PI3Kδ/γ inhibition promotes human CART cell epigenetic and metabolic reprogramming to enhance antitumor cytotoxicity

Abstract

Current limitations in using chimeric antigen receptor T(CART) cells to treat patients with hematological cancers include limited expansion and persistence in vivo that contribute to cancer relapse. Patients with chronic lymphocytic leukemia (CLL) have terminally differentiated T cells with an exhausted phenotype and experience low complete response rates after autologous CART therapy. Because PI3K inhibitor therapy is associated with the development of T-cell-mediated autoimmunity, we studied the effects of inhibiting the PI3Kδ and PI3Kγ isoforms during the manufacture of CART cells prepared from patients with CLL. Dual PI3Kδ/γ inhibition normalized CD4/CD8 ratios and maximized the number of CD8+ T-stem cell memory, naive, and central memory T-cells with dose-dependent decreases in expression of the TIM-3 exhaustion marker. CART cells manufactured with duvelisib (Duv-CART cells) showed significantly increased in vitro cytotoxicity against CD19+ CLL targets caused by increased frequencies of CD8+ CART cells. Duv-CART cells had increased expression of the mitochondrial fusion protein MFN2, with an associated increase in the relative content of mitochondria. Duv-CART cells exhibited increased SIRT1 and TCF1/7 expression, which correlated with epigenetic reprograming of Duv-CART cells toward stem-like properties. After transfer to NOG mice engrafted with a human CLL cell line, Duv-CART cells expressing either a CD28 or 41BB costimulatory domain demonstrated significantly increased in vivo expansion of CD8+ CART cells, faster elimination of CLL, and longer persistence. Duv-CART cells significantly enhanced survival of CLL-bearing mice compared with conventionally manufactured CART cells. In summary, exposure of CART to a PI3Kδ/γ inhibitor during manufacturing enriched the CART product for CD8+ CART cells with stem-like qualities and enhanced efficacy in eliminating CLL in vivo.

Graphical abstract

Figure 1.

Culture with PI3Kis increased the frequencies of CD8⁺ naive and memory T cells and decreased the frequencies of senescent CD8⁺ T cells from untreated patients with CLL. Mononuclear cells collected from CLL donors were analyzed by flow cytometry. Cryopreserved cells were thawed, and T cells were isolated and expanded for 9 days with duvelisib or idelalisib added to the culture medium every 3 days (q3d). (A) T cells from healthy donors (n = 5) and patients with CLL (n = 18) assessed by flow cytometry for expression of costimulatory molecules (CD27 and CD28) on the CD8⁺ and CD4⁺ subsets. (B) The frequencies of CD8⁺CD27⁺CD45RO⁻ T cells in patients with CLL compared with healthy controls. The dashed line indicates the threshold frequency of CD8+ naive T-cells previously associated with complete remission in patients with CLL treated with CD19-targeted CART cells. In vitro expansion of T cells from patients with CLL (n = 8) cultured with idelalisib (C) and duvelisib (D), with the 50% inhibitory concentration (IC₅₀) for PI3Kδ and PI3Kγ shown as vertical dashed lines. Flow cytometric analysis of the frequency of CD8⁺ T cells (E) and CD4/CD8 T-cell ratios (F) in duvelisib-containing and control T-cell cultures. (G) Expression of the immune checkpoint molecules TIM3, LAG3, and PD1 on the CD8⁺ T cells. (H) Frequency of TIM3 on the CD4⁺ T-cell subset. LAG3 and PD1 yielded no significant changes and are not shown on the CD4⁺ subset. (I) Distribution of T cells according to naive and memory phenotypes in control and duvelisib-containing cultures. T-cell populations were defined as follows: naive and T_SCM (CD45RA⁺CD45RO⁻CCR7⁺), central memory (CD45RA⁻CD45RO⁺CCR7⁺), T-memory and effector memory (CD45RA⁻CD45RO⁺CCR7⁻), and terminal effector cells (CD45RA⁺CD45RO⁻CCR7⁻). *P < .05; **P ≤ .01; ***P ≤ .001; ****P ≤ .0001.

Figure 2.

Duvelisib added during CART cell culture enriched CD8+ T_SCM, naive, and central memory CD8⁺ CART cells and modulated CD4⁺ T helper cell polarization. Isolated CD3⁺T-cells from 3 patients with CLL were transduced with anti-CD19 CAR using a multiplicity of infection of 30 and cultured for 14 days, with or without 300 nM duvelisib, to yield control CART cells and Duv-CART cells. Mass cytometry by time of flight analysis was performed, and tSNE was used to reduce all phenotypic data into 2 dimensions, tSNE1 and tSNE2. (A) A representative viSNE plot from 1 of the 3 CLL samples analyzed. (B) Unsupervised FlowSOM showing 10 qualitative clusters. Changes in the size of a given cluster denote differences between control CART cells (0.1% DMSO) and Duv-CART cells. Th1 was defined as CD4⁺CXCR3⁺ and Th2 as GATA3⁺CXCR4⁺. (Confirmatory cytokine expression studies were not performed.) (C) Differences in levels of CART cell phenotypes assessed by CITRUS analysis. **P ≤ .01; ns, not significant. tSNE, t-distributed stochastic neighbor embedding; viSNE, visualization SNE.

Figure 3.

Duv-CART cells had lower expression of immune checkpoint molecules and enhanced antileukemia activity against OSU-CLL cells. Control CART cells and Duv-CART cells from 8 patients with CLL were cultured as described in Figure 2. Number of CART cells (A) and transduction efficiency (B). (C) yields of CD8⁺CART cells and CD4/CD8 CART-cell ratios in duvelisib vs control cultures. (D) Absolute number of naive CD27⁺CD45RO⁻CD8⁺ CART cells. (E) Frequencies of CD8⁺ naive and CD4⁺ central memory CART-cell phenotypes. (F) Expression of TIM3, LAG3, and PD1 on CD4⁺ and CD8⁺ subsets of CART cells. (G) Cytotoxicity of control and Duv-CART cells (effectors) from 4 of the patient CLL samples against OSU-CLL cells (targets) at effector-to-target ratios of 0.5:1, 1:1, and 2:1. Duv-CART dells were tested before (with duvelisib) and after (washed) washing to remove residual duvelisib from the reaction mixture. (H) Data from panel G were analyzed to show CD8 effector-to-target ratios and percentage of specific cell lysis. (I) CD8⁺ control CART and Duv-CART cells were sorted (n = 4 donors) and added to target cells at defined ratios. *P < .05; **P ≤ .01; ***P ≤ .001; ****P ≤ .0001.

Figure 4.

PI3K inhibition during T-cell culture–enhanced TCR and MEK/ERK signaling with alterations in epigenetic regulators that promote T-cell stemness. T cells from patients with CLL (n = 9) were cultured with or without 300 nM duvelisib administered to cultures every 3 days and were harvested after 3, 9, or 15 days of culture. Cells were stimulated with anti-CD3/CD28 beads on days 1 and 9 of culture, with day-9 samples harvested 90 minutes after restimulation with anti-CD3/CD28 beads and addition of duvelisib. Hypothesis-generating NanoString analysis was first performed and followed-up by confirmatory western blot experiments. (A) Differential expression of metabolically relevant genes in duvelisib-cultured cells compared with control cells at day 9 of culture. Changes in messenger RNA levels for 102 genes met Benjamini-Hochberg (BH) false-discovery thresholds (red). The 5 genes with the greatest fold increase or decrease are labeled. (B) Pathway scores based on gene expression levels for TCR, costimulatory, MAPK and mTOR signaling pathways. Western blot of cell lysates probing proteins related to proliferative pathways (C) and epigenetic regulatory pathways (D). Quantifications of select western blots (SIRT1, FOXO1, and TCF1/7) normalized to control (E) with ancillary flow cytometric analysis of frequencies of TCF1/7 expressing cells (F) at day 15 of culture. Results reproduced across 3 patients with CLL with quantitative results shown in the supplemental Data (supplemental Figure 5). *P < .05; **P ≤ .01; ***P ≤ .001.

Figure 5.

Duvelisib added during CART cell culture increased mitochondrial fusion with increased mitochondrial fusion proteins but did not change the OCRs. (A) T cells cultured with or without PI3K inhibitors were stained with NAO and assessed by flow cytometry, with representative NAO fluorescence shown on the CD4⁺ and CD8⁺ T-cell subsets. (B) Transmission electron microscopic images of nontransduced and CAR transduced T cells at day 14 of culture, with cell cross-sectional areas and mitochondrial cross-sectional area as a percentage of total cell area shown below each image. Images with cell size and mitochondrial area representative of the mean for each group are shown. FIJI software measured T-cell sizes across >10 replicate images for each patient and condition. Representative images from >420 acquired images are shown. Red arrows, mitochondria. (C) T-cell size, shown as cross-sectional area, over time in culture without or with duvelisib. (D) Differences in cross-sectional area of cytoplasm and nucleus at day 14 of culture in control and duvelisib T cell cultures. (E) Violin plots showing mitochondrial area as a percentage of T cell cross-sectional area for nontransduced and CART transduced cells. (F) Representative western blot analysis of mitochondrial fusion proteins MFN1 and MFN2. (G) Evaluation of oxygen consumption rates using a Seahorse bioenergetics bioanalyzer of OCR and extracellular acidification rate (ECAR) for CLL T-cell cultures. Cells were plated at 4 × 10⁵ per well for both control CART and Duv-CART. Results from 4 representative CLL samples from 12 replicates are shown. (H) Quantified western blots (MFN1, MFN2, and p.DRP1) from 3 to 5 CLL donors for select proteins. *P < .05; **P ≤ .01; ****P ≤ .0001; ns, not significant.

Figure 6.

Duv-CART cells conferred a survival advantage in mice bearing intermediate disease burden OSU-CLL. OSU-CLL was engrafted in NOG mice and upon OSU-CLL (herein called CLL) reaching a mean of 1.2% (CD28/CD19) or 0.15% (41BB/CD19) of nucleated cell content, mice were treated with 1.0 × 10⁶ control CART or Duv-CART cells on days 15 to 18. Data from anti-CD19 CART with a CD28 costimulatory domain (CD28/CD19 CART) are in the left panels, and data from anti-CD19 CART with a 41BB costimulatory domain (41BB/CD19 CART) are shown in the right panels. (A) Kaplan-Meier survival analysis of CD28/CD19 control- and Duv-CART-cell–treated mice. (B) Frequency of CLL cells in peripheral blood, defined by flow cytometry as CD20⁺CD5⁺ over time after CLL engraftment. The in vivo expansion of total CART cells (C) and CD8⁺ CART cells (D) over time since infusion of CART. (CART cells were gated based on expression of GFP.) (E) Representative flow cytometry plots from peak expansion (day 18 after CART infusion). (F) Kaplan-Meier survival analysis of 41BB/CD19 control- and Duv-CART–treated mice. (G) Frequency of CLL in peripheral blood, defined by flow cytometry as CD20⁺CD5⁺ over time after CLL engraftment. The in vivo expansion of total human T cells (H) and CD8⁺ T cells (I) over time since infusion of CART. (The secondary antibody against the 41BB CAR failed to detect the CAR, and human CD3 was therefore used as a proxy for CAR-expressing cells.) (J) Representative flow cytometry plots from peak expansion (day 21 after CART infusion). In vivo immune checkpoint expression over time for CD28/CD19 CART within the CD4⁺ T- cell subset (K) and CD8⁺ T-cell subset (L). In vivo immune checkpoint expression over time for 41BB/CD19 CART within the CD4⁺ T-cell subset (M) and CD8⁺ T-cell subset (N). *P < .05; **P ≤ .01; ***P ≤ .001; ****P ≤ .0001.

Figure 7.

A proposed mechanism for the increased antileukemia activity of Duv-CART cells. T-cell receptor signaling activates PI3K phosphorylation through phosphorylation of protein lipase C (not shown). Duvelisib inhibits the active sites of PI3Kδ and PI3Kγ at the p110δ and p110γ subunits. Processes that duvelisib upregulates are shown in green, and processes that duvelisib downregulates are shown in red.