Increased T cell proliferative responses to islet antigens identify clinical responders to anti-CD20 monoclonal antibody (rituximab) therapy in Type 1 diabetes

Abstract

Type 1 diabetes mellitus (T1DM) is believed to be due to the autoimmune destruction of β-cells by T lymphocytes, but a single course of rituximab, a monoclonal anti-CD20 B lymphocyte antibody can attenuate C-peptide loss over the first year of disease. The effects of B cell depletion on disease associated T cell responses have not been studied. We compare changes in lymphocyte subsets, T cell proliferative responses to disease- associated target antigens, and C-peptide levels of participants that did (responders) or did not (non-responders) show signs of β-cell preservation one year after rituximab therapy in a placebo-controlled TrialNet trial. Rituximab decreased B lymphocyte levels after 4 weekly doses of mAb. T cell proliferative responses to diabetes –associated antigens were present at baseline in 75% of anti-CD20- and 82% of placebo-treated subjects and were not different over time. However, in rituximab-treated subjects with significant C-peptide preservation at 6 months (58%), the proliferative responses to diabetes associated total (p=0.032), islet-specific (p=0.048), and neuronal auto-antigens (p=0.005) increased over the 12 month observation period. This relationship was not seen in placebo treated patients. We conclude that in patients with T1DM, anti-B cell mAb causes increased proliferative responses to diabetes antigens and attenuated β cell loss. The way in which these responses affect the disease course remains unknown.

Introduction

Type 1 diabetes mellitus (T1DM) reflects the progressive, T cell-mediated destruction of insulin producing β-cells in the pancreatic islets of Langerhans. (1, 2). The current model of pathogenesis has deep roots in studies of the spontaneously diabetic NOD mouse where T cells, reactive to islet and neuronal antigens can adoptively transfer disease in the absence of other accessory cells (3, 4). Similar T cell autoreactivities can be identified in NOD mice and in patients with T1DM(5) (6). Although nearly always measured in the peripheral blood of patients, i.e. the systemically recirculating immune compartment, these T cells likely reflect the diabetic pancreas immune attack, as their specificities are similar in NOD mice where pathogenicity can be confirmed by adoptive disease transfer. Direct demonstration of pathogenicity has been impossible in humans, but in mouse models, β cell killing by T cells has been shown(7).

We previously validated the ability of cellular assays that measured T cell proliferative responses to diabetes-associated target antigens to distinguish responses in patients with T1DM from healthy control subjects(8, 9). Target antigens included those also targeted by diabetes-associated autoantibodies, such as proinsulin, insulin, GAD65, IA-2, ICA69 as well as neuronal antigens that are (e.g. S100β) or are not shared with β cells (e.g. glial fibrillary acidic protein, GFAP)(10). These cellular responses were identified in 60% of patients within the first 12 months after diagnosis of T1DM but in only 31% of healthy controls (8). A possible hierarchy of T cell targeted antigens in T1DM and the relationship of such T cell pools to disease progression are still under study.

Although much attention has been devoted to diabetes-associated (but not tissue toxic) autoantibodies, a pathogenic role for B lymphocytes emerged from the disease protection observed in B cell deficient NOD congenics (11–13), later linked to unique B lymphocyte mediated antigen presentation to T cells (14, 15). Recent interest focused on the role of B cells late in prediabetes, where B cell depletion with anti-CD20 monoclonal antibody (mAb) reversed new onset T1D in a subset of treated NOD mice(16). This observation was translated to patients in the randomized, placebo-controlled TrialNet Rituximab (anti-CD20) trial (17). This trial compared the effects of 4 weekly rituximab or placebo injections, in new onset patients, on C-peptide preservation and metabolic measures, including hemoglobin A1c levels and insulin usage. All parameters were improved by rituximab treatment for as long as 1 year after study entry. The lowest B lymphocyte counts were measured at the earliest time point, 5 weeks after initiation of therapy, and by 12 months the circulating B lymphocytes had recovered to 69% of the baseline levels. Serum levels of IgG were not significantly different in the control and drug treated subjects at that time, but the levels of serum IgM were still reduced 1 year after rituximab therapy.

The mechanisms of anti-CD20 effects are not clear, since, ultimately disease and disease progression are T cell dependent(17). To determine how B cell depletion with anti-CD20 mAb impacted disease associated cellular immune responses, we compared T cell subset distributions and autoreactivity profiles in rituximab recipients that had positive C-peptide responses with those unresponsive to treatment.

Materials and Methods

Subjects and clinical samples

A detailed description of the TrialNet Rituximab trial (NCT00279305) and the baseline characteristics of the study subjects have been published (17). Briefly, 87 subjects were randomized in a 2:1 ratio to rituximab or placebo treatment of which 81 formed the intention to treat population, 78 of whom completed the 1 year mixed meal tolerance test (the basis for the primary outcome, the AUC). Based on the change in the AUC of the C-peptide response from baseline to 6 months, each participant was designated as a C-peptide responder or non-responder.

Of the ITT cohort, 75 (93%) were included in the flow analysis and 80 (99%) in the T cell proliferation studies described herein. Samples were collected for flow cytometry at baseline (prior to study drug dosing), and 5 wks, 3, 6, and 12 months after the first dose of study drug. T cell assays were performed at baseline and 6 and 12 months after the first dose of study drug. Institutional Review Board approval was obtained for these measurements as part of trial participation.

Flow cytometric analysis

Whole blood was shipped overnight at ambient temperature to the Immune Tolerance Network (ITN) Flow Cytometry Core (Roswell Park Cancer Institute, Buffalo, NY). Antibodies used in this study were purchased from BD Biosciences Pharmingen (San Jose, CA). Standard methods were used for staining of cells including blocking Fc receptors with mouse IgG. Cocktails of fluorochrome-labeled mAbs to CD3, CD4, CD8, CD62L, and CD25 were used and the fluorometric analysis was performed using a FACSCanto (BD BioSciences, San Jose, CA) flow cytometer. The flow cytometry data were analyzed using WinList software.

T cell proliferation (TCP) assay

Peripheral blood mononuclear cells were isolated by Ficoll-Hypaque gradient centrifugation from fresh blood samples shipped to the T Cell Core Laboratory at the Hospital For Sick Children, Toronto, Canada. The washed cells were seeded (1×10⁵/well) into flat bottom microplates containing test or control antigens in 200 µL in serum- and protein-free Ex-Cell Hybri-Max medium (Sigma, St. Louis, MO). Test antigens were used at 1–4 concentrations (0.1–10 µg), depending on cell yields. Ten units of recombinant human IL-2 were added to all test wells (8, 9, 18). Depending on the blood volume sent and viable cell yields not all antigens could be used in all tests. Cultures were incubated (37°C, 5% CO₂) for five days, receiving a ³H-thymidine pulse (1µCi) for the last 18 hours. Cultures were then harvested and counted with a β counter. To compare results from different donors, results were transformed into stimulation indices (SI, cpm test antigen/cpm antigen free control cultures). A SI ≥ 1.5 was considered a positive response (10). Positive and negative controls included PHA, tetanus toxoid, and anti-CD3 mAb and actin, Cyc, HH, BSA193, and OVA respectively. Test antigens were classified into three different groupings islet: Tep69, GAD, GAD555, and proinsulin; neuronal S-100, GFAP, MBP, and EX2; and milk: casein, BLG, BSA, and ABBOS (19).

Because of the concern that the in vivo B lymphocyte depletion present only in the rituximab group could have a confounding impact on the in vitro T lymphocyte responses, in preliminary studies, we compared the SIs of samples in which B cells from all samples were depleted ex vivo with magnetic beads prior to culture. Depletion of B cells was confirmed by flow cytometry in a subset of samples. We found that the SIs of samples in which B cells were depleted prior to culture to the undepleted sample, were largely unaffected by B cell depletion (Supplementary Table 1). Therefore, we have reported results from the cultures that were studied without further manipulation of the cells. For each sample, a T cell reactivity score was generated (sum of all positive responses to test antigens). An overall T cell score of 4 or larger was considered evidence for the presence of autoimmunity in a given sample.

Statistical analyses

Study investigators, flow cytometry and T cell laboratories were masked to treatment assignment of each subject. We compared the groups that were treated with rituximab versus placebo and those who were classified as a C-peptide “responder” versus “non-responder” to the drug treatment. The AUC of the C-peptide values over the two hours of the MMTT was calculated using the trapezoidal rule including the time 0 and 2 hour values and the AUC mean C-peptide (pmol/ml) was obtained as AUC/120. The within-subject coefficient of variation (CV) of the AUC mean C-peptide was 0.097 from 2 repeat MMTT assessments conducted within 3–10 days from the MMTT-GST Comparison Study(20). A subject was classified as a C-peptide responder if the AUC mean increased from baseline to 6 months, or decreased by less than the within-subject CV of 0.097. If the subject’s AUC decreased at 6 months and the CV was > 0.097, the subject was classified as a non-responder.

The data from flow cytometry were analyzed by separate ANCOVA models for each cell population at each time point adjusted for baseline flow, age, and sex. SI sums were calculated in groups of antigens that were thematically clustered, and divided by the number of antigens in the group to determine a SI group mean. The T-cell stimulation index (SI) and positivity (reactivity) at 6 months and 12 months were examined using a separate regression model for each antigen or antigen grouping to estimate the change in SI response from baseline by treatment group and by responder status. Logistic regression models were used to examine whether measures of T-cell reactivity at each time point were predictive of responder status with an adjustment for baseline. The association between T-cell reactivity and quantitative C-peptide over time was analyzed using a repeated measures regression model. Least squared means with 95% confidence limits are presented except for baseline continuous variables in which the mean±SD is shown. The %change was calculated by dividing the values at 6 months by the baseline. A Wilcoxon test was used to compare the number of lymphocytes in each group.

Results

Study population

The demographics of the study cohort within treatment groups and those designated as C-peptide responders and non-responders are shown (Table 1). As reported recently (17) the C-peptide responses increased at 3 months in the rituximab treated group whereas placebo treated subjects showed a decline of C-peptide responses(p=0.038). After 6 months, there was a parallel decline in both study groups, but a significant difference remained between the groups in average responses over 12 months (p=0.0013).

Table 1.

Patient characteristics of the rituximab versus control treatment groups in the intention-to-treat cohort, and those classified as responders versus non-responders at 6 months of follow-up^*.

	Drug treated (n=52)*		Placebo treated (n=29)
	Responder (n=30)	Non-responder (n=21)	Responder (n=11)	Non-responder (n=18)
Age (years)	19.1 ± 8.8	19.8 ± 8.9	19.6 ± 9.5	16.0 ± 6.8
Sex (%male)	20 (66.7%)	13 (61.9%)	5 (45.5%)	12 (66.7%)
Time since diagnosis (days)	76±20	87±25	81±20	86±19
Baseline AUC mean C-peptide (pmol/ml)	0.77 ± 0.37	0.79 ± 0.49	0.92 ± 0.48	0.61 ± 0.21
Frequency of biochem autoabs at baseline
1	6 (20%)	3 (14.3%)	2 (18.2%)	1 (5.6%)
2	11 (36.7%)	4 (19.1%)	3 (27.3%)	4 (22.2%)
3	7 (23.3%)	8 (38.1%)	3 (27.3%)	5 (27.8%)
4	6 (20%)	6 (28.6%)	3 (27.3%)	8 (44.4%)
HLA-DR3/4/3,4
Absent	1 (3.3%)	2 (9.5%)	0 (0%)	1 (5.6%)
DR3	8 (26.7%)	4 (19.1%)	5 (45.5%)	2 (1.1%)
DR4	8 (26.7%)	6 (28.6%)	2 (18.2%)	4 (22.2%)
DR3/DR4	13 (43.3%)	9 (42.9%)	4 (36.4%)	11 (61.1%)
Number of CD3+cells at baseline	1469 (1253, 1685)	1654 (1399, 1910)	1448 (1130, 1765)	1543.1 (1300, 1786)

^*One rituximab subject was not evaluated at 6 months and could not be classified as a responder or non-responder.

^†Fisher’s exact test and Wilcoxon rank sum test were used for treatment and responder comparisons.

Means and standard deviations are reported for continuous variables. For categorical variables, the number and percent are reported. AUC denotes the area under the curve. AUC mean denotes the corresponding mean in pmol/ml computed as AUC/(120 min).

Based on the observed change in C-peptide responses relative to the coefficient of variation (CV) of repeated measurements, 58% of the subjects in the rituximab treated group were responders. The age and sex distribution, prevalence of autoantibodies at baseline, HLA genotypes, and baseline MMTT C-peptide responses were similar between responders and non-responder groups (Table 1).

Analysis of lymphocyte subsets by flow cytometry

In rituximab recipients, the total lymphocyte counts had fallen to 87.6% of baseline by 3 months. (Figure 1A, (p<0.001), but recovered later as measured at 6 and 12 months. The decline was accounted for by depletion of CD19+ B cells. With some fluctuation, the placebo group remained at an average of 107.7% of the baseline lymphocyte count over the first year. Using ANCOVA to compare treatment groups, the number of CD3+, CD4+, CD8+, activated CD4+CD25+ cells or Tregs were not significantly different in the rituximab and placebo groups over the 1 year study period (Figures 1B, C, D–F, p values > 0.05).

Figure 1. Analysis of lymphocyte subsets in patients treated with rituximab or placebo.

The number of total lymphocytes, CD3+, CD8+, CD4+, activated CD4+ (CD4+CD25+), and Tregs (CD4+CD25+CD62Lhi) cells and 95% confidence intervals for the mean (dashed vertical bar) are shown for both treatment arms at the indicated time points (dashed line = rituximab, solid line=placebo). The number of lymphocytes were significantly reduced at 5 and 12 weeks in the rituximab treated compared to the placebo treated group (p<0.01). The number of circulating CD3+, CD4+, T cells was higher and lower in the rituximab treated group at 5 (p<0.03) and 12 (p<0.02) weeks, respectively, but overall the differences in the total CD3+ T cell counts in the two treatment arms were not statistically significant (p=0.89), The number of circulating CD4+ T cells was significantly increased in the rituximab treated group at 5 and 12 weeks (p< 0.05) but were overall not different between the treatment arms (p=0.65). The number of circulating CD8+ T cells was significantly lower in the rituximab treated group at 12 weeks (p<0.02) but not statistically different over the 12 months (p=0.68). The numbers of subjects per group at each time point were the following (Rituximab, Control): Baseline (52, 29), Week 5 (47, 22), Week 12 (47, 27), Week 26 (48, 26), Week 52 (44, 26).

T cell studies

We compared ex vivo T cell proliferative responses to an array of test and control antigens as described (8, 9, 21) in responders and non-responders to rituximab treatment as well as placebo recipients. The baseline responses to disease associated test antigens were similar in the two treatment groups and in the drug responders and non-responders (Table 2). Using the definition of a “positive response” in the assay overall as a stimulation index (SI) of 1.5 or greater to at least 4 antigens, 75% of the rituximab and 82% of the placebo treated subjects showed positive T cell autoreactivity at baseline (p=ns), with similar incidence and response amplitudes of positive responses to each of the 12 individual disease-associated antigens (mean SI was 2.02±0.66 in the rituximab treated and 2.28±0.72 in the placebo treated group, p=ns). There were also no differences when test antigens were thematically clustered (Table 2, Figure 2).

Table 2.

Percent positive T cell antigen reactivity by treatment group at month 6 and 12^*.

		Baseline		6 mos		12 mos
Group	Analyte	Ritux (n=52)	Control (n=28)	Ritux (n=49)	Control (n=27)	Ritux (n=46)	Control (n=28)
Neuronal	Ex2	59.6	71.4	74.0	59.0	61.0	46.0
	MBP	53.8	71.4	77.0	58.0	62.0	45.0
	GFAP	69.2	78.6	84.0	66.0	77.0	75.0
	S-100	65.4	78.6	84.0	70.0	75.0	75.0
Islet	GAD	69.2	82.1	86.0	70.0	89.0	84.0
	GAD555	73.1	82.1	85.0	66.0	88.0	89.0
	Proinsulin	80.8	82.1	86.0	74.0	87.0	89.0
	Tep69	69.2	82.1	87.0	73.0	88.0	88.0
Milk	Abbos	69.2	82.1	87.0	73.0	89.0	85.0
	BLG	75.0	82.1	87.0	74.0	89.0	93.0
	BSA	75.0	82.1	87.0	74.0	89.0	93.0
	Casein	71.2	85.7	87.0	73.0	86.0	87.0
Overall		75	82.1	86.7	73.9	88.3	88.8

^*Estimates of percent positive as a function of treatment group for each antigen, and overall, from a separate logistic model at 6 or at 12 months adjusted for baseline positivity comparing treatment group.

^†None of the differences between the rituximab and control groups are nominally statistically significant at p ≤ 0.05.

Figure 2. Average stimulation indices in the TCP assay for each antigen group using cells from rituximab or placebo treated patients.

The average stimulation index (sum of (CPM for each analyte/CPM for media alone)/no of analytes) and 95% confidence intervals for the mean (dashed vertical bar) for the positive control analytes, negative control analytes, total diabetes antigens, and thematically grouped antigens: islet, neuronal, and milk are shown for the rituximab treated (●) and placebo (■) groups at the 3 study time points are shown. None of the differences between the two groups are of statistical significance. The numbers of subjects per group at each time point were the following (Rituximab, Control): Baseline (49, 29), Month 6 (49, 27), Month 12 (46, 28).

At month 6 and 12, T cell autoreactivities in the two treatment groups were also similar (Figure 3, Table 2). While the differences in the percentage of positive responses were not statistically signficant between groups for any specific antigens or overall at either 6 or 12 months, for all 12 antigens the proportion positive was higher in the rituximab versus placebo treated subjects at 6 months, but not at 12 months (Table 2). Likewise, the SI’s were not significantly different for any of the antigen groups or any of the individual antigens.

Figure 3. Average stimulation indices in the TCP assay for each antigen group of clinical responders and non-responders to rituximab treatment.

Subjects were differentiated on the basis of their C-peptide responses during a MMTT at 6 months into responders (●) or non-responders (□). The LS means of the SI, for each antigen group, by responder status over 12 mos were compared by repeated measures mixed model, adjusted for baseline SI within the rituximab group. There was a significant increase in the responses to the total panel of analytes (p=0.032) as well as to the islet (p=0.048) and neuronal (p=0.005) antigens. The numbers of subjects per group at each time point were the following (Responders, Non-responders): Baseline (29, 20), Month 6 (29, 20), Month 12 (27, 19).

However, sizable differences emerged when grouping rituximab recipients into C-peptide responders vs non-responders (Table3, Figure 3). In the rituximab group at 6 months, 86.7% of subjects had a positive response overall, whereas 96.5% of responders versus 75.2% of non-responders were positive (p = 0.028). There was a significant increase in the responses to 9 of the 12 antigens tested in the responders compared to the non-responders. At month 12, these differences were reduced – only the proportion of responses to EX2 and MBP were significantly greater in C-peptide responders (Table 3). There was also a significant (p<0.05) rise in response amplitudes (stimulation indices) over 12 months to 4 of 12 antigens including responses to the islet (p=0.037) and neuronal (0.043) antigen groupings (Figure 4).. Over 6 and 12 months combined the association was statistically significant for 8 of the 12 test antigens (S-100, GAD, GAD555, proinsulin, ABBOS, BLG, BSA, and Casein), for each grouping (islet p=0.01, neuronal p=0.048, milk p=0.006), and overall (p = 0.009). These differences were the result of treatment with rituximab because we did not detect any significant differences in responders and non-responders at baseline (Supplemental Table 2).

Table 3.

Percent positive T cell antigen reactivity among C-peptide responders and non-responders at month 6 and 12.

		6 mos			12 mos
Group	Analyte	Responder (n= 29)	Non- responder (n=20)	P value	Responder (n=27)	Non- responder (n=19)	P value
Neuronal	Ex2	84.6	61.3	.072	78.2	36.5	.004
	MBP	85.0	69.4	.199	79.1	36.6	.005
	GFAP	93.1	72.1	.047	84.2	68.8	.25
	S-100	93.5	71.2	.033	84.2	65.0	.151
Islet	GAD	96.5	70.4	.012	93.1	90.4	.718
	GAD555	92.9	73.4	.067	93.4	91.7	.823
	Proinsulin	96.4	72.2	.023	88.0	92.0	.712
	Tep69	96.7	73.3	.014	94.5	88.1	.394
Milk	Abbos	96.6	73.9	.017	94.1	89.8	.561
	BLG	96.5	75.2	.028	93.4	91.7	.823
	BSA	96.5	75.2	.028	93.4	91.7	.823
	Casein	96.6	74.5	.022	91.6	82.7	.400
Overall		96.5	75.2	.028	93.4	91.7	.823

^*Estimates of percent positive as a function of responder status and p-value for each antigen, and overall, from a separate logistic model at 6 or at 12 months adjusted for baseline positivity comparing responder status.

Figure 4. Association of the change in TCP SI with the change in C-peptide AUC at month 6 in the rituximab and placebo treated groups.

(A): The %change in SI/%change in log(C-peptide AUC)+1 is shown for rituximab (squares) and placebo (triangles) treated subjects. In subjects treated with rituximab, the open squares designate changes that are significantly associated. The changes in the placebo group are all negatively associated with the changes in C-peptide but none are of statistical significance whereas the relationship is positive in the rituximab treated group and the responses to S-100, proinsulin, Abbos, BLG, and BSA are significantly associated (p<0.05). (B) A similar analysis is shown for responders (circles) and non-responders (inverted triangles) in the rituximab group. There is a significant difference between responders and non-responders for each analyte (p<0.0001). However, within the non-responders, none of the slopes are significantly from zero. The slopes for three of the antigens (PI, ABBOS, and BLG) are significantly increased from zero in the responder group (open symbol, p<0.05).

Importantly, the positive and negative proliferative control responses were similar in C-peptide responders and non-responders (p=0.519 and 0.425 respectively). The responses to tetanus toxoid were 12.95 (CI 11.24–14.65) in the responders and 12.04 (CI 9.9 to 14.19) in the non-responders.

The extent of in vivo B cell depletion by rituximab, which was not significantly different in responders and non-responders, did not appear to account for the differences in the T cell responses. When corrected for the number of CD19 cells or the same B cell subsets, there was minimal effect on the p values that describe the differences in the frequency of responses at 6 months (data not shown).

Relationship of cellular and metabolic responses

We regressed the log (C-peptide+1) values on the T cell proliferative responses (SI) at 6 and 12 months, separately and jointly, adjusting for age, sex, and the SI and C-peptide responses at baseline. Within the rituximab group, Figure 4 shows that the C-peptide increased as the SI increased (a positive relationship) at 6 months for all of the antigens and groups of antigens, 6 of which were statistically significant: S-100 (p=0.022) and proinsulin (p=0.033), as well as milk antigens over all (p=0.021), ABBOS (p=0.029), BLG (p=0.0171), and BSA (p=0.029). Conversely, within the placebo group there was in inverse association (C-peptide decreased as the SI increased) for all antigens and antigen groupings, though not significantly so for any one in particular. Neither the negative nor positive controls were associated with C-peptide responses.

In like analyses within the placebo group, there was no significant relationship between 6 and 12 month C-peptide levels and any antigen responses (Figure 4 and data not shown). Further analyses of the two groups combined showed that the relationship between the change in C-peptide/change in SI was significantly different between the rituximab vs placebo groups, i.e. the test of the group*SI interaction, for all 16 of the antigens and antigen groupings. In a similar comparison between responders and non-responders to Rituximab, we found a significant difference between responders and non-responders for each analyte for the slope of % change in C-peptide/unit change in SI (Figure 4B). However, within the non-responders none of these slopes were significantly different from zero. Within the responders, there were 3 analytes with a slope significantly different from zero:_proinsulin (PI), ABBOS, and BLG (p<0.05).

Lymphocyte subsets in subgroups of rituximab recipients

We compared the relative and absolute number of T cell subsets in C-peptide responders and non-responders within the rituximab treated group (Figure 5). By an analysis of variance for repeated measures, the absolute number of CD19+ cells over 12 months was not significantly different in C-peptide responders and non-responders (p=0.41) but the number of CD3+ cells was higher in the responders (1512, CI: 1422, 1608 vs 1346, CI: 1255, 1444; p=0.016). The most significant differences between the groups were accounted for by an increase in the number of CD4+ cells (937, CI: 878, 999 vs 822, CI: 764, 884; p=0.01) since the number of CD8+ cells was only slightly different (423, CI:393, 454 vs 383, CI:353, 417; p=0.08). However to assess this hypothesized effect of CD4 cells among responders and non-responders, we conducted additional analyses adjusting for the CD4 values at baseline, week 5, month 3, and month 6. of the positivity/negativity of each analyte. There was very little change in the p-values, and 9 of the 13 (versus 10 of 13 without adjustment) were still significant at the 0.05 level. The number of activated, CD4+CD25+, cells was not significantly different in responders and non-responders. The number of Tregs (CD4+CD25+CD62L+) was higher at week 12 (48, CI: 41, 57 vs 34, CI:28, 41; p=0.005) in responders, but over the 12 month period with 4 separate measurements, the differences were not statistically significant (p=0.17).

Figure 5. Analysis of lymphocyte subsets in patients treated with rituximab.

The number (±95% CI) of the indicated T cell subsets and 95% confidence intervals for the mean (dashed vertical bar) are shown in rituximab treated study participants who were designated as clinical responders (●) or non-responders (□) based on the C-peptide responses at 6 months. The CD3+ (p=0.02) and CD4+ (p=0.01) cell counts were significantly different over the 12 month study period. In addition, the number of Tregs was increased in the clinical responders at week 12 (p=0.005), but the counts were not significantly different over the 12 months (p=0.17). The numbers of subjects per group at each time point were the following (Responders, Non-responders): Baseline (29, 20), Week 5 (27, 19), Week 12 (26, 20), Week 26 (27, 20), Week 52 (25, 19).

Discussion

This trial discovered that enhanced, T cell proliferative responses to diabetes-associated target antigens are predictive of a positive C-peptide response to rituximab in newly diagnosed patients with type 1 diabetes. Enhanced T cell reactivity in C-peptide responders was not explained by overall elevated T cell reactivity, since neither positive nor negative control responses were changed by therapy, nor could any control responses be associated with stimulated C-peptide levels. The enhanced T cell reactivity cannot be explained by an artifact or assay drift. Background counts were not significantly different between C-peptide responders and non-responders at 6 months (p=0.632), and positive (p=0.519) and negative (p=0.425) controls were unchanged in the same assays. There was also no change in T cell reactivity of placebo recipients. We conclude that the observed elevation of antigen-specific responses to diabetes-associated environmental, islet, and neuronal antigens was caused by rituximab-driven attenuation of post-onset β cell loss over the 1^st year post-therapy. These strictly blinded assays, conducted over a 27 month period, delineated an almost absolute (95.6%) linkage between enhanced T cell reactivity and the presence of definitive signs of β cell preservation. It was conceivable that increased proliferative responses were a function of being a clinical responder per se and not due to the effects of the rituximab treatment. Comparing the change in stimulation indices with C-peptide AUC in both study arms, we found that the positive association between changes in C-peptide responses only occurred in rituximab recipients. In the placebo group, this relationship was more often negative. Thus, the observed changes in T cell responses were a function of the rituximab therapy received rather than a sequelae of the improved metabolic response.

In addition, our findings cannot be explained by differences in the extent of B cell depletion. The number of circulating T cells overall were not significantly different in the rituximab and placebo treated groups. We found modest but statistically significant changes in CD3+ and CD4+ T cells but the mechanistic relevance of these minimal changes in polyclonal populations to the antigen-specific responses is not clear. Even if subtle, a numeric expansion of the T cell compartment might reflect brief homeostatic proliferation into hemopoietic “space”, vacated by rituximab. The observation that this “space” was only 69% re-filled by B lineage cells one year after the single treatment course as well as the growth of the T cell compartment both support the homeostatic expansion scenario. This expansion might include the antigen-specific T cell pools we measured ex vivo. There is precedence for this with rapid diabetes development in lymphocyte-free NOD.scid mice following transfer of small numbers of diabetogenic wild type NOD lymphocytes or following lymphotoxic therapy of wild type NOD mice with cyclophosphamide (22, 23). However, homeostatic expansion alone does not appear to explain the differences in the T cell proliferation assay in responders and non-responders since the responses to control antigens were not different between these two groups yet the responses to the diabetes-related antigens were. This suggests that there has been expansion of disease antigen specific T cells and/or function.

However, in the present trial, loss of β-cells/insulin secretory reserve were not accelerated but significantly slowed. Thus, therapeutic B cell depletion relieved a somewhat counterintuitive, disease progression-associated restraint of T cell expansion. These data demonstrate, for a first time in human T1D, a progression-promoting role of B lineage cells at diabetes onset. This B cell role is amenable to therapeutic intervention and adds a fundamental similarity to T1D pathogenesis in NOD mice and in humans. We conclude therefore, that the statistically significantly different relationships between C-peptide and T-cell responses between the rituximab and placebo groups show that the T-cell responses between responders and non-responders suggest a biological effect of rituximab treatment and not a general response regardless of the treatment received.

Our findings cannot be explained by chance alone, but it is important to note that the statistical relationship that we identified with the T cell assay has an unclear biologic foundation. We do not have a confirmatory assay in which the number of antigen specific cells or the magnitude of their responses to antigens can be confirmed and the mechanism that accounts for the increase in the proliferative responsiveness to disease-associated antigen-specific T cells is not clear. The T cell proliferation assay is performed in the presence of IL-2 and therefore it is also possible that the increased proliferative response that we found was due to an increased proportion of activated or anergic CD4+ or CD8+ T cells (21). Indeed, our flow cytometric analysis of peripheral blood cells did show a modest increase in the proportion of CD4+ and CD8+ T cells that expressed CD25 but the frequency of T cells with modulation of CD62L was not significantly different. It is also possible that B cell depletion altered T cell trafficking rather than purely numeric change in the antigen specific T cells. For example, Piccio et al recently reported that there was a decrease in the number of T cells in the CSF after rituximab treatment of patients with relapsing/remitting multiple sclerosis (24). They suggested that B cells may be critical for T cell trafficking into the CNS inflammatory lesions. An analogous reduction in T cell trafficking into the islet following B cell depletion with the mAb might account for an increase in the number of antigen specific cells in the periphery. A similar mechanism might be operative to explain the finding of increased Tregs that we observed in the peripheral blood of rituximab treated patients. An analysis of cytokine responses to the tested antigens might provide insights into the relationships between the increased proliferative responses and reduced β cell destruction, and determine, for example, whether production of cytokines such as IL-10 and/or TGFβ may have been increased in the antigen specific cells. However, this type of analysis will require further studies with fresh cells from a separate cohort.

We found that the changes in T cell responses correlated with clinical responses in the rituximab treated group and with responses to certain antigens when comparing the drug and placebo groups. One explanation that might account for this finding is that the antigen specific cells that are assessed in the assay have regulatory function. This notion was suggested in preclinical studies of Hu et al who found that the proportion of CD4+CD25+Foxp3+ cells was increased during the recovery phase following depletion of hCD20 expressing NOD mice (16). Interestingly, we found a similar difference in the number of circulating Tregs in C-peptide responders compared to non-responders at week 12. In addition to these conventional Tregs, it is also possible that the antigen specific cells which are increased in the patients have regulatory function, by virtue of production of cytokines such as IL-10 or TGF-β or through other mechanisms (25, 26). Further studies of the cytokines that are made by the T cells in response to antigen might be helpful in addressing this question. The mechanism that would lead to induction of T cells with this phenotype has not been identified but it is of note that IL-10 secreting B lymphocytes have been identified, and the secretion of this cytokine might create an environment that induces differentiation of T cells that produce regulatory cytokines. Induction of regulatory T cells might be postulated to occur through a bystander mechanism in the presence of IL-10 production by B cells that repopulate after CD20 depletion (i.e. Bregs)(26, 27).

Collectively, we have found that proliferative responses to diabetes-associated target antigens are similar in patients who receive anti-rituximab compared to placebo. However, among those treated with rituximab, individuals who are clinical C-peptide responders show an increased proliferative responses to islet, neuronal, and disease-relevant environmental antigens, and the changes in proliferative responses to islet antigens are associated with an increase in insulin secretory function. The perhaps simplest explanation of our data would be that rituximab allowed recovery and/or regeneration of β cells through transient interruption of pathogenic T cell recruitment via surface immunoglobulin-captured autoantigen, but providing new β cell targets during recovery of the CD19+ B cell compartment post treatment (28).