Interleukin-15-armoured GPC3 CAR T cells for patients with solid cancers

Abstract

Interleukin-15 (IL-15) promotes the survival of T lymphocytes and enhances the antitumour properties of chimeric antigen receptor (CAR) T cells in preclinical models of solid neoplasms in which CAR T cells have limited efficacy^1-4. Glypican-3 (GPC3) is expressed in a group of solid cancers^5-10, and here we report the evaluation in humans of the effects of IL-15 co-expression on GPC3-expressing CAR T cells (hereafter GPC3 CAR T cells). Cohort 1 patients ( NCT02905188 and NCT02932956 ) received GPC3 CAR T cells, which were safe but produced no objective antitumour responses and reached peak expansion at 2 weeks. Cohort 2 patients ( NCT05103631 and NCT04377932 ) received GPC3 CAR T cells that co-expressed IL-15 (15.CAR), which mediated significantly increased cell expansion and induced a disease control rate of 66% and antitumour response rate of 33%. Infusion of 15.CAR T cells was associated with increased incidence of cytokine release syndrome, which was controlled with IL-1/IL-6 blockade or rapidly ameliorated by activation of the inducible caspase 9 safety switch. Compared with non-responders, tumour-infiltrating 15.CAR T cells from responders showed repression of SWI/SNF epigenetic regulators and upregulation of FOS and JUN family members, as well as of genes related to type I interferon signalling. Collectively, these results demonstrate that IL-15 increases the expansion, intratumoural survival and antitumour activity of GPC3 CAR T cells in patients.

Extended data figure 1:. Assessment of GPC3 expression: GPC3 expression was measured by Immune-histochemistry.

A. Expression of GPC3 in pediatric tissue array – samples from hepatoblastoma and placenta were used as positive controls. Scale bar=100μm (19). B. Example of an enrollment sample from Patient 15.CAR 2. For A and B, Staining performed once in the clinical pathology laboratory with appropriate positive and negative controls. C. Intensity, extent and cumulative GPC3 expression scores for enrolled patients as previously described (20).

Extended data figure 2.. Patient accrual and enrollment:

Consort diagram summarizing patients referred, enrolled, and treated on GPC3- and 15.GPC3-CAR T cell studies.

Extended data figure 3:. Safety, antitumor activity and peripheral blood and tumor kinetics of patients treated with CAR T cells DL1 and DL2:

A. Total number adverse events and B. Number of adverse events for each patients treated at 1 x 10⁷ / m² (DL1, n=6) and 3 x 10⁷ / m² (DL2, n=6). C. Change in serum AFP levels in patients treated at DL1 dose of GPC3-CAR T cells. D. Change in tumor volume of these patients. E. Peripheral blood transgene copy numbers at indicated timepoints for patients treated at DL1 with GPC3-CAR T cells. F. Comparison of transgene copy numbers in PB (left) and tumor (right) of GPC3-CAR T cell levels treated on DL1 (n=4) and DL2 (n=4). Comparisons by two-tailed, unpaired T test and two-way ANOVA with Šidák correction. Data represented as mean ± SD.

Extended data figure 4.. Serum cytokine and chemokine kinetics post-infusion.

Levels of chemokines and cytokines were quantified on Day −4, Day 0 and weekly until Day 28. Fold change (FC) was calculated from Day 0 (baseline) to assess changes dependent on CAR T and 15.CAR T cell infusions. A. Comparison of FC from baseline to peak concentration for all measured analytes. B. Fold change and peak expansion concentration of IL15 in CAR (n=6) vs 15.CAR (n=12) treated patients. C-D. Differentially expressed cytokines in patients with (n=10) and without CRS (n=8). Overview of all measured analytes (C) and individual cytokines with at least two-fold, statistically significant increase (D). Two-tailed, unpaired T test. Data represented as mean ± SD.

Extended data figure 5.. Antitumor response characteristics in patients treated with CAR and 15.CAR T cells:

A. Long term outcome of treated patients with additional treatments shown for those with Alive with no evidence of disease (ANED). Patients needing the iC9 safety switch indicated. AWD: alive with disease. DOD: Died of disease. B. Serum alpha-feto protein (AFP) was measured in the CLIA certified clinical laboratory before and after CAR T cell infusions. Waterfall plot representing changes in AFP concentration from baseline in patients with AFP secreting tumors. C. Differentially expressed cytokines in non-responders (NR) vs responders (R) according to RECIST criteria. D. Comparison of individual cytokines with at least two-fold, statistically significant increase in responder (R, n=4) and non-responder (NR, n=14) patients. E. Comparison of GPC3 expression of tumors from responder (R, n=4) and non-responder (NR, n=14) patients. Two tailed, unpaired T test. Data represented as mean ± SD.

Extended data figure 6.. Gene expression, cell surface marker phenotype and polyfunctionality of CAR vs 15.CAR T cell infusion products.

A. UMAP projections showing indicated genes for individual cells after combining data from all CAR and 15.CAR T cell infusion products. Region outlined with orange corresponds to 15.CAR enriched cells. Cells were stained for expression of indicated cell surface markers. B. Metabolic profile of CAR and 15.CAR products using scRNAseq. C. CD4 (top) and CD8 (bottom) positive CAR T cells’ expression of exhaustion (LAG3, PD1, TIM3) markers and proportion of CD39/CD69 subsets in all CAR (n=12) and 15.CAR (n=12) T cell infusion products. Comparison by two-way ANOVA with Šidák correction for multiple comparisons. D-E. Manufactured CAR (n=12) and 15.CAR (n=12) cells were evaluated by the Isoplexis, single cell cytokine detection system. (D). Polyfunctionality index of CD4 subset and (E). Polyfunctionality strength index of the indicated products. Two-way ANOVA with Šídák correction. Data represented as mean ± SD.

Extended data Fig 7.. Expansion, persistence and trafficking of CAR vs 15.CAR T cells post-infusion.

Expansion and persistence of infused cell populations were quantified with qPCR. A-B. Transgene copy numbers for the GPC3-CAR (A) and iC9.NGFR.IL15 (B). C. Comparison of CAR and iC9.NGFR.IL15 transgene expression in non-responder (NR, n=8) vs responder (R, n=4) products and peripheral blood samples at indicated timepoints by flow cytometry. D. CAR- negative NK and iNKT subsets in peripheral blood isolated from patients infused with CAR (n=5) or 15.CAR (n=9) measured by flow cytometry. E-F. GPC3-CAR transgene frequencies in tumor biopsies in CAR (n=5) vs 15.CAR (n=10) groups (E) and in R (n=3) vs NR (n=12) groups (F). Two-tailed, unpaired T test. Data represented as mean ± SD.

Extended data Fig 8.. Gene expression evolution in CAR and 15.CAR T cells post-infusion.

The transcriptomic profile of Infusion products and peripheral blood CAR and 15.CAR T cells (A-C) or infusion products and tumor infiltrating 15.CAR T cells (D) were interrogated with single cell RNA sequencing. Differentially expressed genes (DEGs) for indicated groups were determined by comparing the product and post-infusion samples. A. DEGs from infusion product to peripheral blood represented by log2 fold change (log2FC) in CAR (y axis) vs 15.CAR (x axis) T cells. Linear regression. B. Selected gene sets enriched in CAR vs 15.CAR T cells in PB. C. Heatmap representing a subset of DEGs from the pre-infusion product to PB comparison in CAR vs. IL15.CAR. D. Selected cluster specific DEGs in 15.CAR T cells captured in tumors post-infusion. Cluster 9 contains only Non-responder cells. E. CD4/8 T cell subset composition of CAR+ TILs in responder (R, n=5) and non-responder (NR, n=3) tumor biopsies post-infusion in the 15.CAR group. F. CAR-negative lymphocyte subset composition in responder (R, n=5) and non-responder (NR, n=3) tumor biopsies post-infusion in the 15.CAR group. Data represented as mean ± SD.

Extended data Fig 9.. Characteristics of CAR+ and CAR- T cells in patients post-infusion.

A. Expression of indicated genes in CAR-negative, bystander tumor infiltrating lymphocytes (TILs). B. Gene expression signatures in CAR-positive and CAR-negative TILs in tumors of responder (R) and non-responder (NR) patients. C. T cell clones of CAR-positive and CAR-negative subsets from product, blood and tumor samples. Colored lines correspond to VDJ clones trackable between product / peripheral blood and tumor-derived T cells. D. Proportion >1 cells with the same VDJ TCR sequence of responder (R, n=5) and non-responder (NR, n=3) tumor infiltrating CAR-positive and CAR-negative T cell subsets. Two-tailed, unpaired T test, data represented as mean ± SD.

Extended data Fig 10:. Gating strategy for product and peripheral blood phenotyping.

Patient-derived and product samples were processed and stained with indicated fluorochrome-conjugated antibodies. A. After gating on live cells and removing duplets, the T cell population was defined as CD3+ and evaluated for CAR and IL-15 expression based on Anti Fab APC and NGFR PE, respectively. B. Manufactured products were further characterized based on gating strategy in A. The CAR+ subset was further analyzed to determine CD4/CD8, memory, and exhaustion markers. C. Gating strategy to characterize NK and iNKT subsets in peripheral blood.

Fig 1.. Safety characteristics of CAR and 15.CAR infusions:

A. Transgene maps of GPC3-CAR and iC9.NGFR.IL15 constructs used to co-transduce T cells to generate infusion products. B. Schematic representation of patient enrollment and treatment. C. Bubble plots representing frequency of the indicated adverse events for CAR (left) and 15.CAR (right) infused patients. Adverse events (AEs) were collected from Day −4 until Day +28 post-infusion and graded according to the Common Terminology Criteria of AEs v5. Color spectrum corresponds to AE grade 1-5, bubble size corresponds to frequencies. D. Comparison of frequency of AEs between CAR (n=6) vs 15.CAR (n=12) and E. Comparison of AEs between patients with (n=10) and without CRS (n=8) using two-tailed T test and two-way ANOVA with Šidák correction, respectively. Mean ± SD. F-H. Proportions of GPC3-CAR and iC9.NGFR.IL15 expressing T cells quantified by flow cytometry (F) and qPCR (G, red arrows indicate timing of rimiducid administration) and changes in concentrations of indicated serum cytokines (H) in peripheral blood of patients treated with rimiducid, the chemical inducer of the iC9 safety switch.

Fig 2.. 15.GPC3-CAR T cells induce significant antitumor responses in patients.

Antitumor responses were determined by comparing pre- and post-infusion 3D imaging. A. Coronal CT chest (15.CAR 1), axial CT chest (15.CAR 4), PETCT (15.CAR 7), MRI abdomen (15.CAR 9) and axial CT abdomen (15.CAR 12) based images showing pre- and post-CAR T cell infusion. Red arrows and circles represent tumors. B. Waterfall plot representing changes in tumor volumes of patients treated with 3 x 10⁷/m² with CAR or 15.CAR T cells. Red line: 20% increase, Blue line: 20% decrease, Green line: 30% decrease. C. Serum alpha-feto protein (AFP) concentrations at indicated timepoints in responders with AFP secreting neoplasms. D. Pre-and post-infusion tumor biopsy assessed with hematoxylin-eosin staining showing near complete necrosis of patient 15.CAR 9’s liver tumor performed once in the clinical pathology laboratory. Scale bar=50μm.

Fig 3.. Comparison of pre-infusion products and expansion in patients of CAR and 15.CAR T cells.

Products were first assessed with single cell RNA sequencing. A. Uniform Manifold Approximation and Projection (UMAP) identifying unique T cell clusters in integrated CAR and 15.CAR pre-infusion products. B-C. Differential representation of CAR (n=12) vs 15.CAR (n=12) T cells in clusters shown in UMAP projection (B) and proportion for each cluster (C). Center line: median, box limits: first and third quartiles, whiskers: 1.5x the interquartile range, dots: outliers. Percentages of cells in cluster were compared using unpaired two-tailed Student's t test. Shown are adjusted p-values using the Benjamini and Hochberg correction. D. Differentially expressed genes in CAR (n=12) vs 15.CAR (n=12) products. E. Frequencies of CD4/8 and effector / memory subsets in CAR (n=12) vs 15.CAR (n=12) products by flow cytometry. Data represents mean ± SD. Two-tailed, unpaired T test and two-way ANOVA with Šídák correction, respectively. F. Cytotoxicity of CAR (n=12) vs 15.CAR (n=12) products measured by ⁵¹Cr release assay. Two-tailed, unpaired T test, data represented as mean ± SD. G. Polyfunctionality strength index comparing CAR (n=12) vs 15.CAR (n=12) T cell product’s cytokine production by Isoplexis. Two-way ANOVA with Šídák correction. H. Differentially expressed cytokines in CD8 subsets of CAR (n=12) and 15.CAR (n=12) T cell products. Two-tailed, unpaired T test, data represented as mean ± SD. I. Peripheral blood CAR T cell frequencies quantified by qPCR at indicated timepoints for each patient. J. Comparison of peak expansion on dose level 2 of CAR (n=6) vs 15.CAR (n=12) T cells post-infusion. Two-tailed, unpaired T test. Data represented as mean ± SD. K. Comparison of expansion of cells in responders (n=4) vs non-responders (n=14). Two tailed, unpaired Mann-Whitney test, data represented as mean ± SD.

Fig 4.. Comparing the single cell gene expression profile of tumor infiltrating 15.CAR T cells post-infusion in responders vs non-responders.

The transcriptomic profile of Infusion products and tumor infiltrating 15.CAR T cells were interrogated with single cell RNA sequencing. Differentially expressed genes (DEGs) for indicated groups were determined by comparing the product with tumor-derived 15.CAR T cells. A. UMAP projection of tumor infiltrating 15.CAR T cells from responders and non-responders. B. Unsupervised clustering of tumor infiltrating CAR T cells of responders and non-responders from the 15.CAR cohort. C. Differences in cluster proportions for indicated groups. D. DEG comparison (product vs tumor infiltrating 15.CAR T cells) in responders (x axis) vs non-responders (y axis) from the 15.CAR cohort. Linear regression. E. Gene sets enriched in tumor infiltrating 15.CAR T cells. F. Heatmap representing the genes with most differences in change from baseline in responders vs non-responders in 15.CAR T cells.