Long-lived central memory γδ T cells confer protection against murine cytomegalovirus reinfection

The involvement of γδ TCR-bearing lymphocytes in immunological memory has gained increasing interest due to their functional duality between adaptive and innate immunity. γδ T effector memory (TEM) and central memory (TCM) subsets have been identified, but their respective roles in memory responses are poorly understood. In the present study, we used subsequent mouse cytomegalovirus (MCMV) infections of αβ T cell deficient mice in order to analyze the memory potential of γδ T cells. As for CMV-specific αβ T cells, MCMV induced the accumulation of cytolytic, KLRG1+CX3CR1+ γδ TEM that principally localized in infected organ vasculature. Typifying T cell memory, γδ T cell expansion in organs and blood was higher after secondary viral challenge than after primary infection. Viral control upon MCMV reinfection was prevented when masking γδ T-cell receptor, and was associated with a preferential amplification of private and unfocused TCR δ chain repertoire composed of a combination of clonotypes expanded post-primary infection and, more unexpectedly, of novel expanded clonotypes. Finally, long-term-primed γδ TCM cells, but not γδ TEM cells, protected T cell-deficient hosts against MCMV-induced death upon adoptive transfer, probably through their ability to survive and to generate TEM in the recipient host. This better survival potential of TCM cells was confirmed by a detailed scRNASeq analysis of the two γδ T cell memory subsets which also revealed their similarity to classically adaptive αβ CD8 T cells. Overall, our study uncovered memory properties of long-lived TCM γδ T cells that confer protection in a chronic infection, highlighting the interest of this T cell subset in vaccination approaches.

Fig. 5D gdT cells may be primed/trained in a TCR independent manner (first two weeks) and but primed/trained cells require TCR-mediated signals to execute effector functions in the third week (Fig. 5D).This would be in line with a MCMV TCR-specific response.Alternatively, the TCR antibody could provide a TCR-mediated signal interfering with effector functions triggered by innate/cytokine receptors (see above).
We agree with the reviewer and added the following sentence in the discussion (p21): "Our TCRblocking experiment suggest that TCR-signaling is important for viral load control after reinfection.However, we cannot fully exclude that  TCR-antibody also affects effector function or survival of innate or cytokine receptor-dependent activation of  T cells".
Reviewer #4: The manuscript by Yared et al. describes an interesting adaptive-like feature of gd T cells in response to MCMV infection, extending on two previous publications appearing almost at the same time.In their manuscript, the authors elaborate on the crucial question whether immunological memory is generated in the gd T cell compartment.One of the previous publications already suggested a major difference in protecting immunodeficient mice when gd T cells from infected mice were used in transfers.In the current manuscript, the authors then start to characterize the features and functions of the memory-like gd T cells.
The major problem in the characterization is the assumption that well-established markers for memory subpopulations of CD8+ aß T cells, particularly TEM and TCM are transferable to the gd lineage.Without comparing the respective populations side by side, this is a dangerous misconception, in my opinion, as it paves the way for future simplifications in following studies.

Characterization of  T cell memory subsets based on previous knowledge on β T cells has been useful
in the past since it allowed, to some extent, to follow pathogen-responding  T lymphocytes.Indeed, their peculiar mode of Ag recognition and the paucity of knowledge on  TCR ligands preclude the use of tetramers.
Though, we are aware of existing differences between the two T cell subsets and we apologize since we did not want to stress that  TCM and TEM are strikingly similar to their β T cell counterparts.We now added this sentence at the beginning of the results section: "Based on previous knowledge on β T cells, we used CD44 and CD62L expression to characterize  TCM-like (CD44+CD62L+) and TEM-like (CD44+CD62L-) subsets, that will be further on referred to as  TCM and TEM for simplification".
In addition, we have previously compared CMV-specific CD8 αβ T cells and  T cells responding to human CMV and showed that they share a common effector memory phenotype and common expansion kinetics during the course of infection (Couzi et al., J Infect Dis. 2009 Nov 1;200(9):1415-24. doi: 10.1086/644509).
The authors already show in the experiment for Figure 1, for example, that gd cells do not compare to the classical categories, also not regarding expansion and contraction, at least to a much lesser extent."TEM" cells would express CX3CR1 and KLRG1, in naïve mice most of them don´t, however.
As depicted in our previous analysis (Khairallah et al., PLoS Pathog. 2015 Mar;11(3): e1004702. doi.org/10.1371/journal.ppat.1004702), the  T cell-response to primary MCMV-infection shows a rather « classical » (non inflationary) kinetics with expansion of  T cell numbers in organs at d7-14 followed by contraction at d21.In the long-term,  T cells acquire a similar CD44+CD62L-KLRG1+ phenotype as so-called inflationary CD8 αβ T lymphocytes.As reported for CD8 αβ T cells, the initiation and amplitude of the  T cell response, as well as the proportion of CD44+CD62L-KLRG1+ cells in the long-term might depend on the viral dose (Redeker et al. Front. Immunol., 2018, 8: 1953, doi.org/10.3389/fimmu.2017.01953) .In the latter study, analysis in blood of uninfected mice showed that about 42% of CD62L-CD8 αβ T cells were KLRG1-.However, in this study and others, CD44+CD62L-KLRG1+ CD8 αβ T cells are defined as TEM, while we only used CD44 and CD62L.This emphasizes the importance of marker selection for cell-subset definition, as mentioned by reviewer.
A detailed comparison by the transcriptional landscapes would be helpful to dissect the gd subpopulation and compare them to the CD8 aß subpopulation.The analysis in sFig.2 goes in this direction.As this analysis does not differentiate the subpopulations, it is not able to clarify this critical issue, however.
This transcriptional landscape comparison between  TEM and  TCM has been done through single cell analysis.We thank the reviewer for this idea to compare  to CD8 αβ subpopulations.However, as mentioned above, we think that to be informative these comparisons should be done using  and CD8 αβ T cells from the same CMV-infected mice.Since we are using TCR α -/-mice this is unfortunately not possible Strangely, a significant attempt to differentiate the subpopulations is made in Figure 9, most likely because of the previous review process.These "new" data should be shown and discussed after Figure 1 and include a much more detailed bioinformatic analysis, possibly including a comparison with CD8 aß TEM and TCM as they can easily be obtained from deposited data sets.

As suggested by reviewer, a new version of our single cell analysis is now presented (figure 9), and a novel analysis is discussed in the results section. We think that these new data are interesting to describe after our transfer experiment, showing a differential protective function of splenic  TCM-like and TEM-like cells, while Fig 1 displays lung and liver.
These new data have been confined to the GEO repository under the accession number GSE269394.To review GEO accession GSE269394: Go to https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE269394.Enter token oxknemsuppuznwh into the box.
Reviewer #4: The analysis in Fig. 2 is important and the data appear clear in showing that most tissue-based gd T cells 3m after infection appear in the blood.It would be important to strengthen this finding by a control experiment for tissue-resident gd T cells in the same experimental setup, even from the same mice.Dermal or intestinal gd T cells would be very good controls.(Anderson et al. Nat Protoc 9, 209-222 (2014) doi: 10.1038/nprot.2014.005).As a control experiment, when setting up the experiment shown in Fig 2 , we compared flushing or not with buffer, the lung and liver vasculature, for few mice.The proportion of stained CD45+ cells was highly reduced (80%) in the lung and, to a lesser extent (60%) in the liver, sustaining the removal of circulating cells only upon perfusion.

As mentioned in the result section (p9), we used a well-validated labeling technique of circulating cell by intravenous injection of conjugated CD45 mAb
In Fig. 3 the strong reduction of TCM in lung, liver and blood after FTY720 treatment is interpreted as block of lymph node egress.This statement is not valid unless data are shown that TCM are, in fact, residing and, after treatment, accumulating in lymph nodes.They might be otherwise affected.(Smith et al. 2014. PLoS Pathog. 10: e1004233).

FTY720 has been regularly used as a blocker of T cell egress from lymph node. In the setting of MCMV infection, FTY was used to show that maintenance of MCMV-specific effector CD8 T cells does not depend on migration through or antigen recognition within the lymph
In contrast to the strong FTY720-mediated decrease of  TCM observed in liver, lung and blood (Fig 3),  TCM remained unaffected in the lymph nodes (below), although we could not show an accumulation.
We only dissected mesenteric lymph nodes that might only partially reflect what happens in the whole mice.So, in agreement with reviewer's comment, we amended our interpretation in the text and now mention FTY720 only as a product that allows to decrease the number of  TCM in organs (p9).The data in Fig. 3b should not be shown here as there is no clear conclusion possible.
In fact we did a mistake on the p values in the submitted version since the statistics where done on a single experiment whereas the data are pooled from 2 experiments.We corrected this in Fig 3B   We agree with the reviewer that these experiments with the GL3 mAb mainly show that gd T cells are involved in the control of long-term MCMV infection but specific recognition by the gdTCR cannot be drawn by this experiment.The anti-TCRgd mAb might have affected gd T cell functions others than IFNg production, but the extent of functions potentially involved in viral control (cytotoxicity, repair factor production, G or GM-CSF production, chemokine production, etc…) that could be tested would be too large.Thus, we tempered our conclusion of the involvement of a specific recognition by gd TCR by changing the text on page 11.
The positive effects of transfer of gd T cells from long-term infected mice is extending previous findings described in PMID: 25658831 This should at least be mentioned and the additional gain of insight should be worked out.
We agree with the reviewer that adoptive transfer with gd T cells from short-term (between 2 and 3 weeks) infected mice has been performed in (Khairallah et al, PLoS Pathog. 2015    Transfer experiments using  T cells from liver and, above all, from lung are more difficult due to poor  T cell numbers.However, we performed some experiments and, in the case of liver, we were able to sort KLRG1+ and KLRG1-TEM subsets.As shown in the figure below, no protection was evidenced using either subset.Thanks to reviewer for it attentive reading.We have changed this label.

Fig. A. Evolution of the proportion of naïve, TCM and TEM  T cells in organs along the course of MCMV infection in
Reviewer #5: I had a chance to review the article "Long-lived central memory γδ T cells confer protection against murine cytomegalovirus reinfection" by the group of M. Capone that was resubmitted to PLOS Pathogens upon in initial review.Since one of the primary reviewers appears to be unavailable, I was asked to step in during the secondary review.In fairness to the authors, my review does not address questions that were not raised by reviewers during primary review, but in fairness to the process, I had to go through a detailed assessment of the article and raise objections where I identified concerns about data and their interpretation.Overall, I think that the study is interesting and relevant to the readers of PLOS Path., but that several item need to be addressed as major concerns that require additional experimental evidence.The list below is in order of appearance in the text: 1.In Fig. 3B, the authors were requested address the weak stat power of their experiment, resulting in a borderline p value.They addressed this request by altering the interpretation of results, which is not an appropriate course of action: they should instead repeat the assay, generate additional evidence and define if the FTY720 effects on viral loads are significant and reproducible or not.
In fact we did a mistake on the p values in the submitted version since the statistics where done on a single experiment whereas the data are pooled from 2 experiments.We corrected this in Fig 3B where a statistical increase (p= 0.022) is now observed in the liver, while the p value in lung is 0,188.Please find down below the raw data of viral load copy numbers where 1= limit of viral loads detection/PCR Viral loads in organs of treated or untreated mice were quantified by qPCR.Values represents MCMV copy numbers in 1 µg of total DNA.4B, the authors show viral loads as fold changes relative to Day(0) upon primary or secondary infection.They start from a value of 1 in either scenario to show a drastic increase in primary, but not in secondary infection.The problem is that at day 0, the mice were MCMV naïve and the value had to be 0. Therefore, any increase would end up as an infinite-fold increase.The authors have obviously used an arbitrary value for the day 0 of the primary infection, which is not immediately obvious.This needs to be explained more accurately.

In Figure
Thanks for this comment.The arbitrary value that has been used for day 0 of naïve mice is the detection limit of the PCR used to titrate the virus.However, since these viral load results did not convince several reviewers, we decided to delete them from this new version of the manuscript.

3.
In reference to the first comment by Rev#3, the authors have indeed produced no data that the MCMV indeed establishes latency, rather than persisting in their system.Since it is well known that CD4 T cells are required for MCMV control in salivary glands, it is reasonable to assume that their model may be affected by MCMV persistence in this organ (and spurious dissemination of virus to other organs from this focus of persistence The conclusions of this assay will not preclude publication, but they will substantially influence data interpretation (e.g. they may indicate that there is no additional spike in viral loads upon reinfection, because the virus is lytically persistent even before reinfection).Hence, this evidence is critically important.
We agree with the reviewer that our model is not a model of virus latency (as explained in our previous response to reviewer #3) and sorry if there is a misunderstanding because we do not want to claim that this is the case.We know from our viral load data that we can detect up to 10 3 DNA viral copies in organs three months after infection (see e.g. Fig 3B and Fig 5C).This maintenance of viral load is probably due to the absence αβ ab T cells in this model, and more particularly CD4 T cells (Jonjic et al. J Exp Med (1989) 169 (4): 1199-1212. https://doi.org/10.1084/jem.169.4.1199;Zangger et al. Pathogens. 2021 Dec;10(12) Despite the difficulty to compare our results with other studies that don't necessarily use similar mouse strain, infection route, viral strain/origin…, the titer found appears higher than expected in 2 months infected C57BL/6 wild-type mice (Brunel et al. PLoS Pathog 20(1): e1011852. doi.org/10.1371/journal.ppat.1011852 January 18, 2024).So, although γδ T cells were found in SG from MCMV infected mice (Cavanaugh et al. J. Virol., Feb. 2003, p. 1703-1717Vol. 77, No. 3. DOI: 10.1128/JVI.77.3.1703-1717.2003),they are probably not sufficient to fully control the virus.
As mentioned above we decided to delete Figure 4B, and we modified the text (p18).

Minor issues
1.In Figure 5D, the authors now show that the depletion of gd T cells prior to MCMV primary infection results in no effects on viral titer at dpi 7. Nevertheless, the effects are prominent if the depletion occurs later with a similar waiting time of 7 days.How can this be?The authors need to discuss these data.
We would like to clarify that the GL3 mAb does not deplete gamma-delta T cells but induces the internalization of γδ TCR (Koenecke et al, Eur J Immunol. 2009 Feb;39(2):372-9. doi: 10.1002/eji.200838741.).Hence, these results indeed show that this is not the duration of the treatment by GL3 with is important, but the timing when γδ T cells are treated.The data indicate that at least 2 weeks of priming are required for γδ T cell antiviral function to be active.This now better explained in the text (p.11) Also, the marking next to the y axis in Fig. 5D spells ADN, instead of DNA.
Thanks to reviewer for it attentive reading.We have changed this label.
2. The methods used to perform the cluster analysis in Fig. 6C are rather confusing.More detail and careful wording may be sufficient to address this issue.
We modified the paragraph in the material and methods section as such "To evaluate the bias in the γδ TCR repertoires (Miles et al. Immunology and Cell Biology (2011) 89, 375-387;doi:10.1038/icb.2010.139),we computed a distance matrix of pairwise Levenshtein distances (LD) between CDR3aas, using 'stringdist' R package (M.P.J. van der Loo, 2014, The R Journal 6(1):111-122).When two sequences are similar under the defined threshold, LD>1 (i.e., at most one amino acid difference), they were connected and designated as "cluster".CDR3s with more than one amino acid difference from any other sequences are not connected.The number of CDR3aa making up the cluster represents the size of the cluster.The clustering was between d0 and d7R (d0/d7R) and between d21 and d7R (d21/d7R).The mean size of the clusters was compared between d0/d7R and d21/d7R per mouse".
3. In Fig. S11B, it is not clear what is shown by bubble size.Are these cells from infected vs. uninfected mice?Did they compare the nr. of genes per cluster, or the aggregate gene expression levels of genes belonging to a pathway/cluster?Or something else?
We changed the representation and now show (Fig 9D) selected pathways from the Reactome database, that are significantly enriched for up-and down-regulated genes when comparing C5 and C6 clusters (corresponding to γδ TCM and TEM enriched upon MCMV infection,respectively).

Fig S2B .
Fig S2B.Please enlarge the symbols or give them different colors.

Fig.
Fig. Analysis of the proportion (left) and number (right) of  TCM in lymph nodes of FTY-treated and control TCRα-/-mice.

Figure 4
Figure4shows a somewhat faster and to some extend stronger response in secondary infections.The extent is, however, much less pronounced when compared to CD8 ab responses.This should be mentioned somewhere in the manuscript.
Fig. A. Evolution of the proportion of naïve, TCM and TEM  T cells in organs along the course of MCMV infection in TCRα-/-mice.B. Evolution of the proportion of KLRG1+ cells among  TEM.

Figure 8
Figure8and 9 are the most crucial parts in the manuscript as a clear protective role of TCM-like cells is shown by adoptive transfer and analyzed in detail by the scRNA experiment.The weakness is, however, that here only splenic gd T cell populations are used for analysis as well as transfer.Given the potential importance of virus persistence / latency in lung and liver, only "TEM-like cells" from the respective organs might confer protection upon transfer.This would lead to a fundamentally different conclusion, in my opinion.

Fig.
Fig. Survival curve of CD3ε-/-mice receiving 200 000 KLRG1+ and KLRG1- TEM from liver of three months infected TCRα-/-mice.The number of mice for each transfer group is indicated.

Figure 9
Figure 9 needs more attention in the text, and the analysis appears rather rudimentary.At least a description of the 8 UMAP clusters should be presented, for example by displaying the top up-or downregulated genes and signature genes as bubble or violin plots where a statistical increase (p= 0.022) is now observed in the liver, while the p value in lung is 0,188.Please find down below the raw data of viral load copy numbers where 1= limit of viral loads detection/PCR