Polyfunctional T follicular helper cells drive checkpoint-inhibitor diabetes and are targeted by JAK inhibitor therapy

Abstract

Immune checkpoint inhibitors (ICI) have revolutionized cancer therapy, but their use is limited by the development of autoimmunity in healthy tissues as a side effect of treatment. Such immune-related adverse events (IrAE) contribute to hospitalizations, cancer treatment interruption, and even premature death. ICI-induced autoimmune diabetes mellitus (ICI-T1DM) is a life-threatening IrAE that presents with rapid pancreatic β-islet cell destruction leading to hyperglycemia and life-long insulin dependence. While prior reports have focused on CD8+ T cells, the role for CD4+ T cells in ICI-T1DM is less understood. We identify expansion of CD4+ T follicular helper (Tfh) cells expressing IL-21 and IFN-γ as a hallmark of ICI-T1DM. Furthermore, we show that both IL-21 and IFN-γ are critical cytokines for autoimmune attack in ICI-T1DM. Because IL-21 and IFN-γ both signal through JAK/STAT pathways, we reasoned that JAK inhibitors (JAKi) may protect against ICI-T1DM. Indeed, JAKi provide robust in vivo protection against ICI-T1DM in a mouse model that is associated with decreased islet-infiltrating Tfh cells. Moreover, JAKi therapy impaired Tfh cell differentiation in patients with ICI-T1DM. These studies highlight CD4+ Tfh cells as underrecognized but critical mediators of ICI-T1DM that may be targeted with JAKi to prevent this grave IrAE.

Keywords: Autoimmune diseases; Autoimmunity; Cancer immunotherapy; Diabetes; Oncology.

Figure 1. Increased CD4⁺ Tfh cell response in individuals with ICI-T1DM and a mouse model of IrAEs.

(A) Representative flow cytometry of PBMC from ICI-treated patients at baseline and after ex vivo culture under Tfh-skewing conditions (23). (B) Fold change in Tfh cell frequency for individuals with ICI-T1DM versus ICI-treated individuals with no irAEs. Each pair represents 1 individual. (C) DM incidence in NOD mice treated with anti–PD-1 (8 males [M]/8 females [F]) or isotype (Iso) (6M/7F). (D) DM incidence in anti–PD-1 treated NOD mice with a depleting anti-CD4 antibody (5M/5F) or isotype (Mock) (2M/2F). (E) Representative flow cytometry for islet-infiltrating Tfh cells. (F) Quantification of Tfh cells (CD4⁺ICOS⁺PD-1^hiCXCR5⁺) within islets of anti–PD-1–treated (n = 16) versus Iso-treated (n = 8) mice. (G) Quantification of Bcl6⁺Tbet^– and Bcl6⁺Tbet⁺ subsets within CD4⁺ICOS⁺PD-1^hiCXCR5⁺ cells in the islets of anti–PD-1–treated (n = 6) versus Iso-treated (n = 5) mice. (H) Representative flow cytometry and quantification of islet-infiltrating IL-21– and IFN-γ–producing Tfh cells in Iso-treated (n = 7) and anti–PD-1–treated (n = 13) mice. (I) Quantification of IL-21⁺IFN-γ^– and IL-21⁺IFN-γ⁺ subsets within CD4⁺ ICOS⁺PD-1^hiCXCR5⁺ cells in the islets of anti–PD-1–treated (n = 19) versus Iso-treated (n = 8–9) mice. (J) Quantification of BDC2.5-mimotope tetramer⁺ Tfh cells within the islets of Iso-treated (n = 6) versus anti–PD-1–treated (n = 5) mice. (K) Comparison of islet-infiltrating IL-21⁺IFN-γ⁺tetramer⁺CD4⁺ Tfh cells between anti–PD-1–treated (n = 4) and Iso-treated (n = 5) mice. Each point represents data from 1 animal, and data are presented as mean ± SD. Comparisons by 2-way ANOVA for paired samples with subsequent pairwise comparisons (B), log-rank test (C and D), Welch’s t test (G), Brown-Forsythe and Welch ANOVA (I), or Mann-Whitney U test (F, J, and K); *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 2. IL-21 and IFN-γ are key cytokine mediators of ICI-T1DM.

(A) Schematic of cytokine production by Tfh cells. (B) Incidence curve for ICI-T1DM in anti–PD-1 treated NOD WT and NOD.Il21r^–/– (IL-21R KO) mice. WT, Iso (6 males, 7 females); IL-21R–KO, Iso (11 males); IL-21R–KO, anti–PD-1 (8 males, 2 females). (C) Incidence curve for ICI-T1DM in ICI-treated NOD WT and NOD.IFN-γ^–/– (IFN-γ–KO) mice during anti–PD-1 treatment. WT, Iso (6 males, 7 females); IFN-γ–KO, Iso (6 males, 3 females); IFN-γ–KO, anti–PD-1 (6 males, 7 females), WT, anti–PD-1 (3 males, 4 females). (D) Representative H&E-stained pancreas histology sections of Iso- or anti–PD-1–treated WT, IL-21R–KO, or IFN-γ–KO mice (original magnification, 100×). Arrow indicates an islet of Langerhans. (E) Insulitis index determined by histologic analyses of pancreas islet histology across indicated treatment conditions. WT, Iso (5 males, 10 females); WT, anti–PD-1 (6 males, 10 females); IL-21R–KO, anti–PD-1 (4 males, 1 female); IFN-γ–KO, anti–PD-1 (5 males, 5 females). (F) Quantification of CD4⁺ T, CD8⁺ T, and B cells from anti–PD-1–treated WT (2 males, 2 females), IL-21R–KO (2 males, 1 female), and IFN-γ–KO (5 males, 4 females) mice or Iso WT (1 male, 3 females), via multi-immunofluorescence staining. Comparisons by log-rank test (B and C), Fisher’s exact test (E), or Brown-Forsythe ANOVA with Welch’s pairwise comparison test (F). **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 3. JAKi ruxolitinib provides robust protection against ICI autoimmune DM.

(A) Proposed JAK signaling inhibition downstream from IL-21 and IFN-γ to halt autoimmune response. (B) Schematic for treatment of mice with JAKi ruxolitinib (left) and incidence of autoimmune DM (right) in NOD mice treated with anti–PD-1 immunotherapy or Iso, and ruxolitinib or control food gel. Iso (6 males, 13 females); Iso + Ruxo (3 males, 3 females); anti–PD-1 + Ruxo (6 males, 8 females); anti–PD-1 (8 males, 18 females). (C) Representative H&E-stained pancreas histology sections of anti–PD-1 or Iso-treated NOD mice (original magnification, 100×) fed ruxolitinib or control food. Arrow indicates an islet of Langerhans. (D) Insulitis index of anti–PD-1 or Iso-treated NOD mice given ruxolitinib (Iso: 1 males, 1 females; anti–PD-1: 4 males, 4 females); ^data for anti–PD-1 mice given control chow are the same as shown in Figure 2E, reshown here for comparative purposes. (E) Schematic and absolute cell counts of pancreatic islet–infiltrating CD45⁺ cells, as determined by flow cytometry, across Iso + vehicle (n = 12), anti–PD-1 + vehicle (n = 15), and anti–PD-1 + ruxolitinib (n = 5) conditions. Each point represents data from 1 animal. (F) Representative multi-immunofluorescence staining and microscopy images (original magnification, 40×) of CD4, CD8, B220, and DAPI in the islet of Langerhans across experimental conditions. Arrow indicates islet in merge images. (G) Quantification of CD4⁺ T cell, CD8⁺ T cell, and B220⁺ B cell counts per pancreatic islet of indicated treatment condition by immunofluorescence from mice treated with Ruxo + anti–PD-1 (3 males, 2 females); data for Iso or anti–PD-1 treated mice given control chow are the same as shown in Figure 2F, reshown here for comparative purposes. Data are presented as mean ± SD (E and G). Comparisons by log-rank test (B), Fisher’s exact test (D), or ANOVA with Welch’s correction and pairwise comparison by Tukey’s test (E and G). **P < 0.01, ****P < 0.0001.

Figure 4. JAKi treatment reduces CD4⁺ Tfh cell response in mice and humans.

(A) Schematic and quantification of islet-infiltrating, CD44⁺CD4⁺ T cells among mice treated with Iso + vehicle (n = 6), anti–PD-1 + vehicle (n = 8), and anti–PD-1 + ruxolitinib (n = 4), via flow cytometry analysis. (B) Quantification of islet-infiltrating, ICOS⁺PD-1^hiCXCR5⁺CD4⁺ Tfh (left) and IL-21⁺IFN-γ⁺ Tfh cells (right), among mice treated with Iso + vehicle (n = 9–12), anti–PD-1 + vehicle (n = 13–15), and anti–PD-1 + ruxolitinib (n = 5), via flow cytometry. (C) Schematic of proposed action of JAKi on CD4⁺ Tfh cells through blockade of autocrine IL-21 signaling (left). STAT3 phosphorylation in murine CD4⁺ T cells in response to IL-21 (100 ng/mL), ruxolitinib (10 μM), or vehicle in vitro, assessed by flow cytometry (right). (D) Effect of IL-21 receptor genetic deletion in CD4⁺ T cells on Tfh cell induction in vitro, following a 3-day Tfh skew of naive CD4⁺ T cells with anti–PD-1. (E) Frequency of CD4⁺ T cells expressing a Tfh cell phenotype (CD4⁺ICOS⁺PD-1^hiCXCR5⁺) following a 3-day Tfh skew of naive CD4⁺ T cells in the presence of ruxolitinib (10 μM) or vehicle in vitro, assessed by flow cytometric analysis. (F) Expression of canonical Tfh transcription factors Bcl6 and cMAF in murine CD4⁺ T cells following a 3-day Tfh skew in the presence ruxolitinib (Ruxo, 10 μM) or vehicle. (G) Comparison of Tfh cell response in PBMC specimens from ICI-treated individuals at baseline and following a 3-day Tfh skew with JAKi ruxolitinib (Ruxo, 10 μM) or vehicle, measured by flow cytometry. Each point represents data from one replicate (C–F); experiments repeated at least twice or animal (A and B), and data are presented as mean ± SD. For human studies (G), connected points represent data from 1 individual. Comparisons by Brown-Forsythe and Welch ANOVA (A and B), 1-way ANOVA (C and D) with subsequent pairwise comparisons, Welch’s t test (E and F), or 1-way ANOVA for paired samples (G). *P < 0.05; **P < 0.01, ****P < 0.0001.