Dynamic Immune Phenotypes of B and T Helper Cells Mark Distinct Stages of T1D Progression

The natural history of type 1 diabetes (T1D) has been studied extensively with respect to development of islet-reactive autoantibodies, β-cell function, and metabolic markers of disease progression (1–3). This has paralleled an improved understanding of genetic risk factors and environmental factors that influence disease susceptibility. As a result, progress has been made in trials aimed at preserving insulin secretion in established T1D (1,4–6). However, to determine the optimal type and timing of immunotherapies to prevent and treat T1D (7), a better understanding of the immune mechanisms that drive preclinical disease in the at-risk population and interindividual immune heterogeneity will be required.

The strong genetic link with HLA class II alleles and other genes that participate in T-cell function underscores the importance of CD4⁺ T cells in T1D. In the NOD mouse, CD4⁺ effector T cells (Teffs) are required for diabetes progression (8). Increased frequencies of CD4⁺, T helper 17, and follicular T helper (Tfh) cells have been reported in both new-onset and established T1D (9–15). Tfh cells are also increased in autoantibody-positive (autoAb⁺) children with impaired glucose tolerance, suggesting the evolution of a pathogenic Tfh population poised to promote B-cell responses during disease progression (15). Functional consequences likely underlie these phenotypes, as CD4⁺ Teffs are resistant to regulatory T-cell (Treg) suppression (16,17) and exhibit altered responses to cytokines, having a blunted response to interleukin (IL)-2 (18,19) and an enhanced response to IL-6 (20), in established T1D.

Emerging data also implicate B cells in the development of the autoimmune T-cell response in numerous disease settings (21–23). In NOD mice, B cells are required for shaping productive CD4⁺ T-cell responses, via their capacity to process and present islet antigen via MHC class II and as the dominant antigen-presenting cell for self-reactive CD4⁺ T cells (21,24). Furthermore, genome-wide association studies and genotype–phenotype studies in human autoimmune diseases have identified several variant alleles that impact B-cell homeostasis, function, and tolerance checkpoints (25–31). The beneficial outcomes of interventions that target T cells (5,32) and B cells (4,33) provide a mechanistic framework for T- and B-cell–mediated autoimmune pathogenesis. They also indicate that a more complete understanding of the temporal evolution and cooperation of T- and B-cell phenotypes during the natural history of T1D is warranted.

The aim of this study was to define the temporal evolution of CD4⁺ Teff and B-cell phenotypes in T1D development and progression in at-risk subjects. Our results reveal distinct phenotypes in the T- and B-cell compartments at an early stage of autoimmunity, characterized by blunted IL-2 signaling in CD4⁺ Teffs, enhanced responses to IL-21 in the naive B-cell population, and an expansion of transitional B cells. As individuals progress toward clinical disease, we observed the acquisition of Teff resistance, a decrease in the B-cell response to IL-21, and attenuated B-cell receptor (BCR) responses. Our findings suggest that early tolerance checkpoints are altered in B cells, which may predispose to enhanced autoreactivity. This early change in B cells in disease may be potentiated through T-cell help that is driven by blunted responses to IL-2 in Teffs and an enhanced IL-21 response in the B-cell population. By comparison, later in disease, the Teff resistance to suppression is predominant, suggesting an acquired trait.

The study was approved by the Benaroya Research Institute (BRI) institutional review board, and all subjects gave written informed consent. Cohorts are described in Supplementary Tables 1–5. The Type I Diabetes TrialNet Pathway to Prevention (PTP) Trial (TN01 Trial, formerly the TrialNet Natural History Study; NCT00097292) has been previously published (34,35), as have TrialNet’s Rituximab trial (NCT00279305) (33) and the Immune Tolerance Network’s T1DAL trial (32). The subjects with established T1D and healthy individuals were participants in the BRI Immune-Mediated Disease Registry and Repository. All samples were assayed in a blinded manner. Frozen samples were used for all assays. Assays on each thawed sample were prioritized based on required cell number for analysis. Data from samples were excluded when viability was <40% and when data were not reliable due to low event count or responses were not detected in the same-day internal control.

IL-2/phosphorylated (p)STAT5 phospho-flow cytometry was performed as described previously (18). In brief, cells were activated with 25 IU/mL IL-2 for 10 min. For IL-21/pSTAT3 phospho-flow, thawed peripheral blood mononuclear cells (PBMCs) were rested in serum-free X-VIVO 15 medium for 1 h, washed with PBS, and stimulated with recombinant human IL-21 (0.1 ng/mL or 50 ng/mL) (Miltenyi Biotec, Auburn, CA) for 10 min. Cells for both cytokine stimulation assays were fixed with BD Phosflow Buffer I and permeabilized using BD Phosflow Buffer III prior to staining with antibodies against pSTAT5 (pY694), CD25, CD4, and CD45RO or pSTAT3 (pY705), CD3, CD4, CD45RA, CD8, and CD20 for IL-2 and IL-21, respectively. Teffs were defined as CD4⁺CD25^dim.

BCR signaling was performed as previously described (31). For BCR signaling, PBMCs were thawed, rested for 1 h at 37°C in RPMI supplemented with 1% human serum (Gemini Bio-Products, West Sacramento, CA), and then stimulated with F(ab′)2 fragment goat anti-human IgM (10 μg/mL) and F(ab′)2 goat anti-human IgD (10 μg/mL). Cells were fixed, permeabilized as above, and then stained using antibodies against CD27, CD20, and pPLC-γ2 (Y759).

Antibodies are listed in Supplementary Table 6. Data were acquired using the following flow cytometers: BD LSR-II (B-cell immunophenotyping), BD FACSCalibur (IL-2/pSTAT5 phospho-flow and BCR signaling), and BD FACSCanto II (IL-21/pSTAT3 phospho-flow) (BD Biosciences). All flow cytometric analysis was performed with FlowJo software (Tree Star, Inc.).

Teff resistance was determined using an in vitro Treg-mediated suppression assay (36,37). In brief, CD4⁺ T cells depleted of CD25^hi cells were isolated from PBMCs using a no-touch Miltenyi CD4 T Cell Isolation Kit II and positive Miltenyi CD25 Microbeads II prior to staining with carboxyfluorescein succinimidyl ester (Sigma-Aldrich). CD25⁺CD127^lo Tregs from a single healthy donor were sorted, expanded, and frozen as described and used a constant source of Tregs for all suppression assays as previously described (37). CD4⁺CD25^dim T cells (Teff) were cultured at 100,000 cells/well. Tregs were added at ratios of 1:4 and 1:8 (Treg/Teff) and Dynabeads CD3/CD28 T Cell Expander bead (Life Technologies) added at a ratio of 1:25 (beads/Teffs). On day 4, proliferation of live Teff was determined by carboxyfluorescein succinimidyl ester dye dilution after staining cells with CD25, CD4, and LIVE/DEAD Far-Red Dead Cell Stain Kit (Thermo Fisher Scientific).

Statistical analysis was performed using GraphPad Prism 7.04 software or R 3.4.3 software. False discovery rate was controlled using the Benjamini-Hochberg procedure, and P values <0.05 were considered statistically significant. Values were considered outliers when they exceeded 1.5 times interquartile range below the first quartile as determined using R or when detected using the ROUT method in Prism. To assess the significance of each patient feature while controlling for all others, IL-2 response and Teff resistance to Treg suppression were each modeled as dependent variables via simple and multivariable linear regression after each was logit transformed. In the simple regression models, age and number of autoAbs were used as the independent model features. For the multivariable regression, independent variables included: 1) fixed traits including HLA, sex, and age, 2) progression status (nonprogressor or progressor), number of autoAbs, and time point, and 3) clinical parameters: glucose (2-h oral glucose tolerance test [OGTT]) and area under the curve (AUC) C-peptide (2-h OGTT). B-cell immunophenotyping and IL-21 response analyses used multivariable linear regression with log₂-transformed response data and autoantibody-negative (autoAb^neg) first-degree relatives (FDRs) as the reference group and adjusted for age, glucose (2-h OGTT), AUC C-peptide (2-h OGTT), and number of autoantibodies. IL-21/pSTAT3 response was modeled using a linear mixed-effects model in which we modeled the repeated measures using a random effect for individual. Because age is a known covariate for IL-2 response and transitional B-cell homeostasis (38,39), we limited our univariate analyses of these data to subjects ≤18 years of age.

All at-risk individuals were participants in the Type 1 Diabetes TrialNet PTP Trial (34,35) and classified as autoAb^neg or autoAb⁺ FDRs. Samples were obtained at two time points 6–12 months apart. Two cohorts of the autoAb⁺ individuals were selected: autoAb⁺ progressors and autoAb⁺ nonprogressors based on whether they developed clinical T1D at or near the second time point. Due to limited sample volumes, not all assays were conducted on each individual. Thus, three independent PTP cohorts were used: cohort 1 for IL-2 response and Teff resistance (Supplementary Table 1), cohort 2 for B-cell immunophenotyping and BCR signaling (Supplementary Table 2), and cohort 3 for IL-21 response (Supplementary Table 3). Supplementary Table 4 provides a summary of the cohorts for age, sex, number of autoantibodies, and metabolic status, showing that they are well matched for age and sex.

We and others have reported reduced IL-2 response in CD4⁺CD45RO⁺ memory T cells associated with T1D and T1D-associated risk alleles (18,19,40,41). In this study, we assessed whether this phenotype precedes diagnosis in our risk-stratified cohort. We did not observe a diminished IL-2 response in CD4⁺CD25⁺ T cells (data not shown), but did find that the IL-2 response in memory CD4⁺CD25^dimCD45RO⁺ T cells was decreased in autoAb⁺ subjects compared with autoAb^neg FDRs, most significantly between autoAb⁺ nonprogressors versus autoAb^neg FDRs (Fig. 1A and Supplementary Fig. 1A). Within the CD4⁺ memory subset, decreased IL-2 signaling did not correlate with number of autoAbs (Fig. 1B), which was confirmed by a multivariable model using autoAb^neg FDRs as a reference group and adjusting for age. IL-2 response was negatively correlated with the presence of autoAbs, irrespective of the number of autoAbs or progressor status (autoAb⁺ nonprogressors: β coefficient: −0.428, P < 0.05; autoAb⁺ progressors: β coefficient: −0.601, P < 0.05). Together, these data demonstrate a decrease in IL-2 response among Teffs that is present early in the autoimmune process prior to or upon acquisition of autoantibodies.

We and others have previously shown that CD4⁺ Teffs are resistant to Treg suppression in established T1D (16,17,42–44). In this study, we tested whether Teff resistance occurs prior to T1D diagnosis using an in vitro suppression assay (Supplementary Fig. 1B). There was no significant difference in CD4⁺CD25^dim Teff suppression between autoAb^neg FDRs and autoAb⁺ subjects or within the autoAb⁺ cohort based on progression (Fig. 1C). However, we did observe a significant decrease in suppression between time 1 and time 2 in autoAb⁺ subjects with two or more autoantibodies. Stratification by number of autoantibodies across both time points revealed that percent suppression was significantly decreased in subjects with three autoAbs, as compared with autoAb^neg FDRs (Fig. 1D). Multivariable regression analysis using autoAb^neg FDR as the reference group and adjusting for age, glucose, and C-peptide confirmed this significant association between development of autoantibodies and percent suppression and revealed a significant relationship between Teff resistance and autoAb⁺ progressor status. Increased Teff resistance (decreased Teff suppression) was positively correlated with number of autoantibodies (β coefficient: 0.259; P < 0.001). Together, these findings suggest that Teff resistance is an acquired trait that, in part, correlates with increased autoimmunity.

We have previously reported that individuals with long-standing T1D display an expanded population of transitional B cells and anergic (B_ND) B cells, a decrease in CD19⁺CD27⁺ memory B cells, as well as an attenuated response to BCR activation (29). More recently, Smith et al. (45) reported the loss or phenotypic alteration of high-affinity insulin autoantigen-binding B cells from the peripheral B_ND anergic B-cell compartment of autoAb⁺ individuals with prediabetes, subjects with new-onset T1D, and some autoAb^neg FDRs that correlated with a transient loss of total anergic B_ND B cells. Because these phenotypes reflect altered B-cell tolerance, and pathogenic autoantibodies may result from B- and/or CD4⁺ T-cell–driven processes, we first determined whether altered B-cell homeostasis precedes established disease. We found a modest decrease in naive B-cell frequency in autoAb⁺ progressors as compared with autoAb^neg FDRs with no significant differences in the memory and total B-cell frequencies between the cohorts (Supplementary Fig. 2). However, transitional B cells were increased in the autoAb⁺ cohorts as compared with autoAb^neg FDRs at time 1, with a similar trend at time 2 (Fig. 2A and data not shown). In contrast, B_ND anergic B cells were significantly decreased in the autoAb⁺ cohorts compared with autoAb^neg FDRs most prominently at time 2 (Fig. 2B and data not shown). Multivariable regression analyses using autoAb^neg FDRs as the reference group and controlling for age and clinical variables confirmed these results. In this model, transitional B cells were increased in autoAb⁺ nonprogressors (β coefficient: 0.47; P = 0.065) and autoAb⁺ progressors (β coefficient: 1.075; P = 0.001) at time 1, and B_ND anergic B cells were decreased in autoAb⁺ nonprogressors (β coefficient: −1.202; P < 0.01) at time 2. Glucose, C-peptide, and number of autoantibodies were not significantly correlated with transitional B cell frequency in either autoAb⁺ nonprogressors or progressors, when controlling for age (data not shown). In contrast, glucose was a significant variable in the best fit model for B_ND B cell frequency at time 2 and was negatively correlated with percentage of B_ND B cells in autoAb⁺ progressors (β coefficient: −0.628; P = 0.012). Collectively, these data reveal homeostatic alterations in the B-cell compartment at peripheral tolerance checkpoints during progression to T1D, suggesting that B-cell–intrinsic mechanisms may contribute to an increased risk for the development of autoreactive B-cell responses during the natural history of disease.

Our prior studies demonstrated that BCR proximal signals are blunted in naive and memory B cells in established T1D (29). In this study, we assessed proximal BCR responses by measuring phosphorylation of PLCγ2 in the risk-stratified cohorts (Supplementary Fig. 3A–C). Fold change pPLCγ2 in B cells was not different between autoAb^neg FDRs and autoAb⁺ subjects, but a significant decrease was seen in autoAb⁺ progressors compared with autoAb⁺ nonprogressors. This trait was observed at time 1 (Fig. 2C), persisted to time 2 (data not shown), and was most prominent within the naive B-cell compartment (Supplementary Fig. 3B). Taken together, we believe that these data suggest that the BCR response is blunted as one progresses to clinical disease and that this phenotype is more prominent in mature naive B cells.

Given the role of IL-21 signaling in promoting the maturation of B cells and antibody production (46,47), we hypothesized that enhanced B-cell responsiveness to IL-21 may be associated with the development of autoantibodies in T1D. To test this, we analyzed IL-21–mediated STAT3 phosphorylation across all time points and stratified by progressor status or number of autoAbs. We observed a significant increase in fold change pSTAT3 in naive B cells from autoAb⁺ nonprogressors compared with autoAb^neg FDRs and in subjects with two autoAbs versus zero or one autoAb (Fig. 3A–C and Supplementary Fig. 4A and B). A multivariable mixed-effects model, controlled for age and sex, revealed a significant correlation between IL-21 response and progressor status, further confirming the enhanced IL-21 response in at-risk autoAb⁺ nonprogressors (Fig. 3A). This enhancement did not extend to autoAb⁺ progressors or to subjects with three or more autoAbs, suggesting that this phenotype is a dynamic trait over the course of disease development (Fig. 3B and C). In support of this, paired analysis of fold change pSTAT3 in naive B cells showed no differences between times 1 and 2 in autoAb⁺ nonprogressors, whereas a significant decline was observed in autoAb⁺ progressors between time points (Fig. 3C). In order to determine whether this decline was due to changes in IL-21R expression, we assessed IL-21/pSTAT3 and IL-21R surface expression levels in parallel in a separate cohort of autoAb⁺ progressors (Supplementary Table 5), distinct from cohort 3. In this new cohort, the decline in IL-21 response between times 1 and 2 was replicated, but IL-21R expression levels were not significantly different between time points (Supplementary Fig. 4C and D). This suggests that the dynamic response to IL-21 in naive B cells from autoAb⁺ at-risk individuals is not solely due to changes in IL-21R expression. Further, IL-21 response in naive B cells from a cohort of 100 subjects with established T1D was significantly decreased as compared with well-matched healthy individuals (Fig. 3D). This was also evident by a decreased frequency of pSTAT3-positive cells (data not shown). This result implies that decreased response to IL-21 in naive B cells may be an acquired feature that persists in established disease. Thus, these data reveal enhanced naive B-cell responses to IL-21 that occur early in autoimmunity and decline near onset of clinical diabetes.

Given our data suggesting that blunted BCR signaling phenotype and Teff resistance are acquired traits, although IL-2–induced pSTAT5 was fixed, we examined how these traits were influenced by an intervention that slowed the loss of C-peptide. We assessed samples from new-onset T1D treated with rituximab (33). We did not observe any changes in IL-2–induced pSTAT5, but did observe a significant albeit modest decrease in the suppression of Teffs at week 12 in individuals treated with rituximab, the time at which the initial improvement in mean C-peptide levels from baseline was observed (Supplementary Fig. 5A).

As has been reported (33), rituximab treatment resulted in a decreased frequency of total, mature naive, and memory B cells and increased frequency of transitional B cells that did not differentiate responders from nonresponders (Supplementary Fig. 5B and data not shown). BCR response at baseline did not determine clinical response to rituximab, as there was no significant difference in mean basal pPLCγ2 at baseline or after treatment between responders and nonresponders (Supplementary Fig. 5C). However, in rituximab-treated responders, the BCR response in total B cells at week 52 was significantly increased when compared with baseline (Fig. 4A), which did not correlate with IgM/IgD expression (data not shown). Together, these data suggest that BCR signaling “recovered” in responders to rituximab, whereas other immune features remained unchanged.

The source of the recovery of B-cell signaling may be due to direct effects of rituximab on B cells or the result of an altered immune milieu in individuals who respond to treatment. At week 52, we did not observe significant differences in composition of the B-cell compartment between responders and nonresponders (data not shown). To address whether improved BCR response is a function of being a clinical responder per se, independent of the therapeutic intervention, we assessed change in BCR response in the T1DAL trial, which demonstrated that alefacept-mediated depletion of pathogenic T cells could preserve mean 4-h AUC C-peptide and improve glycemic control (32). In contrast to our rituximab findings, BCR response was not significantly improved in B cells from alefacept-treated T1DAL participants who had sustained 2-h AUC C-peptide at 6 months (clinical responders, as defined by the rituximab trial [33,48]) (Fig. 5B). These combined findings indicate that alterations in BCR response in rituximab-treated subjects are likely due to a direct effect of the drug on B-cell depletion and that it may be therapeutically effective in a subgroup of individuals for whom B-cell hyperresponsiveness is a driver of early disease.

The goal of this study was to provide the community with a broader understanding of the immune mechanisms involved in development of T1D. Using three independent longitudinal at-risk cohorts, we provide new insight into how these phenotypes evolve in the context of each other, thus providing a sequential model for better understanding the immune landscape of developing T1D (Fig. 5). The fixed trait of reduced IL-2 response in autoAb⁺ FDRs precedes homeostatic shifts in the immature and mature naive B-cell subsets, and altered IL-21 and BCR signaling, with later acquisition of increased Teff resistance.

In at-risk T1D, it appears that certain B-cell phenotypes are consistent with those found in established T1D, whereas others are unique to the at-risk population. For example, the expanded transitional B-cell compartment in autoAb⁺ subjects as compared with autoAb^neg FDRs may reflect early B-cell selection defect(s) that result in an increase in autoreactive B cells in the periphery, as described in established T1D (28). Notably, our finding of reduced anergic B_ND B cells in autoAb⁺ subjects as compared with autoAb^neg FDRs, together with the significant correlation between B_ND frequency and glucose levels, is consistent with the results of Smith et al. (45,49), who reported transient reduction of B_ND B cells among total B cells in the periphery of subjects with prediabetes and new-onset T1D. More recently, this group found that insulin-binding B cells were polyreactive and exhibited features associated with autoreactivity and autoantibodies (49). Interestingly, this study also reported that high-risk HLA alleles and a subset of non-HLA risk alleles were associated with loss of B-cell anergy. Given these associations, we explored these relationships in our at-risk cohort. The presence of high-risk HLA class II alleles (DR3, DR4, or DR3/DR4) did not associate with total B_ND B cells in autoAb^neg FDRs, autoAb⁺ nonprogressors, and autoAb⁺ progressors at either time point, including when glucose was included in the model. In addition, HLA class II genotype did not affect the difference in frequency of B_ND B cells between autoAb^neg FDRs and autoAb⁺ progressors. For the non-HLA risk alleles, we focused on PTPN22 but were limited by the small number of subjects who carried the risk variant. However, when restricting our analysis to subjects who carry the nonrisk allele of PTPT22, we found trends suggesting an association with increased transitional B cells and blunted BCR signaling in autoAb⁺ progressors (data not shown). Collectively, these findings implicate breach of anergy as a mechanism predisposing at-risk individuals toward autoimmunity, a mechanism that is potentially influenced by extrinsic and/or genetic factors.

Previous human studies have suggested that BCR hyporesponsiveness leads to increased numbers of self-reactive transitional B cells in the setting of T1D (29,50), thus our observation that blunted BCR response is an acquired phenotype of at-risk subjects near clinical diagnosis was unexpected. Early-stage BCR responsiveness in autoAb⁺ nonprogressors, together with enhanced IL-21 response, implies a cooperative mechanism to promote the maturation of autoreactive B cells and production of autoantibodies. Our findings further implicate T-cell–B-cell cooperative interactions in the development of disease. Reduced IL-2 response may impact the balance between tolerance and immunity by impairing Treg fitness while promoting expansion of Tfh and T helper 17 cells, all of which have been described in subjects with long-standing T1D (12,18,51). In this study, the blunted IL-2 response was evident in autoAb⁺ FDRs, indicating that this may be a fixed feature in T1D driven in part by early inflammatory and genetic factors. This phenotype in the at-risk cohort was most significant in CD4⁺ memory T cells and was not observed in Tregs. In addition to its role in Treg development and stability, IL-2 has also been shown to inhibit Tfh-cell differentiation (12,52,53). The presence of impaired IL-2 responses in CD4 T cells at the time of autoAb development suggest that it may promote the development of Tfh and bidirectional B-cell–T-cell interactions in the drive toward islet autoAb development. This is supported by studies in T1D that demonstrate a correlation between IL-2 response and Tfh frequency and studies that have demonstrated increases in circulating Tfh in the at-risk population (12,15). Our study also revealed an enhanced B-cell response to IL-21 that is present in at-risk autoAb⁺ FDRs and attenuated as these individuals progress toward clinical disease. This is consistent with an early priming of naive B cells for the generation of pathogenic B-cell effectors, whereas the dampened response may reflect an exogenous effect of metabolic or inflammatory factors or a compensatory mechanism that influences IL-21 responsiveness. As we found with BCR signaling, augmented IL-21 response is not seen in established T1D. An outstanding question is whether autoAb^neg FDRs display altered IL-21 response over time. Although this was not specifically addressed in this study, we show that individuals with a single autoantibody across both time points did not display enhanced IL-21 response compared with autoAb^neg FDRs. Further, previous studies demonstrate that positivity for two or more autoAbs, but not zero to one autoAbs, identifies a subgroup of patients with increased Tfh activation and greater treatment efficacy with rituximab at disease onset (15). Based on these findings, we would predict that interventions targeting IL-21 might be best used in prevention in autoAb⁺ at risk, but not in established disease.

Tregs have a well-established role in protection from T1D (8,54), and resistance of Teffs to Tregs has been well described in established T1D (16,17). In this study, Treg function was not assessed due to limitations of sample quantities, but the response of Teff to Treg was measured. In addition, although we were unable to match our cohorts for HLA due to the enrichment of HLA DR3 and DR4 alleles in autoAb⁺ subjects as compared with autoAb^neg FDRs in the TrialNet cohort, we did examine the influence of HLA class II genotype with respect to our findings. Using a multivariable logistic model with HLA included in the model, we found no significant association between IL-2/pSTAT5 and HLA DR3, DR4, or DQ0602. However, we did observe that the presence of DQB1*0602 was negatively associated with Teff resistance. Although we were unable to address the role of HLA in disease progression in autoAb⁺ subjects because the majority of HLA DQB1*0602 subjects in our study were in the autoAb^neg FDR cohort, this finding does suggest that the protective genetic role for DQB1*0602 may in part be due to responsiveness of Teff to Treg. Importantly, Teff resistance was still significantly increased in autoAb⁺ progressors when HLA was included in our modeling. In the at-risk cohort, we found that Teff resistance was acquired in FDRs positive for two or more autoAbs, consistent with the acquisition of Teff resistance with age in the NOD mouse (42,43). This finding suggests that the mechanism(s) that drive Teff resistance may be due to increased inflammatory signals, consistent with studies showing that Teff resistance is often most pronounced during disease flares (55) or at sites of inflammation (56–58) in other autoimmune diseases. One factor not tested in our study is a failure of Treg function itself, which, if aberrant early in disease, could contribute to an inflammatory environment, which would promote Teff resistance.

Consistent with our finding that Teff resistance is acquired in T1D, a recent study by Ihantola et al. (44) of autoAb⁺ subjects with prediabetes versus newly diagnosed and long-standing T1D confirmed the presence of the Teff resistance phenotype at later stages of autoimmunity and in established disease, but not in the at-risk population. Importantly, the authors demonstrate that the resistance phenotype in T1D is mediated via STAT3 signals, further linking Teff resistance to the proinflammatory cytokine milieu (20). Together with our findings, these studies suggest that interventions that target Teff resistance mechanisms may be most successful after the development of a single autoantibody, with the additional goal of preventing epitope spreading and the acquisition of additional islet autoAb specificities. Importantly, a source for the development of Teff resistance may be an earlier failure of Treg themselves, which would support the use of Treg therapy in T1D, either prior to the development of Teff resistance or in combination with interventions that target Teff resistance directly.

We also addressed whether the observed B-cell phenotypes predicted response to an intervention targeting B cells. Although we were unable to find a variable that predicted which individuals retained C-peptide, this analysis yielded several important observations. We observed improvement in BCR signaling among the individuals who retained C-peptide with rituximab treatment; by contrast, this was not observed in alefacept-treated subjects, indicating that the change was not due to the improved metabolic parameters, but was instead unique to individuals who responded to B-cell depletion with rituximab. This suggests that the enhanced BCR signaling that occurs after rituximab therapy may promote a reset of the native BCR signaling as the B-cell compartment repopulates. The fact that we see this only in individuals who respond favorably to the therapy may indicate that these individuals have a disease that is more predominantly driven by B cells. Accordingly, the enhanced BCR signaling may not reflect an improvement in the B-cell response; rather, it may return the B cells to a more potentially pathogenic, responsive state.

As a community, we have an ever-growing understanding of the genetic factors that predispose to T1D and features of the immune response that are altered in individuals with T1D. However, in order to develop effective strategies to intervene and prevent disease, we must understand these factors in the context of the trajectory of disease progression. In this study, we have used longitudinal samples from at-risk cohorts to demonstrate that distinct immune phenotypes arise at different times to diagnosis and, importantly, may be transient. Our novel findings provide a more nuanced understanding of the immune processes that precede disease and, importantly, complement the mechanistic insights gained to date from prevention studies (59). For example, early stable changes in IL-2 responsiveness as reported in this study that precede or coincide with transiently altered B-cell responses may inform selection of subjects for sequential immune therapies that first target B cells with rituximab, followed by specific interventions that target enhanced B-cell and Teff responses to promote a more sustained preservation of C-peptide. More focused longitudinal studies of FDRs before and after seroconversion will further distinguish underlying autoimmune mechanisms (genetic risk vs. early autoimmune inflammation), with implications that will inform optimal timing of interventions.

Acknowledgments. The authors thank Dr. Cate Speake (BRI, Seattle, WA) for helpful discussion about the manuscript and longitudinal studies in at-risk populations and Dr. Matt Dufort (BRI, Seattle, WA) for assistance with statistical modeling. The authors also thank the investigators and staff of the BRI Translational Research Program and BRI Diabetes Research Program for recruitment of healthy control subjects and subjects with established T1D, respectively, as well as the BRI Translational Research Clinical Core for sample processing and handling. The authors acknowledge the support of the Type 1 Diabetes TrialNet PTP Study Group, which identified study participants and provided samples and follow-up data for this study.

Funding. This work was supported by National Institutes of Health (NIH) grants RC4-DK-090796, U01-AI-101990-03, and DP3-DK-104466 to J.H.B. and a JDRF Career Development Award (3-2012-205) to S.A.L. The Type 1 Diabetes TrialNet PTP Study Group is a clinical trials network funded by the NIH through the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Allergy and Infectious Diseases, and the Eunice Kennedy Shriver National Institute of Child Health and Human Development through cooperative agreements U01-DK-061010, U01-DK-061034, U01-DK-061042, U01-DK-061058, U01-DK-085465, U01-DK-085453, U01-DK-085461, U01-DK-085466, U01-DK-085499, U01-DK-085504, U01-DK-085509, U01-DK-103180, U01-DK-103153, U01-DK-085476, U01-DK-103266, U01-DK-103282, U01-DK-106984, U01-DK-106994, U01-DK-107013, U01-DK-107014, and UC4-DK-106993 and JDRF. Research specimens from the ITN045AI T1DAL trial reported in this study were provided by the Immune Tolerance Network and supported by the National Institute of Allergy and Infectious Diseases of the NIH under award number UM1-AI-109565.

The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the NIH or JDRF.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. T.H. designed and performed B-cell experiments, analyzed data, and wrote the manuscript. S.A.L. designed T-cell experiments, analyzed data, and wrote the manuscript. P.L.S. performed T-cell experiments and analyzed data. A.B. and A.F. performed B-cell experiments and analyzed data. M.T. performed T- and B-cell experiments and analyzed data. A.M.H. contributed to data analysis and wrote the manuscript. K.C. and C.G. contributed to experimental design and reviewed and edited the manuscript. M.T.M. and E.W. performed data analysis including statistical modeling. D.J.R. conceived the project and contributed to experimental design. J.H.B. conceived the project, designed experiments, analyzed data, and wrote the manuscript. The Type 1 Diabetes TrialNet Study Group provided the samples and associated clinical data and contributed to the writing of the manuscript. J.H.B. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in poster form at the Federation of Clinical Immunology Societies (FOCIS) 2018 Annual Meeting, San Francisco, CA, 20–23 June 2018.