Abstract
Autoimmune therapies aim to inhibit autoreactivity while preserving normal immune function. The transcriptional coregulator OCA-B, also known as Bob.1/OBF-1 (gene symbol Pou2af1) is induced in stimulated naïve CD4+ T cells, where it docks with transcription factor Oct1 to regulate genes such as Il2 and Ifng. OCA-B promotes expression of these targets in cases of repeated antigen exposure, a necessary feature of autoimmunity. Polymorphisms in Ocab itself and binding sites for Oct1/OCA-B complexes are associated with multiple forms of autoimmunity including autoimmune (type-1) diabetes. We hypothesized that T cell-specific OCA-B deletion would protect mice from type-1 diabetes, and that pharmacologic OCA-B inhibition would provide similar protection. We developed an Ocab conditional allele and backcrossed it onto a diabetes-prone NOD/ShiLtJ strain background. T cell-specific OCA-B loss protected mice from spontaneous T1D. Protection was associated with reduced pancreatic T cell and macrophage infiltration and reduced proinflammatory cytokine expression. We profiled prediabetic pancreatic lymph nodes and islets by single-cell RNA sequencing and T cell receptor clonotype analysis. Although lymph nodes showed minimal differences, in the islets CD8+ T cell specificities associated with diabetes pathogenesis failed to emerge in OCA-B deficient activated/memory cells. In contrast, CD4+ clones associated with diabetes were present, but only in anergic cells. The protective effect of OCA-B loss was diminished, or even eliminated, using monoclonal models with high affinity to artificial or neoantigens. Rationally-designed membrane-penetrating OCA-B peptide inhibitors normalized glucose levels, and reduced T cell infiltration and proinflammatory cytokine expression in newly-diabetic NOD mice. Together, the results indicate that OCA-B is a potent autoimmune regulator and a promising target for pharmacologic inhibition.
Introduction
Type 1 Diabetes (T1D) is an autoimmune disease in which the host immune system is directed towards antigens associated with pancreatic beta cells1, 2. Pathologically, T1D is characterized by insulitis, beta cell destruction and inability to produce insulin. The main treatment for T1D, life-long insulin therapy, treats symptoms but not cause. The development of new T1D treatments is limited by an incomplete understanding of disease mechanisms3. Beta cell regeneration is a promising line of therapy, but still requires methods to specifically block T1D autoimmunity. An ideal method of therapy would be to capture patients early in the disease course and block autoimmunity while keeping normal immune function intact. Such a therapy would spare remaining beta cell function.
In CD4+ T cells, the transcription factor Oct1 and its cofactor OCA-B regulate a set of ∼150 target genes, including Il2, Ifng and Csf2 (Gmcsf)4. Upon T cell activation, many of these targets are activated by pathways that converge on transcription factors such as NF-AT, AP-1 and NF-κB. Factors like NF-AT can be thought of as the primary on/off switches for these genes, and drugs that block NF-AT activation effectively block their expression. Such drugs have utility in many contexts, but also have drawbacks including global dampening of immune function and side effects due to expression in other tissues. In contrast, Oct1 and OCA-B insulate silent but previously activated target genes against stable repression. Loss of either protein does not affect CD4+ T cell responses to stimulation in vitro or primary infection in mice4, 5, but causes large defects in target gene expression upon secondary stimulation in vitro and defective CD4+ memory T cell formation and recall responses in vivo4, 6. In addition, OCA-B expression is largely confined to lymphocytes. OCA-B is silent in thymocytes and naïve CD4+ T cells, but is induced and stably expressed after sustained antigen stimulation4. Once expressed, OCA-B recruits the histone lysine demethylase Jmjd1a/Jhdm2a/Kdm3a to Oct1 at target loci, where it locally removes inhibitory H3K9me2 chromatin marks that would otherwise promote repressive chromatin environments4, 6. OCA-B also potentiates polarization of activated helper T cells towards the Th17 phenotype, among the most pro-inflammatory helper T cell subsets7.
Persistent antigen exposure is a common property of autoimmune responses. Among profiled CD4+ T cell subsets, OCA-B levels are elevated in both central memory cells and pancreas-infiltrating, islet-reactive cells (http://www.immgen.org/databrowser/index.html). Both Oct1 and OCA-B are highly conserved to humans, and genome-wide association studies identify Oct1/OCA-B binding site polymorphisms that confer predisposition to a variety of human autoimmune diseases including T1D, multiple sclerosis, lupus, inflammatory bowel disease and rheumatoid arthritis8, 9, 10, 11, 12, 13, 14. Other studies implicate polymorphisms at the Ocab (Pou2af1) locus itself in different forms autoimmunity15, 16. We therefore hypothesized that targeting OCA-B would inhibit diabetogenic T cell responses.
Here, we generate a T cell conditional mouse model of OCA-B deficiency. We show that OCA-B ablation protects animals from spontaneous, polyclonal T1D. Protection was associated with reduced T cell infiltration and proinflammatory cytokine expression. Conditional OCA-B deletion maintained normal pancreatic lymph node cell populations, numbers, gene expression and TCR clonality. In contrast, significant alterations in specific TCR clones were present in the pancreatic islets of OCA-B T cell conditional mice. Protection from OCA-B loss varied among different TCR transgenic and monoclonal antigen models, and was weakest using synthetic antigens and monoclonal transplant models. Protection was stronger, but still incomplete, using a spontaneous CD8 TCR transgenic model. Using a rational-design approach, we developed membrane-penetrating OCA-B peptide inhibitors that are efficacious in vitro and in mice. Collectively, the results indicate that OCA-B is a potent and pharmacologically accessible autoimmune regulator.
Results
Generation of an Ocab conditional allele and NOD backcrossing
We generated a conditional Ocab (Pou2af1) mouse allele using embryonic stem cells provided by the knockout mouse project (KOMP) repository. Deletion of the LoxP sites is predicted to create a null allele by disrupting the first three exons (Fig. 1a). Breeding chimeric mice resulted in germline transmission and a founder mouse (Fig. 1B, lane 2) that was crossed to a FLP deleter mouse strain (FLPRosa26) to remove the reporter cassette and create a floxed conditional allele (Fig. 1a and b, lane 3). Intercrossing these mice generated homozygous fl/fl mice (Fig. 1b, lane 4). The mice were crossed to germline Cre deleter strain (CreRosa26) to generate null (Δ) alleles (lanes 5 and 6). As expected, homozygous fl/fl mice produce normal amounts of both p34 and p35 OCA-B protein isoforms (Supplementary Fig. 1a, lane 6), while no protein of either isoform is present in germline-deleted Δ/Δ spleens (lanes 7-8). These results indicate that the fl allele is OCA-B sufficient and the Cre-deleted allele is a null. Crossing the fl allele onto a CreCD4 driver, which deletes in both CD4 and CD8 compartments, resulted in efficient deletion in splenic T cells (Fig. 1b). These findings indicate that this allele represents a robust system in which to study OCA-B function in T cells.
The Ocab (Pou2af1) conditional allele was generated on a C57BL/6 background. To test the prediction that OCA-B expression in T cells promotes T1D emergence, we conducted speed congenic backcrosses to the NOD strain background. This method allows spontaneous diabetes to be produced rapidly by screening and selecting for 13 microsatellite and SNP markers associated with NOD autoimmune susceptibility17. Backcrossing is facilitated by the fact that Ocab is located on mouse chromosome 9 and is not nearby any of the Idd loci. Following these markers with specific primer pairs (Supplementary Table 1), we produced backcrossed animals with all 13 markers (Supplementary Fig. 1b) that recapitulate spontaneous autoimmunity (not shown). For example, Idd1, which maps to the MHC region18, converted to the homozygous NOD allele after 2 backcross generations (Supplementary Fig. 1B, lane 3), whereas Idd13, which maps near beta-2-microglobulin19 became NOD homozygous after 4 generations (lane 4). Genotyping identified this fourth-generation backcrossed animal as having all 13 microsatellite markers and the Ocab conditional allele (Supplementary Fig. 1c, #122). This founder animal was crossed to NOD.CreCD4 and the floxed allele homozygosed in order to delete OCA-B in T cells to assess the effects on T1D. Similar NOD backcrosses, conducted using Ocab null mice20, resulted in OCA-B germline-deficient NOD mice (data not shown). As with the original C57BL/6 OCA-B germline knockout20, all alleles were viable and fertile with no obvious health issues.
Normal baseline T cell status in NOD mice lacking OCA-B
To study broad effects of T cell-specific OCA-B loss on T cell function, we performed single-cell RNA sequencing (scRNAseq) and TCR clonotype analysis using CD3+ T cells isolated from PLNs of pre-diabetic 8 week-old female NOD.Ocabfl/flCreCD4 or NOD.Ocabfl/fl littermate controls. For each group cells from three animals were combined for microfluidics and subsequent sequencing. T cells were isolated from PLNs of three animals in each group and combined for analysis. ∼6000 knockout and ∼3000 control cells were identified in this analysis, most of which comprise naïve CD4 and CD8 cells (Supplementary Fig. 2a, pink and orange), and regulatory T cells (Tregs, green). There was also a small population of activated CD4++CD8+ cells (green), presumably in the process of exiting the lymph nodes. Other than a small decrease in the percentage of activated CD4++CD8+ cells (4.4 vs 6.1%), few differences in the populations were observed. None of these populations expressed measurable amounts of Ocab (Pou2af1) mRNA (not shown), however Pou2af1 is expressed at low levels in T cells making it difficult capture by scRNAseq. In addition, few changes in gene expression were identified, and the great majority of these were not of immunological interest (Supplementary Fig. 2b, Supplementary Table 2). We also determined TCR repertoires in these T cell populations, identifying few changes as a whole (Supplementary Fig. 2c). However, specific sub-populations did show changes in TCR utilization. For example, NOD CD8+ T cells expressing Trbv13-3 (Vβ8.1) display reactivity to islet-specific glucose-6-phosphatase catalytic subunit related protein (IGRP) and are associated with T1D21. Although this clone was unchanged across the entire population of T cells (Supplementary Fig. 2c), it was over-represented in the effector CD8+ population (Supplementary Fig. 2d). In contrast, other TCR clonotypes such as Trbv13-1, Trbv13-2, Trbv1 and Trbv15 were unchanged (not shown). This result was confirmed with flow cytometric staining of CD8+ T cells using Vβ8.1 antibodies (Supplementary Fig. 2e). This change was coupled with a decrease the presence of this particular clone in pancreatic islets (below). These results indicate that OCA-B loss does not result in gross changes PLN T cell repertoire or status, but results in lymph node retention of potentially autoreactive clones.
NOD mice lacking OCA-B are protected from T1D
T1D onset in NOD mice is spontaneous and acute22. To test the impact of OCA-B loss in T cells on spontaneous T1D, we compared incidence in in female NOD.Ocabfl/flCreCD4 mice, and controls lacking the CreCD4 driver. Glucose was monitored 2-3 times per week. Using a 250 mg/dL blood glucose cutoff, approximately 60% of Ocabfl/fl control mice manifested spontaneous diabetes by 24 weeks of age, while no Ocabfl/flCreCD4 mice became diabetic (Fig. 1d). Blood glucose was stable in Ocabfl/flCreCD4 females but began to spike in control mice beginning at week 16, rising to an average of >300 mg/dL by 24 weeks (Fig. 1e). Similar results were observed using female whole-body knockout NOD.Ocab−/− compared to Ocab+/+ littermate controls. While ∼80% of control females developed T1D within 24 weeks after birth, only ∼20% of NOD.OcabΔ/Δ females did so (Fig. 1f). Cumulatively, the results show that OCA-B expression in T cells is critical for T1D pathogenesis in NOD mice.
NOD.Ocab conditional mice have fewer islet macrophages and IFNγ–expressing cytotoxic T cells, and more Tregs
We sacrificed Ocabfl/flCreCD4 female mice and controls at 8 and 24 weeks, and determined the frequencies of immune cells in the pacreata. We observed more T cells in the pancreata of OCA-B deficient mice (not shown). This result is understandable because the control mice have annihilated their pool of beta cells, resulting in lack of antigen. We also studied pancreatic macrophages in these mice. Islet infiltration of F4/80+CD11b+ macrophages is critical for NOD T1D pathogenesis both in spontaneous and adoptive transfer T1D models23, 24. CCL1 expressed in diabetogenic CD4 T cells is known to recruit CCR8-expressing macrophages25. Interestingly, OCA-B deletion in stimulated, rested and re-stimulated CD4+ T cells decreases Ccl1 mRNA expression compared to controls (Supplementary Table 3). We therefore hypothesized that OCA-B expression in T cells is important for macrophage islet infiltration. Despite the fact that antigen and greater numbers of T cells were present, we observed 5-fold fewer macrophages in 24 week-old NOD.Ocabfl/flCreCD4 islets (Fig. 2a,b). In contrast, macrophage numbers were unaffected by T cell-specific OCA-B loss in pre-diabetic (8 week) mice (Fig. 2a).
IFNγ is a direct OCA-B target gene4 and a strong promoter of T1D. Levels of serum and pancreatic IFNγ gradually rise in young NOD mice, reaching maximum at diabetes onset26. Diabetogenic CD4+ and CD8+ T cells both express IFNγ in humans and NOD mice27, 28, 29. Therefore, we predicted that OCA-B loss would reduce islet IFNγ expression in CD4+ and CD8+ T cells. We compared IFNγ-expressing pancreatic CD4+ and CD8+ T cells in female NOD.Ocabfl/flCreCD4 and littermate control mice lacking CreCD4. In 12 week-old prediabetic mice, CD8+ T cell frequencies were significantly decreased (Fig. 2c), whereas CD4+ T cells at the same timepoint were unchanged (data not shown). At the time of diabetes onset (16 weeks), proportionally fewer IFNγ-expressing CD8+ T cells were present in Ocab conditional-deleted mice compared to controls (Fig. 2d).
CD4+CD25HIFoxP3+ Tregs interfere with T1D development in NOD mice30, 31. Analysis of previously generated RNAseq datasets shows that naïve Ocab null CD4+ T cells express 4-fold more Foxp3 mRNA compared to controls upon in vitro stimulation (Supplementary Table 3). To determine whether the protection conferred by T cell conditional OCA-B is associated with changes in Tregs, we analyzed pancreatic Treg frequencies, observing a trend towards more Tregs in pre-diabetic (8 week-old) NOD.Ocabfl/flCreCD4 mice compared to controls (Fig. 2e). Interestingly, there were no differences late in disease course, at 16 or 24 weeks (data not shown).
If increased Treg numbers help mediate protection in NOD.Ocabfl/flCreCD4 mice, then depletion of these Tregs would be expected normalize T1D pathogenesis in knockout mice relative to controls. Anti-CD25 antibody injection in neonatal NOD mice depletes Tregs (as well as activated T cells), and accelerates T1D onset32. Surprisingly, we observed that anti-CD25 antibody-injected NOD.Ocabfl/flCreCD4 mice were still protected from T1D development relative to controls, though to a lesser degree than in undepleted mice (Fig. 2f). These results support a conclusion that Tregs partly contribute to the T1D protection in mice lacking OCA-B in their T cells.
T cell-specific OCA-B deletion alters the islet-infiltrating TCR repertoire
We performed scRNAseq and TCR clonotype analysis similar to Supplementary Fig. 2, but using older (12 week) but still pre-diabetic NOD.Ocabfl/flCreCD4 or NOD.Ocabfl/fl littermate control female mice, and using total CD45+ cells isolated from pancreatic islets rather than PLNs. Cells were isolated from four animals per group and combined for microfluidics and sequencing. After filtering, ∼6200 knockout and ∼3800 control cells were identified. Overlaying cell from both genotypes, we used their gene expression profiles to definitively identify multiple cell lineages, including naïve CD4+ + CD8+ T cells, memory/effector CD4+ T cells, memory/effector CD8+ T cells, B cells, macrophages and neutrophils (Fig. 3a). The biggest relative change in populations was neutrophils, which decreased from 3.4 to 1.4% in conditional knockout islets (Fig. 3a). B cells were also decreased, albeit more slightly. These changes were accompanied by an increase in the percentages of CD4+ and CD8+ T cells with naïve phenotypes in conditional knockouts (31.7 vs. 25%). Like endpoint diabetic animals (Fig. 2a,b), macrophage numbers were decreased but only slightly. Unlike PLNs, prediabetic islets showed significant changes in gene expression between the two groups (Fig. 3b), the strongest of which occurred in neutrophils (e.g., Cxcl2, down), memory/effector CD8+ T cells (Ccl4, Tcrg-v1, Itga, down; Tcrbv1, Trav8n-2, up), NK T cells (Klra2, Lars2, down) and γδT cells (Gzma, Gzmb, Ccl5, down).
We also profiled TCR clonality in pancreatic T cells. Across total T cells, few changes were observed (Supplementary Fig. 3), however this was largely to do similar representation in the naïve T cell pool. Isolating just antigen-experienced CD4+ and CD8+ effector + memory subpopulations revealed sharp differences. For example, polyclonal NOD mice CD8+ T cells expressing Trbv13-3 (Vβ8.1) display reactivity to IGRP and are associated with T1D21. Relative to controls, this clonotype was under-represented in OCA-B deficient effector + memory CD8+ cells (Fig. 3c, 20.9% vs 8.6%). Approximately half of these cells showed reactivity to IGRP206-214 based on tetramer staining (data not shown). Based on this information, we repeated the clonotype analysis, but only using the effector + memory CD8+ T cell cluster. Within this population, dominant TCR clonotypes used by conditional and control T cells were strikingly different (Fig. 3d). Clonotype 1 for example corresponds to a TCRα V8N-2 chain paired with TCRβ V1 or V15 and has not been associated with T1D. This clonotype was absent in control islets but represented >4% of OCA-B deficient islet CD8+ effector + memory T cells (Fig. 3d). Clonotype 1 cells overlaid with both genotypes is shown in Fig. 3e. Clonotype 2 by contrast is associated with the T1D-associated Vβ13.3 β chain. This clonotype was strongly represented in controls (>3%) but absent from experimental effector + memory CD8+ cells (Fig. 3d). In contrast to CD8-dominant TCR clonotypes, CD4-dominant TCRs associated with T1D were still present in knockouts, but associated with markers of anergy. For example, clonotype 4 (Trav13-1/16N+Trbv15) is associated with pro-diabetogenic, BDC-reactive CD4+ T cells33. This clonotype was enriched in the OCA-B deficient effector + memory CD4+ cells, but only in a CTLA4- and FR4-expressing sub-population (Fig. 3e). Cumulatively, these results indicate that OCA-B loss from T cells significantly skews the landscape of islet effector/memory-like T cells, in particular CD8+ T cells, to an anti-diabetogenic phenotype.
Protection conferred by OCA-B loss is model-specific
T1D in NOD mice originates from defective negative selection in developing T cells, resulting in T cell autoreactivity. OCA-B is not expressed in thymocytes and OCA-B loss does not appear to affect T cell development4. Therefore, the protective effect of OCA-B loss in T cells likely arises from blunting pre-existing autoreactivity in the periphery. To test the hypothesis that OCA-B loss confers protection in simplified systems, we crossed OCA-B deficient mouse models employing TCR transgenes and model antigens.
Depleting Tregs and transferring remaining T cells from BDC2.5 TCR transgenic mice into NOD.Scid mice results in disease after approximately 10 d34, 35. We crossed the germline null Ocab allele to BDC2.5 mice to test the effect of OCA-B in this model. We transferred CD4+CD25− T cells from germline Ocab−/− or Ocab+/+ BDC2.5 mice into NOD.Scid recipients, monitoring diabetes emergence. Transplants from both Ocab−/− and control NOD.BDC2.5 mice resulted in equivalent disease kinetics (Fig. 4a) and severity (not shown). Transferring CD4+CD62L+Vβ+ T cells from Ocab−/− and control NOD.BDC2.5 mice into NOD.Scid recipients also showed no difference between the experimental and control groups (Supplementary Fig. 4a). Negative results were also generated by transplanting total CD4+ T cells into immunodeficient NCG recipients (Fig. 4b). We also tested the effect of T cell conditional OCA-B deletion in a different strain background, C57BL/6J, using the artificial membranous ovalbumin antigen expressed from the rat insulin promoter (RIP-mOVA)36. In this experiment, wild-type OVA-reactive CD8+ OT-I T cells were transferred into either Ocabfl/flCreCD4 or control Ocabfl/fl RIP-mOVA mice. Host CD4+ T cells help transferred OT-I T cells promote T1D pathogenesis in this model37. Mice were then infected with Lm-OVA to induce strong and synchronous beta cell destruction and loss of glucose control (>400 mg/dL). No differences were observed (Supplementary Fig. 4b). These results indicate that OCA-B does not act analogously to transcription factors that act as a primary on/off switches, such as NF-AT. Knockouts of these factors would result in broad immunosuppression and protection from T1D in all models. Rather, our results suggest that the effects of OCA-B loss may protect from autoimmunity in specific contexts, including polyclonal contexts. To test this supposition, we also performed transplant experiments using total splenocytes from polyclonal NOD.Ocabfl/flCreCD4 mice or controls. Using this transplant system, robust T1D protection with T cell-specific OCA-B loss was observed (Fig. 4c). This protection was associated with fewer IFNγ-expressing cytotoxic T cells in the islets of OCA-B T cell conditional NOD mice (Fig. 4d), and less insulitis in histological sections of pancreata (Fig. 4e,f).
NY8.3 transgenic mice express a CD8-restricted TCR directed towards IGRP206-214 and manifest spontaneous T1D38. The NY8.3 clonotype was significantly attenuated in the islets of T cell conditional OCA-B deficient mice (Fig. 3c,d,e). We also crossed the Ocab germline allele NY8.3, passively monitoring Ocab−/− or littermate control Ocab+/+ NY8.3 mice for T1D development. Significant protection was observed (HR=0.4472, p=0.0352), though the degree of protection was less than in polyclonal models (Fig. 4g). Together with the observed depletion of autoreactive CD8 clones from OCA-B deficient pancreatic islets, this result indicates that OCA-B promotes T1D in peripheral auto-reactive T cells in part directly through the CD8+ cells. Using immunoblotting with OCA-B antibodies, we confirmed OCA-B expression in primary splenic C57BL/6 CD8+ T cells. OCA-B is undetectable in naïve splenic CD4+ cells but becomes induced with prolonged activation4. Unlike CD4+ cells (Fig .4h, top panels), OCA-B+ could be detected in naïve cells (Fig. 4h, lane 1). Expression was augmented more rapidly in CD8+ cells upon T cell activation using CD3ε/CD28 antibodies (lane 2). Similar to CD4+ cells, expression was stably maintained when the stimulus was removed and cells were rested with IL-2 (lane 6).
OCA-B deficiency promotes CD4+ T cell anergy in vitro
Our scRNAseq data indicate that CD8-restricted TCR clones associated with T1D are depleted from the islets of OCA-B T cell conditional mice. In contrast, CD4-restricted clones were increased but associated with an anergic phenotype (Fig. 3e). Anergy is a peripheral tolerance mechanism in which T cells that encounter TCR signals without co-stimulation become nonfunctional to protect from autoreactivity39, 40, 41. Normal naïve primary CD4+ T cells can be activated in culture by TCR and costimulatory receptor activation, e.g. with anti-CD3ε/CD28 antibodies. Providing TCR signals (via immobilized anti-CD3ε monoclonal antibodies) to naïve T cells in culture generates anergic responses42. Loss of Oct1, the transcription factor with which OCA-B docks, augments anergic responses in this assay5. Interestingly, OCA-B protein is induced to the same extent in naïve CD4+ T cells stimulated with only anti-CD3ε antibodies as with co-stimulation (Fig.5a). To determine if OCA-B loss promotes anergic CD4+ T cell responses, we stimulated naïve splenic CD4+ T cells from germline knockout Ocab−/− and control Ocab1+/+ animals ex vivo using plate-bound anti-CD3ε antibodies without co-stimulation, and restimulated the cells with PMA/ionomycin. Following restimulation, CD4+ T cells lacking OCA-B generated 2.5-fold less IL-2 (Fig. 5b,c), indicating that OCA-B provides a barrier against anergic responses.
ICOS (Inducible T-cell Costimulator) has an important role in the induction of T cell anergy and is down-regulated by Oct1 loss in T cells43, 44, 45. Similarly in CD4+ cells, elevated expression of FR4 and CD73 is associated with anergy and higher in Oct1-deficient compared to control cells39, 46. To determine whether OCA-B affects in the protein related to T cell anergy, we harvested pre-activated CD4+CD44+ T cells from the spleens of Ocabfl/flCreCD4 and littermate control CreCD4 animals, and re-stimulated them for 24 hr ex vivo using anti-CD3ε antibodies ± co-stimulation using anti-CD28 antibodies. CD4+CD44+FR4hiCD73hi anergic cells were ∼2 fold more strongly induced in T cells lacking OCA-B (Fig. 5d). ICOS levels were similarly elevated in the OCA-B deficient condition (Fig. 5e). Additionally, and consistent with prior data5, the expression of the activation marker CD25 was decreased (Fig. 5f). These results suggest that OCA-B loss creates a requirement for CD28 co-stimulation in secondary stimulation situations, effectively instituting higher immune checkpoint thresholds in order for cells to become fully re-activated.
Generation of an OCA-B peptide inhibitor
The normal T cell developmental and primary immune response phenotypes observed with OCA-B genetic deletion suggested the possibility of a “therapeutic window” in which targeting OCA-B pharmacologically would leave baseline immune function only minimally affected. Drugs that target transcription factors and protein-protein interactions are now entering clinical use47, 48. Pre-clinical membrane-penetrating peptides have also been developed that successfully target transcription factors such as BCL649. Based on the co-crystal structure of the Oct1 DNA-binding domain, the OCA-B N-terminus and consensus binding DNA50, we applied rational-design principles to generate a membrane-permeable peptide inhibitor of OCA-B’s downstream effector functions in order to target OCA-B pharmacologically.
In CD4+ T cells, Oct1 interacts with Jmjd1a in a MEK/ERK-dependent manner4, 6. Oct1 contains two consensus ERK phospho-acceptor serines. The structure of the Oct1 DNA binding domain complexed with consensus octamer binding site DNA51 reveals that these Oct1 serines are located in a flexible, solvent-exposed loop (the linker domain), which connects the two DNA binding sub-domains (Fig. 6a, red dashed line). In contrast with Oct1, OCA-B constitutively interacts with Jmjd1a4. The OCA-B N-terminus has been solved in complex with Oct1 and consensus octamer DNA50. The crystalized region of OCA-B is known to include regions critical for both Oct1 binding and transcription activity52. To identify potential Jmjd1a-interacting OCA-B regions, we aligned the full-length Oct1 and OCA-B amino acid sequences to a region of androgen receptor (AR) known to interact with Jmjd1a53. Human AR mutations that cause androgen insensitivity have been mapped to a cofactor interaction surface spanning residues 698-72154. Aligning this sequence to full-length Oct1 identified the linker domain as the top hit (Fig. 6b), implicating the Oct1 linker as a potential Jmjd1a-interacting surface. The alignment shows conservation of 3 out of 4 sites of deleterious AR mutation (asterisks) with Oct1. The mutation site that is not conserved is a consensus ERK target in Oct1, and a phospho-mimetic glutamic acid residue in AR (blue box). These findings suggest that the Oct1 linker domain constitutes a surface which, when phosphorylated by ERK, interacts with Jmjd1a.
As with Oct1, alignment with OCA-B identifies a potential Jmjd1a interacting surface (Fig. 6b). Unlike Oct1, an OCA-B glutamic acid residue (Glu30) aligns with AR, consistent with the finding that OCA-B constitutively interacts with Jmjd1a in the absence of ERK signaling4. Furthermore, the OCA-B residues aligning to AR lie on one side of a short, solvent-exposed alpha-helix (Fig. 6a). Glu30 (red), Leu31 (yellow), Arg34 (blue) and His38 (blue) may therefore constitute a Jmjd1a docking surface. These residues are conserved to humans (Fig. 6c). We mutated OCA-B L31 and R34 to alanines in a transient expression plasmid to test the effect on the Jmjd1a interaction, and transfected the double-point mutant or parent plasmid control into HCT116 cells, which do not express endogenous OCA-B. Control co-IP using Jmjd1a antibodies and wild-type OCA-B confirms the published interaction (Fig. 6d, lane 5), while the mutant OCA-B protein failed to interact (lane 8). The mutant was expressed equivalently to WT in HCT116 cells (lanes 2, 3). These results show the importance of OCA-B residues 30-38 for the interaction with Jmjd1a.
To identify potential OCA-B competitive inhibitors, we synthesized three overlapping peptides corresponding to the OCA-B N-terminus (Fig. 6e, peptides #1, #2 (hereafter called “JumOCA”) and #3), as C-terminal fusions to the HIV Tat protein for membrane permeability55. Initial preparations were conjugated to FITC to track the peptide. Incubating peptides with total splenic or pancreatic CD3+ T cells in culture with 45 µM Tat-fused peptide for 15 min resulted in significant concentration. Control OVA-fused peptide showed no such effect (Supplementary Fig. 5). We then treated splenic CD4+ T cells with the Tat-fused peptides (lacking FITC) in ex vivo culture. Prior work has shown that 6 hr restimulation of resting but previously activated cells reveals sensitivity to OCA-B loss4. Using this assay with 50 µM JumOCA peptide (every other day with media changes during rest and restimulation) inhibited Il2 mRNA expression relative to β-actin by ∼10-fold (Fig. 6f, compare Tat peptide to JumOCA). Although treated cells showed no obvious changes in viability, morphology or expansion during the course of the assay (not shown), there was a 2-3-fold nonspecific diminution of activity associated with unfused Tat peptide (Fig. 6f, compare Tat peptide with no peptide). The Tat membrane-penetrating peptide has been associated with toxicities49, 56, 57. Similar results were obtained with the larger peptide #3 but not with peptide #1 (not shown). Peptide #1 is missing residues important for the Jmjd1a interaction based on mutagenesis (Fig. 6b). For all further experiments, we exclusively used the JumOCA peptide.
We then synthesized a Tat-conjugated peptide using a scrambled JumOCA sequence. This peptide has the same amino acid composition, mass and pI, but lacks the OCA-B primary amino acid sequence and should not be able to efficiently compete for Jmjd1a. We used the Tat-only, scrambled and JumOCA peptides in assays similar to Fig. 6f, except using flow cytometry to assay endpoint IL-2 production. JumOCA significantly inhibited IL-2 production relative to the scrambled control, however there were significant toxicities associated with both the scrambled and Tat-only peptides relative to vehicle (Fig. 6g). An example mouse from this experiment is shown in Fig. 6h.
The above results are consistent with the JumOCA peptide operating through a Jmjd1a competitive inhibition mechanism. To test this directly, we used scrambled, double point mutant and wild-type (JumOCA) peptides lacking Tat in a Jmjd1a co-IP assay. We subjected lysates from mouse M12 B cells, which expresses endogenous OCA-B, to immunoprecipitation with Jmjd1a antibodies, resulting in efficient OCA-B co-IP (Fig. 6i, lane 3). Incubation of precipitated material with scrambled or double point mutant peptides had no effect on OCA-B co-immunoprecipitation (lanes 4-5), while the same concentration of JumOCA peptide efficiently blocked OCA-B recovery (lane 6).
An OCA-B peptide inhibitor protects NOD mice from newly-arisen diabetes
By the time symptoms arise in NOD mice, 90-95% of beta cells have been destroyed. Remaining beta cells are rendered nonfunctional due to insulitis. Alleviating this inflammation provides a “honeymoon period” during which glucose homeostasis can be restored. We treated 14-18 week-old littermate female NOD mice whose glucose levels were newly-risen above 225 mg/dL but still below 275 mg/dL with three intravenous injections of 10 mg/kg peptide or Tat-only peptide control. Injections were spaced 12 hr apart. 12 hr after the final injection, blood glucose was collected and flow cytometry performed on the pancreas and PLNs. Strikingly, the inhibitor reversed elevated blood glucose (Fig. 7a). Control peptide at the same concentration had no effect. Islet T cell infiltration and pro-inflammatory IFNγ and IL-17A cytokine production were strongly reduced by JumOCA peptide treatment compared to controls (Fig. 7b,c). In contrast to the pancreas, PLNs showed no change in T cell numbers or percentages (Fig. 7d) but similar changes in pro-inflammatory cytokine production (not shown). These data provide evidence that targeting OCA-B is a valid strategy to treat emerging T1D, potentially with minimal side effects, and identify a first-generation inhibitor that is efficacious in vivo.
Discussion
Here we show that targeting OCA-B genetically or pharmacologically can block T1D in mouse models. OCA-B is a transcriptional coregulatory protein named for its strong expression in the B cell lineage, where it is dispensable until after B cell activation, where it is essential for the generation of germinal centers58. OCA-B is not expressed during thymic development or in naïve CD4 T+ cells, but is stably induced upon T+ cell activation. In contrast, OCA-B is expressed in CD8+ T cells, albeit at lower levels compared to activated T cells. In both populations, OCA-B expression is maintained after T cell stimulation has ceased.
In CD4+ T cells, OCA-B associates with the POU-domain transcription factor Oct1 at ∼150 immunomodulatory genes – among them Il2, Ifng, Il17a and Csf2 (Gmcsf). Unlike transcription factors such as NF-AT, AP-1 and NF-kB that act as primary “on” switches via direct gene activation, OCA-B removes inhibitory chromatin modifications to establish permissive chromatin environments that allow for silent but previously activated targets to be robustly expressed later4, 6. More specifically, OCA-B interacts with Jmjd1a/Kdm3a, a histone lysine demethylase that removes inhibitory histone H3 lysine 9 methyl marks. OCA-B target gene expression is unchanged upon primary stimulation of cultured Oct1- or OCA-B-deficient T cells, but upon secondary stimulation (a model of antigen reencounter) the normally observed increased expression does not occur4, 6. In vivo, OCA-B loss leaves T cell development and pathogen response unaffected, but impairs the establishment of new central memory CD4+ T cells. The few cells that are formed respond poorly to antigen re-encounter4. Repeated antigen encounters that drive high levels of proinflammatory cytokines are a key feature of autoimmunity. In both mice and humans, memory or memory-like cells can underlie autoimmunity including T1D59, 60, 61.
The NOD model of T1D is spontaneous but only partially penetrant, much like the human disease62. 60-80% of female NOD mice develop T1D in standard environments22. We find that prediabetic OCA-B T cell deficient NOD mice harbor normal T cell numbers and similar TCR specificities in their PLNs, consistent with prior observations that OCA-B deficient T cells are immunocompetent4. Nevertheless, T cell conditional OCA-B knockouts are fully protected from spontaneous T1D. Protection is associated with reduced islet leukocyte infiltration and reduced proinflammatory cytokine expression. Whole-body NOD Ocab knockouts are also protected from T1D, though to a lesser degree. Prior work showed that OCA-B whole-body knockout protects mice in a MOG/EAE model of multiple sclerosis63. In contrast, a second study showed that OCA-B whole body knockout exacerbates systemic, antibody-driven autoimmunity in Sanroque mice64. The protective vs. exacerbating effects in these models may be due to effects on BCR repertoire65. T cell-specific knockout of Oct1, a transcription factor with which OCA-B docks, blocks EAE but leaves responses to infection with neurotropic viruses intact5. Together these data indicate that in T cells Oct1 and OCA-B promote autoimmunity including T1D, but that their role in B cells may be more complex. OCA-B is typically expressed ∼50-fold higher in B cells compared to T cells. A correctly calibrated dose of competitive inhibitor could therefore be used to selectively blunt T cell-mediated autoimmunity.
scRNAseq experiments reveal decreased neutrophil infiltration and an increase in the percentage of T cells with a naïve phenotype in prediabetic OCA-B conditional knockout islets. Gene expression changes were also identified that were consistent with antidiabetogenic effects of OCA-B T cell knockout. For example, the remaining islet neutrophils in knockout mice show strong reductions in Cxcl2 expression compared to littermate controls. Cxcl2 is a potent neutrophil chemoattractant that promotes T1D in vivo66, 67, and therefore reductions in Cxcl2 may contribute to the reduced neutrophil representation. Strikingly, OCA-B T cell deficient mice also lacked T1D-associated TCR clones in their activated/memory CD8 pool. CD4 clonotypes associated with T1D were present, but associated with anergy. Consistently, OCA-B loss promotes CD4+ T cell anergy in vitro.
BDC2.5 and NY8.3 TCR transgenic mice are simplified monoclonal systems that allow easy tracking of uniform T cell responses to antigens in vivo68. To test if OCA-B loss confers protection in simplified monoclonal systems, we crossed the germline OCA-B null allele to these models, as well as to RIP-mOVA transgenic mice on the C57BL/6 background. The latter mice express synthetic membranous chicken ovalbumin in their thymi and pancreatic beta cells69. We crossed whole-body null mice to the BDC2.5 transgenic line70, which expresses a CD4-restricted TCR specific for a hybrid insulin-chromogranin A peptide71. BDC2.5 autoreactive cells migrate to the islets and cause insulitis beginning at 3-4 weeks. Due in part to Tregs that have escaped allelic exclusion, in many animals insulitis is limited and does not progress to full-blown T1D. However, transferring CD4+CD62L+Vβ+ or CD4+CD25− T cells from BDC2.5 donor mice into NOD.SCID recipients results in rapid (7-10 d) disease onset34. Transplants from both control and OCA-B null NOD/BDC2.5 mice resulted in disease with similar kinetics. Identical results were also generated by depleting CD25+ Tregs, or by transplanting total CD4+ T cells into immunodeficient NCG recipient mice. OCA-B loss also had no effect using the RIP-mOVA model on the C57BL/6 background. This model expresses a synthetic antigen, membranous chicken ovalbumin, in thymic epithelial and pancreatic beta cells36.
In contrast to BDC2.5 and RIP-mOVA, NY8.3 transgenic mice carry a monoclonal CD8-restricted TCR directed towards IGRP206-214 and manifest spontaneous T1D38. Unlike BDC2.5 transplant models, OCA-B T cell conditional NY8.3 mice showed significant protection, though unlike polyclonal NOD mice protection was incomplete. Cumulatively, the results are consistent with a model in which OCA-B loss does not simply block T cell functionality, but rather desensitizes autoreactive T cells. Because transgenic TCR models have fixed affinities to antigens that can be present throughout disease or arise at specific points during disease etiology, they result in graded levels of disease. By shifting the autoreactivity of different T cells along a spectrum of disease potentials, OCA-B loss can result in full or partial protection in models with low and intermediate disease thresholds, whereas models with high disease potentials remain autoreactive. Although protection in some of the more severe models is likely to be inseparable from immunodeficiency, our results show that genetic ablation of OCA-B strongly protects NOD animals from spontaneous T1D, and results in graded protection that drops off with severity of the model. These properties make OCA-B a promising target for pharmaceutical inhibition.
Recent work indicates that the prevailing idea that transcription regulators and protein-protein interactions are “undruggable” is erroneous47, 48, 72. Pre-clinical membrane-penetrating peptides have also been developed that successfully target transcription factors such as BCL649. We applied rational-design principles to generate a membrane-permeable competitive peptide inhibitor of the interaction between OCA-B and a downstream effector, Jmjd1a/Kdm3a. This inhibitor corresponds to a region of the OCA-B N-terminus, which is known to include residues critical for Oct1 binding and transcription activity52. We tested the effect of this inhibitor, termed JumOCA, on T1D onset and severity. The JumOCA sequence does not coincide with other sequences in the human or mouse proteome and is not predicted to be strongly immunogenic. Peptide administration into NOD mice protected animals from newly-arisen T1D and significantly reduced islet-infiltrating T cell numbers and cytokine production, while having more mild effects on the PLNs. OCA-B levels in B cells are at least 50-fold higher than in T cells (immgen.org)73, making it likely that the observed effects are due to inhibition in T cells. The preservation of normal PLN functionality both with peptide treatment and OCA-B genetic ablation suggests a potential therapeutic window in which targeting OCA-B may leave baseline immune function unaffected. While these peptides are unlikely to be used in a clinical setting, they offer proof-of-principle for OCA-B as a therapeutic target, and can be used as tools for the further development of therapeutics.
Methods
Laboratory mice
NOD/ShiLtJ, NOD.Scid (NOD.Emv302/2.CB17-Prkdcscid), BDC-2.5 TCR transgenic, NY8.3 TCR transgenic and all breeder mice were maintained in a specific pathogen-free research colony. NCG (NOD-Prkdcem26Cd52Il2rgem26Cd22/NjuCrl) mice were purchased from Charles River Laboratories. NOD.CreCD4 breeding pairs were a gift from Alexander Chervonsky (University of Chicago). All animal experiments were approved by the University of Utah Institutional Animal Care and Use Committee (IACUC approval 17-05008).
Generation of Ocab (Pou2af1) conditional mice
Pou2af1tm1a(KOMP)Wtsi mice were provided by the knockout mouse project (KOMP) repository (University of California, Davis). Mice were derived from embryonic stem cell clone EPD0761_3_B03, generated by the Wellcome Trust Sanger Institute. These cells were derived from C57BL/6N mice, but subsequent mouse crosses were to C57BL/6J (>20 generations) for the RIP-mOVA model and NOD (four congenic backcrosses plus an additional backcross to NOD.CreCD4). The embryonic stem cells contain a CSD-targeted allele74. The presence of WT (582 bp) and Post-FLP (700 bp) alleles was determined by PCR using CSD-Pou2af1-F and CSD-Pou2af1-ttR primers. The presence of the null (801 bp) allele was determined using CSD-Pou2af1-F and CSD-Pou2af1-R primers. The presence of the floxed (359 bp) allele was determined using CSD-Lox-F and CSD-Pou2af1-R primers. Primer sequences were as follows: CSD-Pou2af1-F, 5’TACAGAGAGACTAGACACGGTCTGC; CSD-Pou2af1-R, 5’GATGAGGACTCTGGGTTCAGAGAGG; CSD-loxF, 5’GAGATGGCGCAACGCAATTAATG; CSD-Pou2af1-ttR, 5’AGAAGGCCTCGTTACACTCCTATGC. Ocab conditional mice were bred with mice expressing Cre under the control of the CD4 promoter.
Backcrossing Ocab germline null and conditional mice to the NOD background
To generate Ocab−/− and Ocab conditional mice on the NOD background, the previously described C57BL/6 Ocab germline null allele20 and the newly-generated Ocab conditional allele were backcrossed to NOD ShiLt/J mice using congenic markers based on Mouse Genome Informatics (http://www.informatics.jax.org) (Supplementary Table 1).
Diabetes development and assessment
Diabetes was monitored using blood glucose test strips (Contour, Bayer). Mice with blood glucose levels >250 mg/dL were considered diabetic.
Leukocyte isolation and intracellular cytokine staining
Pancreatic lymph nodes were grinded and passed through a nylon strainer. Pancreatic leukocytes were isolated as described previously75. Briefly, pancreata were chopped and digested using collagenase IV (1 mg/mL, Gibco) in DMEM containing 1% fetal bovine serum (FBS, Rocky Mountain Biologicals) and 10 units of DNase I for 15 min at 37°C. The digested tissues were passed through a 70 μm strainer. Red blood cells were lysed by ACK (Ammonium-Chloride-Potassium) lysis buffer (150 mM NH4Cl, 10 mM KHCO3, 0.1 mM Na2EDTA). For intracellular cytokine staining, cell suspensions in RPMI medium supplemented with 10% FBS were re-stimulated for 4 hr with phorbol 12-myristate 13-acetate (PMA, Sigma, 50 ng/mL) and ionomycin (Sigma, 1 µg/mL) in the presence of Brefeldin-A (GolgiPlug, BD Bioscience, 1 µl/ml), and fixed by cell fixation/permeabilization solution (BD Cytofix/Cytoperm) according to the manufacturer’s protocol. Antibodies used for flow cytometry were as follows: FITC-conjugated anti-mouse CD4, PE-conjugated anti-mouse CD45, PE-conjugated anti-mouse FR4 (Biolegend), PerCP-conjugated anti-mouse CD8a, APC-conjugated anti-mouse IFNγ, PE-conjugated anti-mouse IL-17, PerCP-conjugated anti-mouse CD11b, APC-conjugated anti-mouse F4/80, PE-conjugated anti-mouse Gr-1, PE-conjugated anti-mouse ICOS (eBioscience), PerCP-conjugated anti-mouse FoxP3, V450-conjugated anti-mouse CD73 (BD Bioscience) FITC-conjugated anti-mouse Vβ8.1/8.2 (Invitrogen).
T cell adoptive transfer
6–8 wk old NOD.Scid recipient mice were injected retro-orbitally with 2×105 of splenic CD4+CD25− T cells from pre-diabetic Ocab−/− or Ocab+/+ NOD.BDC2.5 donors (6–8 wk old, sex-matched). T cells were purified using a CD4+CD25+ T cell isolation kit (Miltenyi). For CD4+CD25+Foxp3+ T cells, 1.5×106 purified splenic CD4+ T cells from pre-diabetic, BDC2.5 transgenic NOD.Ocab−/− or NOD.Ocab+/+ mice were transferred to sex-matched 6-8 wk old NCG recipients (University of Utah pre-Clinical Resource Core) as previously described35. For total splenic transfer experiments, 5×106 splenocytes from pre-diabetic NOD.Ocabfl/fl or NOD.Ocabfl/flCreCD4 mice were transferred into sex-matched 6-8 week-old NOD.Scid recipients mice as described previously35.
In vivo Treg depletion
Anti-CD25 antibody in PBS (1 mg/mouse, Clone PC61, BioXCell) was administered intraperitoneally to NOD.Ocabfl/fl or NOD.Ocabfl/flCreCD4 mice twice (10 and 17 d after birth) as previously described32.
In vitro culture of CD4 T cells
For anergic cell induction, spleens were harvested from NOD.Ocabfl/flCreCD4 or control NOD.Ocabfl/fl animals. Single-cell suspensions were generated by grinding and passage through 70 µm strainers. CD4+ T cells were isolated using a mouse CD4+ T cell isolation kit (Miltenyi Biotec). Isolated CD4+ T cells were cultured and stimulated with 5 µg/ml plate-bound anti-CD3ε (BD Bioscience) and/or 2 µg/ml anti-CD28 antibodies (eBioscience) as described previously5. For OCA-B inhibitor peptide treatments, naïve CD4 T cells were isolated from WT C57BL/6 mice using a mouse naïve CD4+ T cell isolation kit (Miltenyi Biotec) and cultured as described previously6. For the in vitro experiments with OCA-B inhibitor peptides, indicated concentrations of OCA-B inhibitor or control peptides were incubated with cells, with media changes every 2 days after primary stimulation. Activation of C57BL/6 CD4+ cells and subsequent profiling of anergic responses was performed identically to5.
Single cell RNA sequencing
CD3+ T cells were isolated from 8 week-old NOD.Ocabfl/fl or NOD.Ocabfl/flCreCD4 females using a pan T cell isolation kit (Miltenyi). CD45+ pancreatic leukocytes were isolated from 12 week-old NOD.Ocabfl/fl or NOD.Ocabfl/flCreCD4 females were isolated by flow cytometry using a FACS Aria instrument (Becton-Dickinson). For each condition, cells were isolated from three mice and combined. Cells were processed using the 10× Genomics Chromium platform according the manufacturer’s instructions. Paired-end high-throughput sequencing (125 cycles) was performed via an Illumina NovaSeq instrument. Sequencing reads were processed by using 10× Genomics CellRanger pipeline and further analyzed using the Seurat R package. Analysis of cells used a standard filtering protocol removing cells with unique feature counts of >4,000 or <500, as well as cells that have >5% mitochondrial counts (indicative of dead cells). No more than 15% of total cells were removed by this process.
OCA-B mutagenesis and co-IP
For generation of the OCA-B double point mutant, a plasmid transiently expressing human OCA-B, pCATCH-Bob.176, was sequentially mutagenized using QuickChange (ThermoFisher). Correct mutagenesis and lack of additional mutations was confirmed by resequencing. Protein expression was confirmed by transiently transfect HCT116 cells, which do not express endogenous OCA-B. HCT116 lysate preparation, co-immunoprecipitation of OCA-B and endogenous Jmjd1a, and immunoblot detection was performed using the same conditions as described4. For co-IP using M12 cells, lysis buffer consisted of 50 mM Tris-Cl, pH 7.4, 150mM NaCl, 0.5%NP-40, 1 mM EDTA,1 mM EGTA, plus protease/phosphatase inhibitors (PhosSTOP, Roche). Lysates were incubated with 2.5 µg Jmjd1A antibody (Bethyl) and protein-G Dynabeads (ThermoFisher) in lysis buffer containing 20% glycerol overnight at 4°C. After overnight incubation, the indicated concentrations of peptide or PBS vehicle was added and incubated for a further 3 hr at 4°C prior to bead precipitation and washing 3× in lysis buffer plus 20% glycerol. Co-precipitated OCA-B was analyzed by SDS-PAGE and immunoblot.
OCA-B inhibitor peptide synthesis
Unique chemically synthesized peptide sequences were as follows: Peptide#1, ARPYQGVRVKEPVK; Peptide#2/JumOCA, VKELLRRKRGH; Peptide#3, ARPYQGVRVKEPVKELLRRKRGH; Scrambled peptide, VLREKGKRHLR. Peptides were synthesized with and without a covalent C-terminal linker and Tat membrane-penetrating peptide (GSG-GRKKRRQRRRGY). Fmoc protected amino acids were obtained from Protein Technologies Inc. 1-[Bis(dimethylamino)methylene]-1H-1,2,3-triazolo[4,5-b]pyridinium 3-oxid hexafluorophosphate (HATU) was purchased from Chemimpex Inc. H-Rink-Amide-ChemMatrix (RAM) was purchased from Biotage. N,N-diisopropylethylamine (DIEA), dichloromethane (DCM), triisopropylsilane (TIS) were purchased from Sigma Aldrich. Dimethylformamide (DMF), trifluoroacetic acid (TFA), acetonitrile (ACN), and ethyl ether were purchased from Fisher Scientific. Peptides were synthesized using automated Fmoc SPPS chemistry on synthesizer (Syro I). Briefly, 220 mg of RAM resin (Loading = 0.45 mmol/g) was swelled in DCM for 15 min and followed by adding in a solution of specific Fmoc protected amino acid (AA = 0.1 mmol, DIEA = 0.2 mmol in DCM = 4 mL) and incubated at room temperature for 1.5 h. Then the resin was washed with DMF and DCM and incubated with 5 mL of DCM containing 16% v/v MeOH and 8% v/v DIEA for 5 min. This action was repeated for 5 times before thoroughly washed with DCM and DMF. Then the resin was set to synthesizer for automated synthesis. Coupling reactions were performed using HATU (5 eq.), DIPEA (10 eq.) and amino acid (5 eq.) in DMF with 15 min heating to 70 oC (50 oC for Cys). Deprotection reaction was performed using 20% (v/v) Piperidine in DMF, 5 min rt for 2 rounds. Peptides were cleaved from the resin by treatment with a cocktail buffer (3 mL/0.1 mmol, TFA:H2O:TIS:EDT = 95:2:2:1) for 2.5 h. Peptide-TFA solution was then filtered and precipitated in cold ether, and centrifuged and washed with ether twice and vacuum-dried. The crude product was then purified by RP HPLC. Peptides characterization was performed by LC/MS on an Xbridge C18 5 µm (50 × 2.1 mm) column at 0.4 mL/min with a water/acetonitrile gradient in 0.1% formic acid on an Agilent 6120 Quadrupole LC/MS system. Fractions collected from HPLC runs were also analyzed by LC/MS. The purified fractions containing the targeted product were collected and lyophilized using a Labconco Freeze Dryer. All samples were analyzed by the following conditions: Preparative reverse phase HPLC of crude peptides was performed on a Jupiter 5 µ C18 300 Å (250×10 mm) column at 3 mL/min with a water/acetonitrile (ACN) gradient in 0.1% TFA on an Agilent 1260 HPLC system. Purity, isomer co-injection and stability checks were performed on HPLC on Phenomenex Gemini C18 3 µm (110 Å 150×3 mm) column.
OCA-B inhibitor peptide treatment
Anesthetized pre-diabetic WT NOD females with glucose levels newly-risen to between 225 to 275 mg/dL were treated with the inhibitor or control peptides by intravenous (retro-orbital) injection, 3 times every 12 hr. Blood glucose levels were measured at every injection. 4 hr after the last injection, pancreata and pancreatic lymph nodes (PLNs) were collected and cell populations were analyzed by flow cytometry.
Histology
Formalin-fixed pancreatic tissues were embedded in paraffin. H&E-stained sections were scored for islet inflammation based on published precedents77: 0, sparse surrounding sentinel leukocytes, no insulitis; 1, peri-islet leukocytes; 2, some into islet, <50% of islet area; 3, islet insulitis with >50% of islet area occupied by leukocytes; 4, islets destroyed with fibrotic remnants.
Statistical analyses
All error bars denote ±SEM. Two-tailed student T-tests were used to ascribe statistical significance. For all figures, *=p-value ≤0.05; **=p-value ≤0.005.
Author’s contributions
DT conceived the study and designed experiments, supervised the study, and provided administrative, technical and material support. HK, AS, AI and JLJ acquired and interpreted the data. JP performed high-level analysis of scRNAseq and TCR clonotype data. CG helped generate critical reagents. BDE, DH-CC, XH and PJ provided material and intellectual support. All authors were involved in writing, reviewing and revising the manuscript.
Acknowledgements
We thank P. Santamaria, M. Williams, M. Bettini and F. Gounari for critical reading of the manuscript. We thank A. Chervonsky (U. Chicago) for the gift of NOD.CreCD4 mice. We thank J. Marvin and the University of Utah Health Sciences Center Flow Cytometry Core facility for assistance with flow cytometry. We thank M. Hanson and the University of Utah Health Sciences DNA/peptide synthesis core. This work was supported by grants to DT from the Praespero Foundation, Juvenille Diabetes Research Foundation (1-INO-2018-647-A-N) and National Institutes of Health (R01-AI100873).
Footnotes
Pancreatic histology added. Peptide competition experiment in Figure 6i now includes scrambled and double point-mutant controls.