VISTA expression by microglia decreases during inflammation and is differentially regulated in CNS diseases

Abstract V‐type immunoglobulin domain‐containing suppressor of T‐cell activation (VISTA) is a negative checkpoint regulator (NCR) involved in inhibition of T cell‐mediated immunity. Expression changes of other NCRs (PD‐1, PD‐L1/L2, CTLA‐4) during inflammation of the central nervous system (CNS) were previously demonstrated, but VISTA expression in the CNS has not yet been explored. Here, we report that in the human and mouse CNS, VISTA is most abundantly expressed by microglia, and to lower levels by endothelial cells. Upon TLR stimulation, VISTA expression was reduced in primary neonatal mouse and adult rhesus macaque microglia in vitro. In mice, microglial VISTA expression was reduced after lipopolysaccharide (LPS) injection, during experimental autoimmune encephalomyelitis (EAE), and in the accelerated aging Ercc1 Δ/− mouse model. After LPS injection, decreased VISTA expression in mouse microglia was accompanied by decreased acetylation of lysine residue 27 in histone 3 in both its promoter and enhancer region. ATAC‐sequencing indicated a potential regulation of VISTA expression by Pu.1 and Mafb, two transcription factors crucial for microglia function. Finally, our data suggested that VISTA expression was decreased in microglia in multiple sclerosis lesion tissue, whereas it was increased in Alzheimer's disease patients. This study is the first to demonstrate that in the CNS, VISTA is expressed by microglia, and that VISTA is differentially expressed in CNS pathologies.


| INTRODUCTION
Immune checkpoint regulators are a group of molecules expressed on T cells or antigen-presenting cells (APCs), which can provide costimulatory and co-inhibitory signals during T cell activation. This balance between positive and negative signals is essential for mounting antigen specific immune responses, while limiting the risk for autoimmunity. Inhibition of negative checkpoint regulator (NCR) activity has recently entered the clinic as a treatment for cancer, whereas activation of NCRs has potential for the treatment of autoimmunity. Therapeutic inhibition of NCR activity (immunotherapy) in cancer has been associated with the development of central nervous system (CNS) diseases such as encephalitis, myelitis, and exacerbation of multiple sclerosis (MS) (Cuzzubbo et al., 2017;Yshii, Hohlfeld, & Liblau, 2017).
Accordingly, several studies have reported that inhibition of NCRs (CTLA4, PD-1, and PD-L1) in mouse experimental autoimmune encephalomyelitis (EAE), a model of MS, leads to exacerbation of symptoms (Joller, Peters, Anderson, & Kuchroo, 2012). Furthermore, blockade of PD-1 in a mouse model of Alzheimer's disease (AD) improves cognitive performance , suggesting a broad role of NCRs in CNS pathologies. However, the effectiveness of PD-1 blockade in AD is still not resolved (Latta-Mahieu et al., 2018).
V-domain Ig-containing suppressor of T cell activation (VISTA) is a recently discovered NCR, which shares 24% sequence similarity with PD-L1 . VISTA (aliases: PD-1H (Flies, Wang, Xu, & Chen, 2011); DD1α (Yoon et al., 2015); Dies1 (Aloia, Parisi, Fusco, Pastore, & Russo, 2010); Gi24 (Sakr et al., 2010)) is expressed on myeloid and T cells, and can act as both receptor and ligand (Lines, Sempere, Broughton, Wang, & Noelle, 2014). VISTA expressed on APC can function as a ligand, suppressing T-cell activation upon binding to a yet unidentified counter receptor . In addition, ligation of VISTA expressed on T cells also leads to inhibition of T cell activation (Flies et al., 2014). Receptor functions on myeloid cells include regulation of the cytokine response (Bharaj et al., 2014) and uptake of apoptotic cells (Yoon et al., 2015). Deficiency of VISTA in mice increases susceptibility to developing autoimmunity such as EAE  and lupus nephritis (Ceeraz et al., 2016).
In the CNS, expression of several NCRs by microglia has been reported, which is induced under inflammatory conditions (Yshii et al., 2017). Microglia are the principal innate immune cell type of the CNS, which acquire diverse functional phenotypes in response to environmental cues (Salter & Beggs, 2014). During CNS pathologies, microglia lose their homeostatic signature and can shift to an immune-activated state, as evident from transcriptomic studies (Perry & Holmes, 2014;Zrzavy et al., 2017). During immune activation, microglia upregulate inflammatory genes such as genes involved in cytokine production and antigen presentation. It is now appreciated that in conditions such as AD and aging, microglia can acquire a phagocytic phenotype (Galatro et al., 2017a;Krasemann et al., 2017;Varol et al., 2017), thereby facilitating clearance of debris and toxins. In accordance with a more protective phenotype during disease, microglia obtain immunoregulatory characteristics during inflammation. Induction of NCRs (e.g., PD-L1) in immune-activated microglia causes inhibition of T cell activation, and suppression of cytokine production (Duncan & Miller, 2011;Magnus, 2005;Schachtele, Hu, Sheng, Mutnal, & Lokensgard, 2014). Hence, microglial activation is highly heterogeneous and plastic, and can therefore be detrimental or beneficial during disease.
Although the expression and regulation of several NCRs in the CNS and by microglia has been studied, expression patterns of VISTA are unknown as noted recently by Yshii et al. (2017). Studies suggest involvement of VISTA in important functions of monocytes and macrophages, such as cytokine responses and phagocytosis, which exhibit functional similarities to microglia. As microglia express many NCRs and are able to acquire immunoregulatory functions during inflammation, analysis of VISTA expression in these cells could help to understand their role in CNS disease. Furthermore, NCR modulation as a treatment for cancer and autoimmunity impacts CNS biology, which is demonstrated by adverse neurological effects during immunotherapy including encephalitis, myelitis, hypophysitis, and transition from radiologically isolated syndrome to MS (Cuzzubbo et al., 2017;Yshii et al., 2017). Hence, investigating expression and expression changes of VISTA in the CNS might facilitate understanding the impact of NCR modulation on the CNS.
Here, we assessed VISTA expression in mouse and rhesus macaque microglia after immune-activation in vitro and in vivo, and verified our findings using human postmortem tissue. Our results indicate that VISTA is differentially expressed in microglia during inflammation and neurodegeneration. Furthermore, we determined epigenetic changes in the VISTA gene during microglial activation in mice. These findings provide first evidence of a function of VISTA in microglia and during CNS pathology.

| Animals
All animal experiments were approved by the Netherlands Central Committee for Animal Experiments and the University of Groningen.
Mice were housed SPF in groups in macrolon cages with ad libitum access to water and food, and a 12 hr light-dark cycle. Eight-weekold male C57BL/6 mice (bred in-house) were injected with 1 mg/kg LPS (E. coli 0111:B4, Sigma-Aldrich, L4391) intraperitoneally and sacrificed 24 hr later. To generate Ercc1 Δ/− mice, Ercc1 Δ/+ Fvb mice were bred with Ercc1 +/− C57BL/6 mice. Genotype was confirmed using PCR and Ercc1 Δ/− mice were matched with wildtype littermates of the same sex. After 3-4 months of age, Ercc1 Δ/− mice started to develop tremors and aberrant behavior, and were sacrificed. For induction of EAE, 10-week-old female C57BL/6 mice (Harlan, The Netherlands) were immunized with MOG 35-55 in complete Freund's adjuvant (CFA) (Hooke, EK-2110) and injected with pertussis toxin on the day of immunization and 24 hr later. Mice were monitored daily for development of EAE and sacrificed at score 1 (limp tail), score 4 (complete hind leg paralysis), and remission (partial regain of movement in hind legs).

| Immunohistochemistry
Immunohistochemical staining was performed on formalin-fixed paraffin-embedded (FFPE) or paraformaldehyde (PFA)-fixed frozen tissue as indicated. Briefly, FFPE tissue was deparaffinized in xylene (J.T. Baker, 9490) and rehydrated. For human tissue, sodium citrate (pH 6.0) heat-induced antigen retrieval was performed in a microwave using a pressure cooker, whereas Tris-EDTA (pH 9.0) was used for mouse tissue. Endogenous peroxidase activity was blocked in 0.3% hydrogen peroxide for 30 min and mouse tissue was additionally blocked in 10% normal serum. Primary antibodies were applied at 4 C overnight (Supporting Information, Table S1). For human tissue, primary antibodies were diluted in Normal Antibody Green Bright Diluent (ImmunoLogic, BD09-500). Fluorophore-conjugated (Molecular Probes) or biotinylated (Vector) secondary antibodies were applied for 1 hr at room temperature (RT). For fluorescence staining, tissue was incubated 10 min in Hoechst and human tissue was treated with 0.3% sudan black to quench autofluorescence. Tyramide Superboost streptavidin kit (Invitrogen, B40933) was used for VISTA (clone MH5a) according to manufacturer's instructions. For enzymatic immunostaining, tissue was incubated 30 min in Vectastain Elite ABC-HRP (Vector, PK-6100) and immunoreactivity was revealed using 3,3 0 -diaminobenzidine.

| Primary mouse neonatal microglia culture
Primary neonatal mouse microglia cultures were prepared as described previously (Schaafsma et al., 2015) with minor deviations.

| Flow cytometry
Primary microglia were harvested using Accutase (Sigma-Aldrich, A6964) and resuspended in medium (HBSS without phenol red Dapi was used to distinguish dead cells.

| Transcription factor motif enrichment analysis
To determine potential TF binding motifs in ATAC-seq peaks, a motif enrichment analysis was performed using the motif discovery software HOMER (version 4.9) (Salk Institute and University of California San Diego) (Heinz et al., 2010). Peaks on the VISTA locus (eight in total) were identified in Integrative Genomics Viewer (Broad Institute and University of California) and sequences were used for HOMER enrichment analysis (findMotifs, homer2). HOMER analysis determines motifs enriched in sequences compared to scrambled sequences.

| Statistical analysis
Statistical analyses were performed using GraphPad Prism 7 (GraphPad Software, Inc.). For multiple comparison after LPS and TLR experiments, a one-way ANOVA including Dunnett's test for multiple comparison was used. For comparison of gene expression in LPSinjected mice and Ercc1 Δ/− , a (un)paired Student's t test was performed. All error bars indicate mean AE standard deviation (SD).

| VISTA is primarily expressed by microglia in human and mouse CNS
Expression of most NCRs (CTLA4, PD-1, PD-L1, and more), but not VISTA, has been reported in the CNS by microglia, endothelial cells, astrocytes, oligodendrocytes, and/or neurons (Yshii et al., 2017).
Using a combination of immunohistochemistry and flow cytometry, we assessed VISTA expression in mouse and human brain.
In both mouse and human brain tissue without apparent CNS pathology, VISTA immunoreactivity was evident on ramified microglia-like cells and on blood vessel structures (Figure 1a,b). Immunofluorescence co-staining of Iba1 and VISTA (mouse) and CD68 and VISTA (human) revealed a strong co-expression of these proteins ( Figure 1c,d), confirming VISTA expression in microglia.
In accordance with the immunostainings, flow cytometry of whole mouse brain and spinal cord showed that >95% of Cd11b + Cd45 int microglia exhibited surface VISTA expression. Furthermore, the vast majority of Cd11b − Cd45 − cells did not express detectable levels of cell surface VISTA (Figure 1e).
These findings demonstrate that VISTA is primarily expressed by microglia, and to a lesser extent in blood vessels in the CNS.
3.2 | VISTA expression is abundant in adult microglia and expression levels are similar to microglia signature genes To confirm our observations of VISTA expression in microglia and blood vessels, we analyzed published RNA-seq data for VISTA expression in different CNS cell types (Zhang et al., 2014(Zhang et al., , 2016. In mouse brain, VISTA was abundantly expressed by microglia, weakly expressed by endothelial cells, and was not detected in oligodendrocytes, neurons, and astrocytes ( Figure 2a). In human brain, VISTA was expressed by microglia, but also at moderate levels by endothelial cells, and at low levels by astrocytes ( Figure 2b). VISTA expression was very low in oligodendrocytes and neurons. Of note, we did not observe any VISTA immunoreactivity in astrocytes (Figure 1a  In summary, these data demonstrate that VISTA is abundantly expressed by adult human and mouse microglia, and that expression is comparable to microglia signature genes.   Figure S2a). We observed a significant decrease in VISTA expression in all stages of EAE (score 1, 4, and remission) in spinal cord, hindbrain, and forebrain microglia compared to nonimmunized mice ( Figure 4a). In contrast, Pdl1 was upregulated in all conditions (Figure 4b).
To further assess VISTA expression changes during microglial activation, we quantified VISTA mRNA in microglia isolated from Ercc1 Δ/− mice. Ercc1 is a protein essential for nucleotide excision DNA repair and mutant mice display an accelerated aging phenotype (Vermeij et al., 2016). Microglia from whole brain of 4-month-old Ercc1 Δ/− mice exhibited increased Il1β and Axl expression (Supporting Information, Figure S2b), indicating an immune-activated and phagocytic phenotype. VISTA expression in these microglia was significantly reduced compared to wild type (WT) littermates (Figure 4c), whereas Pdl1 expression was increased (Figure 4d).
These data demonstrate that VISTA expression is decreased in microglia in different mouse models of CNS inflammation and during microglial activation, which is in line with our in vitro observations.

| Reduced VISTA expression in LPS-activated microglia is accompanied by chromatin remodeling
To determine if changes in VISTA expression are accompanied by epigenetic alterations, we analyzed a recently generated dataset containing genome-wide transcriptional changes (RNA-seq), histone modifications (ChIP-seq), and chromatin accessibility (ATAC-seq) in isolated microglia after LPS exposure in mice (Zhang et al., in press).  (Johnson et al., 2017). Cdh23 was not altered in response to LPS, demonstrating that changes in VISTA expression were independent of the Cdh23 gene (Supporting Information, Figure S3a).
To determine if decreased VISTA expression is accompanied by epigenetic changes in the gene locus, we analysed ChIP-seq and ATAC-seq datasets. Concomitant with reduced VISTA expression after LPS exposure, we observed decreased H3K27 histone acetylation (H3K27ac) upstream of the VISTA gene (Figure 5c). H3K27ac is enriched on enhancers and associated with active gene transcription.
We next assessed chromatin accessibility using ATAC-seq data.
ATAC-seq provides information about transposase-accessibility of chromatin at specific locations on the genome (Buenrostro, Wu, Chang, & Greenleaf, 2015). Transposase-accessible chromatin is also accessible for transcription factors (TF) and indicated as Peaks (1-8) on the VISTA gene (Figure 5c). Enrichment analysis for putative TF binding motifs revealed that 17 consensus sequences were significantly enriched in the DNA underlying these peaks (Table 1) In DNA sequences of ATAC peaks that were reduced in microglia after LPS injection, we detected consensus binding sites for Pu.1, Rfx6, Elf5, and Sox15 (Peaks 4-6) ( Figure 5c and Table 1). In contrast, Ap4 and Nf1 binding motifs were enriched in DNA sequences of peaks unaltered by LPS stimulation (Peaks 1-3, 7-8) (Figure 5c and Table 1).
Our findings show that reduced VISTA expression is accompanied by altered histone modification enrichments and changes in chromatin accessibility that are associated with transcriptional repression. Furthermore, the presence of consensus binding sites for Pu.1 and Mafb on chromatin accessible DNA on the VISTA gene suggests that VISTA may be regulated by these microglia-specific TF, and that reduced accessibility of Pu.1, Elf5, and Sox15 to consensus binding sites in VISTA underlie reduced expression in response to LPS.

| VISTA expression is differentially regulated in the human CNS
In view of the observed reduction in VISTA expression during microglial activation in vitro and in vivo, we next assessed VISTA expression in human brain tissue of young and old individuals, and in septicemia, MS and AD patients (Supporting Information, Table 3). Based on neuropathological evaluations, one representative patient was selected for analysis of each condition.
Using IBA1 immunostaining, microglial activation was assessed based on morphology and staining intensity, and VISTA immunoreactivity was determined in consecutive tissue sections (Figures 6 and 7).
In tissue of a young individual (27 years), microglia exhibited a typical resting, ramified morphology, and VISTA expression was detected on microglia and endothelial cells (Figure 6). In both the old individual FIGURE 4 VISTA expression is reduced in adult microglia in mouse models of CNS pathology. VISTA (a + c) and Pdl1 (b + d) mRNA levels in acutely isolated adult microglia from spinal cord, hindbrain and forebrain during EAE disease course (C = control, 1 = disease score 1, 4 = disease score 4, R = remission) (a + b), and from whole brain of accelerated aging Ercc1 Δ/− mice and WT littermates (c + d), measured using RT-qPCR and normalized to Hprt1. Statistical analysis conducted was a one-way ANOVA with Dunnett's test for multiple comparisons (a + b) and a paired Student's t test for direct comparisons (c + d). Error bars indicate mean AE SD. WT = wild type, *p < .05, **p < .01, ***p < .001 (70 years) and the septicemia patient, we observed only weakly activated microglia, and VISTA immunoreactivity in microglia was slightly reduced (Figure 6 and Table 2). VISTA staining of endothelium, however, did not seem to be affected. In the AD patient, a strong IBA1 staining intensity suggested microglial activation, which correlated with strong VISTA expression, and was specifically observed in microglia clusters ( Figure 6). Co-staining of IBA1 and β-amyloid revealed that these microglia clusters were surrounding β-amyloid plaque (Supporting Information, Figure S4). No difference in endothelial VISTA expression was observed. In MS normal-appearing FIGURE 5 Decreased VISTA expression after LPS injection is associated with chromatin remodeling in microglia. (a,b) RNA-seq counts per million of VISTA (a) and Pdl1 (b) mRNA expression (n = 3). (c) H3K27ac histone acetylation (top) and ATAC-seq (bottom) peaks corresponding to the VISTA gene (n = 3). ATAC-seq peaks were numbered 1-8 and peaks that were decreased after LPS stimulation are indicated in red (4-6). Data are derived from previously generated datasets (Zhang et al., in press). Transcription factor binding motifs enriched in these peaks are listed in Table 1. Error bars indicate mean AE SD. # = differential expression (DE) based on RNA-seq analysis [Color figure can be viewed at wileyonlinelibrary.com] Motifs enriched in peaks decreased upon LPS (Peaks 4-6). Motifs enriched in peaks unchanged upon LPS (Peaks 1-3, 7, and 8).
Motifs enriched in both unchanged peaks and peaks decreased by LPS.
white matter (NAWM), microglial activation was low and VISTA was highly expressed on microglia and endothelium (Figure 7 and Table 2).
Within and around a chronic lesion in the MS tissue, intermediate to strong IBA1 staining was observed, whereas VISTA staining was almost absent in microglia and endothelial cells.
To establish a correlation between VISTA expression and CNS inflammation in these tissues, we performed a co-staining of VISTA with HLA-DR (Figures 6 and 7). In line with the previous immunostaining, low levels of HLA-DR in the young individual, and positive VISTA signals in microglia and endothelial cells were observed ( Figure 6 and Table 2). In the old and the septicemia donors, HLA-DR expression was only slightly increased, but VISTA expression remained unchanged. In the AD tissue, HLA-DR was highly expressed by microglia, which also abundantly expressed VISTA. In MS, co-staining revealed a negative correlation of HLA-DR and VISTA expression (Figure 7 and Table 2).
VISTA was not detected in microglia/macrophages and endothelial cells at the rim of the lesion; however, a weak VISTA signal was observed in some amoeboid cells expressing high HLA-DR within the lesion.  Yoon et al. (2015) showed that VISTA expression in phagocytic cells is essential for uptake of apoptotic cells. Furthermore, VISTA is involved in the immune response of myeloid cells, as overexpression in human monocytes leads to spontaneous cytokine secretion (Bharaj et al., 2014). Thus, VISTA may be involved in immune surveillance and uptake of apoptotic neurons or other debris by microglia.
Consistent with this argument, our findings indicate that VISTA expression is regulated similarly to known homeostatic microglia genes such as TMEM119 and P2RY12. These signature genes are downregulated upon microglial activation (Grabert et al., 2016).
Expression of P2RY12 is lost in active MS lesions (Zrzavy et al., 2017), The downregulation of VISTA that we observed in activated microglia stands in marked contrast to published studies on expression of other NCRs (Yshii et al., 2017). In microglia, expression of several NCRs is induced or upregulated upon inflammatory stimuli (Yshii et al., 2017), which we confirmed for PD-L1 expression in this study.
Moreover, VISTA expression is upregulated in TLR-stimulated monocytes and macrophages (Bharaj et al., 2014;Wang et al., 2011). This discrepancy underscores our previous argument that VISTA has additional functions in microglia that deviate from other NCRs and other myeloid cells. Considering the function of VISTA in apoptotic cell clearance and cytokine response (Bharaj et al., 2014;Yoon et al., 2015), a loss of VISTA in activated microglia may reduce their ability to clear debris or to mount a cytokine response. However, VISTA may also function as an NCR in microglia, and downregulation likely has consequences for CNS pathologies which involve T cell infiltration.
Several studies have shown that knockout of NCRs including VISTA in mice promotes the development of EAE (Joller et al., 2012;Wang et al., 2014 (Graeber, Li, & Rodriguez, 2011). In AD, microglia are activated by amyloid-β (Aβ) and neuronal debris, contributing to clearance, but also causing tissue damage. A recent study suggests that aggregated Aβ sensed by microglia causes inflammasome activation, which contributes to progression and spreading of inflammation and Aβ pathology (Venegas et al., 2017). Furthermore, single-cell RNAsequence data suggest that plaque-associated microglia in mice obtain a Trem2-dependent phagocytic phenotype (Keren-Shaul et al., 2017).
Hence, VISTA expression changes in activated microglia may depend on environmental cues in CNS pathologies, such as interactions with peripheral immune infiltrates in MS, or activation by Aβ in AD.
Interestingly, we also observed downregulation of VISTA in endothelial cells in chronic MS lesion. The endothelium is involved in MS pathology by recruitment of immune cells to the CNS (Yun, Minagar, & Alexander, 2017). Endothelial cells are involved in antigen presentation of CNS components to antigen-specific lymphocytes (Galea et al., 2007;Lopes Pinheiro et al., 2016;Traugott & Raine, 1985). Blocking co-inhibitory molecules PD-L1 or PD-L2 on human endothelial cells facilitates transmigration responses of lymphocytes in vitro (Pittet, Newcombe, Prat, & Arbour, 2011;Rodig et al., 2003). Reduced VISTA expression on endothelium in MS could therefore promote activation and transmigration of lymphocytes into the CNS. Additional studies are needed to assess the effect of VISTA deficiency in endothelial cells regarding antigen presentation and immune cell infiltration.
In conclusion, we present an elaborate multi-species analysis of VISTA expression in the CNS, including changes during pathology. We demonstrate that VISTA is abundantly expressed by microglia, suggesting a functional role in these cells. Differential expression of VISTA during CNS pathology highlights the importance to further elucidate the function of VISTA in the CNS. Our study is the first to