Immune protection is dependent on the gut microbiome in a lethal mouse gammaherpesviral infection

Immunopathogenesis in systemic viral infections can induce a septic state with leaky capillary syndrome, disseminated coagulopathy, and high mortality with limited treatment options. Murine gammaherpesvirus-68 (MHV-68) intraperitoneal infection is a gammaherpesvirus model for producing severe vasculitis, colitis and lethal hemorrhagic pneumonia in interferon gamma receptor-deficient (IFNγR−/−) mice. In prior work, treatment with myxomavirus-derived Serp-1 or a derivative peptide S-7 (G305TTASSDTAITLIPR319) induced immune protection, reduced disease severity and improved survival after MHV-68 infection. Here, we investigate the gut bacterial microbiome in MHV-68 infection. Antibiotic suppression markedly accelerated MHV-68 pathology causing pulmonary consolidation and hemorrhage, increased mortality and specific modification of gut microbiota. Serp-1 and S-7 reduced pulmonary pathology and detectable MHV-68 with increased CD3 and CD8 cells. Treatment efficacy was lost after antibiotic treatments with associated specific changes in the gut bacterial microbiota. In summary, transkingdom host-virus-microbiome interactions in gammaherpesvirus infection influences gammaherpesviral infection severity and reduces immune modulating therapeutic efficacy.

www.nature.com/scientificreports www.nature.com/scientificreports/ to herpesviruses infections (e.g., zoster) 7,8 . A lethal IVS large vessel arteritis model which closely mimics human Giant Cell Arteritis, Kawasaki's disease and Takayasu's arteritis can be induced by high dose intraperitoneal MHV-68 infection of interferon gamma receptor knockout (IFNγR −/− ) mice 9 . Many infected mice display extensive pulmonary hemorrhage and consolidation, mimicking DIC in viral sepsis, which has a very high attendant mortality 10 . A significant number of mice also have marked colon dilatation reminiscent of toxic megacolon 11 , in addition to aggressive inflammatory cell invasion with hemorrhagic lung consolidation. Herpesvirus infections have been associated with inflammatory bowel diseases [12][13][14] and gastrointestinal involvement in systemic vasculitides has been reported by other groups 15,16 . As other gastrointestinal inflammatory conditions are associated with, or driven by, microbiome changes, the gut microbiome may have a role in the pathogenesis of GHV and specifically MHV-68 infections.
We have previously reported that immune modulating proteins from myxomavirus (MYXV) provide a new class of anti-inflammatory treatments when delivered as recombinant protein or by ectopic expression after Adeno-associated virus (AAV) delivery [17][18][19][20] . MYXV is the causative agent of lethal myxomatosis in the European rabbit (Oryctolagus cuniculus), expressing highly potent immune evasion proteins that act as virulence factors in MYXV infections. MYXV is not a pathogen for mice or humans. When purified as isolated recombinant proteins, these immune modulating biologics can modify disease progression in a wide range of inflammatory diseases in preclinical animals and man 21 . In prior work, the MYXV-derived serine protease inhibitor (serpin) Serp-1 and a Serp-1 reactive center loop (RCL) peptide S-7 (G 305 TTASSDTAITLIPR 319 ) significantly improved survival, reducing pulmonary and aortic inflammation as well as hemorrhagic lung consolidation after MHV-68 infection 10,22 . Serp-1 targets the coagulation pathways, both thrombolytic and thrombotic, and has proven, highly effective anti-inflammatory functions 23,24 . Antibiotic treatment of IFNγR −/− mice prior to MHV-68 infection abrogated the efficacy of S-7 treatment for lethal MHV-68 infections, leading to significantly reduced survival. Therapeutically modified S-7 peptides (MPS7-8 and MPS7-9), designed based upon the serpin crystal structure, maintained efficacy in this model with improved survival after MHV-68 infection 25 . The MPS7 peptides have predicted increased hydrogen bonds when compared to the native S-7 peptide 25 .
The pathophysiologic role of the gut bacterial microbiome in gammaherpesviral infections and on immune modulating treatments has not previously been investigated. We report here a systematic examination of the role of the gut bacterial microbiome in MHV-68 infection and on Serp-1 and Serp-1-derived S-7 peptide treatment in MHV-68 infections. 26 , causing only flu-like symptoms. Conversely, intraperitoneal infection of mice with underlying immune deficiency, as for IFNγR −/− mice, results in a lethal large vessel vasculitis, colitis and hemorrhagic pneumonia with early mortality 9,27 . Intranasal infection, an alternative model for MHV-68, is not a model for severe vasculitic syndromes. Our group and others have used this intraperitoneal infection model of vasculitis to study the pathogenesis of gammaherpesvirus host-pathogen interactions, as well as to investigate the potential for anti-inflammatory treatment of associated inflammatory vasculitic syndromes 10,11,22,25,28,29 . Intraperitoneal infection of IFNγR −/− mice with 12.5 × 10 6 pfu was performed with follow-up for 150 days to determine survival rates. MHV-68 infection with administration of control saline treatment alone for 30 days was lethal, with a median survival of 41 days. Serp-1 (100 ng/g) or S-7 (100 ng/g) intraperitoneal (IP) injections for 30 days improved survival to 60% at 150 days (p = 0.0022 and p = 0.0218, respectively) (Fig. 1B). Thus, Serp-1 and S-7 protect IFNγR −/− mice after lethal MHV-68 infection 10,22 . Loss of gut bacteria accelerates MHV-68 induced disease. We hypothesized that the gut bacterial microbiome may play a role in the host-pathogen interactions of MHV-68 induced disease through transkingdom interactions. We completed a systematic analysis of the effect of microbiome depletion on MHV-68 induced disease progression ( Fig. 2A). Mice were maintained on medicated water containing a broad-spectrum cocktail of four antibiotics: bacitracin, gentamicin, streptomycin and ciprofloxacin (Table 1) in order to suppress gut bacterial populations. During this time the density of the gut bacteria (as measured by colony forming units [CFU] per gram fecal pellet) was markedly reduced, validating a decrease in bacterial load (Fig. 2B). After ten days, mice were returned to normal drinking water for 24 hours, during which time the antibiotic treatment maintained suppression of the microbiome (Fig. 2B). Mice were subsequently infected by IP injection of MHV-68 and injected with control saline daily for 30 days and survival assessed for up to 150 days (Fig. 1B).

Serp-1 protein and S-7 peptide treatments improve survival after MHV-68 infection in ifnγR −/− mice. MHV-68 infections establish latency with chronic infection in wildtype mice
Suppression of gut bacteria by antibiotic treatment markedly accelerated the course of MHV-68 infection. Saline-treated mice exhibited a median survival of 41 days, while antibiotic-treated mice had a median survival of 19 days (p = 0.0411; Fig. 1B). 16 S rRNA gene amplicon sequencing was subsequently performed on DNA extracts from the large intestines of individual mice at 3 days post-MHV-68 infection in order to assess early changes in the gut microbiota. We detected a distinct dysbiosis in mice pre-treated with antibiotics prior to MHV-68 infection versus mice with a stable bacterial microbiome prior to infection (Fig. 2C). This dysbiosis was associated with a non-statistically significant trend towards increased diversity in amplicon sequence variant (ASV) richness ( Fig. 2D; p = 0.329). Sub-population analysis revealed significant changes in candidate ASVs in MHV-68 infected mice that were associated with antibiotic treatment (Fig. 2E, top panels; increased abundance after antibiotics) or with no antibiotic treatment (Fig. 2E, bottom panels; decreased abundance after antibiotics). These results suggest that the gut bacterial microbiome is a significant determinant of MHV-68 induced disease progression.
Serp-1 and S-7 peptide treatment efficacy is dependent on the gut bacterial microbiota. We next investigated the potential for dependence of Serp-1 and S-7 therapeutic efficacy on the gut bacterial microbiome in MHV-68 infected IFNγR −/− mice. Mice were treated with medicated drinking water (Table 1)    placed on normal water for 24 hours, then infected with MHV-68 and concurrently treated for 30 days with either Serp-1 or S-7 by IP injection. Mice were followed for survival or 3-day follow-up ( Fig. 2A). As previously reported, Serp-1 and S-7 peptide improved survival with 60% of mice surviving to 150 days after MHV-68 infection. In contrast, antibiotic-treated mice had accelerated disease, with a median survival of 19 (Serp-1; p = 0.0025; Fig. 1B) and 18 days (S-7; p = 0.0339; Fig. 1B). Serp-1 or S-7 efficacy after antibiotic treatment is not significantly different from saline treatment after antibiotics (Serp-1 + Abx vs Saline + Abx, p = 0.2012; S-7 + Abx vs Saline + Abx, p = 0.9164). Thus, the efficacy of Serp-1 or S-7 treatment and protection in lethal MHV-68 infections was dependent on the gut bacterial microbiome.
Microbiome analysis identifies candidate taxa associated with Serp-1 and S-7 treatments. Based on the findings that (1) antibiotic-treated mice accelerated and exacerbated MHV-68 disease, reducing survival time and increasing mortality, and that (2) immune protection conferred by Serp-1 and S-7 was highly sensitive to antibiotic treatment, we profiled the gut bacterial microbiota of mice treated with Serp-1 or S-7. 16S rRNA gene amplicon sequencing was performed on DNA extracts derived from the large intestines of individual mice 3 days post-infection with MHV-68 and after treatment with Serp-1 or S-7 (Fig. 3). Serp-1 and S-7 treated mice, in the absence of antibiotics, had a higher diversity of microbiota (median = 1.270, p = 0.0079; median = 0.862, p = 0.1508 respectively) than saline treated mice (median = 0.385) (Fig. 3B). Principal coordinate analyses indicate that although the gut bacterial microbiomes of Serp-1 and S-7 treated mice differed from saline treated mice, there was variability within Serp-1 and S-7 treated mice (Fig. 3C, Supplementary Fig. 1B). This finding was consistent with the phenotypic variability observed in vivo. Specifically, there was partial protection by Serp-1 and S-7 treatment with 60% survival (Fig. 1B).
We hypothesized that responsiveness to Serp-1 and S-7 treatment may be driven by the presence of "protective" ASVs (differentially associated with Serp-1 and/or S-7 treatment) or the absence of "potentiator" ASVs (differentially associated with saline treatment). Therefore, to identify potential microbe(s) responsible for the microbiome-mediated protection by Serp-1 and S-7, we analyzed the gut bacterial microbiota for discriminant ASVs that were differentially represented in Serp-1 and/or S-7 treated mice compared to saline treated mice, in the absence of antibiotics (Supplementary Fig. 1C-E). Of the 15 ASVs identified as differentially associated with Serp-1/peptide treatments, ASV4, was consistently associated with both Serp-1 and S-7 treatment (Fig. 3D). ASV4 was most closely-related to the sequence of Bacillus massiliogorillae 30 . Conversely, ASV1 (most closely-related to Frondihabitans peucedani 31 ) and ASV123 (most closely-related to Enterococcus saigonensis 32 ) were identified from 8 ASV candidates as being associated with saline treatment when compared to both Serp-1 or S-7 treatments. Based upon the fact that Serp-1 and S-7 protected mice in an antibiotic-dependent manner, we reasoned that the relative abundance of the discriminant ASVs would be altered in mice that were pre-treated with antibiotics. As predicted, the relative abundance of Serp-1-or S-7-associated ASVs was decreased in antibiotic pre-treated mice, albeit to a more moderate extent in S-7 ( Fig. 3E,F, compare "No Abx" to " + Abx"). These results suggest that the interplay between with bacterial microbiota like ASV4, that are associated with a protective phenotype, and microbiota such as ASV1 and ASV123 that are associated with a patho-exacerbative phenotype, can influence the outcome of immune modulating treatments. In summary, responsiveness to immune modulating therapy in MHV-68 induced disease is associated with specific alterations in the gut bacterial microbiota.
Antibiotic-dependent exacerbation of early lung pathology of MHV-68 induced disease. MHV-68 persistent infection 33 leads to severe hemorrhagic and consolidating pulmonary pathology [34][35][36] . Hence, we investigated whether the pulmonary pathology induced by MHV-68 infection reflected the antibiotic-dependent exacerbation we observed in survival analysis. We performed quantitative morphometric analyses of lungs at an early, 3-day follow-up after infection. In the absence of antibiotics, considerable early pulmonary consolidation and inflammation was observed in MHV-68-infected lung tissue in IFNγR −/− mice treated with control saline alone. This severe pulmonary pathology was considerably reduced by Serp-1 or S-7 peptide treatment (Fig. 4A) 10 . Early lung pathology was markedly worse after antibiotic treatment in all conditions (saline, Serp-1 or S-7), with most animals in the saline group also showing hemorrhage in affected regions.
Previous reports have demonstrated that thickening of the alveolar wall/septa and reduction of alveolar lumen area are reliable indicators of pulmonary inflammation in acute laboratory models [37][38][39] . In this study, quantitative morphometry of the lungs revealed a significant reduction in alveolar wall thickness in MHV-68-infected mice treated with Serp-1 or S-7 versus saline-treated controls that was lost after treatment with antibiotics (Fig. 4B). Without antibiotic pre-treatment and when compared to saline-treated controls, alveolar lumen area was significantly increased in mice treated with S-7 (p = 0.0067) with a trend toward an increase in mice treated with Serp-1. Antibiotic treatment nullified any level of protection by treatment with S-7 (Fig. 4C). No diagnostic pathology nor significant changes in inflammatory cell infiltrates on histological analysis was noted in the gut or aorta at this early, 3-day follow-up time (not shown). These results indicate that immune protection against pulmonary inflammation and early stage hemorrhage promoted by Serp-1 and S-7 treatment are microbiome dependent.

Discussion
Compelling evidence has established an important role for the gut microbiome in immune responsiveness [43][44][45] as well as the pathogenesis of various diseases, including those with primary involvement outside of the gut, such as respiratory 46 , hepatic 47 , renal 48 , neurologic 49 , autoimmune 50 , rheumatoid 51 and cardiovascular 52-54 conditions. Some research groups have demonstrated that targeted manipulation of the gut microbiome to correct dysbiosis, as for fecal transplants, may aid in treatment of certain diseases. For example, a small, randomized clinical trial (NCT02636647) found improved outcomes in recurrent hepatic encephalopathy patients treated by fecal microbiota transplant from a single, rationally selected donor enriched with Lachnospiraceae and Ruminococcaceae (taxa determined to be depleted in hepatic encephalopathy patients) 55 . A second trial (NTR1776) found that fecal microbiota transfer from lean donors to obese donors partially reversed symptoms of metabolic syndrome, including improvements in insulin resistance 56 . Aside from correcting dysbiosis, recent evidence supports the hypothesis that microbiome composition also plays a significant role in determining the efficacy of some treatments. It was recently found that low gut microbiome levels of Akkermansia muciniphila caused non-responsiveness in patients receiving PD-1 blockade immunotherapy for epithelial tumors [57][58][59] , reinforcing the importance of the microbiome in treatment efficacy, as highlighted by a critical earlier study that identified a positive correlation between Bacteroides fragilis and anti-CTLA4 therapy efficacy for melanoma 60 . Alternatively, some treatments may inadvertently alter the gut microbiome and predispose to increased pathogenic infections. For example, it was recently demonstrated that proton pump inhibitors significantly reduced microbial diversity with changes in more than 20% of bacterial taxa, leading to increases in Enterococcus, Streptococcus, Staphylococcus and the potentially pathogenic Escherichia coli species 61 . Thus, an understanding of how the gut microbiome influences, and is influenced by, new therapeutic treatments is crucial.
Trans-kingdom interactions may also play a crucial role in dictating disease pathogenesis in mammalian hosts. For viruses which require enteric bacteria for replication, antibiotics may reduce viral load. Numerous studies have demonstrated this viral-bacterial interaction in laboratory systems, such as murine and human norovirus www.nature.com/scientificreports www.nature.com/scientificreports/ infection 62,63 . For other viruses, however, loss of the bacterial microbiome may lead to worsening clinical disease. A recent study by Thackray et al. reports that oral antibiotic (ampicillin) treatment exacerbates disease severity in flavivirus infections by modulating flavivirus-specific CD8 responses 64 , suggesting a microbiome, immune system and virus interaction. These disparate observations, which depend substantially on (a) the virus being studied, (b) the host system and (c) the antibiotic regimen, highlight how the virome is still a poorly understood, and consequently under-appreciated, component of health and disease 65 .
We recently reported a severe gastrointestinal pathology in a significant fraction of IFNγR −/− mice intraperitoneally infected with MHV-68 11 , a model for systemic arteritis and hemorrhagic pneumonia 9,10 . In this same model, we also previously reported therapeutic efficacy with the MYXV-derived serine protease inhibitor Serp-1 10 as well as peptides derived from the Serp-1 reactive center loop 22,25 . On this basis, we undertook a systematic analysis of the effect of Serp-1 and the RCL-derived Serp-1 peptide S-7 on MHV-68 induced disease.
A broad-spectrum antibiotic cocktail was used to suppress the gut microbiome prior to MHV-68 infection. We found that suppression of the gut microbiota prior to MHV-68 infection accelerated disease, with increased lethality and earlier death occurring at approximately twice the rate of untreated mice (Fig. 2C). This acceleration of MHV-68 disease was accompanied by a dramatic post-antibiotic dysbiosis and a suite of candidate ASVs with increased or decreased relative abundance associated with worsened disease outcomes.
Furthermore, we demonstrate here that bacterial microbiome interactions were required for a protective immune response induced by Serp-1 or S-7 10 . Microbiome ablation completely abolished treatment efficacy and improved survival (Fig. 1B). Probing further, we found that Serp-1 and S-7 facilitated protection against lung inflammation and consolidation associated with an increase in CD3+ cell invasion (with a CD8+ bias) and a reduction of MHV-68 staining in the lungs (Figs. 4 and 5). This finding is in agreement with previous reports that T-cell surveillance (effected by CD8+ cells) is critical for limiting acute GHV infection and that a progressive loss of T-cells is associated with worsened GHV disease [40][41][42] . The loss of Serp-1-based treatment efficacy was not expected to be due to direct interactions between the antibiotics and immune modulating treatments for a number of reasons: (1) the antibiotics used in this study have a relatively short half-life (1-7 hours) and previous studies report negligible concentrations 24 hours after dosing 66-69 -a timeline we incorporated in this study design; (2) we observed a rapid recovery of microbial burden within three days (Fig. 2B); (3) the antibiotics in our cocktail are not known to suppress immune responses on their own, and in some reports are observed to enhance T-cell activity 70 ; (4) it is not expected that the loss of activity is due to direct interactions with the antibiotics because in vitro inhibition of uPA activity by Serp-1 is unaffected by the presence of the same antibiotic cocktail (Fig. S2). A recent report by Yang et al. indicated that in the short-term, antibiotics can induce metabolite changes which have systemic, suppressant effects on immune function in the peritoneal space by altering respiratory activity 44 . Yang et al., however, utilized an E. coli infection and the antibiotic, ciprofloxacin, was given within 4 hours of infection and maintained during the course of the infection. Here, we utilize a gammaherpesvirus infection with longer pathogenic kinetics and remove antibiotics 24 hours prior to infection, wherein any trace of ciprofloxacin would be undetectable 66 , and we are not aware of any other reports suggesting immune suppressant effects of the other antibiotics in our cocktail. Taken together, these data provide the first evidence that microbiome interactions are essential for a protective immune response, induced here by Serp-1 and S-7 treatment. We have previously found Serp-1 to function through the urokinase-type plasminogen activator receptor (uPAR) in systems involving immune cell infiltration (e.g., wound healing 71 , chemokine-induced ascites 72 ). Levels of uPAR have been found to increase substantially in incidences of infection and sepsis 73 . Thus, future work will investigate the possibility that the microbiome is modulating levels of binding partners essential for Serp-1 and S-7 function in MHV-68 infection.
We propose here that transkingdom interactions with the gut bacterial microbiome are required for host innate immunity to mount a protective response against lethal gammaherpesviral disease (Fig. 6). When Interestingly, the gut microbiota of all Serp-1 and S-7 treated mice did not clearly cluster separately from saline treated mice. Thus, the dynamics of selected increased or decreased microbiome ASVs is associated with worsened pathology or alternatively protective immune modulation. Given that not all mice were fully protected by Serpin treatment, we speculate that the gut microbiota heterogeneity of Serp-1-or peptide-treated mice may reflect the true variability in immunological outcomes (i.e. 60% protection). We found specific ASVs that were either increased in response to Serp-1/peptide treatment (e.g. ASV4), or decreased when compared to saline treatment (e.g. ASV1 and ASV123).
Although the current study design could not assess the predictive value of these discriminant ASVs on individual mice survival outcomes, future studies will need to be performed in order to stratify the role of different microbiome sub-populations by specific (i.e., single) antibiotic treatment as well as defined microbial reconstitution (e.g., with the Altered Shaedler Flora community or similar) 74 . Taken together, these findings suggest that the responsiveness to immune modulations in MHV-68 disease involves a combination of interactions with protective microbiota and patho-exacerbative microbiota. Further work will be needed to establish a direct role for bacteria related to these select ASVs to prove cause-and-effect relationships and to define a precise mechanism of bacterial microbiome influence on immune responses in GHV disease.
With this study, we report that the pathogenicity of MHV-68 infection and effective treatment with the immune modulating serpin, Serp-1, and serpin peptide, S-7, is substantially accelerated by suppression of the gut bacterial microbiome with antibiotics. MHV-68 is a model virus for studying GHV pathogenesis and severe viral sepsis 4,75 , therefore this finding may have implications for the pathogenesis and treatments of other GHV, such as KSHV and EBV, where adverse responses to antibiotics are clinically recognized 76 . We further report here the modulation of gut microbiome composition by an immune modulating protein and peptide. In conclusion, (1) microbiome changes may have significant impact on therapeutic immune modulating proteins 77,78 and peptides 79,80 , and (2), antibiotic treatment may increase severity of GHV infections.  Antibiotic treatment. At 4 weeks of age, IFNγR −/− mouse cohorts were transferred from the ABSL1 colony to a separate, ABSL2 colony. Gut microbiome suppression was achieved by replacing standard drinking water with autoclaved reverse osmosis water (obtained from the same animal care facility) containing an antibiotic cocktail (Table 1) composed of Streptomycin (2 g/L), Gentamicin (0.5 g/L), Bacitracin (1 g/L) and Ciprofloxacin (0.125 g/L) for 10 days. One day (24 hours) prior to infection, medicated water was replaced with standard animal care facility water, which was maintained for the remainder of the experiment.

MHV-68 infection.
On day 11 (after 10 days of antibiotics) mice (5 weeks old) were infected with MHV-68 at a dose of 12.5 × 10 6 PFU in 0.1 mL DMEM by intraperitoneal (IP) injection as previously described 22,25,27,29 . Mice were returned to the colony and monitored for signs of distress for the duration of the experiment. Mice were either followed for 150 days to determine survival or euthanized at 3 days post-infection and organs harvested into formalin for histology or RNAlater (Thermo Scientific, USA) for microbiome analyses. Details on the numbers of MHV-68-infected mice in this study are detailed in Table 2.
Gut bacterial microbiome sample processing. Large intestine samples preserved in RNAlater were processed for total genomic DNA isolation using the ZymoBiomics ® miniprep kit (Zymo Research) according to manufacturer's recommended procedure. Isolated DNA was quantified using a NanoDrop 2000C (Thermo Scientific) and stored at −80 °C. Samples were analyzed in the Arizona State University KED Genomics Core for whole-sample 16 S rRNA gene amplicon sequencing. DNA library preparation for Illumina ® MiSeq platform was prepared according to the protocol from Earth Microbiome Project (http://www.earthmicrobiome. org/emp-standard-protocols/16s). The 16 S primer set 515f-806r 81 was used for 2 × 150 pair-ended sequencing.
Histopathology. Mice were euthanized at 3 days follow-up after infection by carbon dioxide asphyxiation followed by cervical dislocation. Organs were harvested and fixed in 10% neutral buffered formalin. Samples were dehydrated through graded alcohol, paraffin-embedded, sectioned into 4-6 µm ribbons, captured on  www.nature.com/scientificreports www.nature.com/scientificreports/ charged glass and dried overnight at 37 °C prior to processing for histopathology. Slides were rehydrated, stained with Gill's hematoxylin No.3 and Eosin Y (H&E) according to standard procedure, dehydrated and mounted in Cytoseal XYL (Thermo Fisher Scientific, USA).
Quantitative morphometry. H&E-stained aorta, lung and colon sections were imaged with a 20 × /0.5NA objective on an Olympus BX51 microscope equipped with an Olympus DP74 camera operated by cellSens Dimensions v1.16 software. Objective-calibrated measurements of alveolar septal thickness and lumen area were collected using cellSens Dimensions. At least 50 measurements were collected and averaged for each mouse, and at least three mice were examined for each group. Images were processed for visualization in figures using ImageJ/ FIJI v1.52i 87 . immunohistochemistry. FFPE blocks containing lung tissue were sectioned into 4-6 µm ribbons, captured on charged glass and dried overnight at 37 °C prior to processing. Slides were rehydrated and epitopes retrieved by boiling in sodium citrated buffer. Endogenous peroxidases were quenched with 3% hydrogen peroxide, slides were blocked in 5% bovine serum albumin in TBS/0.1% Tween 20 and sections were probed with rabbit polyclonal antibody against CD3 (Abcam ab5690; 1:200), CD4 (Abcam ab183685; 1:1000) or CD8 (Abcam ab209775; 1:2000). For MHV-68 detection, 1:500 rabbit anti-serum or pre-immune serum (as control) were used for immunostaining (kind gift of Dr. H.W. Virgin III 9 ). 1:500 goat anti-rabbit HRP was used for secondary staining (Jackson Immuno Research 111-035-144). Antigens were revealed with ImmPACT DAB (Vector Labs, USA) and mounted with Cytoseal XYL. Sections were imaged with a 40 × /0.75NA objective. Viral load determination by qPCR. Total DNA was isolated from each FFPE sample (4 × 5 µm thick sections for each sample was used) using the QIAamp DNA FFPE Tissue Kit according to the manufacturer's instructions. Samples were quantified using a DS-11 series spectrophotometer/fluorometer (Denovix, USA). Quantitative PCR (qPCR) was undertaken to investigate relative viral load using the following reaction per sample: 10 µL SsoAdvanced Universal SYBR Green Supermix (Bio-rad Laboratories Inc., USA), 0.4 µl each of the primers 65 F (5′-GTCAGGGCCCAGTCCGTA-3′) and 65 R (5′-TGGCCCTCTACCTTCTGTTGA-3′), 200 ng of DNA and water up to 20 µL total volume. Reactions were run in triplicate with controls, and a standard curve using pCR2.1 Topo plasmid (Thermo Fisher Scientific, USA) containing the cloned target MHV region 35 . The following cycling conditions used on a CFX96 (Bio-rad Laboratories, Inc., USA) instrument: 95 °C for 20 sec, 40 cycles (95 °C for 15 sec, 60 °C for 20 sec), followed by a melt curve analysis. Statistical analysis. Survival and pathology statistics were analyzed with GraphPad Prism v8.0.1.
Kaplan-Meier survival statistics were calculated using Log-rank (Mantel-Cox) testing. For visualization, individual comparison curves are presented. Lung pathology statistics were compared using a Two-Way ANOVA with a Fisher's LSD or Tukey's post-hoc test. Bars are mean ± standard error. P-value less than 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; n.s. is not significant.

Data availability
Sequence data has been deposited to the NCBI Sequence Read Archive under BioProject accession number PRJNA517927.