Ox40 ligand (Ox40L) locus genetic variants are associated with the risk for systemic lupus erythematosus (SLE); however, it is unclear how Ox40L contributes to SLE pathogenesis. In this study, we evaluated the contribution of Ox40L and its cognate receptor, Ox40, using in vivo agonist and antagonist approaches in the NZB × NZW (NZB/W) F1 mouse model of SLE. Ox40 was highly expressed on several CD4 Th cell subsets in the spleen and kidney of diseased mice, and expression correlated with disease severity. Treatment of aged NZB/W F1 mice with agonist anti-Ox40 mAbs potently exacerbated renal disease, which was accompanied by activation of kidney-infiltrating T cells and cytokine production. The agonist mAbs also induced activation and inflammatory gene expression in splenic CD4 T cells, including IFN-regulated genes, increased the number of follicular helper T cells and plasmablasts in the spleen, and led to elevated levels of serum IgM and enhanced renal glomerular IgM deposition. In a type I IFN–accelerated lupus model, treatment with an antagonist Ox40:Fc fusion protein significantly delayed the onset of severe proteinuria and improved survival. These data support the hypothesis that the Ox40/Ox40L pathway drives cellular and humoral autoimmune responses during lupus nephritis in NZB/W F1 mice and emphasize the potential clinical value of targeting this pathway in human lupus.
Systemic lupus erythematosus (SLE) is a multiorgan autoimmune disease characterized by aberrant cellular and humoral immune responses. Lupus nephritis (LN), one of the most common and severe clinical presentations of SLE, occurs in up to 50% of adults and 70% of children with the disease (1, 2). Despite decades of effort, most clinical trials for SLE have been disappointing, indicating the urgent need to identify and validate new therapeutic targets.
One key aspect of SLE pathophysiology is that immune complexes (ICs), consisting largely of autoantibodies, such as anti-dsDNA and anti-RNA–binding proteins, together with their cognate Ags, deposit in blood vessels and renal glomeruli, leading to vasculitis and nephritis [(3), reviewed in Refs. 4, 5)]. IC deposition results in the recruitment of lymphocytes and myeloid cells to kidney glomeruli, arterioles, and tubular interstitium, which further exacerbates renal damage.
Recent genome-wide association studies indicate that many immune-related pathways contribute to human SLE, and >50 genetic loci are now associated with disease risk (6). Understanding how these loci predispose to disease is critical for understanding disease etiology and for advancing therapeutic hypotheses. Ox40 ligand (Ox40L; TNFSF4) was originally identified as an SLE risk gene by Vyse and colleagues (7); however, only modest progress has been made in understanding its role in SLE pathogenesis.
Ox40L, and its cognate receptor, Ox40 (TNFRSF4), are members of the TNFR superfamily. There is significant evidence that this receptor/coreceptor pair is responsible for regulating adaptive T cell immune responses (reviewed in Refs. 8–11). In brief, Ox40L expression is largely restricted to APCs, whereas Ox40 is expressed primarily on activated or Ag-experienced Th cells and regulatory T cells (Tregs). Ox40 promotes Th cell survival (12–14), cytokine production (15), and memory cell formation and recall responses (16, 17). Ox40 may also have an important role in regulating the balance of effector T cell, Th cell, and Treg functions (18, 19).
Follicular helper T (Tfh) cells make up a specific helper cell subset that promotes humoral immune responses and has been implicated in lupus pathogenesis (20, 21). Tfh cells express Ox40, as well as other functional and identifying markers (e.g., CXCR5 and ICOS), and produce cytokines, such as IL-21. Disruption of Tfh cells or their key functional molecules ameliorate lupus symptoms in animal models (22). For example, in Roquin-deficient mice that develop lupus-like autoimmunity due to derepressed ICOS expression and Tfh cell activity (23), deletion of ICOS, SAP (a signaling molecule required for germinal centers [GCs]), or Bcl6 (a transcription factor required for Tfh cell development) disrupted Tfh cells and disease in these animals (24, 25). In addition to ICOS, Ox40 was recently found to be a direct target of the Roquin repressor protein family (26), highlighting the possibility that the Ox40 pathway might contribute to the disease phenotypes in those animals.
The specific function of Ox40 on Tfh cells is not entirely clear, in part due to conflicting results from various experimental models. Early work demonstrated a role for Ox40 in T-dependent anti-hapten IgG responses (27), but characterization of Ox40- and Ox40L-deficient animals found normal levels of Abs at rest, as well as in response to viral, hapten, or model Ag challenge (28–30). However, more recent studies targeting this pathway have identified defective GC and Ab responses in the context of Ox40 deficiency in response to viral infection (31, 32). Ox40 signaling leads to T cell accumulation in B cell follicles (33), increases the expression of CXCR5 (34–36), and enhances the production of cytokines, such as IFN-γ, IL-4, IL-6, and IL-10 (37, 38). Previous reports have also shown elevated Ox40 and Ox40L in SLE patient blood cells, serum, and kidney biopsies (reviewed in Ref. 11).
In the current study, we characterized the phenotype and function of Ox40-expressing cell populations during spontaneous disease in NZB × NZW (NZB/W) F1 lupus-prone mice. NZB/W F1 mice recapitulate several features of human SLE, including spontaneous onset, female-biased autoimmunity, and IC-mediated disease leading to nephritis and progressive renal failure (39). We assessed the effect of agonist anti-Ox40 mAbs and an antagonist Ox40:Fc fusion protein on disease features in NZB/W F1 mice. These data provide the first in vivo evidence, to our knowledge, that activation of Ox40 signaling promotes murine lupus pathogenesis and that blockade of this pathway may have disease-modifying activity.
Materials and Methods
Mice and disease monitoring
Female NZB/W F1 mice were bred and crossed at The Jackson Laboratory and then maintained at Genentech under specific pathogen–free conditions. All animal experimental protocols were approved by the Laboratory Animal Resources Committee at Genentech. For mouse lupus studies, renal disease severity was monitored by measuring proteinuria from fresh urine collected on a plastic sheet using Multistix 10 SG Reagent Strips and the CLINITEK Status Analyzer (both from Siemens). Animals were considered to have “severe” disease following multiple days of consecutive proteinuria readings ≥ 300 mg/dl.
In vivo treatments
An anti-Ox40 agonist mAb was generated by sequencing the rat anti-mouse Ox40 Ab (clone OX-86) and generating a chimeric fusion using the rat V region from OX-86 and a murine IgG1 Fc region. For anti-Ox40 agonist Ab treatments, cohorts were selected for age and proteinuria status at 12–13 wk (0 mg/dl proteinuria) or 21–27 wk (30–100 mg/dl proteinuria) and randomized by weight into treatment groups, and serum was collected prior to treatment for all experiments. Mice were treated s.c. with anti-Ox40 (IgG1) or an isotype-control IgG1 mAb (10 mg/kg three times per wk for 1–3 wk, depending on the experiment). During treatment and for the subsequent monitoring period, body weight and proteinuria were measured weekly, increasing to daily (if necessary) to monitor clinical symptoms. Animals were anesthetized under isoflurane and euthanized by terminal cardiac puncture or CO2 inhalation.
The Ox40:Fc fusion protein construct was generated by fusing the extracellular domain (residues 20–198) of Tnfrsf4/Ox40 with a linker GVTGSGG to the N terminus (…CKPCIC…) of the H chain of murine IgG1 and by expressing the protein in CHO cells. Purification was performed using protein A affinity medium (MabSelect SuRe), followed by size exclusion chromatography (S200; both from GE Healthcare) to >95% purity by SDS–gel electrophoresis, <5% aggregate (analytical size-exclusion chromatography multi-angle static light scattering; Wyatt Technology), and <0.5 EU/mg. Functional binding of Ox40:Fc was confirmed using a cell line expressing murine Ox40L.
For IFN-α–accelerated lupus studies, mouse IFN-α5 was cloned into a pShuttleX vector and transferred into an Ad5 viral vector (Ad5–IFN-α) by the Baylor College of Medicine Vector Development Lab (Houston, TX). Twelve- to thirteen-week-old NZB/W F1 animals were randomized into treatment groups, and serum was collected prior to treatment. A single i.v. injection of the Ad5–IFN-α vector (1.2 × 108 PFU) was given 3 d prior to treatment with Ox40:Fc or a control Ig s.c. at a dose of 20 mg/kg, twice a week, for the duration of the experiment. Animals were monitored weekly for body weight and proteinuria, increasing to daily, if necessary. For survival studies, animals exhibiting severe moribund symptoms were euthanized. One animal from the control and Ox40:Fc treatment groups were removed from the analysis because they did not develop any symptoms of renal disease.
Serum collection, Ab titers, and autoantibody titers
Serum was collected via the tail vein or via retro-orbital bleeding under isoflurane anesthesia (for live animal experiments) or via cardiac puncture under isoflurane anesthesia (for terminal experiments). Serum was collected following clotting and column separation using Microvette 200 Z-Gel serum separators (order number 20.1291; Sarstedt), per the manufacturer’s instructions. For experiments using anti-Ox40 agonist mAb treatment, serum was collected when animals exhibited proteinuria ≥ 300 mg/dl or at the end of the 3–4-wk experiment. Frozen serum was assayed using mouse anti-dsDNA IgG (catalog number [cat. no.] 5120), anti-ssDNA IgM (cat. no. 5330), anti-dsDNA IgM (cat. no. 5130), and anti–nuclear Ags Total Ig (cat. no. 5210) ELISA kits (Alpha Diagnostic International), per the manufacturer’s instructions. Total Ab titers were measured by Luminex using the MILLIPLEX map Mouse Immunoglobulin Isotyping Magnetic Bead Panel (cat. no. MGAMMAG-300K; Millipore), per the manufacturer’s instructions.
In vitro cultures
Splenic CD4 T cells were isolated from 21–27-wk-old (30–100 mg/dl proteinuria) NZB/W F1 mice using an EasySep Mouse CD4+ T Cell Isolation Kit (cat. no. 19852; STEMCELL Technologies), per the manufacturer’s instructions, and cultured in a sterile round-bottom 96-well culture dish (cat. no. 3799; Costar), at 2.5 × 105 cells per well, with 10 U/ml IL-2 (product number 354043; Corning). Cells were treated with anti-Ox40 (IgG1) or an isotype-control IgG1 (25 μg/ml) along with a 0, 1:10, or 1:1 ratio of T cell activator (anti-CD3/CD28) Dynabeads (cat. no. 11452D; Thermo Fisher). After 48 h, conditioned media were collected, and cytokine titers were measured by Luminex using a MILLIPLEX map Mouse TH17 Magnetic Bead Panel (cat. no. MTH17MAG-47K) or a MILLIPLEX map Mouse Immunoglobulin Cytokine/Chemokine Magnetic Bead Panel (cat. no. MCYTMAG-70K-PX32; both from Millipore), per the manufacturer’s instructions. Cellular RNA was collected for RNA sequencing (RNASeq).
Tissues were harvested into 10% neutral buffered formalin fixative at room temperature, followed by embedding in paraffin for sectioning. Formalin-fixed paraffin-embedded 4-μm histologic kidney sections were stained with periodic acid–Schiff (PAS) for visual evaluation of glomerulonephritis or with H&E for visual evaluation of interstitial inflammation by light microscopy.
Glomerulonephritis severity scoring used a four-point semiquantitative scale: 0, normal or mild global lesions in <50% of glomeruli; 1, global lesions in >50% of glomeruli, <20% of which are severe (defined as more than one segment with less than three patent capillaries); 2, global lesions in >50% of glomeruli, 20–80% of which are severe; and 3, >80% of glomeruli with severe global lesions. Interstitial inflammation was similarly scored based on three separate parameters: tubulointerstitial nephritis (0, normal; 1, mild or focal inflammation not expanding the tubulointerstitium; 2, occasional moderately sized inflammatory infiltrates expanding the tubulointerstitium; and 3, frequent large-sized inflammatory infiltrates expanding the tubulointerstitium), periarteritis (0, normal; 1, sparse or focal moderate-sized inflammatory infiltrates; 2, occasional moderate-sized inflammatory infiltrates, generally not circumferential; and 3, frequent large inflammatory infiltrates, often circumferential), and pyelitis (0, normal; 1, sparse or focal moderate-sized inflammatory infiltrates; 2, occasional moderate-sized inflammatory infiltrates, not expanding the pelvic submucosa; and 3, frequent large inflammatory infiltrates, expanding the pelvic submucosa). A glomerular cellularity score was obtained from PAS/hematoxylin-stained slides using semiautomated digital image analysis after whole-slide scanning on an NanoZoomer HT (Olympus) as follows: 20 random glomeruli per animal were manually traced, and nuclei within each were counted automatically in MATLAB (MathWorks) using color and size criteria for nuclear recognition.
Immunohistochemistry and immunofluorescence
To assess renal T cell infiltration, immunohistochemistry (IHC) using rabbit anti-CD3 (SP7; cat. no. RM-9107-S; Thermo Scientific) was performed on formalin-fixed paraffin-embedded 3-μm kidney sections pretreated with Target Retrieval Solution (DAKO) for Ag retrieval. Naive rabbit clone DA1E (product number 3900S; Cell Signaling Technology) and naive rat IgG2b (A95-1; cat. no. 553986; BD Pharmingen) were used in parallel as nonspecific isotype controls. Detection used goat anti-rabbit biotinylated IgG or rabbit anti-rat biotinylated IgG (both from Vector Laboratories), followed by VECTASTAIN Elite ABC peroxidase (Vector Laboratories) and DAB visualization (Thermo Fisher Scientific). Images were captured on a Leica DM6000 B microscope using a Leica DFC500 camera and Leica Application Suite software (Leica Microsystems, Wetzlar, Germany). To compare renal CD3 T cell frequency in anti-Ox40–treated and control Ig–treated mice, one slide from each treatment group was randomly paired with the other and visually compared in a blinded fashion: the slide with relatively more frequent T cells was assigned a score of 1, and the slide with relatively less frequent T cells was assigned a score of −1 (a score of 0 was assigned when both slides had comparable CD3 T cell frequency). Scoring of tubulointerstitial and glomerular T cell frequencies was performed separately.
Glomerular IC deposits were visualized on 5-μm acetone-fixed OCT-embedded kidney sections by direct immunofluorescence staining using Alexa Fluor 488–conjugated donkey anti-mouse IgG (cat. no. A-21202; Invitrogen), goat anti-mouse IgM (cat. no. A-21042; Invitrogen), or fluorescein-conjugated goat anti-complement C3 (cat. no. 55510; MP Biomedicals). Naive donkey or goat IgG (Jackson ImmunoResearch) was used in parallel as nonspecific isotype controls and were negative. Stained slides were scanned on a NanoZoomer XR (Hamamatsu), and morphologically intact renal cortex regions on each slide were manually selected and analyzed in MATLAB (MathWorks) to determine average pixel intensity in the selected area. A threshold filter was used to eliminate nonspecific background and enrich for glomerular signal, and a normalization factor was applied to combine two independent replicate cohorts.
GCs were visualized on 5-μm paraformaldehyde-fixed OCT-embedded spleen sections by indirect immunofluorescence staining after Ag retrieval with DAKO Target Retrieval Solution (code S1700), using the B cell and T cell activation and GC marker, GL7 (cat. no. 14-5902; eBioscience) and detected with Biotin Mouse Anti-Rat IgM (clone G53-238; cat. no. 550330; BD Pharmingen), followed by Alexa Fluor 546–conjugated streptavidin (cat. no. S11225; Molecular Probes, Thermo Fisher Scientific). GCs were visualized on an epifluorescence-equipped Leica DM6000 B microscope and imaged with a Leica DFC500 camera and Leica Application Suite software. GCs were manually counted on each spleen section, as clusters of at least five adjacent GL7+ cells, in a blinded fashion.
Flow cytometry and cell sorting
Postmortem intracardial perfusion was performed with 20 ml of PBS. Spleen and kidney were harvested, and tissues were harvested into RPMI 1640 media and weighed, followed by mechanical disruption with scissors into 1-mm pieces. Tissues were shaken at 37°C in digestion buffer (67 μg/ml Liberase DL , 20 μg/ml DNase I , 5% FBS [SKU number 2442; all from Sigma-Aldrich] in RPMI 1640) for 15 min. Samples were passed through a 40-μm filter (cat. no. 352340; Corning) into RPMI 1640 supplemented with 5% FBS and 2 mM EDTA. Cells were treated with ACK lysing buffer for 5 min on ice and blocked using FcR Blocking Reagent (order number 130-092-575; Miltenyi Biotec) before surface staining. Cells were independently FcR blocked before intracellular stains following fixation and permeabilization, using a Foxp3/Transcription Factor Staining Buffer Set (cat. no. 00-5523-00; eBioscience), per the manufacturer’s instructions. mAbs used for staining cells in these experiments included CD45 (clone 30F-11; cat. no. 553080; BD), CD4 (RM4-5; cat. no. 560782; BD), CD19 (6D5; cat. no. 115534; BioLegend), CD3 (17A2, cat. no. 100214; BioLegend or cat. no. 56-032; eBioscience), CD44 (IM7; cat. no. 17-0441; eBioscience or cat. no. 560567; BD), CD62L (MEL-14; cat. no. 56-0621; eBioscience or cat. no. 104432; BioLegend), Ox40 (Ox-86; cat. no. 12-1341; eBioscience), Rat IgG1 kappa Isotype Control (eBRG1; cat. no. 12-4301; eBioscience), Ox40 (Ox-86; cat. no. 119410; BioLegend), Foxp3 (FJK-16s; cat. no. 17-5773; eBioscience), CD25 (PC61.5; cat. no. 12-0251-82; eBioscience), CXCR5 (2G8; cat. no. 551960; BD), Rat IgG2a Isotype Control (R35-95; cat. no. 553928; BD), Brilliant Violet 421 Streptavidin (cat. no. 405226; BioLegend), ICOS (15F9; cat. no. 46-9940; eBioscience), Hamster IgG Isotype Control (cat. no. 46-4914; eBioscience), Ki-67 (SolA15; cat. no. 14-5698-80; eBioscience), Rat IgG2a kappa Isotype Control (eBR2a, cat. no. 14-4321; eBioscience), CD21 (7E9; cat. no. 123412; BioLegend), CD23 (B3B4; cat. no. 563200; BD), IgM (RMM-1; cat. no. 406505; BioLegend), IgD (11.26c.2a; cat. no. 405716; BioLegend), CD138 (281-2; cat. no. 142507; BioLegend), and GL7 (cat. no. 144607; BioLegend). Dead cells were discriminated using the LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit (cat. no. L10119; Thermo Fisher). Flow cytometry was performed using an LSR II (BD), and data were analyzed using FlowJo (TreeStar) software. For sorting and RNASeq experiments, cells were prepared as described above, with the exception of the prestain enrichment steps. Kidney tissues were cleaned and enriched for lymphocytes using a 40/60% Percoll gradient (SKU number P4937; Sigma-Aldrich) procedure, per the manufacturer’s instructions. Spleen tissue was enriched for CD4 T cells using a Mouse CD4+ T Cell Isolation Kit (cat. no. 19852; STEMCELL Technologies), per the manufacturer’s instructions. Following staining, cell sorting was performed on a FACSAria II (BD).
RNASeq and pathway analysis
For in vivo treatment experiments, live CD45+CD19−CD3+CD4+ T cells were sorted to high purity from the spleen (n = 4) and kidney (n = 5) after 1 wk (day 8) of anti-Ox40 agonist mAb treatment, followed by lysing with RLT buffer supplemented with 2-ME (Sigma-Aldrich). RNA was extracted using an RNeasy Mini Kit (cat. no. 74104) or an RNeasy Micro Kit (cat. no. 74004; both from QIAGEN), depending on input. For kidney samples, an RNeasy MinElute Cleanup Kit (cat. no. 74204; QIAGEN) was used. For all RNASeq experiments, a Nanodrop 8000 (Thermo Scientific) was used to quantify RNA, and integrity was measured using the Bioanalyzer RNA 6000 Pico Kit (Agilent). Libraries were prepared using the TruSeq RNA Library Prep Kit v2 (Illumina) with 100–500 ng of input and amplified using 10 cycles of PCR. Libraries were multiplexed and sequenced on a HiSeq 2500 System (Illumina), resulting in 15–26 million single-end 50 bp reads per library. Alignment, feature counting, normalization, and differential expression analysis were performed similar to as described previously (40), with few differences, which are listed below. In brief, HTSeqGenie (41) was used to perform filtering, alignment to GRCm38, and feature counting. Normalized reads per kilobase gene model per million total reads (nRPKM) values were computed as a measure of gene expression. Pairwise differential expression analysis was performed using voom and limma (42). For organ-specific differential gene-expression analysis, significant genes were filtered and identified as p < 0.05, nRPKM > 2, and fold change > 2 or < 0.5. For the four-way comparison, significant genes were filtered and identified by the same threshold settings but were included if they were significant in at least one organ. Pathway analysis was performed with Ingenuity Pathway Analysis (IPA) software (QIAGEN) using the Molecular and Cellular Functions module. Heat map euclidean clustering of genes was performed by plotting log 2–transformed fold change values for each replicate sample and each gene (log 2 floor set at −3 for all heat maps). Colored boxes indicate the degree of fold change (unique to each graph). Venn diagrams were generated at bioinformatics.psb.ugent.be/webtools/Venn/. IFN-α–responsive genes (>2 fold change) from splenic CD4 T cells were described previously (43). For the cytokine analysis, genes were included in the heatmap when they reached the following criteria: p < 0.05, nRPKM >2, and fold change >1.5 in at least one organ, for all genes with the cytokine classification by IPA. RNASeq data are available from the National Center for Biotechnology Information Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE99645 and GSE99646.
Graphing and statistics
Data visualization and statistical analyses were performed with GraphPad Prism (GraphPad). For all in vivo cellular and Luminex experiments, p values were calculated using the Mann–Whitney U test. For in vitro culture cytokine RNA experiments (n = 3), a two-tailed unpaired t test was used. For renal disease survival (Kaplan–Meier) graphs, p values were calculated with the log-rank (Mantel-Cox) test, and hazard ratios were calculated using the log-rank test. For ordinal histology scoring experiments, p values were calculated using the χ2 test. The p values were considered significant at *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.
Ox40-expressing CD4+ T cells in the spleen and kidney of NZB/W F1 mice
Using flow cytometry of single-cell suspensions of spleen and kidney from NZB/W F1 mice over a range of disease severities (21–51 wk old), we found that Ox40 was highly expressed on CD4 T cells in both tissues (Fig. 1A) and was not detectable on other lymphoid or myeloid cells (data not shown). Ox40 was upregulated on CD4 T cells prior to the onset of proteinuria, and the frequency and number of these cells correlated with disease severity in both tissues (Fig. 1B).
Ox40+ T cells were enriched for the proliferation marker Ki67, and most cells expressed the CD44+CD62L− activated/memory cell phenotype (Supplemental Fig. 1A, 1B). Ox40+ CD4 T cells were also enriched for Foxp3, a marker for Tregs (Fig. 1C), and CXCR5 and ICOS, markers for Tfh cells (Fig. 1D) (20, 21, 44). Foxp3− conventional helper T (Tconv) cells were also identified in the spleen and kidney (Fig. 1C, Supplemental Fig. 1C).
By IHC, CD3+ kidney T cells were infrequent in NZB/W F1 mice prior to the onset of proteinuria (Fig. 1E), but they increased in abundance in diseased mice with ≥300 mg/dl proteinuria. T cells were found within or near renal glomeruli (Fig. 1F), in periarterial infiltrates, and throughout the tubulointerstitium (Fig. 1G).
Ox40 agonist mAbs exacerbate renal disease of aged NZB/W F1 mice
We next generated a novel anti-Ox40 agonist mouse IgG1 mAb (see 2Materials and Methods) and used it to determine the effect of Ox40 receptor stimulation in vivo in NZB/W F1 mice. Treatment of young mice (13 wk old, prior to the onset of proteinuria) with a 3-wk course (three injections per week) of anti-Ox40 agonist mAbs had no effect on the development of proteinuria compared with control Ig treatment (Fig. 2A). In contrast, renal disease was strongly exacerbated (reaching ≥300 mg/dl proteinuria) in three independent older cohorts with mild-to-moderate underlying disease (21–27 wk old, 30–100 mg/dl proteinuria, n = 47 animals in total) treated with the same regimen (Fig. 2B). A total of 89% of older mice treated with anti-Ox40 mAbs developed severe renal disease (median time to ≥300 mg/dl proteinuria = 14 d) compared with 51% in the control group (median time to ≥300 mg/dl proteinuria = 56 d, p < 0.0001) over the 84-d observation period.
Examination of fixed kidney sections harvested following anti-Ox40 mAb treatment showed increased severity of proliferative glomerulonephritis (Fig. 2C) and a corresponding increase in glomerular cellularity (Supplemental Fig. 1D) compared with control Ig treatment. Worse periarteritis and tubulointerstitial nephritis were also noted. From these data, we conclude that in vivo stimulation of Ox40 rapidly and efficiently exacerbated the onset of severe proteinuria and kidney disease in older NZB/W F1 mice. However, the same treatment had no effect on younger animals prior to the onset of proteinuria, despite those animals harboring Ox40-expressing cells (Fig. 1B).
Anti-Ox40 mAbs induce kidney IgM deposition and T cell activation
We next used immunofluorescent staining of frozen kidney sections to determine glomerular deposition of Igs and C3 following anti-Ox40 agonist mAb treatment. Staining was quantified using whole-slide scanning and digital image analysis (see 2Materials and Methods).
As expected, IgG deposition and IgM deposition in kidney renal cortex were increased during spontaneous progressive disease in control animals (Supplemental Fig. 1E). Of interest, we observed a significant increase in IgM deposits following anti-Ox40 agonist mAb treatment compared with controls, whereas IgG deposition decreased, and C3 remained unchanged (Fig. 2D).
Following 1.5–3 wk of anti-Ox40 mAb treatment, there were no significant changes in kidney mass or infiltrating immune cell numbers, including total CD4 T cells (Supplemental Fig. 1F–J). A mild decrease in the percentage and number of CD4+Foxp3− Tconv cells (Fig. 3A) and a small increase in the frequency of CD4+Foxp3+ Tregs (Supplemental Fig. 1K) were identified.
Despite the drop in cell numbers, Tconv cells were enriched for an activated/memory (CD44+CD62L−) phenotype following treatment (Fig. 3B). Using CD3 IHC staining on fixed kidney sections, we identified a significant increase in the relative frequency of glomerular and tubulointerstitial T cells following treatment (Fig. 3C, 3D).
Elevated serum IgM, splenic Tfh cells, and plasmablasts in Ox40 agonist mAb–treated mice
We next measured serum Ab levels and found a 45% elevation in total serum IgM in the anti-Ox40 mAb–treated group (Fig. 4A), consistent with the Ab deposit data in glomeruli (Fig. 2D). In contrast, serum IgG isotypes were decreased following treatment (Fig. 4B). Autoantibodies to dsDNA, ssDNA, and other nuclear Ags were not significantly affected (Supplemental Fig. 2A–E). Taken together, the observed increase in IgM and decrease in IgG deposition in glomeruli following anti-Ox40 mAb treatment were mirrored in the serum.
Examination of spleens after 1.5–3 wk of treatment showed no large changes in T cell numbers, but there was a slight increase in CD4 T cell frequency that was due, in part, to an expansion of Foxp3+ Tregs (Supplemental Fig. 3A–F). Similar to the kidney T cell response, Tconv cells were enriched for an activated/memory phenotype (CD44+CD62L−) following treatment (Fig. 4C). Notably, we also observed an increase in CXCR5+ICOS+ Tfh cells (Fig. 4D).
B cell subsets were largely unaffected (Supplemental Fig. 3G–I). However, CD138+ plasmablasts roughly doubled in number following anti-Ox40 agonist mAb treatment, including cells expressing IgM (Fig. 4E, 4F). Examination of spleens after 3 wk of treatment showed no increase in the number of GCs or GC phenotype cells, as measured by staining frozen spleen sections or live cell preparations for GL7 (Supplemental Fig. 2F, 2G).
T cell gene expression following Ox40 agonist mAb treatment in vivo
To better understand the molecular response of T cells following Ox40 receptor stimulation in vivo and to gain insight into potential mechanisms driving Ox40-exacerbated renal disease, we performed RNASeq on sorted CD4 T cells from the spleen and kidney following 1 wk of anti-Ox40 agonist mAb treatment. Differentially expressed genes were identified by comparing nRPKM between T cells purified from Ox40 agonist mAb–treated and control Ig–treated mice and applying a strict set of statistical criteria: fold change > 2 (or <0.5), p value < 0.05, and mean nRPKM > 2 (see 2Materials and Methods).
Following anti-Ox40 mAb treatment, 186 genes were significantly induced and 6 were repressed in splenic CD4 T cells (Supplemental Fig. 4A). In the kidney, 188 genes were significantly upregulated, and 31 were repressed (Supplemental Fig. 4B). Ninety-three upregulated genes were shared between the spleen and kidney (Fig. 5A, Supplemental Fig. 4C); however, a larger fraction of genes were not shared (60% of splenic T cell transcripts and 51% of the kidney).
Pathway analysis of the shared genes indicated that the most enriched cellular and molecular pathways were cell cycle, cellular assembly, and DNA replication. Mki67, a marker for T cell proliferation, was enriched in splenic and kidney CD4 T cells (Supplemental Fig. 4A, 4B). Transcriptional evidence for cycling cells was particularly prominent (Proliferation of Cells, 52 genes, p = 2.58 × 10−10) (Supplemental Fig. 4D). Consistent with the accumulation of Tregs observed by flow cytometry, foxp3 and other Treg-related functional genes were also upregulated (Supplemental Fig. 4E).
Pathway analysis of the splenic T cell RNASeq data indicated that Interferon Signaling (p = 2.49 × 10−7) genes were strongly upregulated in the Ox40 mAb–stimulated T cells. We confirmed the enrichment for IFN-regulated genes by comparing the Ox40-induced genes with a CD4 T cell IFN-α response gene dataset (43) and found that 29% of induced genes in the spleen (67/231) were IFN-α regulated compared with only 1% (2/190) in the kidney (Fig. 5A). This finding is particularly interesting given the strong evidence for activation of type I IFN–induced genes in human SLE (45).
Pathway analysis also identified enrichment for Quantity of Interleukins (11 genes, p = 3.03 × 10−8), Movement of T Cells (5 genes, p = 6.44 × 10−7), and Recruitment of Leukocytes (13 genes, p = 1.22 × 10−5). These enriched functional pathways included inflammatory cytokines and chemokines, a number of which have been implicated in SLE pathogenesis (46). We broadened our search for genes of the cytokine classification (>1.5-fold change) and found that Il21, Csf1, and Il10 were significantly induced in spleen and kidney T cells (Fig. 5B). Cxcl10, Ccl5, Ebi3, Nampt, Tnfsf8, and Lif transcripts were upregulated in spleen T cells, and Csf2, Cxcl9, Ccl1, Lta, and Tnfsf11 were elevated in kidney T cells.
TCR-dependent cytokine production following Ox40 agonist mAb treatment
To further explore the relationship between Ox40-stimulated cytokine production and TCR signaling, we cultured purified CD4 T cells in vitro with anti-Ox40 mAbs in the presence or absence of various levels of TCR stimulation (see 2Materials and Methods). Cells were harvested for RNA, and culture supernatants were collected for cytokine protein quantification.
Of the 14 cytokines identified from the in vivo RNA experiments, four mRNAs (Il21, Lif, Csf2, and Lta) showed strong responses in vitro to anti-Ox40 mAb treatment (Fig. 5C, Supplemental Fig. 4F). Il21, Lif, and Csf2 were expressed at background levels (<2 nRPKM) in the absence of TCR stimulation, whereas Lta was moderately expressed at baseline. TCR stimulation alone induced expression of Il21, Lif, and Lta, and higher expression levels were observed for all four cytokine transcripts following combined TCR and anti-Ox40 mAb stimulation.
We next measured supernatant protein for the most highly induced genes and found that TCR stimulation was required for detectable IL-21, LIF, GM-CSF (Csf2), and LTA production (Fig. 5D). LIF and GM-CSF were induced with high TCR alone and were further increased by concomitant anti-Ox40 mAb treatment. IL-21 and LTA production required a combination of high TCR and anti-Ox40 mAb treatment. From these data, we conclude that Ox40-induced inflammatory cytokine production in this in vitro system was dependent on TCR stimulation.
Ox40L blockade delays worsening of renal disease and death
Finally, we tested whether blockade of Ox40/Ox40L interactions would have a therapeutic effect in the IFN-α–accelerated NZB/W F1 lupus model. We used this model, rather than the spontaneous NZB/W F1 model, because it may more accurately model the disease of human SLE patients, who often have prominent type I IFN signatures in blood and tissues (45). The IFN-accelerated model has an added benefit of synchronizing and shortening the normally variable disease course found in NZB/W F1 mice (47). However, the disease found in these animals is more aggressive and resistant to therapy (48). We used a soluble Ox40:Fc fusion protein to block Ox40L (see 2Materials and Methods) and treated 12-wk-old proteinuria-free mice twice a week for the duration of the 15-wk experiment. The initial dose of Ox40:Fc was given 3 d after injection with an IFN-α–expressing adenoviral vector. Across two independent cohorts, Ox40:Fc treatment resulted in a delay in the onset of severe renal disease (≥300 mg/dl proteinuria, median onset = 40 d) compared with control treatment (median onset = 27 d, hazard ratio = 0.5, p < 0.006) (Fig. 6A).
Ox40:Fc extended the lifespan of treated animals by ∼2 wk (Fig. 6B). The median survival was extended from day 55 in the control group to day 70 in the Ox40:Fc group (hazard ratio = 0.48, p < 0.008). These data provide the first in vivo evidence, to our knowledge, that the Ox40/Ox40L pathway has a role in mediating renal disease in a lupus model.
Despite the strong genetic evidence that variants in the Ox40L locus contribute to human SLE risk, there are still significant gaps in our understanding of this pathway’s role in SLE pathogenesis. In the current study, we demonstrate that Ox40-expressing T cells (including follicular, regulatory, and conventional helper subsets) were present in spleens and kidneys of NZB/W F1 lupus-prone mice and increased during progressive disease; in vivo agonism of the Ox40/Ox40L pathway, using an anti-Ox40 mAb, potently exacerbated renal disease; Ox40 stimulation induced kidney-infiltrating and splenic CD4 T cell activation, splenic Tfh cell and plasmablast accumulation, serum IgM elevation, and kidney IgM deposition; Ox40 signaling induced an inflammatory and proliferative transcriptional program in CD4 T cells, including a prominent upregulation of type I IFN–induced genes in the spleen; Ox40 stimulation also led to elevated production of TCR-dependent inflammatory cytokines, including IL-21; and in vivo blockade of Ox40/Ox40L signaling using an Fc-fusion protein showed significant benefit on disease progression and death in an IFN-α–accelerated NZB/W F1 model.
Based on these findings, we conclude that the Ox40/Ox40L pathway contributes to lupus pathogenesis by promoting several specific Th cell functions, such as controlling Tfh cell accumulation and cytokine production in lymphoid tissues and increasing the number of presumably pathogenic Ab-producing plasmablasts. Ox40 stimulation also led to significant worsening of renal disease that was mediated, at least in part, by deposition of IgM in glomeruli and production of inflammatory cytokines and chemokines by T cells residing in the kidney. These data provide new insights into the mechanisms by which Ox40/Ox40L interactions can contribute to systemic and tissue autoimmunity, as well as provide experimental evidence that blockade of this pathway may provide therapeutic benefit in lupus.
Recently, Jacquemin et al. (49) demonstrated that Ox40 stimulation of human T cells in vitro promoted a Tfh cell-like transcriptional program and phenotype. They also described an Ox40L-expressing myeloid APC population in SLE patient blood and kidney tissue, with the frequency of Ox40L-expressing cells correlating with numbers of blood Tfh cells and disease activity. The increased numbers of Tfh cells and the cytokine production that we identified following in vivo anti-Ox40 mAb treatment are consistent with the hypothesis that Ox40 signaling promotes Tfh cell accumulation and function. Tfh cells provide several forms of B cell help (20, 21), but the precise disease-relevant contribution from Ox40 has not been well described for lupus.
We propose that the accumulation of plasmablasts, including IgM+ cells, increased serum IgM levels, decreased IgG levels, and the absence of a change in GCs or GC cell numbers, suggests that Ox40 may stimulate Th cells, resulting in extrafollicular or pre-GC reactions. A similar role has been ascribed for the costimulatory molecule ICOS (50). Ox40 stimulation may promote B cells that might otherwise enter a GC to differentiate into plasmablasts. A recent publication describing Ox40/Ox40L interactions located inside and outside of GCs during infection supports this possibility (32); however, in that study, Ox40 was required for GC formation and anti-viral Ab responses, highlighting the disease-specific nature of this pathway. On the whole, our data support the hypothesis that the Ox40/Ox40L pathway promotes lupus pathogenesis by regulating plasmablast accumulation and serum IgM levels, rather than promoting GCs or class switching.
Ox40-mediated upregulation of IL-21 may be important for the amplified IgM response, given that IL-21 can enhance the generation of IgM-expressing memory B cells, IgM-expressing plasma cells, and production of soluble IgM (51–55). Disruption of the IL-21 pathway has provided benefit in several preclinical lupus models. These studies have included full genetic deletion or B cell–specific deletion of IL-21R in the BXSB-Yaa model (56, 57), IL-21R–Fc treatment in the MRL-FasLpr model (58), anti–IL-21R mAb treatment in the NZB/W F1 model (59), and anti–IL-21 mAb treatment in the B6.Sle.Yaa model (60).
Human genetics evidence further connects the Ox40/Ox40L pathway with Ab responses in the setting of autoimmunity. One of the Ox40L locus polymorphisms associated with SLE risk (rs2205960-T) showed significant correlation with serum autoantibody titers in lupus cases from diverse ancestry (61). Ox40L locus polymorphisms have been associated with systemic sclerosis (62, 63) and, in particular, autoantibody-associated cases (64), as well as Sjögren’s syndrome (65), a disease characterized by autoantibodies. Although Ox40L has not been reported to confer risk for rheumatoid arthritis, one study identified an association between serum Ox40L protein levels and anti-citrullinated protein Abs, as well as IgM rheumatoid factor, in rheumatoid arthritis patients (66).
Stimulation of Ox40 using agonist mAbs was surprisingly potent and resulted in exacerbated renal disease in NZB/W F1 mice with mild-to-moderate pre-existing proteinuria, but not in younger proteinuria-free animals, despite their harboring of Ox40-expressing T cells. This finding suggests that the number or type of Ox40-expressing cells in the spleen or kidney, or the per-cell amount of Ox40, may be important in this system for disease acceleration. Notably, our findings share similarities with other studies using Ox40 agonist treatments. Ox40 agonist treatment exacerbated models of experimental uveitis (67, 68) and promoted inflammatory lung pathology in B6 mice (69). However, in other models, the disease-promoting effects of Ox40 agonism depended upon age and disease severity of the animals (e.g., in experimental autoimmune encephalitis and type 1 diabetes, in which treatment promoted disease during later or effector phases but protected from disease when provided during early or priming phases) (70–72). Evidence from those studies suggests that the disease-promoting versus disease-preventing effects of Ox40 agonism might be explained by how each treatment impacted Tregs in these models. In agreement, coprovision of Treg-boosting treatments, such as IL-2 or JAG1, could reverse the pathogenic effects of Ox40 agonist treatment (69, 73).
We observed an accumulation of Foxp3+ Tregs following anti-Ox40 mAb treatment, despite the progression of autoimmunity. These findings are in line with previous reports of Tregs in aged NZB/W F1 mice, which retain functional regulatory capacity but are unable to control disease (74, 75). Treg-enriched RNA transcripts (Foxp3, Il2ra, Itgae, Lrrc32, Pdcd1, Tigit, and Tgfb) were also upregulated by anti-Ox40 mAb treatment, suggesting functional potential. Although we do not have definitive data explaining why these Tregs do not suppress the pathogenic response, there are several possibilities: the regulatory and effector cells are not in the same anatomical location, the expanded Tregs are not relevant for the specific response, or most likely, the activating effect of the Ox40 agonist mAbs on effector T cells overwhelms the suppressive ability of Tregs.
The transcriptional response in purified T cells following in vivo Ox40 agonism included an IFN signature in the spleens of NZB/W F1 mice; nearly one third of the induced splenic transcripts overlapped with genes induced by IFN-α in CD4 T cells (43). The nature of the splenic IFN response is unclear; however, we suspect that it is indirect (i.e., requiring other cells, likely myeloid cells), because anti-Ox40 mAb–stimulated monocellular in vitro cultures did not result in the same IFN signature. An important future direction for this work will be to understand the chain of events by which Ox40 stimulation leads to a type I IFN response, because the IFN signature is prominent in SLE patients (45).
Ox40-stimulating mAbs are currently in early-stage clinical trials for a variety of human cancers (76, 77). Our results suggest the possibility that stimulation of Ox40 in humans may result in untoward autoimmune or inflammatory responses, particularly for individuals with clinical or subclinical autoimmunity. These data also suggest that targeting the Ox40/Ox40L pathway may result in clinical benefit for SLE and/or LN patients. Although our treatment regimen only resulted in a delay (as opposed to complete prevention) in renal disease and death, the model we used has been demonstrated to be particularly aggressive and resistant to therapeutic intervention (47, 48). It remains possible that Ox40- or Ox40L-blocking therapies would show greater efficacy in a spontaneous disease model. Additional studies are needed to fully understand the pathologic consequences of modulating the Ox40/Ox40L pathway, including determining when and how it can be targeted safely and effectively in cancer and autoimmunity.
We thank Michael Townsend and Andrey Shaw for manuscript feedback; Phil Hass and Ryan Abraham for production of the Ox40:Fc fusion protein; Jovencio Borneo, Terence Ho, Jonathan Paw, and C.K. Poon for cell sorting; Andres Paler Martinez for Luminex; Xiumin Wu, Juan Zhang, and John Liu for assistance with mouse experiments; Debra Dunlap, Felix Chu, and Sreedevi Chalasani for IHC and immunofluorescence; Charles Havnar, Sarajane Saturnio, and Carmina Espiritu for histology; Zora Modrusan and Jeremy Stinson for assistance with RNASeq; and Jennifer Tom and Bill Forrest for statistical advice. We also thank Cara Skon, Nandhini Ramamoorthi, and the rest of the Human Genetics and Diagnostic Discovery departments for scientific discussion.
This work was supported by Genentech, Inc.
The RNA sequencing data presented in this article have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession numbers GSE99645 and GSE99646.
The online version of this article contains supplemental material.
Abbreviations used in this article:
- cat. no.
Ingenuity Pathway Analysis
normalized read per kilobase gene model per million total reads
NZB × NZW
systemic lupus erythematosus
Foxp3− conventional helper T
follicular helper T
regulatory T cell.
All of the authors are full-time employees of Genentech, Inc.