Idelalisib reduces regulatory T cells and activates T helper 17 cell differentiation in relapsed refractory patients with chronic lymphocytic leukaemia

Summary

Phosphatidylinositol 3 kinase (PI3K) inhibitors such as idelalisib have been associated with potentially severe autoimmune toxicity. In the present study, we demonstrate that relapsed refractory patients with chronic lymphocytic leukaemia treated with idelalisib rituximab on the phase III registration trial show uniform decrease in regulatory T cells (Tregs) and increase in CD8 T cells with treatment. Patients who do not develop toxicity show enrichment for T cells expressing multiple chemokine receptors, while those who do develop toxicity have an activated CD8 T cell population with T helper 17 cell differentiation at baseline, which then increases, leading to an increased CD8:Treg ratio that likely triggers autoimmune toxicity.

Idelalisib is the first in class specific inhibitor of the delta isoform of phosphatidylinositol 3 kinase (PI3K) to be approved by the United States Food and Drug Administration (FDA), specifically with rituximab for the therapy of relapsed refractory patients with chronic lymphocytic leukaemia (CLL) and as a single agent for the therapy of relapsed refractory patients with follicular lymphoma. Since then, use of idelalisib and other agents in this class has been limited by idiosyncratic and sometimes severe toxicity, in particular diarrhoea/colitis,¹ but also including transaminase elevation that we have shown to be autoimmune,² as well as interstitial pneumonitis and rash that are also believed to be autoimmune. In previous work, we have shown that approximately three-quarters of patients who received front-line idelalisib on an investigator-initiated trial showed a decrease in regulatory T cells (Tregs) in the first month on therapy, with greater decreases in those patients with early autoimmune hepatotoxicity.² Our study, as well as cross-study comparison,^3,4 suggested that these toxicities were more frequent among previously untreated patients. This observation was recently confirmed by a pooled analysis of all fully enrolled phase II or III Gilead-sponsored studies, in which we clearly demonstrate that the rate of Grade 3–4 transaminitis is higher among previously untreated patients compared to relapsed refractory patients.⁵ We were therefore interested in investigating whether idelalisib has similar T-cell effects in relapsed refractory patients as in previously untreated patients.

We undertook this study in which we used mass cytometry by time-of-flight (CyTOF) to evaluate T cell subsets in peripheral blood mononuclear cell (PBMC) samples from patients who did and did not develop significant autoimmune toxicity while enrolled on the idelalisib arm of studies GS-US-312-0116 (NCT01539512) and GS-US-312-0117 (NCT01539291) (Table S1). Study 116 is a phase III randomised registration trial that compared idelalisib rituximab with placebo rituximab in heavily pre-treated high-risk relapsed/refractory patients with CLL, and study 117 was an extension study for 116 that allowed idelalisib monotherapy for patients on either arm.⁶ These studies were approved by the Human Protections Committee at all participating sites. From samples collected by Gilead during the trial, nine patients had appropriate serial samples from prior to therapy and prior to an adverse event (Table S2). Of the cohort, three patients had Grade ≥3 diarrhoea/colitis; three patients had Grade ≥3 transaminitis; and three patients had no adverse event. All patients who developed toxicity were treated with interruption of idelalisib dosing, and those with diarrhoea/colitis also received steroids. These patients were otherwise representative of the overall trial population, with a median (range) age of 71 (61–87) years, median Cumulative Illness Rating Scale (CIRS) score of 9, median of two prior therapies, four of nine with 17p deletion or tumour protein p53 (TP53) mutation, and eight of nine with unmutated immunoglobulin heavy-chain variable-region (IGHV) (Table S1).

To determine the difference in T-cell subsets in particular between pre- and on-treatment PBMC samples in patients who did or did not develop immune-related adverse events (irAEs), we used mass cytometry to study a total of 34 serial samples from nine patients, at screen (n = 9) and at a variety of on-treatment time-points commonly including study visits six, 11, 12, and visit four on the extension trial (Table S2). For analysis we focused on comparison of patients who did and did not develop toxicity, comparing screening and a key subsequent time-point, either the time-point just before the AE in those with irAEs (or just after if prior was not available), or V11 in the patients without AEs, which was most similar to the AE time-points with respect to time on therapy (Table S2). Unsupervised analysis of the CyTOF staining data identified 39 clusters after performing proportional sampling of single cell events, including all patients and all sample time-points (Figure 1A; Table S3). Based on the expression of antibody markers, all 39 clusters were broadly assigned to seven islands belonging to CD4 T, CD8 T, Tregs, double positive T (DP T), double negative T (DN T), T helper 17 (Th17) and non-T cell subsets (Figure 1B). The event counts of these cell clusters were exported and analysed for statistical significance using the quasi-likelihood negative binomial generalised log-linear model as implemented in Bioconductor library edgeR.⁷ Comparing patients with and without toxicity identified two clusters enriched at baseline and three at time of toxicity in patients with toxicity, and two enriched in patients without toxicity at both time-points (Figure 1C–E).

FIGURE 1.

Overview of mass cytometry by time-of-flight (CyTOF) data from all patients. (A) Unsupervised clustering of CyTOF data from all nine patients studied identifies 39 separate clusters in the cohort. Highlighted in black boxes are the most significantly altered clusters featured later in the manuscript. (B) Seven major islands were identified. DP, double positive; DN, double negative. (C) Volcano plot illustrating significantly different clusters at baseline between patients who did (right panel) and did not (left panel) develop immune-related adverse events (irAEs). Clusters that are significantly different defined as a false discovery rate (FDR) adjusted p < 0.1 are highlighted in red. (D) Volcano plot illustrating significantly different clusters at subsequent time-point at FDR adjusted p < 0.1 between patients who did (right panel) and did not (left panel) develop irAEs. (E) Detail of the subdistribution of individual clusters discussed in the manuscript, among the major islands, across all patients studied. CD8⁺ T cells are circled in blue, and CD4⁺ T cells circled in orange. Clusters that are significantly different defined as adjusted p < 0.1 are highlighted in red [Colour figure can be viewed at wileyonlinelibrary.com]

The two clusters enriched at baseline in patients who developed toxicity were c35 and c36. Clusters 35 and 36 are both activated CD8⁺ T cells expressing granzyme B, Helios, T-box expressed in T cells (T-bet) and T cell immunoglobulin and immunoreceptor tyrosine-based inhibitory motif domains (TIGIT) (Figure 1E and Figure S1a,b). c35 shows CD45RA and CD161, with some retinoic acid receptor-related orphan nuclear receptor gamma T (RORγT) expression (Figure S1a), while c36 shows RORγT (Figure S1B). CD161 is a lectin-like receptor associated with tissue homing and found on CD8 and interleukin 17-expressing T cells,^8,9 while RORγt is a transcription factor associated with Th17 differentiation. Hence, both these clusters suggest increased Th17 activity at baseline in patients who go on to develop toxicity, consistent with our findings also in previously untreated patients with CLL receiving PI3Kδ inhibitors.¹⁰ Both clusters also increase at the time of toxicity (Figure S2A,B and S3A). Clusters nine and 12, that show enrichment at the later toxicity time-point but not at baseline, are also both activated CD8 T cells (Figures S1C,D, S2C,D, and S3B). Cluster 12 is an activated CD8 T cell that clearly increases at the time of toxicity and is defined by the markers CD38, human leucocyte antigen-DR isotype (HLA-DR), T-bet, granzyme B, galectin 9 and CD8a (Figure S1D).

Interestingly both clusters enriched in patients without toxicity (seven and 37) are enriched throughout the idelalisib course (Figures S1E,F and S2E,F). Cluster seven is defined by chemokine (C-C motif) receptor (CCR)6, CCR9, glucocorticoid-induced tumour necrosis factor receptor related protein (GITR), C-X-C motif chemokine receptor 3 (CXCR3), CD146, and CCR4, also with CD4 and CD45RO expression, so a memory CD4 T cell with high expression of chemokine receptors (Figure S1E). Cluster 37 is a CD3⁺ T cell without expression of either CD4 or CD8, but also with expression of multiple chemokine receptors as well as CD161 (Figure S1F). Whether and how these chemokine receptors may be protective is not clear, although a DN CD161⁺ cell has been described to suppress Th17 cells.¹¹

When all CD8 T cells are analysed together by manual gating, we see a trend to an increase at the later time-points in all patients, with or without toxicity (Figure 2A), as we have previously seen also with duvelisib in a front-line setting.¹⁰ Tregs also go down concomitantly in all patients (Figure 2B), with no particular trend to a greater change among those with toxicity (Figure S4). Evaluating the CD8:Treg ratio reveals a significant increase only in the patients with toxicity, albeit with similar trend in those without toxicity (Figure 2C; Figure S5). No definite change in granzyme B expression is seen (data not shown), while cytotoxic T-lymphocyte associated protein 4 (CTLA-4) goes down on CD8s (data not shown), suggesting greater activation potential.

FIGURE 2.

Changes in regulatory T cells (Tregs) and CD8s. (A) Change in CD8⁺ T cells between baseline and on-treatment time-point, in the entire population. The on-treatment time-point is before the adverse event (AE) for those with immune-related AEs (irAEs), or V11 for those without irAEs. (B) Change in Tregs between baseline and on-treatment time-point, in the entire population. (C) Change in the CD8:Treg ratio between baseline and on-treatment time-point, showing a significant difference only in those patients with toxicity. (D) Schematic illustrating the study rationale, hypothesis, and findings. Paired t-test p values for each comparison in panels A–C are included above the line in each panel [Colour figure can be viewed at wileyonlinelibrary.com]

In this study of T-cell subsets in heavily pre-treated patients receiving idelalisib, we see a uniform decrease in Tregs with inhibition of PI3Kδ, which is consistent with our prior data in untreated patients receiving idelalisib² and with mouse models¹² (Figure 2D). The decrease in Tregs was more pronounced with autoimmune toxicity in prior studies but here is seen across all patients. A recent study evaluating the effects of idelalisib, duvelisib and umbralisib on T cells in vitro and in mice also reported that Tregs decrease with all of these drugs, albeit to varying degrees,¹³ and this study associated the level of Treg depletion with autoimmune toxicity in the mice.¹³ All of these data taken together strongly suggest that loss of Tregs, as seen here, contributes to the autoimmune phenotype of PI3Kδ inhibition. In this study, we also find that patients with toxicity have evidence of an activated CD8 T-cell population with Th17 differentiation at baseline that increases at the time of toxicity, and an associated increase in their CD8:Treg ratio. These data are similar to our recent findings among patients with CLL receiving duvelisib with fludarabine/cyclophosphamide/rituximab (FCR) in a front-line setting.¹⁴ This increasing body of work implicates not just alteration in Tregs but also baseline Th17 differentiation and increase in CD8 T cells during therapy among the causes of PI3K inhibitor-induced toxicity. Future work should focus on validation in larger cohorts as well as attempting to mitigate this toxicity by identifying plasma or cellular biomarkers for risk or a combination therapy partner that mitigates these T-cell effects.