Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Research Systems analysis reveals complex biological processes during virus infection fate decisions Jordi Argilaguet,1,12 Mireia Pedragosa,1,12 Anna Esteve-Codina,2,3,12 Graciela Riera,1 Enric Vidal,4 Cristina Peligero-Cruz,1,13 Valentina Casella,1 David Andreu,5 Tsuneyasu Kaisho,6,7 Gennady Bocharov,8,9 Burkhard Ludewig,10 Simon Heath,2,3 and Andreas Meyerhans1,11 1 Infection Biology Laboratory, Department of Experimental and Health Sciences (DCEXS), Universitat Pompeu Fabra, Barcelona, Catalonia 08003, Spain; 2CNAG-CRG, Center for Genomic Regulation (CRG), Barcelona Institute of Science and Technology, 08028 Barcelona, Spain; 3Universitat Pompeu Fabra (UPF), Barcelona, Catalonia 08003, Spain; 4IRTA, Centre de Recerca en Sanitat Animal (CReSA-IRTA-UAB), Campus de la Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Catalonia, Spain; 5Laboratory of Proteomics and Protein Chemistry, DCEXS, Universitat Pompeu Fabra, 08003 Barcelona, Spain; 6Department of Immunology, Institute of Advanced Medicine, Wakayama Medical University, Wakayama 641-8509, Japan; 7Laboratory for Immune Regulation, World Premier International Research Center Initiative, Immunology Frontier Research Center, Osaka University, Osaka 565-0871, Japan; 8Marchuk Institute of Numerical Mathematics, Russian Academy of Sciences, Moscow, 119333, Russia; 9Sechenov First Moscow State Medical University, Moscow, 119991, Russia; 10Institute for Immunobiology, Kantonsspital St. Gallen, 9007 St. Gallen, Switzerland; 11Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona, 08003, Spain The processes and mechanisms of virus infection fate decisions that are the result of a dynamic virus-immune system inter- action with either an efficient effector response and virus elimination or an alleviated immune response and chronic infec- tion are poorly understood. Here, we characterized the host response to acute and chronic lymphocytic choriomeningitis virus (LCMV) infections by gene coexpression network analysis of time-resolved splenic transcriptomes. First, we found an early attenuation of inflammatory monocyte/macrophage prior to the onset of T cell exhaustion, and second, a critical role of the XCL1-XCR1 communication axis during the functional adaptation of the T cell response to the chronic infection state. These findings not only reveal an important feedback mechanism that couples T cell exhaustion with the maintenance of a lower level of effector T cell response but also suggest therapy options to better control virus levels during the chronic infection phase. [Supplemental material is available for this article.] Overwhelming infections with noncytopathic viruses can lead to fected host (Rouse and Sehrawat 2010). This threat is sensed by in- chronic infections. The fate of a virus infection can be fundamen- completely understood mechanisms, and several suppressive tally categorized as acute or chronic according to its temporal rela- immune components are activated that adapt the host response tionship with the host organism (Virgin et al. 2009). In humans, to the viral threat. Among these are T cell exhaustion by deletion acute infections are usually resolved within a few weeks. In con- (Moskophidis et al. 1993) and functional impairment (Zajac trast, chronic infections are not resolved and, instead, develop et al. 1998; Wherry et al. 2003; Barber et al. 2006), the generation when innate and adaptive immune responses are not sufficient of monocyte-derived suppressor cells (MDSCs) (Cai et al. 2013; to eliminate the invading virus during the primary infection Norris et al. 2013) and regulatory cell subsets (Wilson et al. phase. Examples for this latter case are infections with the 2012). They exert their function via distinct processes including Human Immunodeficiency Virus (HIV) or the Hepatitis B and C vi- interaction with inhibitory ligands, the production of soluble im- ruses (HBV, HCV) that can establish persistence in their hosts with munosuppressive factors like IL10 and Indoleamine 2,3-dioxyge- different probabilities and pathogenic consequences (Rehermann nase, and the delivery of suppressive signals by cell–cell contacts and Nascimbeni 2005; Feinberg and Ahmed 2012). to conventional T cells and to antigen-presenting cells, thus influ- A hallmark of an overwhelming infection is the down-regula- encing effector functions directly or indirectly (Barber et al. 2006; tion of immune effector mechanisms to avoid immunopathology. Day et al. 2006; Doering et al. 2012; Ng et al. 2013). Indeed, the simultaneous presence of a widespread virus infection Despite the various suppressive mechanisms induced during and strong cytotoxic effector cell responses can induce massive cell a chronic virus infection, the effector T cell shutdown is only par- and tissue destruction and may directly threaten the life of the in- tial and some T cell functionality remains that restrains the expan- sion of a persisting virus (Schmitz et al. 1999; He et al. 2016; 12 These authors contributed equally to this work. © 2019 Argilaguet et al. This article is distributed exclusively by Cold Spring 13 Present address: Department of Immunology, Weizmann Institute Harbor Laboratory Press for the first six months after the full-issue publication of Science, Rehovot, 76100, Israel date (see http://genome.cshlp.org/site/misc/terms.xhtml). After six months, it Corresponding author: andreas.meyerhans@upf.edu is available under a Creative Commons License (Attribution-NonCommercial Article published online before print. Article, supplemental material, and publi- 4.0 International), as described at http://creativecommons.org/licenses/ cation date are at http://www.genome.org/cgi/doi/10.1101/gr.241372.118. by-nc/4.0/. 29:907–919 Published by Cold Spring Harbor Laboratory Press; ISSN 1088-9051/19; www.genome.org Genome Research 907 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Argilaguet et al. Im et al. 2016; Leong et al. 2016). The processes however that me- infection and d7-d9 of acute plus d5-d7 of chronic infection, re- diate the transition toward a lower level response are not well un- spectively. This indicates that at a tissue level, central host respons- derstood. We therefore sought to analyze on a systems level es are similar in acute and chronic infections albeit, owing to the infection-fate-specific gene signatures and adaptive processes of different inoculum size, they appear at different time points. the host to an overwhelming virological threat. The infection of Finally, d9 and d31 samples from chronic infection, that is once mice with lymphocytic choriomeningitis virus (LCMV) was used CD8+ T cell exhaustion is established, cluster in a separate group as a model system. It enabled reproducibly establishing acute 4 representing a specific chronic infection phase. Thus, distinct vi- and chronic infections, and exhibited several fundamental fea- rus infection phases exhibit separated transcriptome signatures in tures within the virus-immune system’s crosstalk that hold true the lymphoid tissue. for human infections as well (Wilson and Brooks 2010). Taking noninfected mice as a reference point, the analysis of differentially expressed genes after acute and chronic infection showed a similar number of up-regulated and down-regulated Results genes from d5 to d9 (Supplemental Fig. S1D). Appreciable differ- ences between both types of infection were found only at early Spleen gene coexpression networks in acute and chronic viral (d3) and late (d31) time points when acute infected mice showed infections fewer differentially expressed genes. This reflects the different The development of an acute or a chronic virus infection is the re- host response kinetics between the two types of infection at early sult of a coordinated response of the immune system toward the time points and the continuous deviation of the chronic infection invading microbe. Although several of the individual components state from the basal level of transcription of naive mice. Gene that contribute to this fate decision have been described, a system- Ontology (GO) analysis of all genes differentially expressed at ic global view is still lacking. Thus, we characterized transcriptome any time point postinfection showed identical enriched terms in changes at the level of the spleen during the course of viral infec- acute and chronic infections representing general biological pro- tions with different outcomes. C57BL/6 mice were infected with a cesses such as “cell cycle,” “immune response,” and “programmed low-dose (2×102 plaque-forming units; acute infection) or a high- cell death.” dose (2×106 plaque-forming units; chronic infection) of LCMV To explore the dynamics of the spleen transcriptomes during strain Docile (LCMVDoc). Virus titers, virus-specific CD4+ and both infection courses, we applied weighted gene correlation net- CD8+ T cell responses, and spleen-derived total messenger RNA work analysis (WGCNA) and constructed two signed coexpression (mRNA) transcriptomes were determined longitudinally by titra- networks (Zhang and Horvath 2005). Topological overlap was used tion, intracellular cytokine staining (ICS), and RNA-seq, respec- to measure the connection strength between genes based on tively. Because of the different inoculum size, virus titers in shared network neighbors, identifying genes with similar patterns spleen increased later in acute than in chronic infection, however, of connections, and subsequently defining modules of highly similar levels were observed at d5 postinfection (PI) (Fig. 1A). As coexpressed genes (Fig. 1D). Finally, the eigengene was extracted previously described (Moskophidis et al. 1993), low-dose infection for each module to represent the gene expression profile. resulted in virus clearance, whereas high-dose infection led to virus Twenty-three modules were identified in both acute and chronic persistence shown as high viral titers in spleen at d31 PI. The inter- infections (Supplemental Fig. S2). They were color-coded for iden- feron gamma (IFNG)-producing CD8+ and CD4+ T cells expanded tification purposes. Of note is that the same color in modules from at d6-d7, and their percentages remained high at d31 when virus acute and chronic networks does not imply any similarity between was undetectable after low-dose infection (Fig. 1B; Supplemental them. The number of genes per module ranged from 15 to 4952 Fig. S1C). In high-dose infected mice, IFNG-producing CD8+ (Supplemental Fig. S5). T cells also increased at d6-d7 PI, however, their number dropped To obtain a global view on the similarity of the module ex- at d7-d9, which is indicative of CD8+ T cell exhaustion. This latter pression patterns, we generated a hierarchical clustering dendro- conclusion was verified by observing a concomitant drop of TNF- gram based on the module’s eigengenes and defined four clusters production as well as an increase of PDCD1 (also known as PD-1) for each infection outcome (A1 to A4, acute infection; C1 to C4, and HAVCR2 (also known as TIM-3) expression levels that were chronic infection) (Fig. 1E; Supplemental Fig. S2). Clusters A1, maintained at day 15 PI (Supplemental Fig. S1A,B; Wherry et al. A4, and C2 include modules with up-regulated expression profiles 2007). Altogether this strengthens the previous suggestion that ac- at d3-d7 PI reflecting the complexity of the host effector response, cumulative virus loads rather than the virus load per se determine whereas the modules from clusters A3 and C1 show an early down- the observed differences in the immune responses and infection regulation of expression. Clusters A2 and C4 contain modules with fates (Bocharov 1998; Côrtes et al. 2018). bimodal expression profiles having the first peak at an early time To characterize changes of transcriptional profiles from point postinfection when the host response is triggered and a sec- spleens during acute or chronic infection, we isolated total ond peak at d9 when the infection outcome is already determined. mRNAs from naive mice (d0) and from mice at d3, d5, d6, d7, Finally, the expression pattern of the modules from Cluster C3 was d9, and d31 after low-dose or high-dose of LCMVDoc infection, characteristic of chronic infection showing an initial down-regula- and determined time-resolved transcriptomes by RNA-seq. tion but subsequent gradual up-regulation until d31. Hierarchical clustering across all samples revealed four main groups corresponding to different infection phases (Fig. 1C). The first group, naive mice and mice in memory phase, is represented Cellular and humoral adaptive responses are disassociated by samples from acute infected mice at days 3 and 31, and unin- in chronic infection fected mice (d0), indicating that spleen transcriptome profiles re- CD8+ T cells play a major role in the control of an LCMV infection turn to that of naive mice once the infection is resolved. Early as well as in exhaustion during chronic infections (Oldstone effector responses (group 2) and late effector responses (group 3) 2006). We therefore tested whether the behavior of virus-specific are represented by samples from d5-d6 of acute plus d3 of chronic T cells was mirrored by a gene expression pattern of similar 908 Genome Research www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis of virus infection fate A C and the virus-specific CD8+ T cell re- sponse of chronic infection was observed (Supplemental Fig. S3A). This indicates that IFNG production is a good marker to identify pathways involved in the T cell response in acute infection but not sufficient to reveal the underlying gene signature of exhaustion. Hub genes from coexpression mod- ules are defined by a high KIM value, or B alternatively by a high topological over- lap, and are predicted to be central for the biological processes they represent (Jeong et al. 2001; Doering et al. 2012). Thus, we visualized the hub genes of the acute-brown module as a network based on their topological overlap values (Fig. 2C). Two clusters were observed that are enriched in the GO terms TCR signal- ing pathway, vesicle-mediated transport, D inflammatory response, and B cell re- sponse. Cluster 1 contains key genes in T cell activation, and Cluster 2 contains genes from the B cell response. To ana- lyze how these hub genes in acute infec- tion behave during chronic infection, we compared their intramodular con- nectivities in both infection outcomes (Supplemental Fig. S3B). Genes from Cluster 1 related to T cell activation were E found in two different chronic modules, chronic-tan and chronic-greenyellow (Supplemental Fig. S3B). However, only genes in chronic-tan retained a KIM > 0.6 and therefore represented control points of T cell activation in chronic infection. These genes were up-regulated at day 6 and returned to basal levels at day 31, thus having similar kinetics as in acute in- fection (Fig. 2D). In contrast, genes in the chronic-greenyellow module showed an up-regulation at day 6 with a subsequent down-regulation at day 9, thus coincid- ing with T cell exhaustion. To validate Figure 1. Kinetics of virus expansion, CD8+ T cell response, and transcriptome analysis in acute and these observations and to link the day 9 + chronic LCMV infections. Virus titers (A) and percentages of GP33-specific IFNG-producing CD8 T cells down-regulation of the chronic-greenyel- (B) in spleen. The mean ± SEM is shown. (C) Correlation matrix heatmap of gene expression across acute and chronic infection samples. The color scale indicates the degree of correlation. (D) Hierarchical clus- low genes to T cell exhaustion, we per- tering dendrogram for all differentially expressed genes (lines) obtained by WGCNA. The branches cor- formed qPCR of selected genes from respond to modules of highly coexpressed groups of genes. Colors below the dendrogram indicate the spleens of mice at day 9 after acute or module to which each gene was assigned. (E) Hierarchical clustering dendrogram of dissimilarity based chronic infection (Supplemental Fig. on module eigengenes. S3C,D). The qPCR results positively corre- lated with the RNA-seq data. Cd3d and kinetics. For this we correlated the module eigengenes with the ki- Zap70 from the chronic-greenyellow module that play a critical netics of IFNG-producing GP33-specific CD8+ T cells (Supplemen- role in T cell receptor signaling, and that were chosen as examples tal Fig. S3A). The acute-brown module significantly correlated with because of their high KIM values, were significantly down-regulat- the virus-specific CD8+ and CD4+ T cell responses of acute infec- ed in chronic infection thus supporting a role during T cell tion (Fig. 2A). To visualize this correlation at the level of the indi- exhaustion. vidual genes, we compared the intramodular connectivity (KIM) of The B cell response-related genes of Cluster 2 were mainly each gene of this module with its correlation to the T cell response. found in the chronic-green module and retained a high intramod- A correlation was observed indicating a link between the hub ular connectivity value (Supplemental Fig. S3B). Thus, the B cell re- genes (high KIM values) to the T cell response (Fig. 2B). In contrast, sponse during both infection types is controlled by a highly no significant correlation between chronic-module eigengenes coregulated gene expression. The expression kinetics of these Genome Research 909 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Argilaguet et al. A B and chronic LCMV infections were stained for total IgG and analyzed by im- munohistochemistry. IgG production was evident at d9 in both types of infec- tion but was significantly higher in chronic infection. An elevated level of IgG remained visible even at day 48 post- infection when the B cell response of acute infection was comparable to that of naive mice (Fig. 2F,G; Supplemental C Fig S4). Taken together, coexpression net- work analysis from spleen-derived tran- scriptomes permits the identification of biologically relevant molecular pathways that are characteristic for acute or chronic infections. Transcriptional network preservation between acute and chronic infections The obtained gene coexpression modules represent the biological processes partici- pating in acute and chronic infections. To identify the common and the specific features for both infection fates, we first D E determined the gene overlap between the acute and chronic modules by Fisher’s exact test (Supplemental Fig. S5A). Eight acute and 12 chronic mod- ules showed a high degree of overlap (P-value < 10−35) with modules from the other group. To then quantify module preservation (the degree to which coex- pressed genes in acute infection are also coexpressed in chronic infection), we F G computed a connectivity-based preserva- tion statistic (Z-summary) that allows scoring the degree of similarity of the connectivity patterns between genes from two networks. Module preservation scores were inversely correlated with the percentage of differentially connected genes (defined as those with a difference in their KIM between acute and chronic networks higher than 0.4) (Fig. 3A). This indicates that the preserved mod- Figure 2. Genes related to T and B cell responses from the acute-brown module are disassociated in ules, that share similar gene connectivity + chronic infection. (A) Acute-brown module eigengene and kinetics of GP33-specific CD8 and GP61-spe- + + patterns, also share common hub genes. cific CD4 T cells. (B) Intramodular connectivity versus GP33-specific CD8 T cell kinetics correlation sig- nificance for acute-brown module genes. (C) Visualization of the acute-brown module genes. Large Furthermore, module preservation scores nodes represent hub genes with high connection densities (TOM > 0.5). (D,E) Expression kinetics of correlated with the module size (Fig. 3A), acute-brown module genes related to T cell activation (D) and hub genes related to the B cell response indicating that a large number of genes (E) in acute and chronic infections. (F) Representative images of IgG immunohistochemical staining of participate in biological pathways that spleens. Magnification bar: 100 µm. (G) Semiquantification of IgG-positive cells in spleens (score 0– 10). Data shown are mean ± SEM of four mice per group and time point. (∗∗∗ ) P ≤ 0.001 unpaired are common in acute and chronic infec- two-tailed t-test. tion. In contrast, the biological pathways that are specific for an infection fate are governed by few genes. genes during acute and chronic infections were identical, being To further examine the highly preserved modules, we com- up-regulated from d7 to d9 and remaining elevated at d31. pared their expression patterns (Supplemental Fig. S5B,C). However, their expression levels were notably higher in chronic in- Relevant overlapping hub genes between preserved modules are fection (Fig. 2E; Supplemental Fig. S3C), suggesting a causal link to listed in Supplemental Table S1. With the exception of the the previously described hypergammaglobulinemia of chronic in- acute-yellow and the chronic-brown module that shared genes in- fections (Hunziker et al. 2003). To validate this, spleens from acute volved in type I interferon signaling (Irf7, Mx1-2, Oas1a, Ifit1-3), all 910 Genome Research www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis of virus infection fate A B C Figure 3. Module preservation between acute and chronic infections. (A) Module preservation scores were compared to the percentage of differentially connected genes in a module and to module size. Dashed lines indicate the thresholds (Methods). The Spearman’s rank correlation coefficients (r) and the P-values are indicated. (B,C) Eigengene expression profiles of acute- (B) and chronic-specific (C ) modules. the other highly preserved acute modules were decomposed into genta, a module with a decrease in its expression between d7 two modules in chronic infection. For instance, the acute-brown and d9 when T cell exhaustion appears (Fig 1B). Four of these genes module overlapped with the chronic-green and the chronic-tan were related to leukocyte chemotaxis (Itgb2, Itgam, Ccr1, and module, respectively, as described above (Fig. 2). Thirty-six hub C5ar1). qPCR of spleens from additional animals validated the dif- genes from the acute-blue module overlapped with chronic-ma- ferential down-regulation of Ccr1 and C5ar1 between acute and Genome Research 911 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Argilaguet et al. chronic infections at days 7 and 9 PI and showed a tendency to pathway known to play an important role in the regulation of in- decrease for Itgb2 and Itgam (Supplemental Fig. S6A,B). Three of flammation by monocytes and macrophages (Supplemental these four genes (Itgb2, Itgam, and C5ar1) are involved in neutro- Table S3; Burrack and Morrison 2014; Yang and Ming 2014; Liu phil chemotaxis. A sustained expansion of neutrophilic MDSCs et al. 2016). A list of 27 representative genes from this submodule has been described in LCMV chronic infection (Norris et al. with a clear peak of expression at day 6 PI in acute infection, and 2013). Therefore, these results may suggest that prior to the ap- their expression profiles in acute and chronic infections, are shown pearance of neutrophilic MDSCs, there is an alteration in the mi- in Figure 4, E and F, respectively. Il1b, Arg2, and Nos2 expression ki- gration of these cells concomitant with exhaustion of CD8+ T cells. netics was further confirmed by qPCR including additional infect- The modules with low preservation scores represent biologi- ed mice (Supplemental Fig. S7). Together these data show cal processes specific of acute or chronic infections, and therefore differential spleen monocyte/macrophage characteristics during might provide novel mechanisms influencing infection fates. We early time points of acute and chronic virus infections. identified five acute-specific and four chronic-specific modules To test the hypothesis of an early differential polarization of (Fig. 3B,C). The respective genes and their connectivity values monocyte/macrophages during both outcomes of LCMV infec- are listed in Supplemental Table S2. Acute-specific modules tion, we characterized spleen monocyte/macrophages at day 6 showed diverse expression profiles characterized by marked ex- postinfection by intracellular IL6 staining and RNA-seq. The per- pression peaks at different time points. The included genes repre- centage of IL6-producing monocytes/macrophages from chronic sent both innate and adaptive immune pathways highlighting infection were significantly lower than those from acute infection, the need of diverse biological responses to properly resolve a viral whereas the number of pro-inflammatory LY6Chi monocytes/mac- infection. On the other hand, all chronic infection-specific mod- rophages in the spleen were, as described before (Norris et al. ules showed a bimodal expression profile with a second peak or 2013), similar in both infections (Fig. 5A). Transcriptional profil- drop of expression at days 7–9 PI (Fig. 3C) when T cell exhaustion ing from sorted monocytes/macrophages showed that cells from appears. Thus, genes that participate in early events after infection acute infection expressed higher levels of pro-inflammatory cyto- seem also to play a role in critical events governing chronic infec- kines and chemokines, and cellular receptors and transcription tion at later time points. This suggests that the adaptation process factors associated to M1 macrophages. In contrast, monocytes/ to a chronic infection involves not a simple shutdown of genes but macrophages from chronic infection expressed higher levels of is an active process as described at the cellular level for CD8+ T cell transcription factors and genes that exert anti-inflammatory ef- exhaustion (Wherry et al. 2007). Group-specific modules are fur- fects and markers of M2 macrophage differentiation (Fig. 5B). ther analyzed in the next sections. These data demonstrate that inflammatory monocytes/macro- phages are induced early after an acute LCMV infection and, in contrast, during chronic infection these cells shift to an anti-in- Early attenuation of the inflammatory response flammatory profile before T cell exhaustion becomes evident. in chronic infection The monocytic cell population of chronic infection had an Acute infection-specific modules are expected to participate in the up-regulated expression of genes involved in arachidonic acid resolution of the virus infection. From the five acute-specific mod- secretion such as Pla2g2d (phospholipase A2, group IID), an anti- ules (Fig. 3A,B), the acute-gray60 module was most interesting. It inflammatory lipid mediator (Miki et al. 2013). Because these mol- has a marked expression peak at day 6 postinfection (Fig. 3B) just ecules play an important role in the resolution of inflammation before the specific T cell response appeared thus bridging innate (Sugimoto et al. 2016), one might expect a tissue protective role immune responses with the adaptive response. All 31 genes of of these cells in chronic infection. To test this idea, we quantified this module are highly coexpressed only during acute infection the level of fibrosis in spleens from acute and chronic LCMV infec- (Fig. 4A) and enriched for genes involved in the regulation of IL6 tions by Masson’s trichrome staining. Fibrosis, a feature of insuffi- and IL12 production (Fig. 4B). Because IL6 and IL12 are pro-inflam- cient resolution of chronic inflammation, was evident as early as matory cytokines with important roles in antiviral host responses, d10 in acute infection, and remained elevated even at days 15 our observations suggested a differential regulation in both infec- and 48, when the virus infection is resolved (Fig. 5C,D). In con- tion fates before T cell exhaustion becomes apparent. To verify trast, fibrosis was only observed in chronic infection in one of this hypothesis, we plotted the Il6 and Il12b expression kinetics four animals at d48. Together this suggests that besides the known from our RNA-seq data (Fig. 4C) and validated the Il6 expression re- role of suppressive monocytic cells to dampen virus T cell respons- sults by qPCR including additional infected animals. Both RNA-seq es in chronic infection, they also might participate in inflamma- and qPCR gave consistent results showing a peak of Il6 expression tion resolution via expression of pro-resolving lipid mediators at d6 after acute infection that was lacking in chronic infection that finally lead to fibrosis prevention. (Supplemental Fig. S7). Because inflammatory responses are com- plex responses, we next sought to identify whether Il6 and Il12b were part of a broader cluster of coexpressed genes within their The XCL1-XCR1 communication axis links T cell exhaustion acute-blue module. Indeed, we found 1504 genes in the acute- with effector cell maintenance blue module that correlate to the Il6 expression kinetics with a cor- Cytokines and chemokines are major components of the signaling relation value greater than 0.8. Moreover, the level of expression of network that regulates the coordinated immune response against a many of these genes was significantly higher in acute than in viral infection (Cameron and Kelvin 2003). Thus, we next investi- chronic infection at d6 PI (Fig. 4D). KEGG pathway analysis of gated which cytokines/chemokines-encoding hub genes were pre- the 1504 genes showed enrichment of genes involved in inflam- sent in acute- and chronic-specific modules, and therefore might matory pathways such as NOD-like receptor and JAK-STAT signal- participate in processes involved in infection fate regulation. ing pathways, and in cytokine–cytokine receptor interaction From all the identified hub genes (Supplemental Table S2), of par- (Supplemental Table S3). Furthermore, we identified two genes re- ticular interest was the chemokine-encoding gene Xcl1 from the lated to the metabolism of Arginine and Proline (Arg2 and Nos2), a chronic-darkturquoise module. XCL1 is mainly produced by NK 912 Genome Research www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis of virus infection fate A B C D E F Figure 4. Acute-gray60-specific module unveils a differential regulation of genes related to macrophage inflammatory responses in acute and chronic infections. (A) Normalized expression kinetics of genes from the acute-gray60 module and their corresponding kinetics in chronic infection. Red line rep- resents the kinetics of the eigengene. (B) GO terms enriched in genes from the acute-gray60 module (dashed line mark P = 0.05). (C) RNA-seq-derived Il12b and Il6 expression kinetics in acute and chronic infections. (D) Pearson’s correlation between genes of the acute-blue module and Il6 in acute infection is plotted against the log2 fold-change (log2FC) of gene expression between acute and chronic infection at day 6 PI. Dashed line marks a correlation >0.8. (E) List of 27 representative genes with a correlation > 0.8 with Il6 expression kinetics. Gray boxes indicate enriched KEGG pathways to which each gene belongs. (F) Normalized expression kinetics of the 27 genes shown in E. and activated CD8+ T cells, and promotes the recruitment of den- All 17 genes of the chronic-darkturquoise module are highly dritic cells expressing the receptor XCR1 (XCR1+ DCs), which are coexpressed in chronic infection and deregulated in acute infec- implicated in the priming and boosting of cytotoxic responses to tion (Fig. 6A). They show a “two-peak” behavior with an expres- cross-presented antigens (Kroczek and Henn 2012; Brewitz et al. sion peak at d5 and a second up-regulation from d7 to d9 PI. The 2017). second peak of expression of Xcl1 was verified by qPCR using Genome Research 913 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Argilaguet et al. A B C D Figure 5. Early shift from inflammatory to pro-resolving monocytes/macrophages in chronic infection to prevent fibrosis. (A) Percentage of IL6-produc- ing ITGAM+ ITGAX− LY6G− (also known as CD11b+ CD11c− LY6G−) cells and number of LY6C high monocytes/macrophages per spleen at day 6 postin- fection. (B) Heatmap illustrating the relative expression of genes in monocyte/macrophages from acute and chronic infected mice at day 6 PI. (C ) Representative images of Masson’s trichrome staining of spleens. Thickness of the splenic capsules are marked by arrows. Magnification bar: 50 µm. (D) Semiquantification of fibrosis in spleens. (A–D) Data shown are the mean ± SEM of n = 4 to 6 mice. (∗ ) P ≤ 0.05; (∗∗ ) P ≤ 0.01; (∗∗∗ ) P ≤ 0.001 unpaired two-tailed t-test. additional infected mice (Supplemental Fig. S8A). In contrast to staining. Both NK and CD8+ T cells showed a higher expression chronic infection, Xcl1 showed only a single early peak of expres- of XCL1 in chronic infection compared to acute infection or un- sion at d6 in acute infection, returning to basal levels from d7 PI infected mice (Fig. 6C). XCL1 was highly produced by LCMV-spe- (Fig. 6B). To investigate the role of XCL1 during the establish- cific CD8+ T cells once exhaustion was established (Fig. 6D; ment of the chronic infection phase, we first analyzed the chemo- Supplemental Fig. S8B), and particularly abundant in the sub- kine expression by NK and CD8+ T cells at d9 by intracellular population of CXCR5+ cells (Fig. 6D). Because virus-specific 914 Genome Research www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis of virus infection fate A B C D E Figure 6. Chronic-darkturquoise-specific module reveals an important role of XCL1 in chronic infection. (A) Normalized expression kinetics of genes from the chronic-darkturquoise module and their corresponding kinetics in acute infection. (B) Xcl1 expression kinetics obtained from RNA-seq analysis. (C) Percentages of XCL1-producing NK and CD8+ T cells at day 9 PI. Data shown are the mean ± SEM from n = 11 to 13 mice representative of three in- dependent experiments (significance determined using one-way ANOVA). (D,E) MFI of XCL1 in DbGP33-41+ CD8+ T cells (D, right) and percentages of XCL1-producing DbGP33-41+ CD8+ T cells (D, left) and CXCR5- or CXCR5+ DbGP33-41+ CD44+ CD8+ T cells (E) at day 9 PI. Data shown are the mean ± SEM from n = 11 to 13 mice (significance determined using unpaired two-tailed t-test). (ns) not significant; (∗∗ ) P ≤ 0.01; (∗∗∗ ) P ≤ 0.001; (∗∗ ∗∗ ) P ≤ 0.0001. CXCR5+ CD8+ T cells have recently been shown to play a major adaptation to an overwhelming virus threat, and (2) an important role in the control of virus replication during chronic infections component of virus control in the chronic infection phase. (He et al. 2016; Im et al. 2016), our data now show that they also play an important role in the immune adaptation process during exhaustion. Discussion Coinciding with the second peak of Xcl1 expression, the Nonlethal pathogenic viral infections are either acutely resolved or quantification of XCR1+ DCs showed a clear increase of these cells become chronic. Although this fate decision is made during the in the spleens at d9 in chronic infection (Fig. 7A; Supplemental primary infection phase, it is incompletely understood, and a sys- Fig. S8C). To further analyze the impact of this XCL1-XCR1 axis, tems view is lacking. Here, we used time-resolved spleen transcrip- we depleted XCR1+ DC in chronically infected XCR1-DTRvenus tomes from acutely or chronically LCMV-infected mice in mice (Yamazaki et al. 2013) with diphtheria toxin (DT) (Fig. 7B; combination with weighted gene coexpression network analysis Supplemental Fig. S8D). Although the virus-specific CD4+ T cell re- to decipher differentiating host responses and biological path- sponse was not affected, XCR1+ DC depletion led to a significant ways. We demonstrate an early attenuation of inflammatory reduction in the percentage of GP33-specific CD8+ T cells at day monocyte/macrophage responses prior to T cell exhaustion and 15 PI and lower percentages of LAMP1 (also known as CD107a), the involvement of the XCL1-XCR1 communication axis in virus LAMP2 (also known as CD107b), and IFNG-producing cells than containment during chronic infection. Thus, the response of a in untreated animals (Fig. 7C). Moreover, the percentage of host to an overwhelming infection threat is complex and subtle, CXCR5+ CD8+ T cells was also significantly decreased, suggesting involving both immune-stimulatory and immune-suppressive an interdependence between this CD8+ T cell subset and cross-pre- pathways. senting XCR1+ DCs. Consistently, the viral titers in spleen, lung, Acute and chronic infection fates are readily distinguished and kidney were higher in mice lacking XCR1+ DCs (Fig. 7D). To- by their sequential splenic transcriptomes. The whole transcrip- gether these results demonstrate that the XCL1-XCR1 communi- tomes reveal differentiating traits at day 9 postinfection when cation axis is (1) an important component of immune system acute infection starts to be resolved. However, close inspection Genome Research 915 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Argilaguet et al. A B C D Figure 7. XCR1+ dendritic cells maintain antiviral T cell responses and control virus titers. (A) Number of XCR1+ dendritic cells in spleen. (B) Schematic representation of DT treatment regimen for depletion of XCR1+ dendritic cells. (C) Percentages of DbGP33-41+ CD8+ T cells, IFNG-producing CD8+ and CD4+ T cells, LAMP1+ and LAMP2+ CD8+ T cells, and CXCR5+ CD44+ CD8+ T cells at day 15 PI from chronic infected mice nontreated (DT−) or treated (DT +) with DT. (D) Virus titers in spleen, lung, and kidney at day 15 PI from chronic infected mice nontreated (DT−) or treated (DT+) with DT. (C,D) The mean ± SEM is shown. (ns) nonsignificant; (∗ ) P ≤ 0.05; (∗∗ ) P ≤ 0.001; (∗∗∗ ) P ≤ 0.0001 unpaired two-tailed t-test. Data are representative of two or three indepen- dent experiments. of the coexpression modules demonstrates an attenuation of in- leaves a rich comprehensive data set for further, more specific flammatory monocytic cells at day 6 before T cell exhaustion be- analyses. comes apparent. Thus, the decision point to down-regulate Infection-fate-specific module information provides a firm immune responses toward an overwhelming virus threat is al- basis to reveal specific, physiologically relevant changes in the in- ready taken before T cells shut off their effector functions. fected host organism. An intriguing observation was that the infec- Whether this decision point lies even earlier remains to be deter- tion-fate-specific modules contained rather few genes. This notion mined. So far, early innate immune responses represented, that is, may not only help to define diagnostic biomarker for predicting by modules containing type I IFN genes are very similar in acute infection outcomes but also hint to regulators of pathways that and chronic infections, and differences come up only from day 6 may direct phenotype evolution over time. Indeed, the acute- onward. In addition to these global characteristics, we were also gray60 module with its 31 genes that have an expression peak at able to link specific genes and pathways to critical, already estab- day 6, is a good example. Starting from a GO enrichment analysis lished phenotypes associated with both infection fates including of its genes, we subsequently identified important differences in the effective LCMV-specific T cell response as well as T cell ex- early monocytic cell polarization with an M1-type profile in acute haustion and hypergammaglobulinemia. Together this validates infection and cells with an attenuated phenotype during chronic our systems approach for analyzing virus infection fates and infection. Previous studies have described a skewing of monocytic 916 Genome Research www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis of virus infection fate cells toward MDSCs in chronic LCMV infections of mice (Norris nogens to XCR1+ DCs via linkage to XCL1 itself or to a XCR1-tar- et al. 2013) and in HIV, HBV, and HCV infections in humans geting antibody (Fossum et al. 2015; Hartung et al. 2015). (Cai et al. 2013; Qin et al. 2013; Pallett et al. 2015). MDSCs appear Combining such immunotherapeutic vaccination with check- during the primary infection phase and have been characterized, point inhibition may then allow effector cell expansion enhanced for example, by transcriptome analysis at day 14 post-chronic by a positive feedback loop and thus a greater increase in the effec- LCMV clone 13 infection, a few days after T cell exhaustion be- tor cell to virus ratio. comes apparent (Norris et al. 2013). Our own RNA-seq results In conclusion, coexpression network analysis of time-re- from isolated monocytic cells at day 6 postinfection now point solved splenic transcriptome data revealed fundamental features to an even earlier phenotypic shift with an expression of genes of infection fate regulation. Several links between gene transcripts linked to alternative macrophage differentiation and anti-inflam- and immunological phenotypes were established that include an mation. This early monocytic cell skewing was associated with his- early attenuation of inflammatory monocytic cells and an involve- tological changes in the spleen. Acutely infected mice develop ment of the XCL1-XCR1 axis in the functional adaptation of the spleen fibrosis that is maintained even when the virus is already immune response in a chronic virus infection. This not only pro- well controlled, but in chronic infection, fibrosis was almost ab- vides a unique data set to generate and test novel hypotheses about sent (Fig. 5C,D). Although the direct evidence for a causal link be- immune system functioning but also suggests directions for im- tween the splenic monocyte/macrophage phenotypes and fibrosis munotherapeutic strategies to better control or cure persistent in- development is still missing, the known features of tissue macro- fections. An inspection of the results in human infections and phages in tissue injury, fibrosis, and repair are well in line with cancer is clearly indicated. our observations (Sugimoto et al. 2016; Wynn and Vannella 2016). Concomitant with CD8+ T cell exhaustion during chronic in- fection, XCR1+ DC number increase in spleen. Depletion of XCR1+ Methods DCs resulted in reduction of virus-specific CD8+ T cells and a con- Animals, infections, and in vivo cell depletion current increase in viral loads (Fig. 7C,D). However, because down- regulation of T cell effector function by exhaustion is a mechanism Mice were purchased from Charles River Laboratories and main- to avoid immunopathology (Cornberg et al. 2013), as is attenua- tained under specific pathogen-free conditions at in-house facili- ties. All animal work was conducted according to the guidelines tion of inflammatory macrophages and generation of MDSCs from Generalitat de Catalunya and approved by the ethical com- (Medzhitov et al. 2012), why should a host organism at the same mittee for animal experimentation at Parc de Recerca Biomèdica time expand a subtype of antigen-presenting cells that are highly de Barcelona (CEEA-PRBB, Spain). LCMV strain Docile was used potent in priming and boosting novel T cell responses? This funda- for mouse infections. For details on mouse strains, virus quantifi- mental issue may simply highlight two important aspects within cation, and in vivo cell depletion, see Supplemental Information. the in vivo virus-immune system crosstalk. First, the immune sys- tem has to shut down when confronted with an overwhelming vi- ral threat. A high-dose LCMVDoc infection requires an adaptation Flow cytometry, sorting, and immunohistochemistry of the immune response and viral load dynamics such that it levels For flow cytometry analysis and sorting of cells, spleens were har- below a life-threatening situation. This is achieved by a variety of vested, and single-cell suspensions were generated. For immuno- immune-suppressive mechanisms of which T cell exhaustion histochemistry, spleens were formalin fixed and paraffin and the generation of MDSCs are important components. embedded. For details, see Supplemental Methods and Supple- Second, the immune system does not surrender completely to an mental Table S4. overwhelming threat but maintains partial control and restricts vi- rus overloads. One possible way for this seems to be a feedback reg- RNA-seq library preparation and sequencing ulation involving cells from the population of virus-specific CD8+ Total RNA from samples was isolated and submitted for sequenc- T cells that become exhausted. Indeed, the population of LCMV- ing to the Genomics Unit of the Centre for Genomic Regulation specific CD8+ T cells that becomes exhausted contains cells that (CGR, PRBB). For details, see Supplemental Information. produce XCL1 (Fig. 6E) and thus can set up XCR1+ DC recruitment and cytotoxic T cell stimulation and effector cell maintenance. RNA-seq bioinformatic analysis Overall, this argues for the operation of virus-threat-sensing host mechanisms that adapt the immune system effectors to minimize Reads mapping against the Mus musculus reference genome damage and the XCL1-XCR1 axis of being part of a functional ad- (GRCm38) was done using the GEMtools RNA-seq pipeline aptation to the chronic infection phase. (http://gemtools.github.io/docs/rna_pipeline.html) and quanti- fied with Flux Capacitor (http://confluence.sammeth.net/display/ How may these observations be exploited for host benefit? FLUX/Home) with the Mus musculus GENCODE annotation M2 Recent publications have shown that virus-specific CXC5+ CD8+ version (https://www.gencodegenes.org/). Normalization was per- T cells expand most efficiently after immune checkpoint inhibi- formed with the edgeR TMM method (Robinson and Oshlack tion (He et al. 2016; Im et al. 2016). They are, besides NK cells, 2010). Pairwise Pearson correlation coefficients (PCC) were calcu- the main producer of XCL1 in the chronic infection phase (Fig. lated for comparison among transcriptomes of spleens from unin- 6C). Furthermore, XCL1 is elevated in HIV-infected elite control- fected (n = 2, day 0), acute (n = 2 per time point), or chronic (n = 2 lers (Jacobs et al. 2017) and its receptor XCR1 is solely expressed per time point) infected mice. Hierarchical clustering across all on cross-presenting DCs (Dorner et al. 2009) that are specifically samples was based on pairwise Pearson correlation coefficients up-regulated during the chronic infection phase (Fig. 7A; among RNA-seq libraries. Differential expression analysis was per- Supplemental Fig. S8C). Therefore, enforcing the XCL1-XCR1 formed with the “robust” version of the edgeR R package (Zhou axis in an immunotherapeutic intervention seems a promising et al. 2014). Genes with a false discovery rate (FDR) < 5% were con- strategy to better control chronic virus infections. One option in sidered significant. Differentially expressed genes in acute and this direction would be the already established targeting of immu- chronic infection time series (n = 13,971) were used to construct Genome Research 917 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Argilaguet et al. a coexpression network with the WGCNA R package for each data Acknowledgments set (Langfelder and Horvath 2008). First, a signed weighted adja- cency matrix was calculated with the “blockwiseModules” func- We thank Mónica Pérez (CReSA-IRTA), Rita De Giuli-Buehrer tion using the following parameters: power = 30, TOMtype = (Kantonsspital St. Gallen), and the Flow Cytometry unit “signed”, minModuleSize = 15, mergeCutHeight = 0.25, reassign (Universitat Pompeu Fabra) for their excellent technical support. Threshold = 0, networkType = “signed”, numericLabels = TRUE, This work was supported by a grant from the Spanish Ministry of pamRespectsDendro = FALSE, nThreads = 7, maxBlockSize = Economy, Industry and Competitiveness and FEDER Grant No. 17000. The power law of 30 was selected to meet the scale-free to- SAF2016-75505-R (AEI/MINEICO/FEDER, UE), the Russian pology assumption with the pickSoftThreshold function. Then, Science Foundation Grant 18-11-00171, the Uehara Memorial genes were clustered into network modules using average linkage Foundation (to T.K.), Grant-in-Aid for Scientific Research B hierarchical clustering and the topological overlap measure 17H04088 from the Japan Society for the Promotion of Science (TOM) as proximity. Each of the identified modules was summa- (JSPS), (to T.K.), and the grant of the National Bioinformatics rized by its module eigengene (the first principal component), Institute (INB), PRB2-ISCIII (PT13/0001/0044 to A.E.-C.). which represents the weighted average expression profile of all Author contributions: A.M., G.B., and J.A. conceived the pro- module genes (Langfelder and Horvath 2008). To identify intra- ject; J.A., M.P., G.R., E.V., V.C., and C.P.-C. performed experi- modular hub genes inside a given module, the intramodular con- ments; A.E.-C. and S.H. performed bioinformatics’ analysis; D.A., nectivity (KIM) was calculated for each gene and VisANT (http:// B.L., and T.K. provided tools; A.M. and J.A. designed and super- visant.bu.edu/) was used for network visualizations (TOM > 0.3). vised the study; J.A., A.E.-C., and M.P. analyzed the data; A.M., Module preservation and module overlap were calculated with J.A., and M.P. wrote the manuscript. functions “modulePreservation” and “userListEnrichment”, re- spectively. Viral loads, CD4+, and CD8+ T cell levels were correlated with the module eigengenes with Pearson correlation coefficients. References Gene Ontology (GO) enrichment analysis was performed with Barber DL, Wherry EJ, Masopust D, Zhu B, Allison JP, Sharpe AH, Freeman DAVID (http://david.ncifcrf.gov/) (Huang et al. 2009). GJ, Ahmed R. 2006. Restoring function in exhausted CD8 T cells during chronic viral infection. Nature 439: 682–687. doi:10.1038/nature04444 Bocharov GA. 1998. Modelling the dynamics of LCMV infection in mice: conventional and exhaustive CTL responses. J Theor Biol 192: 283– Module preservation analysis 308. doi:10.1006/jtbi.1997.0612 Z-summary, implemented within WGCNA, with 100 random Brewitz A, Eickhoff S, Dähling S, Quast T, Bedoui S, Kroczek RA, Kurts C, permutations of the data, was used as a connectivity-based pre- Garbi N, Barchet W, Iannacone M, et al. 2017. CD8+ T cells orchestrate pDC-XCR1+ dendritic cell spatial and functional cooperativity to opti- servation statistic able to determine whether the connectivity pat- mize priming. Immunity 46: 205–219. doi:10.1016/j.immuni.2017.01 tern between genes in a reference network is similar to that in a .003 test network. A Z-summary score < 2 indicates no evidence of pre- Burrack KS, Morrison TE. 2014. The role of myeloid cell activation and argi- servation, 2 < Z-summary < 10 implies weak preservation, and nine metabolism in the pathogenesis of virus-induced diseases. Front Immunol 5: 428. doi:10.3389/fimmu.2014.00428 Z-summary > 10 suggests strong preservation. Highly preserved Cai W, Qin A, Guo P, Yan D, Hu F, Yang Q, Xu M, Fu Y, Zhou J, Tang X. 2013. modules were defined as those with a preservation Z-summary val- Clinical significance and functional studies of myeloid-derived suppres- ue greater than 10 and a percentage of differentially connected sor cells in chronic hepatitis C patients. J Clin Immunol 33: 798–808. genes less than 20%. Group-specific modules were defined as those doi:10.1007/s10875-012-9861-2 Cameron MJ, Kelvin DJ. 2003. Cytokines and chemokines—their receptors with a preservation Z-summary value less than 2 and a percentage and their genes: an overview. Adv Exp Med Biol 520: 8–32. doi:10.1007/ of differentially connected genes greater than 35%. 978-1-4615-0171-8_2 Cornberg M, Kenney LL, Chen AT, Waggoner SN, Kim SK, Dienes HP, Welsh RM, Selin LK. 2013. Clonal exhaustion as a mechanism to protect against severe immunopathology and death from an overwhelming Quantitative real-time PCR CD8 T cell response. Front Immunol 4: 475. doi:10.3389/fimmu.2013 Total RNA from spleens was used to perform quantitative real-time .00475 Côrtes FH, de Paula HHS, Bello G, Ribeiro-Alves M, de Azevedo SSD, Caetano PCR using SYBR select master mix (Thermo Fisher Scientific). DG, Teixeira SLM, Hoagland B, Grinsztejn B, Veloso VG, et al. 2018. Primers for all genes were designed with Primer Express 3.0 Plasmatic levels of IL-18, IP-10, and activated CD8+ T cells are potential (Applied Biosystems). For details, see Supplemental Information. biomarkers to identify HIV-1 elite controllers with a true functional cure profile. Front Immunol 9: 1576. doi:10.3389/fimmu.2018.01576 Day CL, Kaufmann DE, Kiepiela P, Brown JA, Moodley ES, Reddy S, Mackey EW, Miller JD, Leslie AJ, DePierres C, et al. 2006. PD-1 expression on Statistical analysis HIV-specific T cells is associated with T-cell exhaustion and disease pro- gression. Nature 443: 350–354. doi:10.1038/nature05115 Two-tailed t-test and ANOVA analyses were performed using Doering TA, Crawford A, Angelosanto JM, Paley MA, Ziegler CG, Wherry EJ. GraphPad Prism 6.0. P-values below 0.05 were considered signifi- 2012. Network analysis reveals centrally connected genes and pathways cant and were indicated by asterisks: (∗ ) P ≤ 0.05; (∗∗ ) P ≤ 0.01; involved in CD8+ T cell exhaustion versus memory. Immunity 37: 1130– (∗∗∗ ) P ≤ 0.001; (∗∗∗∗ ) P ≤ 0.0001. Nonsignificant differences were 1144. doi:10.1016/j.immuni.2012.08.021 Dorner BG, Dorner MB, Zhou X, Opitz C, Mora A, Güttler S, Hutloff A, indicated as “ns.” Fisher’s exact test was used to quantify overlap Mages HW, Ranke K, Schaefer M, et al. 2009. Selective expression of between modules. Correlation coefficients (r) were calculated us- the chemokine receptor XCR1 on cross-presenting dendritic cells deter- ing the Spearman rank correlation test. mines cooperation with CD8+ T cells. Immunity 31: 823–833. doi:10 .1016/j.immuni.2009.08.027 Feinberg MB, Ahmed R. 2012. Born this way? Understanding the immuno- logical basis of effective HIV control. Nat Immunol 13: 632–634. doi:10 Data access .1038/ni.2351 Fossum E, Grødeland G, Terhorst D, Tveita AA, Vikse E, Mjaaland S, Henri S, All raw and processed sequencing data generated in this study Malissen B, Bogen B. 2015. Vaccine molecules targeting Xcr1 on cross- presenting DCs induce protective CD8+ T-cell responses against influen- have been submitted to the NCBI Gene Expression Omnibus za virus. Eur J Immunol 45: 624–635. doi:10.1002/eji.201445080 (GEO; http://www.ncbi.nlm.nih.gov/geo/) under accession num- Hartung E, Becker M, Bachem A, Reeg N, Jäkel A, Hutloff A, Weber H, Weise ber GSE107880. C, Giesecke C, Henn V, et al. 2015. Induction of potent CD8 T cell 918 Genome Research www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis of virus infection fate cytotoxicity by specific targeting of antigen to cross-presenting den- atitis B immunopathology by myeloid-derived suppressor cells. Nat Med dritic cells in vivo via murine or human XCR1. J Immunol 194: 1069– 21: 591–600. doi:10.1038/nm.3856 1079. doi:10.4049/jimmunol.1401903 Qin A, Cai W, Pan T, Wu K, Yang Q, Wang N, Liu Y, Yan D, Hu F, Guo P, et al. He R, Hou S, Liu C, Zhang A, Bai Q, Han M, Yang Y, Wei G, Shen T, Yang X, 2013. Expansion of monocytic myeloid-derived suppressor cells damp- et al. 2016. Follicular CXCR5- expressing CD8+ T cells curtail chronic vi- ens T cell function in HIV-1-seropositive individuals. J Virol 87: 1477– ral infection. Nature 537: 412–428. doi:10.1038/nature19317 1490. doi:10.1128/JVI.01759-12 Huang DW, Sherman BT, Lempicki RA. 2009. Systematic and integrative Rehermann B, Nascimbeni M. 2005. Immunology of hepatitis B virus and analysis of large gene lists using DAVID bioinformatics resources. Nat hepatitis C virus infection. Nat Rev Immunol 5: 215–229. doi:10.1038/ Protoc 4: 44–57. doi:10.1038/nprot.2008.211 nri1573 Hunziker L, Recher M, Macpherson AJ, Ciurea A, Freigang S, Hengartner H, Robinson MD, Oshlack A. 2010. A scaling normalization method for differ- Zinkernagel RM. 2003. Hypergammaglobulinemia and autoantibody ential expression analysis of RNA-seq data. Genome Biol 11: R25. doi:10 induction mechanisms in viral infections. Nat Immunol 4: 343–349. .1186/gb-2010-11-3-r25 doi:10.1038/ni911 Rouse BT, Sehrawat S. 2010. Immunity and immunopathology to viruses: Im SJ, Hashimoto M, Gerner MY, Lee J, Kissick HT, Burger MC, Shan Q, Hale What decides the outcome? Nat Rev Immunol 10: 514–526. doi:10 JS, Lee J, Nasti TH, et al. 2016. Defining CD8+ T cells that provide the .1038/nri2802 proliferative burst after PD-1 therapy. Nature 537: 417–421. doi:10 Schmitz JE, Kuroda MJ, Santra S, Sasseville VG, Simon MA, Lifton MA, Racz .1038/nature19330 P, Tenner-Racz K, Dalesandro M, Scallon BJ, et al. 1999. Control of vire- Jacobs ES, Keating SM, Abdel-Mohsen M, Gibb SL, Heitman JW, Inglis HC, mia in simian immunodeficiency virus infection by CD8+ lymphocytes. Martin JN, Zhang J, Kaidarova Z, Deng X, et al. 2017. Cytokines elevated Science 283: 857–860. doi:10.1126/science.283.5403.857 in HIV elite controllers reduce HIV replication in vitro and modulate HIV Sugimoto MA, Sousa LP, Pinho V, Perretti M, Teixeira MM. 2016. Resolution restriction factor expression. J Virol 91: e02051-16. doi:10.1128/JVI of inflammation: What controls its onset? Front Immunol 7: 160. doi:10 .02051-16 .3389/fimmu.2016.00160 Jeong H, Mason SP, Barabási AL, Oltvai ZN. 2001. Lethality and centrality in Virgin HW, Wherry EJ, Ahmed R. 2009. Redefining chronic viral infection. protein networks. Nature 411: 41–42. doi:10.1038/35075138 Cell 138: 30–50. doi:10.1016/j.cell.2009.06.036 Kroczek RA, Henn V. 2012. The role of XCR1 and its ligand XCL1 in antigen Wherry EJ, Blattman JN, Murali-Krishna K, van der Most R, Ahmed R. 2003. cross-presentation by murine and human dendritic cells. Front Immunol Viral persistence alters CD8 T-cell immunodominance and tissue distri- 3: 14. doi:10.3389/fimmu.2012.00014 bution and results in distinct stages of functional impairment. J Virol 77: Langfelder P, Horvath S. 2008. WGCNA: an R package for weighted correla- 4911–4927. doi:10.1128/JVI.77.8.4911-4927.2003 tion network analysis. BMC Bioinformatics 9: 559. doi:10.1186/1471- Wherry EJ, Ha SJ, Kaech SM, Haining WN, Sarkar S, Kalia V, Subramaniam S, 2105-9-559 Blattman JN, Barber DL, Ahmed R. 2007. Molecular signature of CD8+ T Leong YA, Chen Y, Ong HS, Wu D, Man K, Deleage C, Minnich M, Meckiff cell exhaustion during chronic viral infection. Immunity 27: 670–684. BJ, Wei Y, Hou Z, et al. 2016. CXCR5+ follicular cytotoxic T cells control doi:10.1016/j.immuni.2007.09.006 viral infection in B cell follicles. Nat Immunol 17: 1187–1196. doi:10 Wilson EB, Brooks DG. 2010. Translating insights from persistent LCMV in- .1038/ni.3543 fection into anti-HIV immunity. Immunol Res 48: 3–13. doi:10.1007/ Liu C, Rajapakse AG, Riedo E, Fellay B, Bernhard M-C, Montani J-P, Yang Z, s12026-010-8162-1 Ming X-F. 2016. Targeting arginase-II protects mice from high-fat-diet- Wilson EB, Kidani Y, Elsaesser H, Barnard J, Raff L, Karp CL, Bensinger S, induced hepatic steatosis through suppression of macrophage inflam- Brooks DG. 2012. Emergence of distinct multiarmed immunoregulatory antigen-presenting cells during persistent viral infection. Cell Host mation. Sci Rep 6: 20405. doi:10.1038/srep20405 Microbe 11: 481–491. doi:10.1016/j.chom.2012.03.009 Medzhitov R, Schneider DS, Soares MP. 2012. Disease tolerance as a defense Wynn TA, Vannella KM. 2016. Macrophages in tissue repair, regeneration, strategy. Science 335: 936–941. doi:10.1126/science.1214935 and fibrosis. Immunity 44: 450–462. doi:10.1016/j.immuni.2016.02 Miki Y, Yamamoto K, Taketomi Y, Sato H, Shimo K, Kobayashi T, Ishikawa Y, .015 Ishii T, Nakanishi H, Ikeda K, et al. 2013. Lymphoid tissue phospholi- Yamazaki C, Sugiyama M, Ohta T, Hemmi H, Hamada E, Sasaki I, Fukuda Y, pase A2 group IID resolves contact hypersensitivity by driving antiin- Yano T, Nobuoka M, Hirashima T, et al. 2013. Critical roles of a dendritic flammatory lipid mediators. J Exp Med 210: 1217–1234. doi:10.1084/ cell subset expressing a chemokine receptor, XCR1. J Immunol 190: jem.20121887 6071–6082. doi:10.4049/jimmunol.1202798 Moskophidis D, Lechner F, Pircher H, Zinkernagel RM. 1993. Virus persis- Yang Z, Ming XF. 2014. Functions of arginase isoforms in macrophage in- tence in acutely infected immunocompetent mice by exhaustion of an- flammatory responses: impact on cardiovascular diseases and metabolic tiviral cytotoxic effector T cells. Nature 362: 758–761. doi:10.1038/ disorders. Front Immunol 5: 533. doi:10.3389/fimmu.2014.00533 362758a0 Zajac AJ, Blattman JN, Murali-Krishna K, Sourdive DJ, Suresh M, Altman JD, Ng CT, Snell LM, Brooks DG, Oldstone MBA. 2013. Networking at the level Ahmed R. 1998. Viral immune evasion due to persistence of activated T of host immunity: immune cell interactions during persistent viral in- cells without effector function. J Exp Med 188: 2205–2213. doi:10.1084/ fections. Cell Host Microbe 13: 652–664. doi:10.1016/j.chom.2013.05 jem.188.12.2205 .014 Zhang B, Horvath S. 2005. A general framework for weighted gene co-ex- Norris BA, Uebelhoer LS, Nakaya HI, Price AA, Grakoui A, Pulendran B. pression network analysis. Stat Appl Genet Mol Biol 4: Article17. doi:10 2013. Chronic but not acute virus infection induces sustained expan- .2202/1544-6115.1128 sion of myeloid suppressor cell numbers that inhibit viral-specific T Zhou X, Lindsay H, Robinson MD. 2014. Robustly detecting differential ex- cell immunity. Immunity 38: 309–321. doi:10.1016/j.immuni.2012.10 pression in RNA sequencing data using observation weights. Nucleic .022 Acids Res 42: e91. doi:10.1093/nar/gku310 Oldstone MBA. 2006. Viral persistence: parameters, mechanisms and future predictions. Virology 344: 111–118. doi:10.1016/j.virol.2005.09.028 Pallett LJ, Gill US, Quaglia A, Sinclair LV, Jover-Cobos M, Schurich A, Singh KP, Thomas N, Das A, Chen A, et al. 2015. Metabolic regulation of hep- Received July 3, 2018; accepted in revised form May 14, 2019. Genome Research 919 www.genome.org Downloaded from genome.cshlp.org on February 15, 2021 - Published by Cold Spring Harbor Laboratory Press Systems analysis reveals complex biological processes during virus infection fate decisions Jordi Argilaguet, Mireia Pedragosa, Anna Esteve-Codina, et al. Genome Res. 2019 29: 907-919 originally published online May 28, 2019 Access the most recent version at doi:10.1101/gr.241372.118 Supplemental http://genome.cshlp.org/content/suppl/2019/05/31/gr.241372.118.DC1 Material References This article cites 48 articles, 9 of which can be accessed free at: http://genome.cshlp.org/content/29/6/907.full.html#ref-list-1 Creative This article is distributed exclusively by Cold Spring Harbor Laboratory Press for the Commons first six months after the full-issue publication date (see License http://genome.cshlp.org/site/misc/terms.xhtml). After six months, it is available under a Creative Commons License (Attribution-NonCommercial 4.0 International), as described at http://creativecommons.org/licenses/by-nc/4.0/. Email Alerting Receive free email alerts when new articles cite this article - sign up in the box at the Service top right corner of the article or click here. To subscribe to Genome Research go to: https://genome.cshlp.org/subscriptions © 2019 Argilaguet et al.; Published by Cold Spring Harbor Laboratory Press