Study approach and pharmacokinetics. In 2 independent single-arm, open-label clinical studies, a single dose of either anti–IL-6 (siltuximab) or anti–IL-6R (tocilizumab) therapy was administered to individuals with established T1D. Both studies were designed to assess predetermined mechanistic immunological endpoints, not clinical outcomes. There were 10 individuals in the siltuximab study and 9 individuals in the tocilizumab study. Importantly, 5 individuals participated in both studies, allowing an assessment of intraindividual responses to the 2 therapies; these individuals participated first in the siltuximab study and then in the tocilizumab study, and the time between administration of each drug ranged from 166 to 370 days. Blood samples were collected at 6 visits: screening at 2 weeks before infusion; a baseline visit with intravenous administration of the drug; and 4 follow-up visits at day 1, day 14, week 4, and week 12 after infusion. There were no significant differences between screening and day 0 for any of the assays, so for simplicity, only data for day 0 are presented. Given limited sample volumes, follow-up assays were not conducted on every individual or at all time points. Table 1 summarizes demographics, clinical characteristics, and serum drug concentrations. The individual patients and assays performed for each patient are listed in Supplemental Table 1 (supplemental material available online with this article; https://doi.org/10.1172/jci.insight.159436DS1). Supplemental Table 1 also lists an additional 3 patients treated with tocilizumab enrolled after completion of the primary cohort; samples from these individuals were used for secondary endpoint assays.

Study participant demographics and clinical characteristics

The half-lives of siltuximab and tocilizumab are 20.6 and 11 days, respectively. Peak drug concentration for both was measured at day 1 after infusion (Table 1). Neutrophil counts were monitored throughout the period of the study because neutrophil decline was expected as a consequence of these therapies (15–18). The frequency, magnitude, and kinetics of neutrophil decline differed considerably by drug (Supplemental Figure 1). At day 1 after siltuximab infusion, 5 individuals exhibited a slight decline in neutrophil counts, but none dropped below the normal range (Supplemental Figure 1A). However, at 28 days after infusion, 9 of the 10 individuals exhibited lower values compared with baseline. Of these 9 individuals, 5 showed clinically significant reductions, dropping below the lower limit of normal (LLN, 2 × 109/μL); 3 had grade 1 neutropenia (1.5 × 109/μL–2 × 109/μL), 1 had grade 2 neutropenia (1.0 × 109/μL–1.5 × 109/μL), and 1 had grade 4 neutropenia (0 × 109/μL–0.5 × 109/μL). In contrast, both the magnitude and kinetics of neutrophil decline differed for tocilizumab. At day 1 after tocilizumab infusion, neutrophil levels declined for all individuals, with 5 individuals dropping below LLN; of these 5 individuals, 2 had grade 2 neutropenia, 2 had grade 3 neutropenia (0.5 × 109/μL–1.0 × 109/μL), and 1 had grade 4 neutropenia (Supplemental Figure 1B). Neutrophil count decline on day 1 after tocilizumab was modestly correlated with peak drug concentration (Pearson correlation, r = –0.66, P = 0.053). No infections or other adverse events occurred contemporaneously with low neutrophil counts in either study, and neutrophil counts normalized without intervention in all individuals by week 12.

Comparison of IL-6–induced phosphorylated STAT3 in PBMCs isolated from tocilizumab-treated patients and siltuximab-treated patients. We first examined how each drug influences IL-6 signaling on T cells by measuring both membrane-bound IL-6R (mbIL-6R) expression and IL-6–induced phosphorylation of STAT3 in cryopreserved PBMCs from individuals treated with each drug. After siltuximab treatment, we found no change in IL-6R expression on the cell surface of naive CD4+ T cells (Figure 1A and Supplemental Figure 2A); also, there was no change in soluble IL-6R serum levels (Supplemental Figure 2B). In contrast, after tocilizumab treatment, we detected less IL-6R on the cell surface of naive CD4+ T cells at day 1, day 14, and week 4 with a return to preinfusion levels by week 12 (Figure 1A). Since this loss of detection may be due to tocilizumab bound to the mbIL-6R and blocking binding of the anti–IL-6R antibody, we tested side by side 2 monoclonal anti–IL-6R antibodies: clone UV4, which was used in Figure 1A, and clone M5. In this experiment, PBMCs from a healthy control individual were preincubated in the presence or absence of tocilizumab and then stained with UV4 or M5. In the absence of tocilizumab, both clones were able to detect cell-surface IL-6R, although M5 staining was more robust than UV4 (Supplemental Figure 3A, left). However, in the presence of tocilizumab, only M5 but not UV4 was able to bind and hence detect IL-6R (Supplemental Figure 3A), indicating that M5 binds to a different epitope on mbIL-6R than tocilizumab, whereas UV4 competes with tocilizumab binding to mbIL-6R. Therefore, the reduction of mbIL-6R levels shown in Figure 1A was likely due to the presence of tocilizumab bound to mbIL-6R, thus preventing binding of the UV4 monoclonal antibody. These results also suggest that after a single dose, tocilizumab was bound to mbIL-6R on T cells for at least 4 weeks after treatment but was no longer present at week 12 (Figure 1A). To confirm that M5 binds to a different IL-6R epitope than tocilizumab, we stained with both M5 and fluorescently labeled tocilizumab and demonstrated dual staining (Supplemental Figure 3B). We then examined PBMCs still available from our in vivo study using the M5 monoclonal antibody to assess IL-6R expression in 3 patients treated with tocilizumab (Supplemental Figure 3, C and D). With this new antibody, we found no change in mbIL-6R levels after tocilizumab, yet we did see a change in phosphorylated STAT3 (p-STAT3) consistent with tocilizumab binding at week 2 (Supplemental Figure 3, C and D). Together, these results suggest that although tocilizumab remained bound to mbIL-6R on CD4+ T cells for at least 4 weeks after the single-dose treatment, it did not alter the levels of mbIL-6R over 12 weeks.

Suppression of IL-6–induced p-STAT3 persists in tocilizumab-treated but not siltuximab-treated patient PBMCs. Each line represents an individual patient; n = 10 for siltuximab and n = 9 for tocilizumab. (A) Frequency of mbIL-6R+ cells in naive CD4+ T cell compartment at baseline. (B) Frequency of IL-6–induced p-STAT3+ cells in the naive CD4+ T cell compartment. Thawed and rested PBMCs from siltuximab-treated or tocilizumab-treated patients with T1D were stimulated with recombinant IL-6 (2 ng/mL) for 10 minutes.

Siltuximab and tocilizumab had differential effects on the frequency of IL-6–induced p-STAT3+ naive CD4+ T cells isolated from PBMCs from treated individuals. There was no change in the frequency of IL-6–induced p-STAT3+ naive CD4+ T cells after siltuximab; this may be explained by siltuximab being washed out when PBMCs are isolated. In contrast, there was a substantial reduction after tocilizumab (Figure 1B, Supplemental Figure 3F and Supplemental Figure 4), which was rapid and almost complete at day 1 after tocilizumab infusion, consistent with the persistent binding of tocilizumab to the mbIL-6R on PBMCs. Of note, there was heterogeneity in the rate of return to preinfusion levels among tocilizumab-treated patients, where some individuals began to return to preinfusion levels at week 2 while other individuals still had significantly reduced levels at week 12 compared with day 0 levels (Figure 1B). Moreover, this heterogeneity did not correlate with peak drug concentration or with most proximal drug concentration at 4 weeks (data not shown); no serum tocilizumab was detectable for any individual by 12 weeks after infusion. Further investigation demonstrated that tocilizumab but not siltuximab reduced IL-6–induced p-STAT3 in other T cell subsets, including memory CD4+ T cells, memory CD8+ T cells, and CD4+ Tregs (Supplemental Figure 5) with similar kinetics and individual-by-individual heterogeneity in rate of return to responsiveness as the naive CD4+ T cell population. Likewise, IL-6–induced p-STAT1 was strongly modified after tocilizumab, not siltuximab, in in vitro assays of each of these populations (Supplemental Figure 6). Together, these findings indicate that tocilizumab achieved global suppression of IL-6–induced p-STAT3 signaling in T cells within 24 hours after a single dose, which was maintained after PBMC isolation due to persistent binding of tocilizumab to the mbIL-6R for at least 4 weeks after infusion.

Tocilizumab but not siltuximab decreases ICOS expression on T follicular helper cells. We next investigated the impact of both drugs on T cell subsets for which IL-6 signaling is known to play a role in determining phenotype. These include follicular T (Tfh) cells, peripheral T helper (Tph) cells, Th17 cells, and Tregs (19–24). We first assessed the frequency of CD4+ and CD8+ naive and memory T cells, CD4+ Tregs (CD4+CD25hiCD127lo), and Th17 cells (CD4+FOXP3–IL-17+) at each time point and found no change in frequency in response to either drug for any of these populations (data not shown, Supplemental Figure 7, A and B). We also confirmed that there was no change to FOXP3 MFI on Tregs (Supplemental Figure 7C). We performed a more comprehensive assessment of T cell subsets, analyzing PBMCs from days 0 and 14, and observed that siltuximab treatment had either no effect or resulted in an increased frequency of IL-17+ memory CD4+ Teffs (Figure 2A). In contrast, although not statistically significant, tocilizumab resulted in a decreased frequency of IL-17+ memory CD4+ Teffs for 5 of 6 individuals tested (Figure 2A), suggesting that a single dose of tocilizumab can impact Th17 function or lineage development. Additionally, intracellular cytokine staining after PMA/ionomycin stimulation showed that tocilizumab reduced the frequency of memory CD4+ Teffs capable of producing IL-21 (Figure 2B) as well as IL-21+FOXP3+ Tregs (Supplemental Figure 8A). This suggests that tocilizumab but not siltuximab may impair Tfh and/or Tph differentiation or function, both of which produce IL-21 (25). To address this further, high expression of PD-1 was used as a marker of both Tfh and Tph; however, we were unable to differentiate Tfh from Tph because of a failure of CXCR5 in our panel. After tocilizumab treatment, the frequency of PD-1hi+ memory CD4+ Teffs was lower in 4 of the 6 individuals (Supplemental Figure 8B), whereas there was no consistent pattern for frequency of PD-1hi+ memory CD4+ Teffs in the siltuximab-treated patients (Supplemental Figure 8B). Since IL-6R expression has been shown to be required for Tfh expression of ICOS (24), we examined ICOS expression on PD-1hi+ memory CD4+ Teffs. We found that in response to tocilizumab but not siltuximab, there was a significant decrease in ICOS expression among PD-1hi+ memory CD4+ Teffs (Figure 2, C and D), and this decrease in ICOS expression correlated significantly with the peak drug concentration measured at day 1 after drug infusion (Figure 2E). ICOS expression was modestly but not significantly decreased among FOXP3+ Tregs after tocilizumab, while overall the frequency of FOXP3+ Tregs remained unchanged (Supplemental Figure 8, C and D). To confirm that siltuximab did not modify the homeostatic relationship between these cell populations, we next assessed the correlations between fold change in frequency for ICOS+ PD-1hi memory CD4+ Teffs, IL-17+ memory CD4+ Teffs, and FOXP3+ Tregs after siltuximab exposure. There was an inverse correlation between ICOS+PD–1hi memory CD4+ Teffs and IL-17+ memory CD4+ Teffs (Figure 2F). Additionally, FOXP3+ Tregs correlated inversely with ICOS+PD-1hi+ memory CD4+ Teffs (Supplemental Figure 8E) but correlated positively with IL-17+ memory CD4+ Teffs (Supplemental Figure 8F). Collectively, these findings suggest that direct blockade of IL-6 signaling with tocilizumab had a greater impact on Th17 and Tfh/Tph lineages and ICOS expression than siltuximab.

Tocilizumab but not siltuximab decreases ICOS expression of T follicular helper cells. Thawed and rested PBMCs from siltuximab-treated or tocilizumab-treated patients with T1D were stimulated with PMA/ionomycin for 1 hour followed by an additional 3 hours in the presence of Brefeldin A. Each line represents an individual patient; n = 10 for siltuximab and n = 6 for tocilizumab. Solid circles represent d0 prior to drug infusion, and open circles represent d14 after drug infusion. (A) Frequency of IL-17+ cells in memory CD4+ Teffs. (B) Frequency of IL-21+ cells in memory CD4+ Teffs. (C) Representative histograms showing PD-1hi ICOS+ memory CD4+ Teffs at d0 and d14 after tocilizumab infusion from a single patient. (D) Frequency of ICOS+ cells in PD-1hi memory CD4+ Teffs. (E) Linear regression for tocilizumab cohort showing negative correlation between peak drug concentration on d1 and fold change d14 versus d0 for frequency of ICOS+ cells in PD-1hi memory CD4+ Teffs. (F) Linear regression for siltuximab cohort showing negative correlation between fold change d14 versus d0 for frequency of ICOS+ cells in PD-1hi memory CD4+ T cell compartment and fold change d14 versus d0 for frequency of IL-17+ cells in memory CD4+ T cell compartment. Statistical tests: (A, B, and D) Wilcoxon matched-pairs signed-rank test; and (E and F) linear regression.

Siltuximab but not tocilizumab reverses Teff resistance to Treg-mediated suppression. We and others have previously shown that CD4+ Teffs are resistant to Treg suppression in established T1D (11–13). IL-6 has been implicated in the resistance of T cells to Treg-mediated suppression (13, 26–28). To determine whether Teff resistance to Treg suppression could be modified in vivo by either drug, we used an in vitro suppression assay (29) to assess the suppression of Teffs isolated from PBMCs at day 0, day 14, and week 12 by a standard pool of Tregs derived from a healthy control individual. Strikingly, siltuximab but not tocilizumab enhanced suppression of Teffs by Tregs (Figure 3). At day 14 after siltuximab treatment, Teff resistance was reversed in 9 out of 10 individuals (Figure 3A). These improved levels of response to Treg suppression were sustained to 12 weeks after infusion in 5 of the 9 individuals (Figure 3A). Notably, the percentage of suppression levels at day 14 correlated with peak drug concentration at day 1 (Figure 3B). This was not seen after tocilizumab, where there was no change in resistance to suppression at either day 14 or week 12 after drug infusion (Figure 3A). Markedly, these drug-specific differences were seen in the individuals who participated in both studies (Figure 3C). These findings suggest that the influence of these 2 IL-6–targeted therapies resulted in drug-specific alterations in response to activation that were T cell intrinsic and influenced their response to regulation. To address these differences, we interrogated the changes in the response to T cell receptor activation.

Siltuximab but not tocilizumab reverses Teff resistance to Treg-mediated suppression. Expanded allogeneic Tregs from a healthy control donor were cocultured at a ratio of 1:4 with Teffs from either siltuximab-treated or tocilizumab-treated patients with T1D in the presence of anti-CD3/anti-CD28–coated beads for 2 days. The percentage of suppression was determined by measuring the frequency of activated CD25+CD134+ Teffs. Each line represents an individual patient; n = 10 for siltuximab and n = 9 for tocilizumab. (A) Percentage suppression for siltuximab-treated patients and tocilizumab-treated patients. (B) Linear regression for siltuximab cohort showing positive correlation between peak drug concentration on d1 and percentage suppression at d14. (C) Percentage suppression for cohort that participated in both studies (n = 5); note these individuals are also included in A and B. Statistical tests: (A and C) Wilcoxon matched-pairs signed-rank test; (B) linear regression.

Tocilizumab and siltuximab have opposing effects on TCR-induced p-STAT3 signaling. TCR-mediated phosphorylation of STAT3 has previously been shown to be enhanced in CD4+ Teffs from patients with T1D compared with healthy control individuals, suggesting altered kinetics of STAT3 activation in patients with T1D (13). To determine the effect of tocilizumab or siltuximab on TCR-induced STAT3 activation at days 0 and 14, we stimulated enriched CD3+ T cells with anti-CD3/CD28 beads and measured STAT3 phosphorylation across CD4+ and CD8+ naive and memory T cell subsets (Supplemental Figure 9A), comparing day 0 and day 14. Strikingly, TCR-induced p-STAT3 MFI was increased in 8 of 10 individuals who received siltuximab but decreased in 7 of 10 individuals who received tocilizumab (Figure 4A). TCR-induced p-STAT3+ cell frequency was increased in 8 of 10 patients treated with siltuximab but decreased in 7 of 10 patients treated with tocilizumab (Figure 4B). Again, these drug-specific effects were seen in individuals participating in both studies (Figure 4C). Total STAT3 MFI was decreased in tocilizumab-treated patients but was not significantly changed in those treated with siltuximab (Figure 4, D and E). The opposing effects of tocilizumab and siltuximab on TCR-induced STAT3 activation was most prominent in naive CD4+ T cells but also observed in memory CD4+ T cells (Supplemental Figure 9, B–D) as well as in naive CD8+ T cells (Supplemental Figure 9, E–G). Further investigation showed that the alteration in p-STAT3 signaling in response to siltuximab appears to be specific to TCR stimulation, as there was no enhancement in IL-10–induced p-STAT3 MFI (Figure 4F). Conversely, IL-10–induced p-STAT3 MFI was decreased in tocilizumab-treated individuals (Figure 4F), consistent with the observed reduction in total STAT3 expression (Figure 4E), and indicating that the decrease was not due solely to blockade of IL-6R. We also assessed TCR-induced p-STAT1 and p-STAT5. Tocilizumab treatment resulted in a decrease in both TCR/p-STAT1 and TCR/p-STAT5, whereas no significant alterations in TCR/p-STAT1 or TCR/p-STAT5 were found in T cells from siltuximab-treated patients (Supplemental Figure 10). Tocilizumab also decreased total STAT5 levels but had no effect on total STAT1 (Supplemental Figure 10). Collectively, these findings suggest that the effect of tocilizumab on STAT activation after TCR stimulation was driven by a direct effect of treatment on total STAT expression, whereas the response to TCR stimulation after siltuximab treatment resulted in an alteration that was specific to STAT3 signaling.

Tocilizumab and siltuximab have opposing effects on TCR-induced p-STAT3 signaling in naive CD4+ T cells. Enriched pan T cells from siltuximab-treated or tocilizumab-treated patients with T1D were stimulated or not with anti-CD3/anti-CD28–coated beads for 4 hours. Cells were stained for p-STAT3 and total STAT3. Each line represents an individual patient; n = 10 for siltuximab (except for F where n = 7; gated CD4+CD45RA+ naive CD4+ T cells) and n = 10 for tocilizumab (gated CD4+CD27+CD45RA+ naive CD4+ T cells). Solid circles represent d0 prior to administration of the drug, and open circles represent d14 after drug administration. (A) p-STAT3 geometric MFI (gMFI). (B) Frequency of p-STAT3+ cells. (C) p-STAT3 MFI for cohort that participated in both studies (n = 4); note these individuals are also included in A, B, E, and F. (D) Gating strategy for total STAT3 gMFI of gated naive CD4+ T cells using unstimulated enriched pan T cells: histograms for representative siltuximab-treated patient and representative tocilizumab-treated patient. (E) Total STAT3 MFI of unstimulated cells. (F) p-STAT3 MFI after stimulation with IL-10 (20 ng/mL for 30 minutes). Statistical tests: Wilcoxon matched-pairs signed-rank test.

Siltuximab but not tocilizumab enhances T cell production of regulatory cytokines. To further explore the differential effects of siltuximab and tocilizumab on the response to TCR activation, we examined cytokine production using a multiplexed assay to measure IL-6, IL-10, IL-21, IL-22, and IL-27, since these cytokines have been reported to regulate both immunogenic and tolerogenic responses (30–32). Siltuximab treatment resulted in increased production of IL-10, IL-21, and IL-27 in 9 of 10 individuals but had no effect on IL-6 and IL-22 production (Figure 5). Conversely, tocilizumab had no effect on any of these cytokines (Figure 5). To further investigate the link between enhanced TCR-induced production of IL-10, IL-21, and IL-27 and the enhanced STAT3 phosphorylation observed in response to siltuximab, we calculated Pearson correlations between 4-hour cytokine production and frequency of p-STAT3+CD4+ T cells from day 0 and day 14. At day 14, TCR-induced p-STAT3 had a positive correlation with IL-10 (R2 = 0.4726, P = 0.04) and IL-27 (R2 = 0.4490, P = 0.0483) but a slightly weaker correlation for IL-21 (R2 = 0.3933, P = 0.07). There was no correlation at either time point between TCR-induced p-STAT3 and IL-22, an IL-10 family cytokine that shares the IL-10R2 receptor with IL-10 (33–35), demonstrating the specificity of our findings.

Siltuximab but not tocilizumab enhances T cell production of regulatory cytokines. Meso Scale Discovery assays were used to measure cytokine production by enriched pan T cells from siltuximab-treated or tocilizumab-treated patients with T1D stimulated with anti-CD3/anti-CD28–coated beads for 4 hours. Each line represents an individual patient; n = 9 for siltuximab (except for A where n = 8) and n = 10 for tocilizumab. Solid circles represent d0 prior to administration, and open circles represent d14 after drug administration (d14). (A) IL-6, (B) IL-10, (C) IL-21, (D) IL-22, and (E) IL-27. Statistical test: Wilcoxon matched-pairs signed-rank test.

To exclude the possibility that altered cytokine receptor expression is associated with the enhanced TCR-induced p-STAT3 signaling observed in response to siltuximab, we assessed cell-surface levels of IL-10R, IL-21R, IL-27R, and gp130. Although limited by sample availability for this question, no differences in expression were observed between days 0 and 14 for any of these receptors (Supplemental Figure 11). Taken together, these findings further distinguish the impact of tocilizumab and siltuximab on T cell function in vivo, suggesting indirect effects of IL-6 blockade with siltuximab, as opposed to the direct effect of blunted IL-6R signaling seen with tocilizumab, which leads to enhanced regulatory cytokine expression and p-STAT3 responses.