Macrodomain catalytic activity modulates Chikungunya virus dissemination and transmission potential in Aedes mosquitoes

Reviewer #1 (Public review):

Summary:

This paper from Bardossy et al. explores whether viral macrodomains in dual-host viruses contribute to infection in the mosquito vector. Using the CHIKV Caribbean strain, the authors generated nsP3 macrodomain catalytic site mutants (N24A or N24D) and identified a compensatory mutation site at position 31 during virus propagation in Vero cells. They then assessed the impact of these mutations on viral growth kinetics in A549 (human) and U4.4 (Ae albopictus cells), as well as on infectivity and dissemination in vivo in Ae. aegypti and Ae. albopictus. Biochemical and structural analyses of recombinant macrodomain proteins (alone or in combination) revealed effects on stability, catalytic activity, and ADP-ribose binding. Overall, the study demonstrates that CHIKV macrodomain catalytic activity plays an important role in virus infectivity and dissemination within the mosquito vector.

Strengths:

A complete set of experimental approaches spanning generation of recombinant viruses, in vitro characterization, in vivo studies in mosquitoes, and detailed biochemical and structural characterization.

Weaknesses:

(1) The sequence analysis of the generated stocks revealed the emergence of a second-site mutation at position 31 of the nsP3 macrodomain when (N24A or N24D) CHIKV mutants were generated on Vero cells. However, it is not clear from the text or the experimental design how many independent replicates were performed. Based on the current description, it appears this was done only once, which raises the question of whether mutations at position 31 represent a reproducible outcome of infection. This is particularly important because experiments in A549 cells did not reveal emergence of mutations at position 31. To strengthen this finding, the experiment should be performed at least three independent times.

(2) Based on the primer information used to generate amplicons for sequencing, the amplicons evaluated do not span the full nsP3 gene as stated in the text (Line 105). Instead, they cover only the first 119 amino acids of the macrodomain (160 aa long). Thus, the current data do not rule out the emergence of other compensatory mutations elsewhere in the nsP3 macrodomain or in the full-length protein. Additional sequencing is recommended, or the text should clearly state that only a portion of the macrodomain was sequenced.

(3) Another key question is whether this is a specific feature of the Caribbean strain or a feature conserved across different CHIKV lineages.

(4) The use of A549 cells (interferon-competent) to study CHIKV infection is somewhat surprising, as the current literature indicates that this cell line is not efficiently infected by Asian or ECSA lineages of CHIKV (PMID: 17604450) unless the Mxra8 receptor is overexpressed (PMID: 29769725) or IFN signaling is inhibited (PMID: 31682641). The data presented here are compelling and suggest specific features of the Caribbean strain that enable efficient infection of this cell line (Do the authors observe detectable cytopathic effect (CPE) in CHIKV-infected A549 cells?).

However, to further support the authors' claim related to human immunocompetent cells, it would be important to demonstrate the phenotype in an additional interferon-competent cell line that is well-established as highly permissive to CHIKV, such as human fibroblasts.

(5) To fully support the conclusion stated in lines 234- 237, the authors should fully sequence the virus stock used to demonstrate that no additional mutations (beyond N24D-D31H/N) are present that could contribute to the enhanced dissemination phenotype. This is especially important if the experiment was performed with only one stock of virus, given justified gain-of-function concerns.

(6) The authors did not assess transmission but transmission potential (only viral dissemination to heads was measured). The sentence at line 360 should be modified to accurately reflect the data-supported conclusion.

Reviewer #2 (Public review):

Summary:

To address how the CHIKV macrodomain contributes to replication dynamics in mammalian and insect hosts, the authors initially created two separate mutations in the highly conserved N24 residue, which is known to be critical for the CHIKV macrodomain's ability to erase ADP-ribose from target proteins. Interestingly, they could not produce a virus with a mutation in this residue without second-site mutations in an aspartic acid residue nearby (D31). However, when tested biochemically, these second-site mutations did not enhance the enzymatic activity of the protein, indicating that other enzyme dynamics, such as substrate binding, may be impacting these mutations. Mutations at this residue allowed the CHIKV to replicate in Vero cells and in mosquito cells, but they replicated poorly in IFN-competent human cells, indicating clear IFN-specific impacts on these viruses. Interestingly, they found unique impacts on virus dissemination and replication in live mosquitoes. While the N24A/D31N virus did poorly in vivo in all accounts, the N24D/D31H/N virus tended to infect both the bodies and heads of the mosquitoes better than the WT virus, though titers were reduced. The authors claimed, based on a DSF assay, that there were no real differences in ADP-ribose binding and thus suggested that these differences could be due to changes in substrate specificity, as the D31 residue resides in the substrate exit path, potentially tuning the virus to unique substrates in different species. The authors also produced crystal structures of the mutants to demonstrate the changes in the binding pocket caused by these mutations.

Strengths:

The authors have done a rigorous job of evaluating CHIKV macrodomain mutant viruses and the proteins' biochemical activities. The use of live mosquitoes is highly unique and provides important insights into the importance of the macrodomain in different species.

Weaknesses:

It is not clear if the interpretation of the ADP-ribose binding data is correct. It appears there are notable differences that could explain the results, though the authors chose to minimize the impact that these differences had on the results. The N24D-D31H/N proteins had at least a 1C degree difference in the thermal shift assay when compared to the N24A/D31N, single D31 mutants, and WT proteins, which is likely significant and could explain the dichotomous results between the two viruses in mosquito cells. Even the single N24D mutant had enhanced binding compared to the WT protein. Furthermore, as this virus has no enzymatic activity, one could hypothesize that enhanced binding to a substrate that is normally cleaved by the protein could certainly lead to alterations in phenotypic effects, whether good or bad. The authors should test the binding activity in a separate assay, such as an ITC assay, to determine if there are, in fact, binding differences or not. Having said this, it is likely that the impacts of these mutations on replication and transmission in human and mosquito cells are multi-factorial and could include both enhanced binding with altered substrate specificity amongst other activities.

Additionally, as both mutants had no detectable enzymatic activity but had quite different phenotypes in mosquitoes, I don't agree with the title stating that catalytic activity modulates dissemination and transmission potential in mosquitoes. It seems more likely that alterations in binding activity or substrate recognition (even suggested by the authors) impact these phenotypes in mosquitoes.

Reviewer #3 (Public review):

Summary:

The authors investigated the role of the nsP3 macrodomain catalytic activity in the replication and transmission of CHIKV in mosquito vectors. The conserved dual-host alphavirus catalytic site N24 has previously been shown to be essential for ADP-ribosylhydrolase activity. Despite this, mosquito-specific alphaviruses do not share this catalytic site. To assess whether the macrodomain catalytic activity of a dual-host virus was essential in insect hosts, the authors targeted the N24 site to abolish catalysis while maintaining binding capacity. The loss of ADP-ribosylation led to the emergence of compensatory mutations at site D31 that impact viral infectivity, dissemination, and transmission in Aedes sp. mosquitoes in vivo. The conclusions are well supported by the results and provide insight into the importance of nsP3 macrodomain activity in the mosquito vector, which hasn't been explored before.

Strengths:

The main strength of this study is the use of Aedes sp. mosquito models to investigate the selective pressure of macrodomain mutations in vivo. The functional characterization as well as the structural analysis of the mutants provide supporting evidence of a potential role of the compensatory mutations at site D31 in substrate recognition.

Weaknesses:

A considerable part of this study relies on the use of N24 mutant viral stocks generated in Vero cells, which yields an additional mutation at site 31 and consequently doesn't allow the authors to properly dissect the effect of mutation of N24 and D31 independently. It would be recommended to generate stocks with individual mutations in both A549 and U4.4 cells, pooling and concentrating them if needed. Replication of the N24A mutant in A549 cells does not lead to mutation at residue 31. Yet surprisingly, there is no reversion from N back to D at site 31 when the double mutant Vero stocks are passaged in A549. Since they are double mutants, it isn't possible to assess whether the defects in the growth of mutants N24A/T-D31N and N24D-D31H/N compared to WT are due to site 24 or 31, or both (Figure 2, panel c). Even though the authors emphasize that the compensatory mutation could have additional roles that impact viral infectivity and transmission in mosquito cells, it would strengthen the work to show that these mutations would spontaneously appear in stocks generated directly in mosquito cells. As a corollary, is it known whether insect-specific alphaviruses that lack macrodomain catalytic activity have corresponding mutations at site 31?

Additionally, there is a lack of consistency in the prevalence of WT virus at days 5 and 7 in in vivo experiments with Ae. albopictus and Ae. aegypti (Figure 3 and Supplementary Figure 2). This raises concern about the reproducibility of these experiments.

The inability to tease apart the roles of N24 and D31 in mosquito hosts partially prevented the authors from fully achieving their aims, but the work is nonetheless of interest to the field and suggests that more work is necessary to fully understand the role of the nsP3 macrodomain and its catalytic activity in the two disparate but obligate hosts for CHIKV and other dual-host alphaviruses.

Author response:

Public Reviews:

Reviewer #1 (Public review):

Summary:

This paper from Bardossy et al. explores whether viral macrodomains in dual-host viruses contribute to infection in the mosquito vector. Using the CHIKV Caribbean strain, the authors generated nsP3 macrodomain catalytic site mutants (N24A or N24D) and identified a compensatory mutation site at position 31 during virus propagation in Vero cells. They then assessed the impact of these mutations on viral growth kinetics in A549 (human) and U4.4 (Ae albopictus cells), as well as on infectivity and dissemination in vivo in Ae. aegypti and Ae. albopictus. Biochemical and structural analyses of recombinant macrodomain proteins (alone or in combination) revealed effects on stability, catalytic activity, and ADP-ribose binding. Overall, the study demonstrates that CHIKV macrodomain catalytic activity plays an important role in virus infectivity and dissemination within the mosquito vector.

Strengths:

A complete set of experimental approaches spanning generation of recombinant viruses, in vitro characterization, in vivo studies in mosquitoes, and detailed biochemical and structural characterization.

Weaknesses:

(1) The sequence analysis of the generated stocks revealed the emergence of a second-site mutation at position 31 of the nsP3 macrodomain when (N24A or N24D) CHIKV mutants were generated on Vero cells. However, it is not clear from the text or the experimental design how many independent replicates were performed. Based on the current description, it appears this was done only once, which raises the question of whether mutations at position 31 represent a reproducible outcome of infection. This is particularly important because experiments in A549 cells did not reveal emergence of mutations at position 31. To strengthen this finding, the experiment should be performed at least three independent times.

We thank the reviewer for raising this important point. The emergence of second-site mutations at residue D31 in Vero cells was actually observed in two independent experiments, each initiated by transfection of viral RNAs encoding either the N24A or N24D mutant. In the second experiment, additional timepoints were sampled for sequencing. Comparison of the two experiments revealed that, in the second replicate, only the D31N variant was recovered, whereas D31H was not detected. We will include the results of both independent experiments in the updated Figure 1 and will modify the text accordingly to improve clarity.

In addition, we will present new experiments performed in other cell types that further confirm the reproducibility of D31 mutation emergence across different cellular contexts. Specifically, we will include results from new viral RNA transfections in BHK-21 mammalian cells and C6/36 mosquito cells, which will be included in the updated Supplementary Figure 1.

(2) Based on the primer information used to generate amplicons for sequencing, the amplicons evaluated do not span the full nsP3 gene as stated in the text (Line 105). Instead, they cover only the first 119 amino acids of the macrodomain (160 aa long). Thus, the current data do not rule out the emergence of other compensatory mutations elsewhere in the nsP3 macrodomain or in the full-length protein. Additional sequencing is recommended, or the text should clearly state that only a portion of the macrodomain was sequenced.

We thank the reviewer for this correction. The sequencing was indeed focused on the region of the macrodomain surrounding the introduced mutations, covering the first 119 amino acids of nsP3. This region was selected to confirm the stability of the introduced mutations at position 24 and to monitor the emergence of potential second-site mutations in its immediate vicinity. We acknowledge that this approach does not rule out the emergence of compensatory mutations elsewhere in the macrodomain or in the full-length nsP3 protein. We will correct the text accordingly to accurately reflect the region that was analyzed.

(3) Another key question is whether this is a specific feature of the Caribbean strain or a feature conserved across different CHIKV lineages.

This is an excellent point. To address it, we introduced N24A and N24D mutations into an infectious clone of the Indian Ocean strain, a representative of the ECSA lineage, and assessed the emergence of D31 second-site mutations during viral stock production. As observed with the Caribbean strain, the D31N secondary mutation consistently emerged in both N24A and N24D Indian Ocean mutant viruses. These results will be included in the updated Supplementary Figure 1 and in the main text.

(4) The use of A549 cells (interferon-competent) to study CHIKV infection is somewhat surprising, as the current literature indicates that this cell line is not efficiently infected by Asian or ECSA lineages of CHIKV (PMID: 17604450) unless the Mxra8 receptor is overexpressed (PMID: 29769725) or IFN signaling is inhibited (PMID: 31682641). The data presented here are compelling and suggest specific features of the Caribbean strain that enable efficient infection of this cell line (Do the authors observe detectable cytopathic effect (CPE) in CHIKV-infected A549 cells?).

However, to further support the authors' claim related to human immunocompetent cells, it would be important to demonstrate the phenotype in an additional interferon-competent cell line that is well-established as highly permissive to CHIKV, such as human fibroblasts.

We thank the reviewer for raising this point. We acknowledge that previous studies have reported limited infection of A549 cells by certain CHIKV strains. However, in our hands, with our viral stocks and under our experimental conditions, we observe an increase in viral titers following infection of A549 cells with WT Caribbean strain virus, indicating productive viral replication. We also confirmed that the Indian Ocean strain replicates in A549 cells under the same conditions, and we will include growth curve data for this strain in the updated version of the manuscript. Importantly, we did not observe detectable cytopathic effects in A549 cells infected with either WT or N24 mutant viruses, which is consistent with the notion that CHIKV replicates less efficiently in this cell line compared to other mammalian cell lines. We will include a sentence acknowledging this in the discussion of the revised manuscript.

Regarding the suggestion to use human fibroblasts, we respectfully note that the primary focus of this study is the role of macrodomain catalytic activity in the mosquito host, and the experiments in A549 cells were performed to confirm the known importance of the macrodomain in interferon-competent mammalian cells. We therefore consider the current data in A549 cells sufficient to support this conclusion within the scope of the manuscript.

(5) To fully support the conclusion stated in lines 234- 237, the authors should fully sequence the virus stock used to demonstrate that no additional mutations (beyond N24D-D31H/N) are present that could contribute to the enhanced dissemination phenotype. This is especially important if the experiment was performed with only one stock of virus, given justified gain-of-function concerns.

We acknowledge that the full viral genome was not sequenced, and we cannot rule out the presence of additional mutations elsewhere in the genome that could contribute to the observed phenotype. However, we note that the enhanced dissemination phenotype was also observed in independent experiments performed with the Indian Ocean strain mutant viruses in Ae. albopictus, which were generated independently from the Caribbean strain stocks. The consistency of the phenotype across two independently generated sets of mutant viruses from different CHIKV lineages strongly supports the conclusion that the enhanced dissemination phenotype is linked to the macrodomain mutations. These new data will be included in the updated manuscript.

(6) The authors did not assess transmission but transmission potential (only viral dissemination to heads was measured). The sentence at line 360 should be modified to accurately reflect the data-supported conclusion.

We thank the reviewer for pointing this out. The text will be modified accordingly to accurately reflect that we assessed transmission potential, based on viral dissemination to heads, rather than actual transmission.

Reviewer #2 (Public review):

Summary:

To address how the CHIKV macrodomain contributes to replication dynamics in mammalian and insect hosts, the authors initially created two separate mutations in the highly conserved N24 residue, which is known to be critical for the CHIKV macrodomain's ability to erase ADP-ribose from target proteins. Interestingly, they could not produce a virus with a mutation in this residue without second-site mutations in an aspartic acid residue nearby (D31). However, when tested biochemically, these second-site mutations did not enhance the enzymatic activity of the protein, indicating that other enzyme dynamics, such as substrate binding, may be impacting these mutations. Mutations at this residue allowed the CHIKV to replicate in Vero cells and in mosquito cells, but they replicated poorly in IFN-competent human cells, indicating clear IFN-specific impacts on these viruses. Interestingly, they found unique impacts on virus dissemination and replication in live mosquitoes. While the N24A/D31N virus did poorly in vivo in all accounts, the N24D/D31H/N virus tended to infect both the bodies and heads of the mosquitoes better than the WT virus, though titers were reduced. The authors claimed, based on a DSF assay, that there were no real differences in ADP-ribose binding and thus suggested that these differences could be due to changes in substrate specificity, as the D31 residue resides in the substrate exit path, potentially tuning the virus to unique substrates in different species. The authors also produced crystal structures of the mutants to demonstrate the changes in the binding pocket caused by these mutations.

Strengths:

The authors have done a rigorous job of evaluating CHIKV macrodomain mutant viruses and the proteins' biochemical activities. The use of live mosquitoes is highly unique and provides important insights into the importance of the macrodomain in different species.

Weaknesses:

It is not clear if the interpretation of the ADP-ribose binding data is correct. It appears there are notable differences that could explain the results, though the authors chose to minimize the impact that these differences had on the results. The N24D-D31H/N proteins had at least a 1C degree difference in the thermal shift assay when compared to the N24A/D31N, single D31 mutants, and WT proteins, which is likely significant and could explain the dichotomous results between the two viruses in mosquito cells. Even the single N24D mutant had enhanced binding compared to the WT protein. Furthermore, as this virus has no enzymatic activity, one could hypothesize that enhanced binding to a substrate that is normally cleaved by the protein could certainly lead to alterations in phenotypic effects, whether good or bad. The authors should test the binding activity in a separate assay, such as an ITC assay, to determine if there are, in fact, binding differences or not. Having said this, it is likely that the impacts of these mutations on replication and transmission in human and mosquito cells are multi-factorial and could include both enhanced binding with altered substrate specificity amongst other activities.

We agree that the mutants may indeed have stronger binding for modified substrates than the WT protein; however, given that DSF is not a quantitative measure of binding affinity and that free ADPr is not the relevant ligand (in fact we do not know the relevant ADPr-modified molecule), we have refrained from speculating further than saying in the discussion:

“The progressive selection of D31H over D31N in the mosquito host further suggests that subtle differences in ADP-ribose substrate recognition may influence viral fitness in the mosquito environment in ways that are not yet understood.” (Lines 368-370)

Additionally, as both mutants had no detectable enzymatic activity but had quite different phenotypes in mosquitoes, I don't agree with the title stating that catalytic activity modulates dissemination and transmission potential in mosquitoes. It seems more likely that alterations in binding activity or substrate recognition (even suggested by the authors) impact these phenotypes in mosquitoes.

Regarding the title, we agree with the reviewer that it could be misleading, as both mutants lack catalytic activity yet show distinct phenotypes in mosquitoes. We will therefore modify the title to: "Loss of macrodomain catalytic activity modulates Chikungunya virus dissemination and transmission potential in Aedes mosquitoes", which more accurately reflects that the observed phenotypes arise as a consequence of the loss of catalytic activity at position N24.

Reviewer #3 (Public review):

Summary:

The authors investigated the role of the nsP3 macrodomain catalytic activity in the replication and transmission of CHIKV in mosquito vectors. The conserved dual-host alphavirus catalytic site N24 has previously been shown to be essential for ADP-ribosylhydrolase activity. Despite this, mosquito-specific alphaviruses do not share this catalytic site. To assess whether the macrodomain catalytic activity of a dual-host virus was essential in insect hosts, the authors targeted the N24 site to abolish catalysis while maintaining binding capacity. The loss of ADP-ribosylation led to the emergence of compensatory mutations at site D31 that impact viral infectivity, dissemination, and transmission in Aedes sp. mosquitoes in vivo. The conclusions are well supported by the results and provide insight into the importance of nsP3 macrodomain activity in the mosquito vector, which hasn't been explored before.

Strengths:

The main strength of this study is the use of Aedes sp. mosquito models to investigate the selective pressure of macrodomain mutations in vivo. The functional characterization as well as the structural analysis of the mutants provide supporting evidence of a potential role of the compensatory mutations at site D31 in substrate recognition.

Weaknesses:

A considerable part of this study relies on the use of N24 mutant viral stocks generated in Vero cells, which yields an additional mutation at site 31 and consequently doesn't allow the authors to properly dissect the effect of mutation of N24 and D31 independently. It would be recommended to generate stocks with individual mutations in both A549 and U4.4 cells, pooling and concentrating them if needed. Replication of the N24A mutant in A549 cells does not lead to mutation at residue 31. Yet surprisingly, there is no reversion from N back to D at site 31 when the double mutant Vero stocks are passaged in A549. Since they are double mutants, it isn't possible to assess whether the defects in the growth of mutants N24A/T-D31N and N24D-D31H/N compared to WT are due to site 24 or 31, or both (Figure 2, panel c). Even though the authors emphasize that the compensatory mutation could have additional roles that impact viral infectivity and transmission in mosquito cells, it would strengthen the work to show that these mutations would spontaneously appear in stocks generated directly in mosquito cells. As a corollary, is it known whether insect-specific alphaviruses that lack macrodomain catalytic activity have corresponding mutations at site 31?

We thank the reviewer for this important comment. Regarding the generation of viral stocks in A549 cells, transfection of N24A and N24D viral RNAs into A549 cells did not yield sufficient viral titers to produce usable stocks. However, as described in our response to Reviewer #1, we confirmed the reproducible emergence of D31 second-site mutations in C6/36 mosquito cells and BHK-21 mammalian cells, which will be included in the updated Supplementary Figure 1. These results demonstrate that D31 mutations spontaneously emerge in stocks generated directly in mosquito cells, addressing the reviewer's concern.

Regarding the question about insect-specific alphaviruses, we examined the sequence at position 31 using a multiple sequence alignment of 14 alphaviruses with diverse host ranges, including dual-host, insect-specific, and aquatic alphaviruses. We observed that Yada Yada virus (GenBank: QGR15362.1) has an asparagine (N), Tai Forest alphavirus (GenBank: YP_009333615) has an aspartic acid (D), Mwinilunga alphavirus (GenBank: BBC45634.1) has an aspartic acid (D), Eilat virus (GenBank: QBG67155.1) has an aspartic acid (D), and Agua Salud alphavirus (GenBank: QEV83787.1) has a lysine (K) at this position. These results suggest that insect-specific alphaviruses do not share a conserved residue at position 31, and therefore no clear conclusion can be drawn regarding a direct correspondence with the compensatory mutations observed in our study. This sequence alignment with the corresponding text will be included as supplementary data in the revised manuscript.

Additionally, there is a lack of consistency in the prevalence of WT virus at days 5 and 7 in in vivo experiments with Ae. albopictus and Ae. aegypti (Figure 3 and Supplementary Figure 2). This raises concern about the reproducibility of these experiments.

We thank the reviewer for this observation. We acknowledge that the prevalence of WT virus infection shows variability between experiments and timepoints. Based on our experience with infectious blood meal experiments, this variability is sometimes observed between independent experiments even under identical experimental conditions and with the same virus. In this particular case, each set of experiments was performed with independently produced viral stocks. Specifically, for the experiments shown in Figure 3, WT, N24A-D31N, and N24D-D31H/N viral stocks were produced in parallel from transfection of viral RNAs, and the same stocks were used for sequencing, growth curves, and mosquito infections. Subsequently, when we generated single D31H and D31N mutant viruses, a new WT viral stock was produced in parallel with the D31 mutant stocks, and these independently produced stocks were used for the experiments shown in Supplementary Figure 2. Differences in absolute infection rates between experiments are therefore expected, as they reflect both the use of independently produced viral stocks and the inherent variability in the efficiency of midgut infection and dissemination between mosquito cohorts. Importantly, the comparisons between WT and mutant viruses are always made within the same experiment, using stocks produced in parallel.

The inability to tease apart the roles of N24 and D31 in mosquito hosts partially prevented the authors from fully achieving their aims, but the work is nonetheless of interest to the field and suggests that more work is necessary to fully understand the role of the nsP3 macrodomain and its catalytic activity in the two disparate but obligate hosts for CHIKV and other dual-host alphaviruses.