illustrate the difference between the two isoforms.
In Figure B, the SSB writing is not clear enough visible. Corrected.
The single molecule argument might be misleading see e.g. ZAP interacts with translating ribosome, line as not a single molecule ZAP was studied, but a single RNA molecule.
We apologize for the confusing statement regarding the ZAP ribosome interaction due to word limits for the abstract. Indeed, the ribosome interaction refers to bulk analysis and not to single molecule experiments. We have modified this sentence to better reflect this point. This point is well taken. Following the suggestion of the reviewer, we have changed the representation from single example traces to show representative curves for folding and unfolding for each sample in Fig. D I. We have also calculated the work from the change in free energies for the wild type and frameshifting RNA variants used in the study. This data is plotted in Fig. J, K and histograms for change in contour length and work for each RNA variant is also added in Supplementary Fig. and .