VISTA deficiency attenuates antibody-induced arthritis and alters macrophage gene expression in response to simulated immune complexes

DBA/1 J mice were purchased from Jackson Labs. VISTA-deficient (V-KO) mice were produced on the C57Bl/6 genetic background as described previously [13]. V-KO mice were bred at the VA Medical Center. All procedures were approved by the VA Medical Center IACUC, and comply with federal guidelines.

Hamster anti-mouse VISTA antibody MH5A was purchased from Biolegend. Anti-mouse VISTA antibody (8G8) was produced in hamster by the Noelle Lab at Geisel Medical School. The specificity of 8G8 was confirmed using VISTA-specific ELISAs, assessing the distribution of antigen across cell lineages and the absence of detection on cells from V-KO mice. CAIA was induced in DBA/1 J mice by IP injection of 1.5 mg 5-clone Cocktail (Chondrex, Redmond, WA, USA) with an IP injection of 50 μg LPS given 3 days afterward. Mice were treated by IP injections of monoclonal antibodies (100 μg MH5A or 300 μg 8G8) or control hamster IgG beginning 3 days before induction of CAIA and every 3 days thereafter. CAIA was induced in V-KO and C57Bl/6 WT mice by IP injection of 5 mg of 5-clone Cocktail (Chondrex) and an IP injection of 50 μg LPS given 3 days afterward. Animals were scored daily for inflammation in each paw using a scale of 0–4; the total scores for each mouse were summed and the mean of these sums for each experimental group was plotted as the arthritis score with standard error and statistical analysis by nonparametric Mann–Whitney test. The data shown are representative of multiple experiments performed with V-KO mice and antibody treatment.

Histologic assessment of joint damage was performed using tissue sections of formalin-fixed decalcified joints of mice (knees, ankles, paws) collected 13 days after initiation of CAIA. Histologic analyses in preliminary experiments indicated that arthritis initially develops in the knee joints in wild-type mice several days prior to the appearance of overt paw/ankle swelling; consequently, knee joints were scored and analyzed. A cumulative scoring method was used since many distal joints of V-KO mice were found to be completely devoid of signs of arthritis, making the calculation of mean scores and statistical analyses for the experimental groups particularly difficult. Three tissue sections representing 500 μm of tissue at the approximate center of joints were examined and scored (0–4) for extent of pannus formation, cartilage damage/loss, and bone erosions.

Bone marrow-derived macrophage (BMDM) cultures were prepared from V-KO and C57Bl/6 (WT) mice as described previously [22]. Briefly, whole bone marrow was flushed from the femurs and cultured in BMDM medium (RPMI supplemented with 10% FBS, 50 ng/ml recombinant mouse CSF-1, and Pen/Strep). Cultures were fed on day 3 and then harvested on days 5 and 7 for total cell protein and RNA. Splenocytes were harvested from 7-week-old male C57/Bl/6 and V-KO mice and inflammatory monocytes, F4/80⁺Gr1⁺CD11b⁺ cells expressing C5aR, were sorted on a BD FACSAria III (BD Bioscience). Cells were stimulated with 10 ng/ml LPS (Sigma-Aldrich) for 24 hours and supernatants were collected and screened for cytokines/chemokines by Luminex assay.

Localization of VISTA was performed on tissue sections of formalin-fixed and paraffin-embedded human synovial tissue selected from a collection of de-identified, archived materials (RAF) or were purchased from a commercial source (Origene, Rockville, MD, USA). Tissue sections were deparaffinized and subjected to antigen retrieval with citrate buffer, and endogenous peroxidase was quenched with hydrogen-peroxide in methanol. Nonspecific binding was blocked with 2.5% goat serum and slides were incubated at 4 °C with rabbit monoclonal anti-VISTA (catalog number D1L2G XP; Cell Signaling), without primary antibody or with an equal concentration of isotype control (catalog number SP137; Abcam). Signal amplification was performed with Biotin-XX Tyramide SuperBoost (catalog number B40921; Invitrogen) followed by reaction with streptavidin horseradish peroxidase (catalog number S911; Invitrogen) and insoluble substrate development with diamino-benzidine (ImmPact DAB, catalog number SK-4103; Vector Laboratories). Slides were counterstained with Hematoxylin2 (catalog number 7231; ThermoFisher).

Whole cell extracts from BMDMs were harvested in 200 ml of 2× Sample Buffer (Sigma Chemical, St. Louis, MO, USA) and heated to 95 °C for 5 minutes. Twenty microliters of each sample were resolved on 10% SDS-PAGE gels and transferred to Immobilon-P membranes (Millipore, Inc., Billerica, MA, USA). Blots were probed with antibodies specific for mouse VISTA (R&D Biosystems, Minneapolis, MN, USA), or with antibodies for phosphorylated AKT, total AKT, phosphorylated ERK, total ERK, beta actin or tubulin (Cell Signaling Technology, Danvers, MA, USA) and were visualized by chemiluminescence.

Splenocytes were incubated with Tris-buffered ammonium chloride (ACT; Sigma-Aldrich) to lyse red blood cells. Cells were then washed in PBS for 5 minutes at 1500 rpm followed by staining with live/dead Fixable Dead Cell Stain (Life Technologies) at RT for 20 minutes. Cells were subsequently washed twice in PBS for 5 minutes at 1500 rpm at 4 °C and stained with the appropriate directly labeled antibodies on ice for 20 minutes. Cells were then washed twice in PBS for 5 minutes at 1500 rpm at 4 °C and fixed until acquisition. Samples were run on the MACSQuant Analyzer 10 (Miltenyi Biotec) and analyzed using the FlowJo software (Treestar). For whole bone marrow, BMDM cultures, purified neutrophils, and PBMC were analyzed by flow cytometry using a Becton Dickinson FACSCanto II cell analyzer and FACSDiva software. Fluorescently labeled antibodies against mouse VISTA (MH5A) C5aR, Cd11b, Ly6G, CD45, and F4/80 were purchased from BioLegend (San Diego, CA, USA), FcRI and FcRIV from BD Biosciences (San Jose, CA), FcRIII from R&D Systems (Minneapolis, MN), and FcRIIb from eBiosciences (ThermoFisher).

Graphs were prepared using GraphPad Prism software versions 6 and 7 (GraphPad, San Diego, CA, USA). Statistical analysis is shown using the Student t test (two-tailed) at ***=P < 0.005, **=P < 0.025, and *=P < 0.05.

Six-well tissue culture plates were coated overnight at 4 °C with the 5-clone monoclonal antibody cocktail containing IgG2a and IgG2b isotype monoclonal antibodies (100 μg/ml, 2 ml/well). Wells were washed twice with PBS prior to the addition of cells. BMDMs from WT and V-KO mice were cultured for 6 days on nontreated tissue culture plates and then harvested by incubating for 10 minutes at 4 °C and scraping with a Cell Lifter (Thermo-Fisher Scientific). Cells were washed, counted, and then added to either coated or uncoated wells at 3 × 10⁵ cells/well. Following an additional 24 hours of culture, RNA was isolated using the RNeasy Minikit (Qiagen, Germantown, MD, USA) for analysis by NanoString assay.

RNA was isolated from the frozen spleen tissue blocks using the PureLink RNA Mini Kit (Ambion/Invitrogen) and PureLink DNase Set (Ambion/Invitrogen). To isolate RNA samples from formalin-fixed paraffin-embedded knee joints, the Qiagen RNeasy FFPE Kit (Qiagen) was used. All samples were run on a Bioanalyzer to determine purity. Gene expression was measured using the nCounter^® GX Mouse Immunology, Mouse Inflammation, and Mouse Myeloid Cell codesets (NanoString Technologies), run and read on an nCounter^® Analysis System (NanoString Technologies).

To analyze the NanoString data, gene expression data from NanoString were normalized in nSolver and log₂-transformed for further analysis for differential expression. Data from joint samples were analyzed in R using unpaired t tests followed by Benjamini and Hochberg multiple hypothesis correction. Data from spleen samples were analyzed in R/Bioconductor using the limma package followed by Benjamini and Hochberg multiple hypothesis correction. Boxplots were made using the R package ggplot2. Heat maps were constructed by UPGMA hierarchical clustering of gene expression using 1 – Pearson’s correlation coefficient as the distance, followed by Z-score row normalization for plotting in Excel or the gplots package in R [23‐27]. Gene expression values from immune-complex-stimulated BMDM cultures were normalized to the geometric mean of housekeeping genes using nSolver software (NanoString Technologies). Gene expression values between V-KO and WT BMDM cultures were compared by multiple two-tailed t tests and discoveries were identified by the Benjamini and Hochberg method, with a Q value of 1% (GraphPad Prism 7). Discovered genes that showed at least a 2-fold change between V-KO and WT BMDM cultures, either under basal or IgG-stimulated conditions, were chosen for hierarchical clustering. A heat map was generated using nSolver software, with a Genes z-score transformation, average linkage, and Pearson’s correlation as the distance metric.

BMDMs in complete media were plated in the upper wells of invasion chambers fitted with 8-μM FluoroBlok inserts (BD Biosciences, Billerica, MA, USA) according to the manufacturer’s instructions. Medium containing 50 ng/ml C5a was added to the lower chamber to act as a chemoattractant. After an additional 2 hours, invaded cells were stained with 4 μM Calcein AM and imaged using a Nikon ECLIPSE TS100 inverted microscope fitted with 4× objective and a Spot digital camera. Migrated cells were counted in triplicate wells using NIS Elements Advanced Research imaging software.

A procedure for localization of immunoreactive VISTA was developed and optimized by preliminary staining of sections of human tonsil in which VISTA-expressing cells were stained predominately in the interfollicular zone in a pattern similar to that of CD68-expressing macrophages/dendritic cells (data not shown). In human synovium, the highest expression of VISTA protein was by the synovial lining membrane cells. This was the case in both normal human synovium (Fig. 1b) and synovium of patients with RA (Fig. 1e, h). Virtually every cell of the synovial lining reacted strongly with anti-VISTA (Fig. 1b, e, h) but not with isotype control antibody (Fig. 1c, f, i). Also, cells identified by their polymorphic nuclei as neutrophils were strongly stained for VISTA immunoreactivity (Fig. 1k). No staining occurred in controls performed with nonspecific IgG of the same isotype (Fig. 1l), ensuring that nonspecific IgG binding or staining due to unquenched endogenous peroxidase activity was not a confounding factor. A few cells present in lymphoid follicles from a subset of synovial tissue specimens stained for VISTA, although the majority of cells were not stained (Fig. 1h). Note that it is possible for expression of VISTA protein by some cell types in the synovium to be below the limits of detection by this method.

VISTA was also detected by western blot analyses of total protein lysates of synovium of RA patients (n = 3) as well as normal synovium (n = 2) (Fig. 1m), confirming by a second method that VISTA is present in cells in normal synovium and in the inflamed synovium of patients with RA as well. The significance of VISTA expression by resident cells of synovium to the pathogenesis of arthritis remains unknown.

DBA/1 J mice were treated with either nonspecific hamster IgG or one of two different anti-VISTA mAbs; mAb clone MH5A (Fig. 2a) or mAb clone 8G8 (Fig. 2b). With each of the antibodies, prophylactic treatment of mice with systemic anti-VISTA monoclonal antibody greatly reduced paw-swelling arthritis scores (Fig. 2a, b). CAIA was also performed using mice deficient in VISTA gene expression, V-KO mice, to verify that the inhibition of CAIA by treatment of DBA/1 J mice with anti-mouse VISTA antibodies was due to specific binding of antibodies to VISTA and not nonspecific effects. Compared to wild-type C57Bl/6 mice, V-KO mice were highly resistant to induction of arthritis and developed significantly lower arthritis scores and disease incidence (Fig. 2c). Examples of the swelling observed in paws of wild-type mice and a lack of overt paw swelling observed for many joints of V-KO mice are shown (Fig. 2d). Representative photomicrographs of the extensive pannus formation observed in knee joints of WT mice compared to minimal synovial pannus in knee joints of V-KO mice are also shown (Fig. 2e). Cumulative histologic scores for pannus formation, cartilage damage, and bone erosion were greater in WT mice than in V-KO mice in knees, ankles, and paws (Fig. 2f).

To further characterize the attenuation of joint inflammation in V-KO mice, we isolated RNA from knee joints at day 13 of CAIA and performed a gene profiling analysis using the NanoString nCounter^® Mouse Inflammation Kit. Knee joints were chosen for this analysis because histologic analyses showed that the incidence of inflammation in knees was more uniform for V-KO mice and WT mice than it was in paws (inflammation in knees actually precedes inflammation in distal joints by several days), since many V-KO mice completely lacked inflammation of paws and digits. One noteworthy finding was that the knee joints of V-KO mice had nearly 4-fold less MMP-3 gene expression compared to their WT counterparts on day 13 of arthritis (Fig. 3), consistent with the reduced cartilage and bone erosion scores (Fig. 2f). Interestingly, MMP-3 expression was reduced in V-KO joints even though IL-1B, IL-6, or TNF was not reduced at this point of disease (data not shown), suggesting that MMP-3 expression in joints may not have been regulated primarily by inflammatory cytokines, as is often the case in the much less complex cell culture model systems. Surprisingly, certain genes implicated in arthritis pathogenesis were more than 2-fold higher in arthritic V-KO joints, including NOS2, IL-23a, and IFNA1 (Fig. 3). Among the genes that were overexpressed in V-KO arthritic joints were several typically associated with the M2 macrophage phenotype, usually associated with resolution or antagonism of inflammation, including CCL11 and CCL24 (Fig. 3) (a comprehensive gene list is included in Additional file 1: Figure S1).

Using NanoString technology, gene expression analysis of spleens from WT and V-KO mice undergoing CAIA was performed. This analysis of total splenocytes revealed significant reductions in genes associated with macrophage function, including CD163, CD36, Cd1d1, and CD14 in spleens from V-KO mice (Additional file 2: Figure S2).

Since phagocyte responses to the complement-derived peptide C5a are critical for induction of the CAIA model, we investigated the plasma concentrations of C5a during CAIA induction, the expression of the cell surface C5a receptor, and selected in-vitro responses to C5a for WT versus V-KO mice [21]. Comparable levels of C5a were detected in the plasma of WT and V-KO mice on day 6 after CAIA initiation, rendering it unlikely that attenuated induction of disease in V-KO mice was due to defective generation of complement fragment C5a (data not shown).

Interestingly, FACS analysis of neutrophils and monocytes showed that cell surface expression of C5a receptor was consistently reduced for V-KO mice compared to cells from WT mice, both on cells in the peripheral blood and on cells in the bone marrow (Fig. 4a, b). This difference in MFI for WT versus V-KO cell surface expression of C5aR was statistically significant both in vivo and in cultured BMDMs (Fig. 4b, d). Reduced C5a receptor was also observed in a monocyte subset of particular interest in joint inflammation, the F4/80⁺/Gr1⁺/CD11b⁺ “inflammatory” monocyte (Additional file 3: Figure S3), which was further examined. Inflammatory monocytes that expressed C5aR were reduced in abundance in spleens of V-KO mice compared to WT mice and this subset also had reduced cell surface C5aR expression as evidenced by a difference in mean MFI values in FACS analyses (WT (n = 6) MFI = 1192 ± 212; V-KO (n = 6) MFI = 428 ± 76.3; p = 007). The inflammatory monocyte subset was isolated by FACS analysis and cells were assayed for the production of inflammatory cytokines both with and without LPS stimulation. Cells from V-KO mice produced less TNF, IL-17, and IL-12p70 in response to LPS compared to cells from WT mice (Additional file 3: Figure S3). These results indicate that the inflammatory subset of monocytes in V-KO mice has important functional differences that could contribute to the attenuation of CAIA.

To further examine the role of VISTA on monocyte and macrophage function, BMDM cultures were established and the expression of VISTA protein was assessed during in-vitro differentiation of macrophages. VISTA was detected by western blot analysis as a 55-kDa protein with maximal expression observed on day 6 of culture of BMDMs from WT mice. The 55-kDa protein is the glycosylated form of VISTA [4]. This 55-kDa protein was completely absent from the lysates of V-KO BMDM cells as expected, as shown for day 5 of culture (Fig. 4c, right panel). Western blot analysis confirmed that other important macrophage proteins such as the C5a receptor were expressed in BMDM cultures from WT and V-KO mice, and total cellular C5aR was expressed in similar amounts for WT and V-KO BMDMs as shown for day 5 of culture (Fig. 4c, right panel). The differentiation/maturation marker F4/80 was expressed similarly by both WT BMDM cells and V-KO BMDM cells when analyzed by flow cytometry, suggesting that, at least by some criteria, in-vitro monocyte differentiation in culture was not greatly altered by VISTA deficiency. However, cell surface expression C5aR was significantly reduced in V-KO BMDM cells (Fig. 4c). The reason for the reduced amount of C5a receptor on the surface of cells from V-KO mice is not yet understood, since total protein levels for C5aR in WT and V-KO BMDM cultures assessed by semi-quantitative western blot analysis of whole cell extracts appeared comparable (Fig. 4c, right panel). Thus, although the underlying mechanism is not understood, C5aR surface expression was reduced in splenic inflammatory monocytes, cultured BMDMs, peripheral blood monocytes, and neutrophils of V-KO mice, possibly contributing to the attenuation of CAIA in these mice.

The effect of the reduction in C5aR expression on the surface of BMDMs upon chemotactic responses to C5a in vitro was examined. Compared to BMDMs from WT mice, BMDMs from V-KO mice exhibited a significantly reduced chemotactic response to C5a at doses of 10 and 50 ng/ml (Fig. 5a). To determine whether C5a receptor signal transduction was also affected in BMDMs from V-KO mice, we assayed the phosphorylation of AKT and ERK, two members of protein kinase signaling pathways normally activated very rapidly by C5a [28]. AKT phosphorylation was maximally induced within 5 minutes by C5a in BMDMs from WT mice, while phosphorylation of AKT was reduced in BMDMs from V-KO mice (Fig. 5b). As expected, AKT was not rapidly phosphorylated by LPS stimulation in WT cells, or in V-KO cells. C5a also induced robust phosphorylation of ERK in WT BMDMs as expected, while ERK phosphorylation was attenuated in V-KO cells (Fig. 5c). In contrast to C5a, LPS activation of ERK phosphorylation was comparable in BMDMs from the two strains. Taken together, the reduced basal cell-surface expression of C5aR in vivo and the reduced responses to C5a observed in cultured cells from V-KO mice suggest the possibility that altered C5aR responses contribute to the attenuation of CAIA in V-KO mice, although not yet definitively demonstrated in vivo.

Immune complexes generated in the CAIA model trigger myeloid cell responses through Fc gamma receptor signaling (Fcgr) [29]. Using the NanoString Myeloid Cell codeset, we performed NanoString gene expression analysis on WT and V-KO BMDMs exposed to plate-bound Chondrex 5-clone antibodies (a mixture of IgG2a and IgG2b antibodies) to simulate the type of immune complex binding to Fc receptors that occurs in CAIA [30]. FACS analysis revealed small but significant reductions in the levels of cell surface expression of FcRI, FcRII, and FcRIII on V-KO cultured macrophages used for this analysis, while there was no difference in FcRIV expression (Fig. 6a). Unstimulated V-KO and WT BMDMs displayed very similar basal gene expression; however, after stimulation with plate-bound IgG, the gene expression \pattern differed greatly for V-KO macrophages and WT macrophages. In fact, 32 genes were differentially expressed between V-KO BMDMs and WT BMDMs after they were stimulated by plate-bound simulated IgG immune complexes (Fig. 6b, c), several more than 10-fold. By comparison, only three genes were differentially expressed by the unstimulated BMDMs. These differences were detected with a 1% false discovery rate (Benjamini and Hochberg) and focusing on 2-fold changes or greater. For example, IgG-stimulated V-KO BMDMs had much higher levels of interleukin-1 receptor antagonist (Il1rn), Mmp13, Fcgr1, and Ifit1bl1 mRNA compared to IgG-treated WT BMDMs. An ELISA performed for IL-1 receptor antagonist confirmed that the protein was made and secreted into culture medium (data not shown). In contrast to induced genes, IgG-treated V-KO BMDMs had less than half the levels of S100a4 and Hist2h2aa1 mRNA observed in the WT IgG-treated cultures. These data demonstrate that V-KO macrophages have very different gene expression responses to simulated IgG immune complexes in vitro, and these differences are likely to have contributed to the overall attenuation of inflammation in V-KO mice that was observed in the CAIA model.