Integrin Alpha E (CD103) Limits Virus-Induced IFN-I Production in Conventional Dendritic Cells

Early and strong production of IFN-I by dendritic cells is important to control vesicular stomatitis virus (VSV), however mechanisms which explain this cell-type specific innate immune activation remain to be defined. Here, using a genome wide association study (GWAS), we identified Integrin alpha-E (Itgae, CD103) as a new regulator of antiviral IFN-I production in a mouse model of vesicular stomatitis virus (VSV) infection. CD103 was specifically expressed by splenic conventional dendritic cells (cDCs) and limited IFN-I production in these cells during VSV infection. Mechanistically, CD103 suppressed AKT phosphorylation and mTOR activation in DCs. Deficiency in CD103 accelerated early IFN-I in cDCs and prevented death in VSV infected animals. In conclusion, CD103 participates in regulation of cDC specific IFN-I induction and thereby influences immune activation after VSV infection.


INTRODUCTION
Type I interferon (IFN-I) is the strongest antiviral cytokine and prevents severe outcome of virus infection (1,2). IFN-I-dependent and -independent antiviral effector programs target several stages of virus life cycle, including viral genome replication, transcription and translation (3). Most important antiviral effector molecules are the myxovirus resistance (Mx) protein, which targets incoming capsids prior to replication (4), 2,5-oligoadenylate synthetase (OAS) and RNase L degrading viral RNA (5), while Protein kinase R (PKR) and eukaryotic initiation factor 2 (eIF2a) limit translation of viral proteins (6). In addition to its antiviral activity, IFN-I is involved in several immune regulatory networks like innate immune activation, NK cell activation, CD8 T cell priming, but also immunosuppressive mechanisms i.e., interleukin-10 (IL-10) and programmed cell death ligand 1 (PD-L1) expression (7,8). Enhanced and prolonged IFN-I activity results in over activation of the immune system which may lead to severe autoimmune disease (9).
During virus infection, signaling via pathogen recognition receptors (PRRs), such as Toll-like receptors (TLRs) or retinoic acid inducible gene I (RIG-I) induce IFN-I resulting in production of IFN-I (10)(11)(12). Several cell types are able to produce IFN-I, however specialized plasmacytoid dendritic cells (pDCs) produce high amounts of IFN-I, and the IFN-I induction in these cells is different from other innate cell types. In early endosomes of pDCs, contact with viral RNA or DNA leads to activation of TLR7 and/or TLR9, IRF7 activation and fast production of IFN-I (13). Conventional dendritic cells (cDCs) are also equipped with TLR7/9 and molecules of the IRF7 signaling cascade, however compared to pDCs, cDCs produce reduced amounts of IFN-I upon activation of TLR7/9 (14). Indeed, during virus infection, cDCs mainly produce IFN-I after activation of RIG-I (15). One explanation for this difference between cDCs and pDCs for IFN-I production is signaling of FcϵR1g and DAP12, which in cDCs inhibits IRF7 activation after TLR7/9 triggering (13). More recently, the involvement of the PI3K-AKT pathway and mTOR signaling was shown to be a major innate immune regulator, which can regulate IFN-I induction and switches pDCs into high IFN-I producer (16)(17)(18). Yet, mechanisms underlying a potential differential AKT/mTOR activation linked to distinct abilities to produce IFN-I in cDCs have to be defined.
The integrin alpha E (Itgae, aE, and CD103) interact with Integrin beta 7 (ITGB7 and b7) to form heterodimer aEb7, which is known to interact with E-cadherin (19). CD103 expression is mainly described on migrating and resident T cells and DCs in many organs such as skin, gut, lung, spleen, lymph nodes or tumors (19)(20)(21). The expression of CD103 is restricted to type I dendritic cells (CD8a+ cDC1) and absent on type II dendritic cells (CD11b+ cDC2). CD103+ cDC1 are efficient in cross presentation by sampling antigens from infected sites or tumors and migrate into the draining lymph node to mount antigen-specific T cell response. CD103+ cDC1s also cross-present self-antigens to maintain self-tolerance. CD103+ DC1s are more crucial cell type for T cell immunity than cDC2 which is necessary for host survival during respiratory viral and bacterial infections (22). Previous reports demonstrated that CD103+CD11b+ DCs respond strongly during airway infections (23)(24)(25). Further intestinal CD103+ CD11b-dendritic cells are essential to prevent severe colitis (26). This regulatory function of CD103+ CD11b-dendritic cells was linked to their ability to induce IDO1 and interleukin-18 in epithelial cells. CD103+ cDC1s support influenza replication as compared to CD103-cDC2 in the lung and are associated with dissemination of virus in the draining lymph node (27). The role of CD103 molecules on DCs is only considered to define the migratory cDC1. However, its direct role for regulating the antiviral immunity in DC1 remains unknown. E-cadherins are surface adhesion molecules and can modulate the PI3-kinase/Akt signaling cascade (28). In fact interaction of E-cadherin with CD103 promotes anti-tumor CTL activity (29,30). In addition, the interaction of E-cadherin with CD103 can modulate several signaling pathways, including Wnt/bcatenin, PI3K/Akt, Rho GTPase, and NF-kB signalling (31). Such interaction will influence the function of monocytes, macrophages and DCs. Whether such modulation can influence the ability of DCs to produce IFN-I remains mainly unknown.
Here, we performed a genome-wide association screen in inbred mouse strains to determine mechanisms that regulate early virus replication and IFN-I induction in the spleen. We identified CD103 as blocker of IFN-I induction in cDCs. Mechanistically CD103 limited AKT phosphorylation and mTOR activation in cDCs.

GWAS and Genphen
The GWAS was performed for 4,000,000 SNPs found in different inbred mouse strains (publicly available from http://mouse.cs. ucla.edu/mousehapmap/full.html ), and the total amount of PFU from spleen of 17 VSV infected inbred mouse lines (three individuals per line) as a quantitative trait. First, the phenotype was log10 transformed to make it conform to normality, followed by an association test between each SNP and the phenotype with EMMA as previously described (33,34). The strength of association was summarized with p-values.
The results obtained with EMMA were tested by a more detailed analysis with genphen (freely available at https://www. bioconductor.org/packages/genphen). To characterize the strength of association between each SNP and the phenotype, genphen computes several association metrics, including classification accuracy (how well the alleles of a given SNP are predicted from the quantitative trait), Cohen's kappa (degree of classification accuracy above expected), Cohen's d (effect size estimated with Bayesian inference model for testing the difference of a continuous trait between two groups (alleles of a SNP). The model was implemented in STAN (35), and is robust against outliers. It enables the computation of complete distribution of credible values for the Cohen's d effect size. For each metric, we estimated its mean, and 99% highest density interval (HDI). The SNPs which scored high with respect to all three metrics were considered as strong associations.
Virus, Plaque Assay, and Cell Line VSV (Indiana strain, Mudd-Summers isolate), was originally obtained from Prof. D. Kolakofsky (University of Geneva, Switzerland). Virus was propagated on BHK-21 hamster kidney fibroblasts at a multiplicity of infection 0.01 and was plaqued onto Vero cells. VSV-EBOV was provided by Prof. Marylyn M. Addo (co-author in this paper), grown on BHK-21 cells, and quantified using Vero cells.

Surface Staining and Flow Cytometry
For surface staining, samples were incubated in antibody solution prepared in FACS buffer for 30 min at 4°C. After incubation samples were washed one time with FACS buffer and processed for FACS analysis.

Cell Preparation and Staining for Dendritic Cell Subsets Study and FACS Sorting
Spleens were injected with DMEM media containing 0.1 mg/ml DNAse (grade II, Sigma Aldrich) and 1 mg/ml collagenase D (Sigma Aldrich) and incubated at 37°C for 40 min. After incubation, spleens were minced with scissor, mixed using pipette to obtain single cell suspension, and cells were washed one time with FACS buffer. Cell suspension was passed through 100-mm strainer, and cells were treated with ACK buffer for 2 minutes to get rid of red blood cells. After one more washing step, cells were processed for antibody staining.
Cells were first blocked for Fc receptors using anti-mouse CD16/CD32 antibody (2.4G2) for 10 min at 4°C. Next, cells were stained for biotin conjugated antibodies against lineage markers CD3e (145-2C11), CD19 (6D5), N1.1 (PK136), and Ly6G (1A8) for 15 minutes. For FACS sorting, in next step, cells were incubated with anti-biotin micro beads (Miltenyi Biotec-130-090-485) and passed through column to get rid of undesired cells. Unbounded cells were washed one time in MACS buffer and incubated with a cocktail of antibodies against CD11c, CD11b, B220 (RA3-6B2), CD8a, PDCA1 (eBio927), CD103, and fluorescent labeled Streptavidin and DAPI for 20 min. After one time washing with MACS buffer, cells were processed for FACS sorting. For normal DC subset analysis, samples were processed for antibodies staining without magnetic sorting. FACS gating strategy for different DC populations is shown in Figure S2A.

Bone Marrow Derived Dendritic Cell (BMDC)
Bone marrow cells were harvested from tibiae and femurs of mice, erythrocytes were lysed with ACK lysis buffer, filtered and cultured in VLE-DMEM media (FG1445; Merck Millipore) supplemented with 5% culture supernatants containing murine GM-CSF (prepared in house from X63-GM-CSF cell line culture), or 20% of supernatant containing murine FLT3L (prepared in house from CHO FLT3-L FLAG cell culture; CHO FLT3-L FLAG cell line was kindly provided by Tim Sparwasser Institute of Infection Immunology, TWINCORE, Hannover, Germany), heat inactivated 10% fetal calf serum (10270; Gibco), 1% of penicillin, streptomycin, L-glutamine solution (G6784 Sigma), and 0.1% 50mM beta mercaptoethanol (D4551, Sigma), which constitute a complete media (CM) and cells were incubated at 37°C in a humidified atmosphere with 5% CO 2 . Bone marrow cells seeded with 3 × 10 6 cells per petri dish (58 cm 2 ) in 10 ml of CM or in 24well plates with 2 × 10 5 cells/well in 1 ml CM. On day 3, extra CM was added with volume equal to initial volume. On days 5-6, cells were either harvested for seeding in 24 well plates or directly treated in petri dishes.

ELISA
Interferon a (IFN-a) enzyme-linked immunosorbent assay (ELISA) was performed according to the manufacturer's protocol (PBL Interferon source).

Statistical Analysis
Data are expressed as means ± SEM. Student's t-test was used to detect statistically significant differences between groups. Statistical analysis for survival between two groups was analyzed by log-rank (Mantel-Cox) test. The level of statistical significance was set at P < 0.05.

GWAS Screen Reveals CD103 as Regulator of Viral Replication
To determine new host factors that modulate virus replication, we performed a genome-wide association study (GWAS) in a mouse model of vesicular stomatitis virus infection (VSV). We selected 17 different commercially available inbred mouse lines, which have been genotyped for single nucleotide polymorphisms (SNPs) and are distinguished by overall 1,846,759 mutations in total of 41,017 genes (36). All 17 mouse strains were infected with 2x10 8 PFU of VSV and after 7 hours we determined infectious replicating viral particles in the spleen. We found very diverse replication of VSV within different mouse strains ( Figure 1A), suggesting that different variants in these strains influence replication of VSV and/or innate immune activation. EMMA-analysis of publicly available 4 million SNPs revealed that among 207,760 SNPs located in exons of coding genes, 5 top SNPs in the genes Ckap2l, Sirpa, Lypla2, Itgae and Ablim1 correlated significantly with VSV replication (Figures 1B and Table S1). To test this first statistical survey using an independent approach, we applied the genphen method (described in material and methods), which classifies all SNPs to their corresponding phenotype. Using this approach, we found SNPs in 28 known different genes which could explain the differences in VSV replication ( Figure 1C and Table S2). Among these 28 genes, 2 genes, Itgae and Sirpa had a missense variant ( Figure 1C). Because these two genes were also among the top hits of the EMMA analysis, we considered them as most probable for explaining the differences found in the different inbred mouse strains. To validate the direct role of Sirpa and Itgae on VSV replication, we infected Sirpa −/− and Itgae −/− mice with VSV and analyzed viral replication after 8 hours. Viral titers were not affected in Sirpa −/− mice, but were reduced in spleen and liver of Itgae −/− mice ( Figure 1D). From these results we concluded that the missense SNP rs26903412 in the Integrin alpha-E (Itgae, CD103) gene on chromosome 11 position 73,115,640 (Supplementary Figure 1) could explain the different VSV replication in the different mouse strains and, more important, that CD103 is one important regulator, which influence early VSV propagation.

CD103 Limits Early Interferon Production and Enhances Susceptibility to Virus Infection
Next, we aimed to analyze the mechanisms leading to different viral replication in the absence of CD103. We considered that early innate immune activation and induction of antiviral programs was one likely explanation for differences in viral replication. Indeed, CD103 was mainly expressed in conventional DCs (Figure 2A), suggesting this cell type was mainly responsible for different VSV replication in Itgae −/− mice and inbred mice with the missense SNPs. Since fast innate immune activation and production of IFN-I is most effective to block VSV replication, we considered that IFN-I response may be modulated by CD103. During VSV infection first pDCs produce IFN-I which is hardly systemically detectable, but still important for viral control (37). In a second line of defense, conventional dendritic cells produce systemically measurable IFN-I (37). To get insights on the role of CD103 on IFN-I induction we first analyzed systemic IFN-alpha in WT and Itgae −/− mice. Itgae −/− mice showed increased systemic IFNalpha already 4 hours after infection ( Figure 2B). At later time points IFN-alpha levels were similar or even enhanced in WT mice ( Figure 2B). This suggests that either other innate immune cells such as pDCs produce higher amounts 4 hours after VSV infection, or that in addition to pDCs, conventional DCs produce IFN-I already 4 hours after VSV infection. To get insights we infected WT and Itgae −/− mice with VSV and then sorted splenic cDC1s, cDC2s and pDCs and analyzed their capacity to produce IFN-I. After 4 hours of VSV infection, pDCs of WT versus Itgae −/− mice showed similar levels of IFN-I mRNA expression ( Figure 2C). Therefore, we conclude, that IFN-I mRNA expression by pDCs is not significantly influenced by CD103 expression ( Figure 2C). In contrast cDC1s and cDC2s of WT versus Itgae −/− mice showed enhanced levels of IFN-I mRNA expression ( Figure 2C), which suggests that absence of CD103 enhances the capacity of these cell type to produce IFN-I, albeit the overall increase of IFN-I mRNA expression in pDCs compared to cDC1 and cDC2 was much higher in infected versus uninfected mice. CD103 expression on cDC2 was absent in naïve state (Figure 2A), however it was induced upon VSV infection (Supplementary Figure 2B) which explains the enhanced IFN-I response in cDC2s after VSV infection ( Figure 2C). The enhanced IFN-I induction in cDCs could not be explained by enhanced replication of VSV in these cells, as VSV replication was reduced in Itgae −/− cells ( Figure 2D and Supplementary Figure 2C). Next, we analyzed whether the different IFN-I induction and viral replication impacted on virus control. Survival experiments revealed enhanced survival of Itgae −/− mice ( Figure 2E), suggesting an important role of CD103 during VSV infection. Together these data show that lack of CD103 enhanced IFN-I production in conventional dendritic cells, which impacts on outcome of the infection.

CD103 Limits IFN-I Response in Bone Marrow Derived Dendritic Cells
To validate our hypothesis that CD103 regulates IFN-I response in cDCs, we generated dendritic cells by culturing bone marrow cells (BMDCs) in presence of GMCSF which are most close to cDCs (38). BMDCs expressed CD103 ( Figure 3A and Supplementary Figure 3A). To test their ability to produce IFN-I in response to VSV, we infected BMDC cultures with VSV and measured IFN-I in RT-PCR. In line with the data from ex vivo analyzed cDCs, we found elevated IFN-I induction in Itgae −/− BMDCs at early time point ( Figure 3B). Again, enhanced IFN-I induction could not be explained by enhanced viral replication, as VSV replication was strongly reduced in Itgae −/− BMDCs ( Figure 3C and Supplementary Figure 3B).
Also, production of infectious virus was strongly reduced in in Itgae −/− BMDC cultures ( Figure 3D).   Figure 3C). We could find similar suppression pattern for virus replication in Itgae −/− BMDCs from theses cultures (Supplementary Figure 3D). Further, we excluded that enhanced virus replication in presence of CD103 was due to a direct interaction of VSV-GP with CD103, as the recently licensed Ebola vaccine VSV-EBOV showed similarly reduced replication in Itgae −/− BMDCs cultures ( Figure 3E). Rather, we considered that the reduced VSV replication in BMDC cultures was due to enhanced IFN-I induction in cultures of Itgae −/− BMDCs. Indeed, blockage of interferon-a/b receptor (IFNAR) enhanced replication of VSV in Itgae −/− BMDCs to WT levels after 72h of infection, although in earlier time points CD103 still accelerated VSV propagation ( Figure 3F). Vice versa treatment with IFN-a4 limited VSV replication in WT BMDCs to levels of Itgae −/− BMDCs ( Figure 3G). In conclusion we found that intrinsic expression of CD103 on dendritic cells limited induction of IFN-I and thereby was responsible at least partially for accelerated VSV replication.

CD103 Modulates AKT Activation and mTOR Signaling
In a next step, we addressed the molecular mechanisms leading to high IFN-I production in Itgae −/− BMDCs. IFN-I is produced upon activation of pattern recognition receptors which recognize viral nucleic acids and positively regulated by various molecules such as AKT/mTOR (18,40), glycogen synthase kinase 3 (GSK3), NF-kB, histone deacetylase 3 (HDAC3) and tyrosineprotein phosphatase non-receptor type 22 (PTPN22) (41). To get insights we generated BMDCs from WT and Itgae −/− mice and analyzed for known pathways that regulate IFN-I signaling.
Interestingly we found increased phosphorylation of AKT in Itgae −/− BMDCs ( Figure 4A). Next, we analyzed for GSK-3b which is a prime target of AKT in downstream signaling and get degraded upon phosphorylation. We detected higher phosphorylation of GSK-3b and reduced total GSK-3b protein level ( Figure 4A). One downstream target of AKT is mTOR (42). mTOR was shown to be one major contributor to IFN-I production in pDC (18). Indeed, phosphorylated mTOR was overexpressed in Itgae −/− BMDCs when compared to WT ( Figure 4B). In line with these findings, inhibition of AKT and mTOR signaling enhanced VSV replication in Itgae −/− BMDCs, but not in WT BMDCs ( Figures 4C, D), suggesting that enhanced AKT/mTOR activation is one crucial parameter underlying the reduced virus replication observed upon CD103 deficiency. Further to confirm the direct role of mTOR inhibition with rapamycin in regulation of IFN-I response, we treated BMDCs with rapamycin and analyzed early IFN-I response upon VSV infection ( Figure 4E). As expected, rapamycin treatment inhibited the virus induced early IFN-I response in Itgae −/− BMDCs similar to the level of WT BMDCs. These findings suggest that CD103 inhibits IFN-I response in AKT/ mTOR dependent manner which promotes virus replication in DCs.

DISCUSSION
Here, we found, using a mouse model of vesicular stomatitis virus (VSV) infection, that Integrin alpha-E (Itgae, CD103) modulated IFN-I induction in cDCs. Mechanistically CD103 limited AKT signaling and mTOR activation, one pathway which is essential for early IFN-I response. In vivo, lack of CD103 resulted in enhanced early IFN-I production and improved virus control and survival during VSV infection. pDCs are specialized IFN-I producing cells. Especially during VSV infection, production of IFN-I by pDCs a few hours after infection is essential for virus control (37,43). Conventional dendritic cells (cDCs) are also equipped with TLR7. However, in contrast to pDCs, cDCs produce hardly any IFN-I after uptake of VSV into TLR7 containing endosomes, and only extensive replication of VSV in cDCs leads to RIG-I activation and IFN-I induction (37,44,45). Our data indicate that expression of CD103 is one factor, which explains this difference between pDCs and cDCs. Lack of CD103 in cDCs accelerated the production of IFN-I. The enhanced IFN-I production in CD103-deficient cDCs could not be explained by extensive replication of VSV in cDCs, as we found limited replication of VSV in CD103-deficient cDCs. From our data it remains unexplained whether the enhanced IFN-I production during VSV infection was specific for either TLR7 or RIG-I. As we found increased mTOR signaling in CD103-deficient cDCs, we consider that both pathways might be affected by CD103. cDC1 are known to express higher level of TLR3 than TLR7. For this study we used VSV, which is a single stranded RNA virus. VSV can activate several innate receptors including TLR4, TLR7, and RIG-I. Therefore, we cannot conclude on the specific role of CD103 in TLR3 signaling. Further studies have to be done to clarify this point.
We consider it likely that the enhanced IFN-I induction is not specific for VSV. This would suggest that cDCs are hyperreactive in terms of IFN-I production, which might accelerate autoimmune disease. Indeed, it was reported previously that mice lacking CD103 start to develop spontaneous inflammatory skin disease after 6 months of their age on 129/Sv x BALB/c background (46). Whether this autoimmunity is linked to overactivated cDCs and is dependent on IFN-I remains unknown so far. Further studies have to be undertaken to address the possible role of CD103, IFN-I signaling and autoimmunity.
We found that high CD103 expression in the spleen is mainly restricted to naïve CD8 + T cells and CD8a + DC (cDC1). CD103 expression on cDC1 modulated IFN-I and promoted virus replication. We have not identified how CD103 influences IFN-I induction and whether certain ligands modulate IFN-I response in cDCs. E-cadherin is one interaction partner of CD103 which can modulate functional activity in CD103 + CD8 T cells (30). Therefore, we could consider that the interaction of E-cadherin with CD103 influences IFN-I induction in DCs.
Van Prooyen et al. published that during fungal infection CD103+ conventional DCs are the major producer of IFN-I and IFN-gamma in the lungs of wild-type mice (25). While these data may look contradictory to our work, the authors did not directly analyze the role of CD103 in IFN-I expression, but rather used expression of CD103 to identify the specific dendritic subset. We could speculate that either, these cell types have limited interaction with E-cadherin or that absence of the CD103 signaling in their system might still enhance the production of IFN-I.
We found that CD103 affected also replication of VSV-EBOV which might have an impact on the outcome of vaccination with VSV-EBOV, which was recently licensed as the first Ebola vaccine and is currently administered large scale in the ongoing Ebola outbreak in the Democratic Republic of Congo. In humans SNPs in the CD103 genes are known, however whether these modulate the interaction of CD103 with Ecadherin and thereby modulate IFN-I responses remains to be studied. If indeed SNPs in CD103 would modulate the binding to E-cadherin such SNPs would accelerate IFN-I response and adaptive immune activation after vaccination. Because VSV-EBOV can persist in joints after vaccination (47), and therefore can lead to side effects, such SNPs could correlate with reduced incidence of side effects. More data have to be generated to shed light in this interaction.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by Landesamt für Natur, Umwelt und Verbraucherschutz (LANUV) Nordrhein-Westfalen.

AUTHOR CONTRIBUTIONS
VD and VK designed and performed the experiments, analyzed data, and wrote the manuscript. SK executed GWAS and genphen analysis, and wrote and corrected the manuscript. TH performed some in vitro and in vivo experiments and analyzed the data. AK was involved in initiation, analysis, and discussion of GWAS. JL helped for performing FACS sorting and RT-PCR for DC subsets. MA, TA, and HB performed immunohistology for VSV-EBOV, serum IFN ELISA and helped in other assays. S-KF and FL helped for in vitro BMDCs culture experiments. PK was involved in the design of the genetic approach, data discussion and identifying CD103, and corrected the manuscript. AF was involved in data discussion. MMA provided VSV-EBOV vaccine, reviewed the manuscript, and discussed the data. CH was involved in experiment design and data discussion. DH initiated and discussed the GWAS and Genphen study, and wrote and corrected the manuscript. PL discussed the data, study design and the GWAS approach and corrected the manuscript. KL initiated, proposed, and organized the study, wrote, and finalized the manuscript. All authors contributed to the article and approved the submitted version.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2020. 607889/full#supplementary-material SUPPLEMENTARY FIGURE 1 | Association between mouse genotype and splenic virus replication of inbred mouse strains. Table showing the details for a top SNP in the Itgae gene associated with virus replication correspondence to its genotype and average virus particles (log 10 transformed, n = 3) per spleen in different strains. 'A' represents Adenine and 'G' represents Guanine.   Figure 1C.