A Precision B Cell–Targeted Therapeutic Approach to Autoimmunity Caused by Phosphatidylinositol 3-Kinase Pathway Dysregulation

This article is featured in In This Issue, p.3337

Multiple mechanisms are involved in the maintenance of B cell tolerance to autoantigens. In the bone marrow, receptor editing and clonal deletion ensure that B cells undergoing high avidity interactions with self-antigens are removed from the repertoire (1–4). However, B cells recognizing lower avidity self-antigens do not undergo receptor editing, but, instead, are released into the periphery, where they are maintained transiently in an unresponsive state called anergy (5–7). Anergy is rapidly reversible, requiring chronic receptor stimulation by self-antigen (8, 9), suggesting maintenance by nondurable biochemical mechanisms. Anergy is therefore a fragile state, and these cells represent a pool of autoreactive cells that may participate in pathogenic autoimmune responses under circumstances of immunological stress, such as inflammation. Increasing evidence indicates that a number of genetic alleles that confer increased risk of autoimmunity may act by weakening intrinsic mechanisms that maintain the unresponsiveness of anergic B cells (10–16).

Genome-wide association and candidate studies have revealed more than 100 genetic polymorphisms that confer increased risk of developing systemic lupus erythematosus (SLE) (17), several of which encode molecules thought to function in regulation of BCR signaling (reviewed in Ref. 18). Precise regulation of BCR signaling is key to ensuring that protective responses are mounted against potential pathogens while preventing responses to self-antigens or endogenous Ags. Maintenance of the anergic state of peripheral autoreactive B cells involves multiple regulatory mechanisms that operate proximally in BCR signaling. Among these are inositol lipid phosphatases, PTEN and SHIP-1, which, in anergic cells, prevent the BCR-mediated accumulation of PI (3,4,5)P3, which is crucial for recruitment and activation of PH-domain–containing signaling intermediaries such as Bruton's tyrosine kinase (BTK) and phospholipase Cγ (PLCγ) (19–21). Acting in concert with parallel signaling pathways, these effectors function in B cell activation and differentiation. Certain alleles of genes that encode or regulate expression of components of this axis, including PTEN (22), SHIP-1 (23), SHP-1 (24, 25), Csk (16), PTPn22 (10–13), and Lyn (14, 15), have been shown to confer risk of autoimmunity (26). We, and others, have shown that acute deletion of SHIP-1 or PTEN and expression of a constitutively active catalytic subunit of PI3K in anergic B cells leads to immediate loss of anergy, followed by cell proliferation, differentiation, and production of autoantibodies, thus demonstrating the importance of these proteins and their regulation of the PI3K pathway in maintaining B cell anergy (19, 27, 28). Importantly, B cells from SLE, type 1 diabetes (T1D) and autoimmune thyroiditis (AITD) patients express reduced levels of PTEN, consistent with a possible role in autoimmunity (22, 29). The apparent inability to regulate the PI3K pathway in these patients suggests that inhibition of PI3K could, by compensating for reduced inositol lipid phosphatase activity, be an affective therapeutic.

PI3Ks regulate numerous biological functions via generation of inositol lipid second messengers. Class IA PI3Ks are heterodimeric proteins comprised of a regulatory subunit (p85α, p85β, or p55γ) and a catalytic subunit (p110α, p110β, or p110δ) that function in Ag, costimulatory, and cytokine receptor signaling. Class IB PI3Ks consist of a regulatory subunit, p101, and a catalytic subunit, p110γ, and are activated by chemokine receptor signaling. Both p110δ and p110γ are restricted in expression to the lymphoid compartment with nonredundant, nonoverlapping roles, whereas p110α and p110β are ubiquitously expressed, and removal of these subunits results in embryonic lethality (30, 31). There is a growing body of evidence indicating that p110δ is the functionally dominant isoform used in BCR signaling (32–34). Mice deficient in p110δ show marked phenotypic changes in the B cell compartment, with defects in BCR-mediated calcium mobilization, decreased germinal center formation, and reduced Ab responses to both T-dependent and T-independent Ags (35). To eliminate potential confounding compensation from other isoforms, Okkenhaug and colleagues (36) introduced a point mutation in p110δ that resulted in an amino acid change, p110δ^D910A, rendering the enzyme catalytically inactive. The p110δ^D910A mice have drastically decreased B cell responses, both in vivo and in vitro, with slight reduction in T cell populations. There is conflicting evidence regarding the functional importance of p110δ in T cells, as no defects, or only mild defects, are observed in knockout mice (35–37). This topic is expertly reviewed elsewhere (38). The minimal effect of p110δ deficiency on the T cell compartment is presumably due to the compensation by redundant p110 isoform function, which enables normal T cell function in the absence of p110δ (39). It is noteworthy that p110δ^D910A T cells have a more naive-like phenotype in the periphery, suggesting T cells develop, but fail to mature normally (33). Conversely, p110γ knockout mice have marked reductions in development and function of the T cell compartment, both in vitro and in vivo, whereas B cell responses and populations are unaltered (40). Recent in vitro data using isoform-specific inhibitors of p110δ, p110γ, and dual p110δ and p110γ attempt to tease apart the contribution of individual isoforms in B cell signaling and B cell responses. Inhibition of both p110δ and p110γ did not suppress B cell responses more than inhibition of p110δ alone (41). Additionally, treatment with p110γ inhibitor alone did not have an effect on B cell proliferation, survival, or plasmablast formation, suggesting that p110γ plays a minor role in B cell function (42). Thus, p110δ is critical for B cell, but not T cell, function.

We hypothesized the autoimmunity caused by failed regulation of PI (3,4,5)P3 levels in B cells might be corrected by compensatory inhibition of p110δ using low dosages of pharmacologic inhibitor that preserves the ability of naive B cells to mount protective immune responses. Idelalisib, or CAL-101, is a reversible p110δ inhibitor that noncovalently binds the ATP binding pocket of the catalytic subunit (32). Idelalisib targets the p110δ isoform with 110–453-fold more selectivity than other class 1 isoforms. Idelalisib was approved by the Food and Drug Administration for the treatment of lymphocytic lymphoma, chronic lymphocytic leukemia, and non–Hodgkin lymphoma in 2014 (43) and has completed a phase I clinical trial for treatment of allergic rhinitis (44). It is noteworthy that dosages of idelalisib (30 mg/kg) used in these applications do indeed significantly deplete B cells, and a black box warning has been issued for fatal and/or severe colitis, pneumonitis, and infection (45, 46). However, we reasoned that, at lower dosages, idelalisib may specifically prevent/treat autoimmunity caused by defective PI3K pathway regulation.

In this study, we demonstrate that compensation of failed PI3K pathway regulation using low dosages of PI3K inhibitor is sufficient to delay development of autoimmunity in VH125.NOD mice, a murine model of T1D. Chronically treated animals remain immunocompetent, as indicated by production of class-switched, high-affinity Abs in response to immunization. We show that low-dosage p110δ inhibition can selectively inhibit participation in autoimmunity of autoreactive B cells that have lost anergy because of defective PI3K pathway regulation, whereas autoreactive B cells that have lost anergy because of loss of a regulatory tyrosine phosphatase, SHP-1, still develop autoimmunity. We report that low-dosage idelalisib treatment does not affect in vitro or in vivo T cell responses. This study supports the principle of effective generation of precision therapies based on predisposing genetic factors and should provide a precise therapeutic approach for patients that possess risk alleles that compromise PI3K pathway regulation.

Except where otherwise indicated, 6–16-wk-old mice were used in all experiments. Both male and female mice were used, but experiments were sex matched, and both sexes gave identical results, with the exception of only female mice being used in the VH125.NOD experiments, as female mice develop accelerated disease. Dr. J. W. Thomas (Vanderbilt) generously provided VH125.NOD animals. hCD20-TamCre animals (47) were intercrossed with mice carrying the rosa26-flox-STOP-YFP allele (48), generating mice in which YFP is expressed in B cells upon Cre activation. These mice were crossed with Ars/A1 (49) BCR transgenic mice to generate hCD20-TamCre × rosa26-flox-STOP-YFP × Ars/A1 mice. B cells from these mice will be referred to as wild-type (WT) Ars/A1. These mice were also crossed with SHIP-1^flox/flox mice [gift from J. Ravetch and S. Bolland, The Rockefeller University, New York, NY (50)]. hCD20-TamCre × rosa26-flox-STOP-YFP × Ars/A1 mice were also crossed to PTEN^flox/flox mice [gift from R. Rickert, Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA (51)] and SHP-1^flox/flox mice (52) to generate mice in which SHIP-1, PTEN, and SHP-1 deletion can be induced in anergic B cells. hCD20-TamCre × rosa26-flox-STOP-YFP × PTEN^flox/flox × Ars/A1 were crossed to hCD20-TamCre × rosa26-flox-STOP-YFP × SHIP-1^flox/flox × Ars/A1 to generate hCD20-TamCre × rosa26-flox-STOP-YFP × PTEN^flow/wt × SHIP-1^flox/wt × Ars/A1 to allow for the double haploinsufficiency of both PTEN and SHIP-1 within B cells. Mice were housed and bred at the Biological Resource Center at National Jewish Health or at the University of Colorado Anschutz Medical Center Vivarium, with the exception of C57BL/6 mice and CD45.1 mice (B6.SJL-Ptprca Pepcb/BoyJ), which were purchased from Jackson ImmunoResearch Laboratories All experiments with mice were performed in accordance with the regulations and approval of National Jewish Health and the University of Colorado Denver Institutional Animal Care and Use Committees.

Two hours before adoptive transfer, C57BL/6 recipient mice received 200 rad irradiation. For MD4 transfers, recipients did not receive prior irradiation. B cells from donor mice were isolated via depletion of CD43⁺ cells with anti-CD43–conjugated magnetic beads (MACS anti-mouse CD43; Miltenyi Biotec). Alternatively, CD4⁺ T cells were isolated via CD4-positive selection (MACS anti-mouse CD4 [L3T4]; Miltenyi Biotec). Resultant populations were >97% pure based on flow cytometric analyses. Donor B cells were labeled with either CellTrace Violet (Molecular Probes) or CFSE (Molecular Probes) at 5 μM for 5 min at room temperature prior to transfer. Donor CD4 T cells were labeled with CFSE (Molecular Probes) at 5 μM for 5 min at room temperature prior to transfer. 1–2 × 10⁶ donor cells in 200 μl PBS were adoptively transferred via i.v. injection. Twenty four hours posttransfer, tamoxifen was administered to activate Cre. Tamoxifen (T-5648; Sigma-Aldrich) was dissolved in 100% corn oil (Sigma-Aldrich) at 20 mg/ml. Recipient mice were injected i.p. with 100 μl (2 mg) on two consecutive days.

Idelalisib (LC Laboratories) was shipped to Research Diets for blending the compound homogenously into modified open source diet with 24% kcal protein, 16% kcal fat, 60% kcal carbohydrate, 100 g of cellulose, and 25 g inulin. Diet dosage is calculated by multiplying the single daily dosage by the body weight of a mouse and dividing that by the daily food intake [diet dosage = (single daily dosages × body weight of a mouse) per daily food intake]. Chow used includes base chow, as described above with: +0 mg idelalisib per kg diet (vehicle control), +600 mg idelalisib per kg diet (30 mg/kg ingested idelalisib dosage), +75 mg idelalisib per kg diet (3.75 mg/kg ingested idelalisib dosage), and +18.75 mg idelalisib per kg diet (0.9375 mg/kg ingested idelalisib dosage).

For PTEN^fl/wt × SHIP-1^fl/wt and SHP-1^fl/fl adoptive transfers, control chow and idelalisib-containing chow was given to mice on day 7 after tamoxifen administration. For MD4 and OT-II adoptive transfers, control chow and idelalisib-containing chow was given to mice 2 d posttransfer. For VH125.NOD experiments, immediately following weaning, animals are placed on control chow. Following two consecutive diabetic blood glucose readings (between 150 and 200 mg/dl), animals either remain on vehicle control chow or are placed on 0.9375 mg/kg idelalisib-containing chow and disease progression is monitored. For NP₄OVA-alum experiments, animals were placed on either vehicle control chow or 0.9375 mg/kg idelalisib chow for 28 d and subsequently immunized.

HEL conjugated to SRBCs was used to produce Ag for experiments with MD4 B cells. SRBCs were purchased from the Colorado Serum Company and stored in Alsever solution at 4°C. SRBCs are washed three times in PBS prior to use. One milliliter of 50 mg/ml of the chemical cross-linker N-(three-dimensional dimethylaminopropyl)-N-ethylcarbodiimide hydrochloride (Sigma-Aldrich) was added to 1 ml packed SRBCs and 15 ml of 5 mg/ml HEL (Sigma-Aldrich), mixed and rotated at room temperature for 45 min. Mice were immunized i.p. with 200 μl of a 5% HEL-SRBC in PBS. For NP₄OVA-Alum immunizations, 10 mg/ml alum and 5 mg/ml NP₄OVA were mixed to a final concentration of 250 μg/ml NP₄OVA and 2.5 mg/ml alum and rotated for 3 h at room temperature. NOD mice were placed either on vehicle control chow or 0.9375 mg/kg idelalisib chow for 28 d. Mice were then immunized with NP₄OVA-alum i.p. in 200 μl per mouse, and responses were measured on day 7 and day 14 postimmunization.

Spleens were mechanically disrupted, single-cell suspensions were generated, and RBCs were lysed with ammonium chloride Tris. Cells were resuspended in PBS containing 1% FBS and incubated with indicated Abs. For analysis of cell surface markers, Abs against the following molecules were used: B220-PE (RA3-6B2; BioLegend), B220-BV786 (RA3-6B2; BD Biosciences), B220-BV510 (RA3-6B2; BioLegend), CD4-BV711 (GK1.5; BioLegend), CD8 BV421 (53-6.7; BioLegend), CD138-PECy7 (281-2; BioLegend), CD69-BV786 (1H.2F3; BD Biosciences), and CD86-PerCPCy5.5 (GL-1; BioLegend). After cell surface staining, the cells were fixed and permeabilized with Cytofix/Cytoperm (BD Biosciences) (as per the manufacturer’s instructions) and stained with Dylight650-E4 anti-Ars/A1 Id (produced and conjugated in our laboratory) or HEL-650 (produced and conjugated in our laboratory), and Alexa Fluor 488/anti-GFP (rabbit polyclonal; Life Technologies). Events were collected on a CyAn ADP (Dako) and subsequent analysis using FlowJo software (Tree Star).

For measurements of intracellular free calcium concentration ([Ca²⁺]_i), RBC-depleted, single-cell suspended splenocytes were simultaneously stained with CD8-PE (53-6.7; BD Biosciences), CD4-APC (GK1.5; BioLegend), or B220-PE (RA3-6B2; BioLegend) and loaded with Indo-1 acetoxymethyl (Indo1-AM; Molecular Probes), as described previously (8). For analysis of ([Ca²⁺]_i), cells were suspended at 10 × 10⁶ cells/ml in warm IMDM plus 2% FBS in a 500 μl volume. Cells were acquired for 30 s to establish a baseline and then stimulated with 5 μg/ml of F (ab′)₂ rabbit anti-mouse anti-IgG^+/− (H chain and L chain [H&L]; Invitrogen) indicated dosages of idelalisib (LC Laboratories) and acquired for 3 min. For CD4 and CD8 T cells, cells were acquired for 30 s to establish a baseline and then 10 μg/ml of anti-CD3/biotin^+/− (145-2C11; BD Biosciences) indicated dosages of idelalisib was added; 60 s later, 20 μg/ml streptavidin (Sigma-Aldrich) was added and acquired for 3 min. Mean relative ([Ca²⁺]_i) was monitored over time using an LSRFortessa X-20 (BD Biosciences) with analysis using FlowJo software (Tree Star).

RBC-depleted, single-cell suspended splenocytes were suspended at 10 × 10⁶ cells/ml in serum-free IMDM^+/−-indicated dosages of idelalisib (LC Laboratories) (as indicated in Fig. 4) and rested for 1 h at 37°C. Cells are then washed twice in serum-free IMDM and stimulated with 5 μg/ml of F (ab′)₂ rabbit anti-mouse anti-IgG (H&L; Invitrogen) or 10 μg/ml anti-CD3/biotin (145-2C11; BD Biosciences) and 20 μg/ml streptavidin (Sigma-Aldrich) for 2 min. Signaling was stopped by addition of 20% paraformaldehyde to a final concentration of 2%, incubated at 37°C for 15 min, and resuspended in 100% ice-cold MeOH (directly from −80°C). Cells were then placed on ice for 30 min and placed at −20°C for storage. For analysis, cells were stained with B220-BV786 (RA3-6B2; BD Biosciences), CD4-BV711 (GK1.5; BioLegend), CD8-BV421 (53-6.7; BioLegend), and/or p-AKT–Alexa Fluor 647 (pS373; M89-61; BD Biosciences), p-PLCγ-PE (pY759; K86-689.37; BD Biosciences), p-BTK-BV421 (pY180; N35-86; BD Biosciences), and pSyk/Zap70-PE (pY352/pY319; BD Biosciences) at room temperature for 1 h. Cells were washed three times, and samples were acquired in triplicate on an LSRFortessa X-20 (BD Biosciences) with analysis using FlowJo software (Tree Star).

For detection of IgM^a anti-Ars Abs, microtiter plates were coated with 10 μg/ml Ars-BSA16 in PBS and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. For detection of IgM^a anti-HEL Abs, microtiter plates were coated with 10 μg/ml HEL in PBS and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. For detection of total NP-specific IgM and IgG, microtiter plates were coated with 20 μg/ml NP₂₇BSA and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. For detection of high-affinity, NP-specific IgM and IgG, microtiter plates were coated with 20 μg/ml NP₂BSA and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. Serial dilutions of mouse serum in PBS were added and incubated overnight at 4°C. Ars/A1-derived IgM^a Abs and MD4-derived HEL IgM^a Abs were detected with biotinylated DS.1 anti-IgM^a (BD Pharmingen) in PBS, followed by streptavidin-HRP (Thermo Fisher Scientific). For NP-specific IgM, Abs were detected using goat anti-mouse IgM-HRP (SouthernBiotech). For NP-specific IgG, Abs were detected using goat anti-mouse IgG-HRP (SouthernBiotech). Between all steps, plates were washed three times with PBS 0.05% Tween 20. The ELISA was developed with TMB single solution (Invitrogen), and the reaction was stopped with 1 M HCl. OD was measured at 450 nm using a VersaMax Microplate Reader (Molecular Devices), and data were analyzed with SoftMax Pro6 software.

For detection of IgM^a anti-Ars Abs, microtiter plates were coated with 10 μg/ml Ars-BSA16 in PBS and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. For detection of IgM^a anti-HEL Abs, microtiter plates were coated with 10 μg/ml HEL in PBS and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. For detection of total NP-specific IgM and IgG, microtiter plates were coated with 20 μg/ml NP₂₇BSA and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. For detection of high-affinity, NP-specific IgM and IgG, microtiter plates were coated with 20 μg/ml NP₂BSA and blocked with 2 mg/ml BSA in PBS 0.05% Tween 20. Plates were washed three times prior to use with PBS 0.05% Tween 20. RBC-depleted, single-cell suspension of splenocytes in complete medium were added in 2-fold serial dilutions starting at one-hundredth of a spleen in the first well. Plates were incubated overnight at 37°C. Ars/A1-derived IgM^a Abs and MD4-derived HEL IgM^a Abs were detected with biotinylated DS.1 anti-IgM^a (BD Pharmingen) in PBS, followed by streptavidin-AP (Southern Biotech). For NP-specific IgM, Abs were detected using goat anti-mouse IgM-AP (SouthernBiotech). For NP-specific IgG Abs were detected using goat anti-mouse IgG-AP (SouthernBiotech). Between all steps, plates were washed three times with PBS 0.05% Tween 20. The plates were developed by being incubated with ELISPOT developing buffer (25 μm 5-bromochloro-3-indolyl phosphate p-toluidine, 100 nm NaCl, 100 mM Tris, and 10 mM MgCl₂; pH 9.5) for 1 h. The reaction was stopped by washing the plates three times with PBS 0.05% Tween 20. The number of spots at a cell dilution in the linear range was determined, and the number of Ab-secreting cells (ASCs) was calculated.

Data were analyzed using Prism GraphPad Software. Statistical analyses were performed using the indicated statistical tests in figure legends. All p values ≤0.05 were considered statistically significant. Throughout, asterisks were used to denote the following p values: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.005, ****p ≤ 0.0001.

At the outset of our studies, we attempted a home-run experiment, exploring the ability of low dosages of p110δ inhibitor to arrest progression of T1D in a genetically complex model of autoimmunity. The most commonly used mouse model of T1D, the NOD mouse, reflects disease progression in the human (53). Female NOD mice develop overt diabetes at ∼20 wk of age, with lymphocytic infiltration of the islets and autoantibody production preceding hyperglycemia and diabetes. NOD mice are protected from disease development upon deletion of the B cell compartment (NOD.uMT^−/−) (54, 55) or upon skewing of the BCR repertoire away from insulin reactivity (VH281.NOD) (56), but not upon removal of autoantibody (mIgM NOD) (57). In these studies, we used the VH125.NOD mouse model of disease in which mice carry an IgH transgene specific for insulin, the dominant autoantigen in T1D (58). Importantly, the transgenic H chain can pair with any endogenous L chain, resulting in a frequency of peripheral B cells reactive with insulin of 1–3%. In WT female NOD mice, disease penetrance only reaches ∼70%, but skewing of the B cell repertoire toward insulin reactivity leads to 100% penetrance of disease in female mice and earlier disease onset (56).

Although multiple insulin-dependent T1D (Idd) loci contribute to disease development in NODs (59), B cells in these mice exhibit a marked reduction in PTEN levels in both insulin-reactive B cells and total B cells (29) compared with closely related, autoimmunity-resistant VH125.C57BL/6-H2g7 mice. On this background, but not VH125.NOD, high-affinity, insulin-reactive B cells are anergic. NOD mice have increased susceptibility to additional autoimmune diseases, such as rheumatoid arthritis, SLE, and the experimental autoimmune encephalomyelitis mouse model of multiple sclerosis (60). Reduced B cell expression of PTEN has been reported in lupus patients (22), and we have observed reduced PTEN levels in the B cells of both T1D and AITD patients (29, 61). Thus, loss of B cell tolerance in both human and mouse may be driven in part by PI3K pathway dysregulation. We therefore postulated VH125.NOD mice and T1D patients, both of which have PI3K pathway regulation defects, may benefit from low-dosage idelalisib, a p110δ inhibitor, to reinstate anergy of autoreactive B cells.

To test this possibility, immediately postweaning, female VH125.NOD mice were placed on vehicle control chow to allow habituation to the diet. Upon two consecutive blood glucose readings in the prediabetic range, some mice were maintained on the vehicle control chow, whereas others were fed 0.9375 mg/kg idelalisib-containing chow (Fig. 1A). Disease progression was monitored based on blood glucose levels and visible signs of disease (i.e., hunching, scruffy fur, and excessive urination). In mice receiving 0.9375 mg/kg idelalisib, disease progression was significantly delayed, and survival was extended (Fig. 1B, Supplemental Fig. 1).

As mentioned previously, B cell–depletion therapies are somewhat efficacious in T1D, but are not without safety concerns. Removal of an arm of the adaptive immune system can leave patients susceptible to infection and prevent proper response to immunization. We therefore sought to determine if low-dosage idelalisib treatment would spare responsiveness to immunization. Nondiabetic female NOD mice were placed on vehicle control or 0.9375 mg/kg idelalisib-containing chow for 4 wk, immunized with NP₄OVA plus Alum, and their Ab response was assessed (Fig. 1C). Fourteen days postimmunization, the anti-NP IgM response in the spleen was not different between the vehicle control and 0.9375 mg/kg idelalisib–treated cohorts. Further, the number of total anti-NP IgM ASCs per spleen was similar, as were the numbers of high-affinity IgM anti-NP ASCs per spleen (Fig. 1D, left panel). We also observed no difference between the vehicle control and 0.9375 mg/kg idelalisib–treated cohorts in either total IgG anti-NP or high-affinity IgG anti-NP ASCs per spleen (Fig. 1D, right panel). The levels of IgM anti-NP and IgG anti-NP found in the periphery were also comparable (Fig. 1E) (for preimmune anti-NP serum levels, see Supplemental Fig. 2). There was no difference in B cell numbers in the spleens of animals receiving vehicle control and 0.9375 mg/kg idelalisib–containing chow (data not shown). These data show low-dosage idelalisib treatment does not affect the ability to respond to immunization, as evidenced by similar levels of class switching and affinity maturation between the treated and untreated groups. Thus, animals receiving dosages of idelalisib sufficient to slow progression of disease and prolong survival in VH125.NOD mice remain immunocompetent, alleviating potential adverse outcomes inherently associated with B cell–depletion therapies.

We next sought to confirm the specificity of the idelalisib effect for autoimmunity driven by dysregulation of the PI3K pathway. We and others have previously shown that regulation of the PI3K pathway by the inositol phosphatases PTEN and SHIP-1 is required for maintenance of B cell anergy (27). B cells in T1D, AITD, and SLE patients express reduced PTEN and SHIP-1, presumably due to an increase in the miRNAs that regulate them (e.g., mir-7 and mir-155) (23). Studies in animal models using B cell–targeted conditional deletion of either of these molecules is sufficient to drive autoreactive B cells out of anergy, leading to rapid proliferation and differentiation into ASCs (27). Removal of a single allele of both PTEN and SHIP-1 is also sufficient to allow for loss of anergy because both degrade PI(3,4,5)P3. To best approximate physiologic conditions, we used B cell–targeted conditional deletion (huCD20cre^tam) of one allele of PTEN (PTEN^flox/+) and one allele of SHIP-1 (SHIP-1^flox/+), coupled with a YFP reporter to determine cre-activity, and crossed onto an anti-DNA (Ars/A1) transgenic background that renders B cells anergic (detailed in 2Materials and Methods) (49).

Anergic B cells were adoptively transferred into C57BL/6 recipients, as shown diagrammatically in Fig. 2A. Seven days after tamoxifen treatment, anergy was lost (27), and idelalisib treatment was begun. Fourteen days after adoptive transfer, mice on 0.9375, 3.75 mg/kg, and the clinically prescribed dosage of 30 mg/kg idelalisib–containing chow had significantly decreased serum autoantibody relative to untreated controls (Fig. 2B, quantified in Fig. 2C). Furthermore, the number of autoreactive ASCs per spleen was significantly reduced in cohorts receiving 0.9375 and 3.75 mg/kg idelalisib–containing chow, with undetected ASCs in cohorts receiving 30 mg/kg idelalisib–containing chow (Fig. 2D). The reduction in peripheral autoantibody, as well as ASCs per spleen was not due to a differential recovery of transferred cells among treatment groups, as we found no significant difference in total Ars/A1 Id⁺YFP⁺ B cells in the spleens of animals on day 14 posttransfer regardless of treatment (Fig. 2F). Adoptively transferred PTEN^−/+ × SHIP-1^−/+ B cells from cohorts receiving tamoxifen and 0.9375, 3.75, and 30 mg/kg idelalisib–containing chow underwent significantly decreased proliferation (Fig. 2E, top panel, quantified Fig. 2G), and plasmablast differentiation as measured by CD138 positivity (Fig. 2E, bottom panel, quantified Fig. 2H) relative to mice receiving vehicle control chow. Thus, 0.9375 mg/kg idelalisib, as well as higher dosages, is sufficient to constrain an autoreactive B cell response driven by haploinsufficiency of the inositol phosphatases that regulate the PI3K pathway.

SHP-1 is a regulatory SH2-domain–containing tyrosine phosphatase that mediates the function of inhibitory receptors, such as CD22, PD1, and FCγRIIB, and is necessary for maintenance of B cell tolerance (27, 52, 62). Allelic variants of SHP-1 have been shown to increase risk of developing SLE (24), with studies indicating a subset of SLE patients having reduced SHP-1 protein in their B cells (25). Additionally, reductions in SHP-1 mRNA and protein have been observed in peripheral blood B cells of multiple sclerosis patients (63). Studies in viable motheaten mice have revealed that a mutation in a splice site in Ptpn6, the gene that encodes SHP-1, resulting in an 80–90% reduction in enzymatic activity, leads to severe B cell immunodeficiency and autoantibody production (64). Our laboratory has shown B cell–targeted conditional deletion of SHP-1 from anergic B cells in vivo leads to proliferation and autoantibody production (27). SHP-1 is required to maintain B cell anergy, acting through a pathway distinct from SHIP-1 and PTEN (27).

To test the specificity of idelalisib effects on non-PI3K pathway dysregulation-mediated autoimmunity, we used the adoptive transfer system described in Fig. 3A in conjunction with SHP-1^flox/flox B cells. Idelalisib-containing chow is administered on day 7 posttransfer upon removal of SHP-1 protein from transferred cells. Unlike its enforcement of anergy caused by PI3K pathway dysregulation, low- and intermediate-dosage idelalisib treatment had no effect on loss of anergy caused by SHP-1–induced deficiency. However, autoimmunity was blocked by high-dosage idelalisib (Fig. 3B, quantified in Fig. 3C, 3D). Although we saw a trend of reduced recovery of adoptively transferred SHP-1^−/− B cells from the spleens of animals receiving idelalisib-containing chow, this was NS (Fig. 3F). In animals receiving vehicle, 0.9375 or 3.75 mg/kg idelalisib–containing chow, SHP-1^−/− B cells proliferated (Fig. 3E, top panel, quantified Fig. 3G) and differentiated (Fig. 3E, bottom panel, quantified Fig. 3H) comparably. Only in animals receiving 30 mg/kg idelalisib–containing chow did SHP-1^−/− B cells undergo decreased proliferation and differentiation (Fig. 3E, 3G, 3H). As illustrated in the visual abstract, these findings lead to the conclusion that the particular risk allele-mimetic conditions at play determine the ability of partial p110δ inhibition to enforce tolerance. Specifically, p110δ inhibition compensates for defects in PI3K pathway regulation, but not defects in regulation by the tyrosine phosphatase SHP-1.

In naive B cells, Ag receptor stimulation leads to phosphorylation of the two conserved tyrosines in the ITAMs of CD79a/b, leading to recruitment of Lyn and Syk to the receptor complex. Lyn phosphorylates CD19, allowing its interaction with Lyn and PI3K and subsequent activation of p110δ. p110δ Converts PI(4,5)P2 to PI(3,4,5)P3, generating docking sites for PH-domain–containing downstream BCR effectors, such as PLCγ, AKT, and BTK (21). Multiple parallel pathways emanate from this signalosome, ultimately leading to cell activation, differentiation, proliferation, and migration.

To determine whether p110δ inhibition blocks proximal BCR signaling events, we stimulated splenic B cells from C57BL/6 mice with polyclonal F (ab′)₂ anti-Ig heavy and L chain Abs with simultaneous addition of 0, 15, 60, and 490 nM idelalisib. It is noteworthy that these approximate dietary dosages of 0, 0.9375, 3.75 and 30 mg/kg idelalisib, respectively, were used in vivo; although, because of complex pharmacodynamics, we do not know the precise concentration of idelalisib in B cells in vivo. All dosages of idelalisib tested suppressed calcium mobilization (Fig. 4A, quantified in Fig. 4B). Similarly, phosphorylation of AKT, BTK, and PLCγ are significantly reduced following exposure to all dosages of inhibitor (Fig. 4C). It is interesting that although we can achieve a complete abrogation of calcium mobilization with idelalisib, we observe only modest, albeit significant, decreases in phosphorylation of PLCγ. We investigated this phenomenon further to reveal a direct correlation between the strength of BCR stimulus used and the dosage of idelalisib required to suppress this response (Supplemental Fig. 3). These results may seem counterintuitive because of the dependence of calcium responses on PLCγ. However, previous studies indicate that PLCγ phosphorylation is not affected by SHIP-1 activation following FcγRIIB coaggregation with the BCR (J. C. Cambier, unpublished data). Additionally, previous work demonstrated that PLCγ interacts directly with the Lyn SRC family kinase during BCR signaling (65). This suggests that PLCγ activation and hydrolysis of PI(4,5)P2 may be a two-step process involving phosphorylation at the BCR receptosome, followed by PI(3,4,5)P3-dependent translocation to substrate rich sites in the plasma membrane. In this scenario, PLCγ phosphorylation would not be PI3K dependent. These findings lead to the conclusion that low dosages of a p110δ inhibitor that enforce anergy while sparing the Ab response has an inhibitory effect on early events in BCR signaling that are predicted to be dependent on PI3K activation.

As shown in Fig. 1, low-dosage idelalisib does not inhibit Ab responses in NOD mice. To investigate this further, we examined the effect of a range of idelalisib dosages on in vivo B cell responses to immunization. We adoptively transferred MD4 B cells loaded with dilution dye into C57BL/6 recipients, and, after allowing the cells to rest for 48 h, placed recipient mice on varying dosages of idelalisib-containing chow. Twenty four hours later, we immunized with HEL conjugated to SRBC, and analyzed the B cell response 5 d later (Fig. 5A). MD4 B cells from mice receiving vehicle control chow and 0.9375 mg/kg idelalisib–containing chow mounted similar IgM anti-HEL Ab responses and generated comparable HEL-specific ASCs per spleen (Fig. 5D–F). MD4 B cells in mice receiving 3.75 and 30 mg/kg idelalisib–containing chow prior to immunization mounted significantly reduced IgM anti-HEL Ab responses and generated a reduced number of ASCs per spleen in comparison with MD4 B cells from vehicle control cohorts (Fig. 5D–F). This dosage-dependent reduction in HEL-specific Ab is further reflected in the recovery of MD4 B cells in the spleens of mice from the various cohorts. Mice receiving 3.75 and 30 mg/kg idelalisib–containing chow had significant decreases in recoverable transferred cells, whereas recovered cells proliferated less than in controls (Fig. 5B, 5G, quantified in Fig. 5C). These data allowed clearer definition of dosages that enforce anergy of autoreactive B cells while sparing B cell responses to exogenous Ag.

T cells are essential components of Ab responses to most proteinaceous Ags, including autoantigens. T cells use the p110δ isoform, but the role of low-dosage inhibition of this isoform on T cell function has not been studied. Studies of p110δ knockout mice or functionally inactive p110δ mice have yielded conflicting results with respect to the requirement of this isoform for T cell responses (35, 36, 39, 40, 66–68). Autoantibody responses caused in this study by compromise of Ars/A1 anti-chromatin B cells is T cell dependent. Upon transfer into TCRα^−/− recipients, these B cells fail to proliferate, differentiate, and secrete autoantibody (A. Getahun and J. C. Cambier, unpublished data). This raises the possibility that idelalisib is mediating its effect by inhibiting T cell function. It is noteworthy, however, that if this were the case, the inhibitor should have been equally effective in inhibiting autoimmunity caused by B cell–targeted PTEN^−/+ × SHIP-1^−/+ and SHP-1^−/− conditions. Nonetheless, we set out to determine the consequence of low-dosage p110δ inhibition on T cell responses in vitro and in vivo.

Because we see a therapeutic effect of the inhibitor on VH125.NOD disease progression and survival, reduction in autoantibody responses in PTEN^−/+ × SHIP-1^−/+cells (yet comparable autoantibody responses by SHP-1^−/−), and MD4 B cells in mice receiving 0.9375 mg/kg idelalisib–containing chow, we chose to focus our analysis on effects of this dosage on T cell function. Fifteen nanomolar idelalisib (comparable to 0.9375 mg/kg in vivo DD), failed to inhibit calcium mobilization of CD4 T cells stimulated by TCR aggregation with biotin/anti-CD3 and avidin (Fig. 6A, quantified in Fig. 6B). To analyze in vivo CD4 and CD8 T cell responses, we adoptively transferred OT-II CD4 T cells or OT-I CD8 T cells into congenically mismatched recipients and immunized with OVA⁺P:IC, as represented diagrammatically in Fig. 6C, 6G, respectively. Five days postimmunization with OVA⁺P:IC, mice that received vehicle control chow and mice that received 0.9375 mg/kg idelalisib–containing chow proliferated [Fig. 6E (OT-II), 6I (OT-I), left panel; quantified Fig. 6E (OT-II), 6I (OT-I), right panel] and upregulated the activation marker CD44 to similar levels [Fig. 6F (OT-II), 6J (OT-I), left panel; quantified Fig. 6F (OT-II)/6J (OT-I), right panel]. The transferred OT-II and OT-I cells that proliferated most underwent the greatest upregulation of CD44 [Fig. 6D (OT-II), 6H (OT-I)]. Dosages of idelalisib that enforce anergy of autoreactive B cells, delay disease progression, and prolong survival in VH125.NOD mice do not affect CD4 or CD8 T cell responses in vitro or in vivo.

In this study, we report that autoimmunity caused by failed regulation of PI(3,4,5)P3 levels in B cells can be prevented by treatment with low dosages of PI3K p110δ inhibitor, dosages that neither inhibit autoimmunity caused by altered expression of the tyrosine phosphatase SHP-1 nor block Ab responses to exogenous Ag. Further, we show that, in VH125.NOD mice( which express reduced PTEN), low-dosage p110δ inhibitor therapy delays progression from the hyperglycemic state to diabetes while sparing the ability to mount Ab responses following immunization. Results speak to the likely success of precision approaches to therapy in which effects of specific risk alleles are mitigated by both qualitative and quantitative targeting of the offending gene(s).

Personalized or precision medicine aims to subset, and subsequently treat, patients based on the underlying etiology of disease (i.e., genetic polymorphisms), rather than grouping patients based on symptom similarity. This approach would harness genetic and expression information to stratify patients as predictable responders or nonresponders to a particular therapy based on the possession of specific risk-conferring alleles. It is estimated that only 10–20% of Genome-Wide Association hits are attributable to coding-region germline mutations, with the remainder driven by differential expression of proteins governed by noncoding-region mutations (69). Thus, heritable differences in promotors and enhancers can, by affecting protein levels, affect risk of disease. It appears that by affecting protein levels, differences in miRNA expression can also affect risk of disease. Increased expression of mir-7 and associated reduced expression of its target PTEN are associated with lupus (22). Altered expression of mir-155 and its target SHIP-1 have also been implicated in SLE (70, 71). We have recently reported decreased expression of PTEN along with increased regulator miRNAs in B cells from new-onset T1D patients. Finally, PTEN expression is reduced in B cells from NOD mice. PTEN and SHIP-1 are critical regulators of the PI3K pathway, which is important in signaling by a number of immune system receptors, most notably Ag and activating receptors for IgG Igs. Anergic autoreactive B cells are characterized by increased expression of PTEN relative to naive B cells (19), and as discussed below, PTEN is required to maintain anergy.

Our laboratory and others have shown that regulation of the PI3K pathway by the inositol phosphatases PTEN and SHIP-1 is crucial for maintenance of B cell tolerance. Browne et al. showed that PTEN is upregulated on anergic B cells in the MD4.ML5 HEL/anti-HEL mouse and is required to maintain anergy in that model. We subsequently showed that an adoptive transfer model that acutely induced deletion of PTEN or SHIP-1 in anergic anti-DNA B cells leads to rapid cell activation, proliferation, and differentiation into autoantibody-secreting cells. Similar effects are seen upon acutely induced deletion of SHP-1 (27). Acute induction of haploinsufficiency of either of these genes in anergic B cells has no adverse effects; however, haploinsufficiency of both PTEN and SHIP-1 leads to autoimmunity, suggesting that these inositol phosphatases function in the same regulatory pathway. This is not surprising given the fact that they attack the same substrate, PI(3,4,5)P3. Although acutely induced deletion of the tyrosine phosphatase SHP-1 leads to loss of anergy, induced haploinsufficiency does not. Most importantly, haploinsufficiency of both SHP-1 and SHIP-1 does not lead to autoimmunity, suggesting that they function in distinct pathways, both of which are critical for maintenance of anergy. Given this result, it is not surprising that compensatory inhibition of PI3K blocks autoimmunity caused by defective PI3K regulation, but not by defective regulation of tyrosine phosphorylation by SHP-1. This result demonstrates the principle that autoimmunity can be treated by specifically targeting the predisposing genetic defect.

PI3K exists as a heterodimer, comprised of a regulatory and catalytic subunit. The catalytic subunits p110α and p110β are ubiquitously expressed, whereas p110δ and p110γ are restricted to the lymphoid compartment. p110δ Knockout and catalytically inactive mice have demonstrated p110δ being the dominant isoform used within B cells, with redundant p110 isoforms capable of compensating in T cells for the lack of p110δ (40). Thus, it can be expected that idelalisib, a p110δ isoform-specific inhibitor with 110–453-fold more selectivity over other class I isoforms, should mediate its biologic effects by blocking PI3K p110δ activation in B cells, as indicated by its effectiveness when autoimmunity is driven by a B cell–specific defect and lack of inhibition of T cell function.

It is curious that idelalisib blocks development of autoimmunity under dosing conditions far below those needed to block growth of B cell tumors and Ab responses to exogenous immunogens. This presumably reflects the fact that autoimmunity can result from only partial dysregulation of PI(3,4,5)P3 levels and, therefore, be compensated by partial inhibition of PI3K.

Therapeutic effects of idelalisib in the NOD mouse model would not have been anticipated prior to our observation of NOD B cells expressing reduced levels of PTEN (29). Genetic analyses of NOD mice has revealed more than 30 Idd susceptibility loci, with the NOD-derived H-2^g7 MHC as a necessary, but not sufficient, component of disease. Although many dominant Idd loci have been described, the causal genes are technically difficult to identify, as even small intervals contain a large number of candidate genes. Further, disease risk is likely determined by concerted effects of multiple Idd loci. Interestingly, PTEN and SHIP-1 do not fall into identified Idd loci, nor do the miRNAs that regulate their expression, mir-7 and mir-155, respectively. Thus, we hypothesized this inherent inability to regulate the PI3K pathway in the NOD mouse would render our p110δ inhibitor effective in the VH125.NOD mouse. Results suggest that Idd loci may contain a previously unrecognized regulator of PTEN expression, as the known regulators of PTEN do not map to identified loci.

Future therapeutic approaches for T1D must have as their goal prevention of development of life-long dependence on insulin. More specifically, intervention must begin prior to establishment of insulin dependence and prevent progression to insulin dependence. With this in mind, studies described in this paper were designed to test the ability of low-dosage PI3Kδ inhibitor administration to prevent progression from the modestly hyperglycemic state to disease of such severity that insulin therapy is required (i.e., prior to irreversible β cell destruction). Results are consistent with utility of initiation of therapy when hyperglycemia and other markers indicate that clinical disease is imminent.

Traditional medicine has operated under the premise that established therapies will work for most patients, but perhaps not all, with a given disease. Personalized medicine intends to target disease while simultaneously reducing or alleviating collateral damage and risk. The future goal of medical professionals involved in treatment of autoimmune diseases should be induction of tolerance/enforcement of anergy and restoration of protective immune function. This will be achieved by harnessing genetic and expression information to predict patient response rate to a particular therapy. The studies performed in this study provide evidence that enforcement of the anergic state in a risk allele-dependent manner while sparing protective immunity in response to immunization can be achieved.

We thank Dr. Soojin Kim for managing our mouse colony. We thank the University of Colorado Immunology and Microbiology Department Flow Cytometry Core.

The authors have no financial conflicts of interest.