Robust differentiation of NK cells from MSLN.CAR-IL-15-engineered human iPSCs with enhanced antitumor efficacy against solid tumors

Abstract

Human induced pluripotent stem cells (iPSCs) offer a promising source for chimeric antigen receptor (CAR)-engineered natural killer (NK) products. However, complex iPSC-NK (iNK) manufacturing challenges clinical use. Here, we identified LiPSC-GR1.1 as a superior iPSC line for iNK production. By engineering LiPSC-GR1.1 with a mesothelin (MSLN)-targeting CAR and interleukin-15 (IL-15), we achieved robust differentiation of iPSCs into mature activated iNK cells with enhanced tumor killing efficacy, superior tumor homing, and vigorous proliferation. Single-cell transcriptomic analysis revealed that transforming growth factor-β (TGF-β)-producing tumor cells up-regulated major histocompatibility complex molecules and down-regulated MSLN post-CAR-IL-15 iNK treatment. Tumor-infiltrating CAR-IL-15 iNK cells exhibited high levels of CAR, IL-15, and NK-activating receptors, negligible checkpoint exhaustion markers, and extremely low levels of NK suppressive factors CISH, TGFBR2, and BATF, enabling them to sustain activation, metabolic fitness, and effective tumor killing within TGF-β-rich hypoxic tumor microenvironment. Overall, we developed MSLN.CAR-IL-15-engineered GR1.1-iNK therapy with enhanced antitumor efficacy for solid tumor treatment.

Fig. 1.. Differentiation of iNK cells from six human iPSC lines.

(A) Scheme of NK differentiation from iPSCs. EBs were generated using the spin method from a single-cell dissociated human iPSCs seeded on day −6 and maintained under feeder- and serum-free condition in a U-bottom 96-well plate. Six days later (after hematopoietic progenitor cells generated), EBs were transferred into NK cell differentiation conditions on day 0. NK cells were then harvested 4 weeks postdifferentiation. (B) The expression of pluripotency markers SSEA-4 and TRA-1-60 was detected in the iPSC lines via flow cytometry. (C) Left panel: morphology of day 0 EBs and 4 weeks of NK differentiation from six different iPSC cell lines. Scale bars represent 1000 μm. Right panel: representative magnified pictures of differentiated iNK cells 4 weeks post-NK differentiation. Scale bars represent 200 μm. (D) iNK differentiation yield per EB from different iPSC lines (n = 2 to 3 independent differentiation experiments performed for each iPSC line). ND means that no differentiation was observed in these lines. *P < 0.05 and ***P < 0.001, analyzed using unpaired Student’s t test. (E) NK differentiation from the three iPSC lines LiPSC-GR1.1, NCRM5, and NCRM6 was validated by staining for CD45 and CD56. (F) The median fluorescence intensity (MFI) of CD56 was measured by flow cytometry on PBNK cells (from three healthy donors) and iNK cells (derived from LiPSC-GR1.1, NCRM5, and NCRM6).

Fig. 2.. Robust iNK differentiation yield from MSLN.CAR-IL-15–transduced LiPSC-GR1.1.

(A) MSLN.CAR-IL-15 piggyBac transposon vector construct encoding the CAG promoter, MSLN.CAR (consisted of MSLN-targeted hYP218 scFv, CD8a hinge spacer, NKG2D transmembrane domain, 2B4 ICD, and CD3ζ chain stimulatory domain), EGFRt, and human IL-15. GMCSFRss, GM-CSF receptor-α chain signal sequence directing cell surface expression. (B) Genetic modification of LiPSC-GR1.1 cells with MSLN.CAR and MSLN.CAR-IL-15 piggyBac transposon systems. MSLN.CAR expression was examined by measuring EGFRt expression in transfected iPSCs. (C) Human IL-15 production in the supernatant of iPSC culture. ****P < 0.0001. (D) NK differentiation yield per EB from MSLN.CAR-IL-15–modified iPSCs and control iPSCs. ***P < 0.001. (E) iNK differentiation yield achieved with different concentrations of IL-15 (10, 20, 40, and 80 ng/ml) added to the mock LiPSC-GR1.1-iNK differentiation media. *P < 0.05. (F) NK differentiation and CAR expression in NK were validated by detecting CD45, CD56, and EGFRt expression in the cells. (G) Cryopreserved differentiated iNK cells were thawed and expanded for 7 days. The fold increase of expansion is shown. **P < 0.01. The statistics were analyzed using unpaired Student’s t test.

Fig. 3.. Characterization of freshly differentiated GR1.1-iNK cells.

GR1.1-iNK cells were harvested 4 weeks post-EB differentiation. (A) The cell surface receptor expression in NK cells was detected via flow cytometry. CD56⁺CD3⁻ PBNK and CD56⁺ GR1.1-iNK cells were gated for the indicated biomarker analyses. The dark gray area represents biomarker staining. The light gray area represents isotype control. (B to D) RNA-seq of PBNK, mock iNK, and MSLN.CAR-IL-15 iNK cells. Data are from one RNA-seq analysis with three biological replicates per group. (B) PCA of DEGs. Each dot represents one sample. Each color represents an NK population. (C) Bubble plot of GSEA analysis showing the highest-ranked top 10 KEGG and Hallmark gene sets significantly up-regulated (P < 0.05, adjusted P < 0.1) in mock iNK versus PBNK (top panel) and MSLN.CAR-IL-15 iNK cells versus mock iNK cells (bottom panel). The x axis represents the enrichment score, and the size of the bubble represents the number of genes in the gene set. (D) The heatmap shows the gene expression [log₂(TPM + 1)] (where TPM means transcripts per million) of NK phenotypic markers in PBNK, mock iNK, and MSLN.CAR-IL-15 iNK cells, including NK signature, activating receptors, inhibitory KIR, checkpoint, cytokine and receptors, chemokine receptors, and cytolytic molecule–related genes.

Fig. 4.. In vitro tumor killing function of MSLN.CAR-IL-15 GR1.1-iNK cells.

(A) Cytotoxicity of mock iNK, MSLN.CAR iNK, and MSLN.CAR-IL-15 iNK cells against MSLN⁺ KLM1-WT and MSLN⁻ KLM1-KO cells. (B) Cytotoxicity of all the iNK cells against multiple MSLN⁺ solid tumor cell lines. (C) CD107a degranulation signature induced by the coculture of iNK cells with KLM-1 tumor cells.

Fig. 5.. Antitumor efficacy of MSLN.CAR-IL-15 GR1.1-iNK cells in the NCI-meso63 mouse model.

(A) Schematic of tumor inoculation and iNK treatments in the NCI-meso63 tumor model. ip, intraperitoneally; D, day. (B) Tumor growth monitored through BLI. P values were analyzed using unpaired Student’s t test. (C) Tumor growth based on BLI measurements. The statistics were analyzed using unpaired Student’s t test. (D) The median overall survival of different groups is shown [P = 0.07, mock iNK versus untreated; P < 0.005, CAR-IL-15 iNK versus mock iNK, analyzed using the log-rank (Mantel-Cox) test].

Fig. 6.. Increased tumor infiltration of MSLN.CAR-IL-15 iNK cells in NCI-meso63 tumor.

(A) Schematic of iNK treatments and tissue harvest in the NCI-meso63 tumor model. (B) Tumor growth monitored using BLI. (C) Flow cytometric analysis of hCD45⁺mCD45⁻ cells in the tumors on day 7 post-iNK treatment (top panel). The hCD45⁺mCD45⁻ cells were further gated to analyze hCD56 and hEGFRt expression (bottom panel). (D) Percentage of hCD45⁺mCD45⁻ cells among total live single cells isolated from spleens and tumors on day 7 (left panels); percentage of hEGFRt⁺hCD56⁺ cells among hCD45⁺mCD45⁻ cells in spleens and tumors harvested from MSLN.CAR-IL-15 iNK–treated mice on day 7 (right panel). n.s., not significant. (E) Multiplex immunofluorescence imaging of tumors harvested from mice treated with mock iNK and MSLN.CAR-IL-15 iNK cells on day 7 post-iNK treatment. Scale bars represent 150 μm. (F) hCD45⁺ cell density in iNK-treated tumor. Three independent whole tissue images were analyzed. (G) High-magnification images showing Ki67⁺ hCD45⁺ NK cells colocalized with hMSLN⁺ tumor cells in mice treated with MSLN.CAR-IL-15 iNK. Scale bars represent 40 μm. (H) Percentage of Ki67⁺ cells among total hCD45⁺ cells in iNK-treated tumor. Three independent whole tissue images were analyzed. P values were analyzed using unpaired Student’s t test.

Fig. 7.. scRNA-seq analysis of MSLN.CAR-IL-15 iNK cell–treated NCI-meso63 tumors.

(A) UMAP plots visualizing transcriptome-defined clusters of NK cells and tumor cells. (B) Representative signature genes across NK and tumor cells. The size of the dots indicates the percent of cells expressing the gene, while the color of the dots indicates the average gene expression level. (C) UMAP for total cells from untreated and treated groups. Cell types are annotated with the same color scheme as in (A). (D) Fractions of defined tumor subclusters among tumor cells in treated and untreated groups. (E) UMAP of identified aneuploid and diploid cells using copyKat analysis (top) and their fractions in each tumor cluster (bottom). (F) Expression of representative tumor signature genes across tumor subclusters. (G) MSLN expression in tumor subclusters in untreated and treated groups. (H) Top up-regulated and down-regulated Hallmark pathways in treated versus untreated tumor subclusters (adjusted P < 0.25). The x axis represents the enrichment scores, and the size of the bubble represents the number of genes in the gene set. JAK, Janus kinase; UV, ultraviolet. (I) Expression of HLA type I and HLA type II genes in untreated and iNK-treated tumor subclusters.

Fig. 8.. Single-cell transcriptional profiles of NCI-meso63 tumor–infiltrating CAR-IL-15 iNK cells.

(A) Fractions of defined NK subclusters in NK cells in preinfusion and postinfusion groups. (B) Expression of representative signature genes across NK subclusters. (C) Expression of NK-associated biomarker genes (activation, inhibitory receptors, cytokines, chemokines, and cytolytic-related categories). (D) Up-regulated and down-regulated Hallmark pathways in total postinfusion versus preinfusion iNK cells (adjusted P < 0.1). (E to G) Heatmap indicating the expression of selected gene sets in NK subtypes.