Collagen-binding IL-12-armoured STEAP1 CAR-T cells reduce toxicity and treat prostate cancer in mouse models

Chimaeric antigen receptor (CAR)-T cell therapy has demonstrated remarkable efficacy against previously incurable blood cancers^1,2,3. However, the therapy has not yet been very effective against solid cancers. The immunosuppressive tumour microenvironment (TME) limits infiltration and activation of CAR-T cells and endogenous anti-tumour immune cells⁴. Antigen heterogeneity in solid tumours poses additional difficulties for CAR-T cells, which are designed to recognize specific antigens to lyse cancer cells^4,5.

Interleukin 12 (IL-12) is a cytokine that can remodel the immunosuppressive, ‘cold’ TME in multiple cancer types for favourable therapeutic outcomes⁶. IL-12 is a potent inducer of interferon-γ (IFNγ)^6,7, which promotes trafficking and infiltration of T, natural killer (NK) and NKT cells into tumours through induction of chemokines, such as C-X-C motif chemokine ligand 9 (CXCL9)^8,9. IL-12 induces proliferation of T and NK cells and enhances their cytotoxic capacity¹⁰. IFNγ increases major histocompatibility complex (MHC)-I expression¹¹ and matures cross-presenting dendritic cells (DCs)^10,12 to promote cellular immunity against diverse tumour-associated antigens. IL-12 also reduces myeloid-derived suppressor cell (MDSC) numbers by facilitating their maturation to DCs and macrophages¹³.

Despite great promise, the clinical application of IL-12 has been hampered by dose-limiting immune-related adverse events (irAEs)^14,15, strongly suggesting the need for optimized drug delivery systems. Several studies have attempted to use the ability of tumour-reactive T cells to migrate into tumour tissue for delivery of IL-12 (refs. ^16,17). The nuclear factor of activated T cells (NFAT)-responsive promoter enables the engineered T cells to express payloads (that is, anti-tumour biologics) upon recognition of tumour antigen, thereby restricting systemic concentration of the payload¹⁸. NFAT promoter-driven induction of IL-12 was reportedly safer than its constitutive expression¹⁹. However, in a Phase I trial, 50% of metastatic melanoma patients who received ≥300 million NFAT-IL-12-equipped tumour-infiltrating lymphocytes exhibited grade 3 or 4 hepatotoxicity, which hampered clinical development²⁰. Additional technological advances to counteract the toxicity of IL-12 are required to combine IL-12 with CAR-T cells for clinical use.

We previously showed that intravenously injected A3 collagen-binding domain (CBD) of von Willebrand factor (vWF) effectively accumulates in tumours, due to tumour-specific collagen accessibility²¹. A recombinant fusion of CBD to IL-12 (CBD-IL-12) improved both safety and anti-tumour efficacy²². Intravenous CBD-IL-12 induced long-lasting elevation of intratumoural IFNγ levels compared with unmodified IL-12. CBD fusion to IL-12 significantly reduced liver damage marker alanine transaminase (ALT) activity and serum IFNγ.

Metastatic castration-resistant prostate cancer (mCRPC) is a largely incurable solid tumour, with a median overall survival of only 3 years²³. We have recently identified a promising tumour-associated antigen, six transmembrane epithelial antigen of prostate 1 (STEAP1), which is more broadly expressed than prostate-specific membrane antigen in mCRPC²⁴. We have developed second-generation CAR-T cells against human STEAP1 (hSTEAP1), which showed significant, yet non-curative efficacy in multiple prostate cancer models in mice²⁴. Intravenously injected CBD-IL-12 protein and STEAP1 CAR-T cells synergized to extend the survival of prostate cancer-bearing mice²⁴. We recently initiated a first-in-human trial of STEAP1 CAR-T cells for men with mCRPC (NCT06236139).

Here we further engineered STEAP1 CAR-T cells to express CBD-IL-12 upon STEAP1 recognition to enhance tumour-specific immune responses. We hypothesized that fusion of CBD to IL-12 as a cargo would decrease the systemic toxicity associated with armouring T cells with unmodified IL-12.

CBD-IL-12-armoured STEAP1 CAR-T cells efficiently kill cancer cells in vitro

We designed single-chain (sc) mouse IL-12 variants with different positions of CBD fusion, all of which showed potent half-maximal effective concentration (EC₅₀) values (Fig. 1a–c and Supplementary Table 1). We next designed gamma-retroviral vectors for constitutive expression of the mouse CAR targeting hSTEAP1 (ref. ²⁴) and NFAT-driven expression of an IL-12 variant (Fig. 1d). Mouse primary T cells were transduced with IL-12 variant-armoured CAR constructs; we noted some reduction in transduction efficiency in proportion to the length of the transgenes (Fig. 1e and Extended Data Fig. 1a). A vector for constitutive expression of enhanced green fluorescent protein (EGFP) and NFAT-driven expression of CBD-IL-12-CBD (CBD on both ends) was also prepared as a control to test IL-12 production triggered by CAR–antigen interaction (Extended Data Fig. 1b).

All the STEAP1 CAR-T variants lysed RM9-hSTEAP1-firefly luciferase (Fluc) mouse prostate cancer cells, but not RM9 wildtype (WT) (Fig. 1f and Supplementary Fig. 1a). Expression of the IL-12 variants significantly increased cytotoxicity against RM9-hSTEAP1 cells, suggesting that autocrine IL-12 contributes to the enhanced cytolytic capacity¹⁰. CAR-T cells armoured with one CBD fused to IL-12 (CBD-IL-12 or IL-12-CBD) secreted the payload upon stimulation (Fig. 1g). However, the amount of secreted CBD-IL-12-CBD was low. Although CBD-IL-12-CBD was produced in HEK293F and Jurkat cells (Fig. 1b and Extended Data Fig. 2a), its protein and mRNA expressions in primary mouse T cells were inefficient compared with other IL-12 variants (Extended Data Fig. 2b,c). Nonetheless, all the armoured CAR-T cells showed enhanced IFNγ production (Fig. 1h). These results demonstrate that the IL-12 variant-armoured hSTEAP1 CAR-T cells can lyse hSTEAP1-positive cancer cells more potently and secrete the payload in an antigen-dependent manner.

CBD-IL-12-armoured hSTEAP1 CAR-T cells demonstrate anti-tumour efficacy without pre-conditioning

We tested the therapeutic efficacy of the conventional hSTEAP1 CAR-T cells against subcutaneous RM9-hSTEAP1-Fluc tumour (Fig. 2a–e). As we previously reported, hSTEAP1 CAR-T cells showed significant anti-tumour effects (Fig. 2b–e). However, complete response (CR) was not observed. Next, we treated the same tumour model to screen the collagen-binding IL-12-armoured CAR constructs (Fig. 2f–i). All the armoured CAR-T treatments clearly delayed tumour growth (Fig. 2g and Extended Data Fig. 3a) and prolonged survival compared with untransduced (UTD) T cells (Fig. 2h). One of five mice treated with CBD-IL-12-armoured CAR-T (CBD-IL-12 CAR-T) cells showed CR, as did two of six mice treated with IL-12-CBD CAR-T cells, whereas tumours were not eradicated in any mice treated with CBD-IL-12-CBD CAR-T cells (Fig. 2h). Body weight loss was not observed (Extended Data Fig. 3b). There was a transient and non-statistically significant trend towards increased serum IFNγ levels in mice treated with CBD-IL-12 or IL-12-CBD CAR-T cells on day 11 (Fig. 2i). CBD-IL-12-CBD CAR-T treatment did not increase serum IFNγ levels, consistent with poor expression of CBD-IL-12-CBD (Fig. 1g and Extended Data Fig. 2). These results indicate that fusing one CBD to IL-12 is more therapeutically effective than fusing two CBDs.

CAR-T-cell-mediated delivery of CBD-IL-12 balances therapeutic efficacy and safety

We selected N-terminal fusion CBD-IL-12-armoured CAR-T cells for detailed comparisons with IL-12-armoured CAR-T cells given the similar characteristics of N- and C-terminal fusions in vitro and in vivo. CBD fusion to IL-12 significantly decreased serum IFNγ levels in the treated mice (Fig. 2j,k). Body weight loss was not observed (Extended Data Fig. 3c). Albeit not statistically significant, IL-12 CAR-T cells showed stronger anti-tumour efficacy compared with CBD-IL-12 CAR-T cells (Fig. 2l,m). Immunogenicity of the CBD (human origin) might have disturbed long-term therapeutic efficacy of CBD-IL-12 CAR-T in the immunocompetent mouse model because we detected anti-CBD-IL-12 IgG in plasma (Extended Data Fig. 3d).

Next, we compared therapeutic efficacy of CBD-IL-12 hSTEAP1 CAR-T cells with the combination therapy of hSTEAP1 CAR-T cells and CBD-IL-12 protein (Fig. 2n)²². The monotherapy of CBD-IL-12 CAR-T cells outperformed the combination therapy and showed 80% of CR rate (Fig. 2o,p). This suggests that CAR-T-cell-mediated, local and continuous delivery of CBD-IL-12 to the TME is more efficacious than a single, large dose of the protein.

To investigate whether CBD-IL-12 CAR-T cells developed anti-tumour immune memory to hSTEAP1 and other antigens, we rechallenged the complete responders with RM9-hSTEAP1 or parental RM9 WT cells (hSTEAP1⁻). All the long-term survivors rejected the second challenge with RM9-hSTEAP1, demonstrating anti-tumour immune memory against hSTEAP1 (Extended Data Fig. 3e,f). RM9 WT tumours were also suppressed when implanted to the complete responders, and the survivors previously treated with CBD-IL-12 CAR-T cells showed 40% of rejection rate (Fig. 2q and Extended Data Fig. 3g,h). The results indicate that CBD-IL-12 CAR-T cells induce antigen spreading to counteract antigen heterogeneity.

Secreted CBD-IL-12 demonstrates enhanced localization to the tumour

We quantified the IL-12 and CBD-IL-12 secreted from CAR-T cells in vivo (Fig. 3a). Without the administration of IL-12-armoured CAR-T cells, IL-12 was barely detectable in the tumour (Extended Data Fig. 4a,b). On day 8, a significantly higher level of intratumoural IL-12 was detected in mice treated with CBD-IL-12 CAR-T cells (Fig. 3b). In contrast, serum concentrations of IL-12 were significantly lower in CBD-IL-12 CAR-T-treated mice on day 8 (Fig. 3c). Significantly less IL-12 was detected from spleens of CBD-IL-12 CAR-T-treated mice on day 6 (Fig. 3d). The levels of IL-12 detected from the lung were comparable between the two groups (Fig. 3e). IL-12 was barely detectable in the heart, liver and kidney (Fig. 3f–h). Given that the detected concentrations of IL-12 in serum were low in the immune competent model, we next quantified CAR-T-derived IL-12 using NSG mice, in which endogenous immune cells that consume IL-12 are scarce. MyC-CaP-hSTEAP1-bearing male NSG mice received C57BL6/J-derived CAR-T cells intravenously (Fig. 3i). Both CBD-IL-12 CAR-T cells and IL-12 CAR-T cells showed anti-tumour efficacy (Fig. 3j). Importantly, serum CBD-IL-12 level was significantly lower than serum IL-12 level (Fig. 3k). These results indicate that CBD-IL-12 secreted from CAR-T cells are significantly more localized to the tumour and less distributed to other parts of the body compared with IL-12.

CBD-IL-12 CAR-T therapy decreases systemic toxicity

irAEs are one of the major obstacles for clinical translation of IL-12 therapies. Hepatic toxicity is a common side effect in patients who receive systemic administration of recombinant human IL-12 (ref. ²⁵). Serum concentrations of liver damage-associated ALT and alkaline phosphatase (ALP) were significantly increased by IL-12 CAR-T cells, but not by CBD-IL-12 CAR-T cells (Fig. 4a,b and Supplementary Fig. 2a). IL-12 CAR-T cells significantly facilitated infiltration of CD3⁺ T cells into non-target organs (liver, lung and kidney) compared with CAR-T cells, whereas CBD-IL-12 CAR-T cells did not (Fig. 4c–e, Supplementary Fig. 2b and Extended Data Fig. 5). No structural abnormalities were found in organs of any treatment group in haematoxylin and eosin (H&E) staining (Extended Data Fig. 5). These results demonstrate that the CBD fusion to IL-12 reduces the toxicity of CAR-T cells armoured with IL-12.

CBD-IL-12 CAR-T cells inflame tumours and induce anti-tumour immune infiltrates

Tumour cell heterogeneity is the major cause of recurrence and inefficacy of CAR-T cells. We observed CRs in some of the animals (Fig. 2h,m,p) as well as rejection of RM9 WT rechallenge in a fraction of the complete responders (Fig. 2q), suggesting that CBD-IL-12 CAR-T cells induce antigen spreading. We analysed the TME to investigate the underlying mechanisms. IL-12-induced IFNγ promotes cross-presentation by upregulating expressions of key proteins, such as MHC-I¹¹. CXCL9 recruits T cells and NK cells into the TME²⁶. The granulocyte macrophage colony-stimulating factor (GM-CSF) is also induced by IL-12, facilitating the maturation of cross-presenting DCs¹² that have a major role in T cell-mediated anti-tumour immunity^27,28. We found that IL-12 CAR-T cells and CBD-IL-12 CAR-T cells upregulate intratumoural IFNγ, CXCL9 and GM-CSF compared with UTD and unarmoured CAR-T cells (Fig. 5a–c and Supplementary Fig. 2a).

We also characterized the immune cell infiltrates in the RM9-hSTEAP1 tumour by flow cytometry (Fig. 5d–l and Supplementary Fig. 2b). In accordance with the increase in intratumoural CXCL9, CBD-IL-12 CAR-T cells increased the frequency of CD45⁺ leucocytes, CD3⁺ T cells, CD8⁺ T cells and CD3⁻NK1.1⁺ NK cells compared with CAR-T cells (Fig. 5e–h), and the increased frequencies correlated with a lower tumour burden (Extended Data Fig. 6a–d). CBD-IL-12 CAR-T cells increased cross-presenting conventional type 1 DCs (cDC1s), including CD103⁺ cDC1s and CD8α⁺ cDC1s within the CD11c⁺MHC-II⁺ cDC population (Fig. 5i,j), in accordance with the increase in GM-CSF. CBD-IL-12 CAR-T cells significantly increased the frequency of neutrophils, which play a crucial role in eliminating antigen-negative cancer cells following adoptive transfer of antigen-specific T cells²⁹ (Fig. 5k). CBD-IL-12 CAR-T cells significantly reduced CD11b⁺Ly6C^hiLy6G⁻ monocytic MDSCs (M-MDSCs), an especially suppressive subset of MDSCs³⁰ (Fig. 5l), and M-MDSC frequencies positively correlated with tumour burdens (Extended Data Fig. 6e). Splenic T cells derived from mice treated with CBD-IL-12 CAR-T cells produced a significantly higher amount of IFNγ when co-cultured with RM9 WT (Fig. 5m), suggesting T cell responses to antigens other than hSTEAP1. Collectively, CBD-IL-12 CAR-T cells change the composition of immune infiltrates in the tumour towards an anti-tumour state, inducing antigen spreading to fight against heterogeneity of the prostate cancer.

CBD-IL-12 CAR-T cells remodel the intratumoural transcriptional landscape to enhance anti-tumour immunity

Limited trafficking and infiltration of CAR-T cells in solid tumours and immunosuppressive TME are major challenges to the efficacy of current therapies. We performed spatial transcriptomics to assess whether the entire TME can be remodelled by CBD-IL-12 CAR-T cells (Supplementary Fig. 2b). We observed infiltration of CD45⁺ leucocytes throughout the tissues and reduction of tumour regions (pan-cytokeratin⁺) in IL-12 CAR-T or CBD-IL-12 CAR-T-treated samples (Fig. 6a–c). Immunohistochemistry (IHC) staining confirmed enhanced infiltration of CD3⁺ T cells in the armoured CAR-T-treated tumours. Principal component analysis (PCA) showed that tumour tissues treated with the three different therapies displayed markedly different characteristics (Fig. 6d). Gene set enrichment analysis (GSEA) revealed that CBD-IL-12 CAR-T cells activated the IL-12 pathway and activated antigen processing and presentation by MHC-I (Fig. 6e,f). Enrichment of genes in the IL-12 pathway was not statistically significant (P = 0.059), presumably because only 17 genes in the dataset were annotated to this pathway. Indeed, Fisher’s exact test showed statistical significance of IL-12 pathway enrichment in differentially expressed genes (P = 0.01241). Because we observed a tendency for an increase in B cells in tumours treated with IL-12 CAR-T cells and CBD-IL-12 CAR-T cells (Extended Data Fig. 7a), we analysed gene expression changes related to tertiary lymphoid structure (TLS) (Extended Data Fig. 7b). We observed increased expression of a variety of TLS-related chemokines and chemokine receptors including Cxcr4, Cxcr5, Cxcl12 and Cxcl13 in both CBD-IL-12 CAR-T and IL-12 CAR-T-treated tumours. Interestingly, CBD-IL-12 CAR-T cells, but not IL-12 CAR-T cells, clearly upregulated Bcl6 (master regulator for T follicular helper cells^31,32,33) and Aicda (essential for class switch recombination and somatic hypermutation^34,35). Consistent with the upregulation of Bcl6, Prdm1 (master regulator of plasma cells³⁶ and known to reciprocally regulate Bcl6 (ref. ³¹)) was downregulated by CBD-IL-12 CAR-T cells. These data suggest that CBD-IL-12 CAR-T cells enhanced the maturation and function of intratumoural immune aggregates resembling TLS. We also observed specific enrichment of co-stimulatory molecules such as Cd80, Cd86, Cd40 and Tnfsf4 in the CBD-IL-12 CAR-T-treated tumours. These co-stimulatory signals may facilitate local T cell priming and promote immune synapse formation, thereby enhancing anti-tumour immunity. Collectively, the data suggest that CBD-IL-12 CAR-T cells reprogramme the TME towards a more immunologically active, TLS-supportive niche.

CBD-IL-12 CAR-T cells combined with immune checkpoint inhibitors (CPIs) eradicate large established RM9-hSTEAP1 tumours

Since CBD-IL-12 CAR-T cells alone did not constantly induce tumour-free CRs (Fig. 2h,m,p) and tumour-infiltrating T/NK cells had elevated programmed cell death protein 1 (PD-1) and cytotoxic T lymphocyte associated protein 4 (CTLA-4) expression (Extended Data Fig. 8a–f), we examined the anti-tumour efficacy of CBD-IL-12 CAR-T cells, anti-PD-1 and anti-CTLA-4 immune checkpoint inhibitors (CPIs) combination therapy. When the therapeutic intervention was started on day 4 (tumour volumes were ~60 mm³), armoured CAR-T cells in combination with the CPIs achieved a 100% CR rate (Extended Data Fig. 9a–d). Even when we delayed the initial CAR-T treatment until day 6 (tumour volumes were ~120 mm³), CBD-IL-12 CAR-T cells + CPIs cured 80% of RM9-hSTEAP1-bearing mice and significantly extended their survival compared with CAR-T cells + CPIs (Fig. 7a–c). These data show strong anti-tumour efficacy of the combination immunotherapy against the large established tumours. All the tumour-free survivors rejected subcutaneously rechallenged RM9-hSTEAP1 cells (Fig. 7d). None of the treatments caused body weight loss in the animals (Fig. 7e and Extended Data Fig. 9e). A previous study using IFNγ receptor-knockout (KO) mice indicated that IL-12-induced IFNγ plays a central role in irAEs³⁷. In this study, significantly increased serum IFNγ was observed in mice treated with IL-12 CAR-T + CPIs, but not in mice treated with CBD-IL-12 CAR-T + CPIs (Extended Data Fig. 9f), consistent with the toxicity evaluation of the CAR-T monotherapies (Figs. 2k and 4a–e).

Fully human CBD-IL-12 CAR-T cells suppress human prostate cancer xenograft and minimize circulating IL-12

After confirming that unmodified and CBD-fused human IL-12 (CBD-hIL-12) proteins could be recombinantly expressed (Supplementary Fig. 3), we generated lentiviral vectors for human T cells. We successfully transduced Jurkat cells and primary human T cells with the hSTEAP1-hBBζ + NFAT-CBD-hIL-12 vector (Fig. 8a, Supplementary Fig. 4a,b and Extended Data Fig. 10a–d,g). Human CBD-hIL-12 CAR-T cells and hIL-12 CAR-T cells showed enhanced expression of hIL-12 upon co-culture with hSTEAP1⁺ 22Rv1 human prostate cancer cells (Fig. 8b and Extended Data Fig. 10e,h). These hIL-12-armoured CAR-T cells showed better in vitro killing against 22Rv1 compared with CAR-T cells without killing 22Rv1 hSTEAP1 KO cells (Fig. 8c and Extended Data Fig. 10f,i). Finally, we evaluated the human armoured CAR-T cells in subcutaneous (Fig. 8d–h) and disseminated metastatic 22Rv1 tumour models (Fig. 8i–l). STEAP1 CAR-T cells were unable to control the subcutaneous tumour growth (Fig. 8e,f). In contrast, both hIL-12 CAR-T cells and CBD-hIL-12 CAR-T cells induced tumour regression and prolonged survival. Strikingly, serum concentrations of CBD-hIL-12 were more than 10× lower than those of IL-12 at all timepoints tested (Fig. 8g). Body weight loss was not observed (Fig. 8h). Consistent with the subcutaneous model, serial bioluminescence imaging (BLI) demonstrated that hIL-12 CAR-T cells and CBD-hIL-12 CAR-T cells exhibited superior anti-tumour efficacy compared with CAR-T cells in the 22Rv1 disseminated tumour model (Fig. 8i–k). CBD fusion to hIL-12 significantly decreased serum IL-12 levels (Fig. 8l). Collectively, these results indicate that the CBD-IL-12-expression platform technology is applicable to human T cells.

The goal of cancer immunotherapy is to achieve safe and long-term effectiveness. The next milestone for CAR-T therapies is to overcome solid tumour unresponsiveness and/or tumour recurrence. We have previously reported that second-generation STEAP1 CAR-T cells induce tumour regression but face recurrence due to STEAP1 antigen loss. This could be overcome by combining with a potent cytokine, such as IL-12, which would efficiently activate the host immune system. We previously reported that conjugation of vWF-derived CBD to CPI antibodies (PD-L1, CTLA-4), cytokines (IL-2 and IL-12), chemokine CCL4 or serum albumin improves their localization to the tumour stroma, enhancing therapeutic efficacy and decreasing side effects^21,22,38,39. Using our drug delivery approach that combines protein engineering and cell engineering, CBD-IL-12 expression from CAR-T cells showed stronger anti-tumour efficacy than just combining CAR-T with CBD-IL-12 protein, presumably due to local and sustained production of CBD-IL-12.

CBD-IL-12 expression from our novel CAR-T cell therapy improved the potency of CAR-T cells against mouse and human prostate cancer models by targeting the extracellular matrix. Compared with unmodified IL-12, the CBD fusion enhanced intratumoural retention of IL-12 from the CAR-T payload and reduced systemic IL-12 exposure, reducing irAEs. irAEs of IL-12-expressing adoptive T cell therapy is the major reason for clinical trial failures. The tumour-homing capacity of CBD-IL-12 (ref. ²²) may contribute to toxicity reduction even when CBD-IL-12 is secreted from CAR-T cells in healthy tissues, due to the leakiness of the NFAT-responsive promoter and/or low-level STEAP1 expression in irrelevant tissues. Our data demonstrate that CBD fusion to IL-12 enhances the safety of IL-12 as a cargo of adoptive T cell therapies, and CBD-based drug delivery systems could be applied to other cell therapies.

Chemotherapy pre-conditioning is widely used in combination with CAR-T therapy to remove pre-existing immune cells and enable CAR-T cells to expand, infiltrate and remodel the TME⁴⁰. While effective, chemotherapy often causes haematological and systemic toxicities and increased risks of frequent lethal infections⁴¹. In accordance with a previous study¹⁷, CBD-IL-12 CAR-T cells + CPI antibodies eradicated large established RM9-hSTEAP1 tumours without pre-conditioning. Given that IL-12 CAR-T cells combined with pre-conditioning failed because of toxicity in the clinic, it is worth investigating whether CBD-IL-12 CAR-T cells show objective response with tolerable toxicity in patients in the future.

One of the greatest challenges in the treatment of solid malignancies is antigen heterogeneity. IL-12 is a suitable cytokine for tackling this issue by enhancing the capacity of cellular immunity to fight diverse antigens. Our flow cytometric, transcriptomic and splenic T cell restimulation data suggest that CBD-IL-12 CAR-T cells can induce multifaceted activation of anti-tumour immunity and antigen spreading to fight antigen heterogeneity in solid tumours. The differences in gene expression patterns observed in the IL-12 CAR-T and CBD-IL-12 CAR-T groups in the transcriptomics analysis are intriguing. Investigating the mechanism by which these differences arise due to localization of IL-12, and how these differences can be leveraged to treat solid cancers including prostate cancer, would be our future work.

Other approaches to reduce toxicity of IL-12 as a payload include tethering IL-12 to the CAR-T cell surface by fusing a transmembrane domain (as we and others previously reported^42,43,44) or an antibody fragment targeting cell surface receptors to IL-12 (ref. ⁴⁵) and conjugating IL-12-loaded nanoparticles onto CAR-T cells through click chemistry⁴⁶. Although these methods can be powerful in reducing toxicity, they might reduce the chance for IL-12 to act in trans by directly binding to an IL-12 receptor on endogenous immune cells. It will be important to see how these methodologies may differ from secreted IL-12 in terms of levels of therapeutic efficacy and mechanisms of action.

Our matrix-binding technology has several additional advantages for clinical applications to a wide range of CAR-T-cell therapies against solid tumours. First, collagen is abundantly expressed across patients with multiple tumour types, probably making this technology useful for a wide variety of patients. Second, identifying a promising target antigen for each solid tumour type is a major challenge in the field^4,5, while our versatile, collagen-targeting approach enables a payload to avoid competition with CARs. Third, CBD is unlikely to confer immunogenicity to CAR-T cells because it is of fully human origin (the anti-CBD-IL-12 IgG induction observed in the RM9-hSTEAP1 syngeneic mouse model is less likely to happen in humans).

In summary, this study demonstrates a protein-engineering approach to make IL-12 safer for use with CAR-T therapy. The use of CBD-IL-12 lies in its capacity to overcome the immunosuppressive TME and antigen heterogeneity while avoiding irAEs and competition with CARs for tumour targeting. CBD-hIL-12-armoured human CAR-T cells showed potent anti-tumour efficacy while minimizing IL-12 in the circulation in 22Rv1 human prostate cancer, demonstrating a high potential for clinical translation.

Cell lines

HEK293T (CRL-3216), RM9 (CRL-3312), MyC-CaP (CRL-3255), 22Rv1 (CRL-2505) and Jurkat (TIB-152) cells were obtained from ATCC. HEK293T, RM9 and MyC-CaP cells were maintained in DMEM medium (Gibco) supplemented with 10% FBS (Gibco), 2 mM GlutaMAX (Gibco) and 1% penicillin/streptomycin (Gibco). 22Rv1 and Jurkat cells were maintained in RPMI 1640 medium (Gibco) supplemented with 10% FBS, 2 mM GlutaMAX and 1% penicillin/streptomycin. PLAT-E (RV-101) was obtained from Cell Biolabs and maintained in DMEM medium supplemented with 10% FBS, 2 mM GlutaMAX, 1% penicillin/streptomycin, 1 µg ml⁻¹ puromycin (Merck) and 10 µg ml⁻¹ blasticidin (Stratech). RM9-hSTEAP1-Fluc, 22Rv1-Fluc-RFP, 22Rv1 hSTEAP1 KO and 22Rv1 hSTEAP1 KO + rescue cells were generated in our previous work²⁴. Myc-CaP-hSTEAP1 cells were generated using a FU-hSTEAP1-CGW vector²⁴.

Validation of hSTEAP1 expression on prostate cancer cells

Cells were resuspended in cold phosphate-buffered saline without calcium and magnesium (PBS (−)) and stained with BD Horizon Fixable Viability Stain 510 (BD). After a wash with PBS (−) supplemented with 2% FBS, the cells were stained with Vandortuzumab (Invitrogen, 1:200), followed by Alexa Fluor 594 AffiniPure F(ab’)₂ Fragment Donkey anti-human IgG (H + L) (Jackson ImmunoResearch, 709-586-149, 1:200). Cells were acquired using a BD FACSymphony A3 flow cytometer and data were analysed using FlowJo (BD).

Mice

All animals used in Imperial College London were handled in accordance with the 1986 Animal Scientific Procedures Act and under a United Kingdom Government Home Office-approved project licence and overseen by ethics committees of Imperial College London. Male C57BL/6J (6–9 weeks), FVB/N mice (6–9 weeks) and NSG (NOD-SCID-IL2Rγ-null) mice (3–5 weeks) were obtained from Charles River UK and housed at the Hammersmith Central Biomedical Services facility of Imperial College London. A 12-h light/dark cycle was used, room temperature was maintained between 20 and 21 °C, and room humidity was maintained at 50 ± 10%. All mouse studies performed in UCLA were in accordance with protocols approved by the UCLA Institutional Animal Care and Use Committee, known as the Chancellor’s Animal Research Committee (ARC). Male NSG mice (4–6 weeks) were obtained from The Jackson Laboratory and housed in UCLA, California NanoSystems Institute (CNSI) animal housing facility. A 12-h light/dark cycle was used, room temperature was maintained at 23 ± 3 °C, and room humidity was maintained at 50 ± 20%. BLI was performed using the IVIS Spectrum (PerkinElmer) at the Preclinical Imaging Technology Center at the Crump Institute for Molecular Imaging, CNSI.

Production and purification of recombinant scIL-12 variants

Optimized sequences encoding p35 and p40 connected with a glycine-serine linker were synthesized and subcloned into pcDNA3.1(+) vector by GenScript. The collagen-binding domain (A3 domain of human vWF)⁴⁷ was fused to the N terminus, C terminus or both termini of the protein to create CBD-IL-12 variants. His-tag was added to the C terminus after the CBD. Recombinant IL-12 variants were produced by transient expression in HEK293F cells and purified as described previously²². The purity of the proteins was evaluated using SDS–PAGE as described previously²². Protein concentration was quantified by measuring absorbance at 280 nm using a NanoDrop One spectrophotometer (Thermo Fisher). Full amino acid sequences of the IL-12 variants are available in the Supplementary Information.

STAT4 phosphorylation assay

Primary mouse T cells were treated with IL-12 variants to evaluate phosphorylation of STAT4 as described previously²². In brief, CD3⁺ T cells were purified from spleens of C57BL/6J mice using the MagniSort Mouse T cell Enrichment kit (Invitrogen) according to manufacturer instructions. Purified T cells were activated using the T cell Activation/Expansion kit, mouse (Miltenyi Biotec) according to manufacturer instructions and cultured in the mouse T cell culture medium (RPMI 1640 medium supplemented with 10% FBS, 2 mM GlutaMAX, 1% MEM non-essential amino acids (Gibco), 1 mM sodium pyruvate (Gibco), 50 µM 2-mercaptoethanol (Gibco), 1% penicillin/streptomycin and 10 ng ml⁻¹ recombinant human IL-2 (Peprotech)) for 3 days. T cell activation/expansion beads were magnetically removed, and the T cells were rested in fresh culture medium (without IL-2) overnight. T cells were seeded into 96-well V-bottom plates at 1 × 10⁵ cells per well; stimulated with the indicated concentrations of IL-12 variants at 37 °C for 15 min; fixed with BD Phosflow Lyse/Fix buffer and permeabilized with BD Phosflow Perm Buffer III according to manufacturer instructions; stained with anti-pSTAT4 Alexa Fluor 647 (pY693, BD, 1:100) and anti-mouse CD8α BV510 (53-6.7, Biolegend, 1:200); and then acquired using a BD FACSymphony A3 Cell Analyzer.

CAR expression plasmids

pSIRV-NFAT-eGFP was a gift from Peter Steinberger (Addgene, plasmid 118031). pHR_SFFV was a gift from Wendell Lim (Addgene, plasmid 79121). hSTEAP1-BBζ CAR sequences were prepared as described previously²⁴. SFFV promoter, hSTEAP1-mBBζ CAR, mouse scIL-12 variant and WPRE were subcloned into the pSIRV-NFAT-eGFP vector via in-fusion cloning (Takara Bio) as illustrated in Fig. 1d. The eGFP sequence was replaced with the IL-12 sequence during the process. MSCV promoter, NFAT-responsive promoter, human scIL-12 variant, WPRE and hSTEAP1-hBBζ CAR sequences were cloned into pRRLSIN vector backbone⁴⁸.

Production of gamma-retroviral vectors

A day before transfection, PLAT-E cells were resuspended in antibiotic-free medium with 10% FBS and GlutaMAX, seeded and cultured for 24 h. PLAT-E cells were transfected with a retroviral transfer vector using GeneJuice (Merck Millipore). Supernatant was collected at 48 h and 72 h after transfection and passed through a 0.45 µm PES filter (Merck). Supernatant collected at 48 h was stored at 4 °C overnight and pooled with supernatant collected at 72 h. Pooled supernatant was concentrated by high-speed centrifugation (24,000 g at 4 °C for 2 h) using an Avanti JXN-30 centrifuge (Beckman Coulter) with a JS-24.15 rotor (Beckman Coulter) and resuspended in DMEM medium supplemented with 10% FBS⁴⁹. Concentrated viral vector was immediately added to a non-treated cell culture plate pre-coated with 20 µg ml⁻¹ RetroNectin (Takara Bio) and centrifuged (2,000 g at 32 °C for 1.5 h) before transduction of T cells.

Producion of lentiviral vectors

psPAX2 (Addgene, plasmid 12260) was a gift from Didier Trono. HEK293T cells (ATCC) maintained in DMEM medium (Gibco) supplemented with 10% FBS (Gibco), 2 mM GlutaMAX (Gibco) and 1% penicillin/streptomycin (Gibco) were used. HEK293T cells were transfected with a lentiviral transfer vector, psPAX2 and pVSVg²⁴.

Manufacture of primary mouse CAR-T cells

Splenocytes were collected from spleens of male C57BL/6J or FVB/N mice (Charles River UK). A spleen was put on a 70-µm cell strainer pre-wet with DMEM supplemented with 2% FBS and mashed using the plunger of a 2.5-ml syringe (Terumo). The strainer was washed with DMEM medium twice and centrifuged at 300 g for 5 min. After removing the supernatant, red blood cells were lysed using ACK lysing buffer (Gibco) at room temperature for 5 min. Cells were washed with excess PBS (−) and resuspended in PBS (−) supplemented with 2% FBS and 2 mM EDTA. Mouse T cells were purified as described above. Purified T cells were activated using T cell Activation/Expansion kit, mouse, and cultured in the mouse T cell culture medium as described above. After 24 h, the activation/expansion beads were magnetically removed and 1 µg ml⁻¹ anti-mouse IL-12 p40 (C17.8, Bio X Cell) was added for IL-12 CAR-T cells and CBD-IL-12 CAR-T cells. T cells were seeded into a RetroNectin-treated viral vector-coated plate, centrifuged (300 g at 32 °C for 10 min) and cultured for 48 h. After 48 h of culture, the cytokine supplementation was changed from recombinant human IL-2 to 10 ng ml⁻¹ recombinant human IL-7 (Peprotech) and 10 ng ml⁻¹ recombinant human IL-15 (Peprotech)⁴⁹. %CAR⁺ was assessed using either a combination of biotinylated recombinant protein L (Thermo Scientific, 1:3,000) and streptavidin-Alexa Fluor 647 conjugate (Invitrogen, 1:1000) (for Fig. 2a–i) or a combination of Biotin-SP-conjugated AffiniPure F(ab’)₂ Fragment Goat Anti-Human IgG, F(ab’)₂ Fragment Specific (Jackson ImmunoResearch, 109-066-006, 1:200) and phycoerythrin (PE)-streptavidin (Biolegend, 1:200) (for all other figures) on day 4 by flow cytometry. CAR-T cells were resuspended in cold PBS (−), stained with acridine orange/propidium iodide (DeNovix) and counted using a CellDrop automated cell counter (DeNovix) to adjust the concentration before use in experiments. CAR-T cells were intravenously injected to mice on day 5 after viral transduction.

CAR-T-cell-mediated cytotoxicity assay (calcein-AM)

CAR-T cells (1 × 10⁵ per ml) in the culture medium without supplementation of any interleukins were seeded into a 96-well V-bottom plate. Target cells were labelled with 1 µM calcein-AM as described previously⁵⁰. Labelled target cells (1 × 10⁵ per ml) in the culture medium with 5 mM probenecid were mixed with the CAR-T cells at 10,000 target cells per well. The cells were centrifuged (300 g for 5 min at 37 °C) and incubated for 24 h at 37 °C in the presence of 5% CO₂. The plate was centrifuged and 100 µl per well of the supernatant was transferred to a 96-well black F-bottom plate (Greiner Bio-One). Fluorescence intensity (Ex 490 nm, Em 530 nm) was measured using a CLARIOstar Plus microplate reader (BMG Labtech). After subtracting the autofluorescence derived from medium, % lysis was calculated as:

F₁: Fluorescence intensity of the mixture of CAR-T cells and labelled target cells

F_E: Fluorescence intensity of CAR-T cells

F₀: Fluorescence intensity of labelled target cells

F₁₀₀: Fluorescence intensity of labelled target cells lysed with 1% Triton X-100

Enzyme-linked immunosorbent assay (ELISA) for detection of IL-12 and IFNγ secreted from CAR-T cells

CAR-T cells (5 × 10⁵ per ml) in the culture medium (without supplementation of any interleukins) and 5 × 10⁵ target cells per ml in the same medium were seeded into a 96-well V-bottom plate (50,000 CAR-T cells per well, CAR-T/Target ratio = 1), centrifuged (300 g for 5 min at 37 °C) and incubated for 24 h at 37 °C in the presence of 5% CO₂. The plate was centrifuged again and the culture supernatant was collected for quantification of IL-12 and IFNγ by ELISA. Invitrogen ELISA kits were used for mouse proteins. Human IL-12 was quantified using IL-12 p70 DuoSet (R&D Systems). eBioscience Cell Stimulation Cocktail (Invitrogen) was used as a cocktail of phorbol 12-myristate 13-acetate (PMA) and ionomycin. For quantification of scIL-12 variants secreted from CAR-T cells, the IL-12 variants with His-tag were used as standards.

Intracellular IL-12 staining

Primary mouse T cells were stimulated with a cocktail of PMA and ionomycin in the presence of 5 µg ml⁻¹ Brefeldin A overnight. Cells were fixed and permeabilized using the eBioscience intracellular fixation and permeabilization buffer (Invitrogen) and stained with rat anti-mouse IL-12 PE (C15.6, BD, 1:100).

RT–qPCR of mouse IL-12 transcripts

CAR-T cells (1 × 10⁶ per sample) were stimulated with PMA and ionomycin for 4 h. Total RNA was purified using RNeasy Mini kit (Qiagen) according to manufacturer instructions, quantified using Nanodrop One and stored at −80 °C until use. Reverse transcription was performed using TaqMan Reverse Transcription Reagents (Invitrogen) using random hexamers. qPCR was performed using SYBR Green Universal Master Mix. Expression levels of mouse p35 and p40 in each sample were normalized by an internal control (EGFP for EGFP + NFAT-CBD-IL-12-CBD; WPRE for hSTEAP1-mBBζ + NFAT-CBD-IL-12 and hSTEAP1-mBBζ + NFAT-IL-12-CBD) and relative gene expression changes with PMA and ionomycin stimulation were calculated. The following primers were used (primer sequences for IL-12 subunits are optimized for the gamma-retroviral vectors used. Thus, their sequences are not completely matched to mouse genomic sequence): WPRE F: 5’-GTGGATACGCTGCTTTAATGCCT-3’, WPRE R: 5’-GTTGCGTCAGCAAACACAGT-3’; EGFP F: 5’-AAGGGCATCGACTTCAAGG-3’, EGFP R: 5’-TGCTTGTCGGCCATGATATAG-3’; p40 F: 5’-TCGAGCTGGCCCTGGAGG-3’, p40 R: 5’-CGACCTGGCTGTTCTTCAGGG-3’; p35 F: 5’-ACGAGTCTTGTCTGGCCACC-3’, p35 R: 5’-CCTCGTAGATGCTGCCCAGG-3’.

Anti-tumour efficacy of CAR-T cells against RM9-hSTEAP1

A total of 5 × 10⁵ RM9-hSTEAP1-Fluc cells resuspended in 50 µl of PBS (−) were subcutaneously injected on the left side of the back of each male C57BL/6J mouse on day 0. Primary mouse CAR-T cells were intravenously injected on day 4 or day 6. The dose of CAR-T cells in each study is described in figure legends. Heterodimeric CBD-IL-12 protein²² (10 µg or 25 µg) was intravenously injected on day 8 in Fig. 2n–p. In some studies, anti-CTLA-4 (9H10, Bio X cell) and anti-PD-1 (RMP1-14, Bio X cell) were intraperitoneally injected (100 µg each) 3 times starting at 3 days after the CAR-T administration with 4 days interval. Tumours were measured using a digital caliper and volumes were calculated as ellipsoids (V = 4/3 × 3.14 × depth/2 × width/2 × height/2). Blood samplings from tail veins were performed at the indicated timepoints to assess serum concentration of IFNγ by ELISA. Serum samples that showed values below the detection limit were plotted at the detection limit value in the logarithmic graph. In some studies, complete responders received subcutaneous tumour rechallenge (5 × 10⁵ cells) as indicated in experimental timelines and figure legends. Mice were euthanized when tumour volume had exceeded 500 mm³, or tumour ulceration of more than 5 mm in diameter had been observed.

Detection of serum anti-IL-12 IgG

Blood samples were collected in K3 EDTA-coated tubes (Sarstedt). The samples were centrifuged at 2,000 g for 10 min at 4 °C. Plasma was collected in protein lobind tubes (Eppendorf) and stored at −20 °C until use. A 96-well F-bottom plate (medium binding, Greiner Bio-One) was coated with 5 µg ml⁻¹ of mouse scIL-12-His or molar equivalent of CBD-mouse scIL-12-His in PBS overnight at 37 °C. Following three washes with PBS-T (0.05% Tween 20), the plate was blocked with 2% BSA in PBS-T for 1 h at room temperature. Plasma samples diluted 10× in PBS (−) were added and incubated for 2 h at room temperature. After three washes, Peroxidase AffiniPure F(ab’)₂ Fragment Goat Anti-Mouse IgG (H + L) (Jackson ImmunoResearch, 115-036-003) diluted in PBS (1:10,000) was added and incubated for 1 h at room temperature. After four washes, eBioscience TMB solution (Invitrogen) was added, followed by 0.13 M H₂SO₄ to stop the reaction. The values obtained by subtracting the absorbance at 620 nm from the absorbance at 450 nm are shown.

Quantification of IL-12 in serum, tumour and major organs

Tumour-bearing mice received intravenous injection of CAR-T cells as indicated in experimental timelines and figure legends. Blood samples, tumours and major organs were collected at the indicated timepoints. Serum was collected in 1.5 ml protein lobind tubes (Eppendorf) and stored at −80 °C until use. Parts of the removed tumours and organs were immediately put into pre-weighed Lysing Matrix D tube (MP Biomedicals) containing 1 ml of T-PER Tissue Protein Extraction reagent (Thermo Scientific) supplemented with cOmplete EDTA-free protease inhibitor cocktail (Roche). Tubes with tissue samples were weighed again and tissues were cut into small pieces using surgical scissors. The samples were lysed using FastPrep-24 5G (MP Biomedicals) and stored at −80 °C until use. IL-12 was quantified using mouse IL-12 p70 ELISA kit (Invitrogen) or Human IL-12 p70 DuoSet ELISA (R&D Systems) and normalized to total tissue weight. Recombinant CBD-IL-12 and IL-12 (Supplementary Table 1) produced in-house were used as standards. Tumour and organs of RM9-hSTEAP1-bearing mice receiving 10 million UTD T cells were used to obtain background IL-12 signals for Fig. 3a–h.

A total of 2 × 10⁶ MyC-CaP-hSTEAP1 cells resuspended in 50 µl of 50% PBS (−) and 50% Phenol Red-free Matrigel (Corning) were subcutaneously injected on the left side of the back of each male FVB/N mouse. MyC-CaP-hSTEAP1 tumour-bearing FVB/N mice received 5 million CAR-T cells when average tumour volume was ~230 mm³. Tumours were collected 4 days after CAR-T cell administration and 3 fragments were obtained from each tumour tissue for the experiment. Intratumoural IL-12 was quantified as described above and normalized to total protein content. Total protein content was quantified using the Pierce BCA Protein Assay kit (Thermo Fisher). Bovine serum albumin was used as standard.

Subcutaneous MyC-CaP-hSTEAP1 tumours were established in male NSG mice as described above. Five million CAR-T cells (derived from male C57BL/6J mice) were intravenously injected on day 8. Blood was collected from the tail vein on day 12 for serum IL-12 quantification. The study was terminated on day 16 due to rapid body weight loss indicating graft-versus-host disease.

Systemic toxicity of armoured CAR-T-cell therapy

RM9-hSTEAP1-Fluc tumour-bearing mice received 10 million CAR-T cells on day 4 after tumour inoculation. Terminal blood sampling by cardiac puncture for blood chemistry analysis was performed on day 12. Serum ALT and ALP were quantified using the Skyla VB1 veterinary chemistry analyser. Skyla pre-anaesthetic panel discs were used according to manufacturer instructions.

Quantification of cytokines in CAR-T-treated tumours

RM9-hSTEAP1-Fluc tumour-bearing mice received 10 million CAR-T cells on day 4 after tumour inoculation. Tumours were collected on day 12 and lysed as described above. All samples were stored at −80 °C until use. Intratumoural cytokines were quantified using a LEGENDplex kit (Biolegend) according to manufacturer instructions and normalized to total protein content.

Histological analysis of immune infiltrates in tumour, lung, liver and kidney

RM9-hSTEAP1-Fluc tumour-bearing B6 mice received 5 million CAR-T cells on day 4 after tumour inoculation. Tumours, lungs, livers and kidneys were collected on day 14 and fixed with 4% paraformaldehyde solution in PBS overnight at 4 °C. Formalin-fixed, paraffin-embedded tissue samples were sectioned, followed by H&E staining and immunohistochemistry analysis as previously described²⁴. Briefly, tissue slides were deparaffinized and rehydrated, followed by antigen retrieval in Citrate-Based Antigen Unmasking Solution (Vector Labs). Slides were blocked and stained with rabbit anti-CD3 antibody (Thermo Fisher, MA5-14524, 1:100) for 1 h at 37 °C. Slides were washed three times with TBST (pH 8.0) and incubated with PowerVision Poly-HRP anti-rabbit IgG (Leica Biosystems, 1:100) at 37 °C for 30 min. After three washes, slides were incubated with 3,3′-diaminobenzidine (DAB) (Sigma Aldrich) at room temperature for 10 min. Slides were stained for H&E, followed by dehydration steps and mounting. Stained slides were scanned at the Translational Pathology Core Laboratory (TPCL) at UCLA. QuPath 0.5.1 was used to quantify CD3⁺ area in the IHC images⁵¹. The average channel was used for tissue detection, followed by the DAB channel for detection of CD3⁺ stained area.

Analysis of immune cells in tumour using flow cytometry

RM9-hSTEAP1-Fluc tumour-bearing mice received 5 million CAR-T cells on day 4 after tumour inoculation. Tumours were collected on day 14 and cut into small pieces using surgical scissors and digested in DMEM medium supplemented with 2% FBS, 2 mg ml⁻¹ collagenase D (Sigma Aldrich) and 40 µg ml⁻¹ DNase I (Roche) for 30 min at 37 °C. Single-cell suspensions in DMEM supplemented with 2% FBS were prepared from digested tumours using a 70-µm cell strainer (Thermo Fisher). Red blood cells were lysed with ACK lysing buffer (Gibco) for 5 min at room temperature and neutralized with PBS (−). Cells were stained with BD Horizon Fixable Viability Stain 510 (BD). Fc receptors were blocked using purified anti-mouse CD16/32 antibody (93, Biolegend, 1:200). Cells were stained with a cocktail of anti-mouse antibodies, fixed with eBioscience IC Fixation Buffer (Invitrogen) and acquired using a BD FACSymphony A3 flow cytometer, and data were analysed using FlowJo (BD). Simple linear regression was performed to analyse the correlation between tumour burden and immune cell infiltrates using GraphPad Prism 10. The following anti-mouse antibodies were used: CD45.2 APC-Cy7 (30-F11, Biolegend, 1:200), CD3 BUV395 (145-2C11, BD, 1:100), CD4 BUV805 (GK1.5, BD, 1:200), CD8 Alexa Fluor 700 (53-6.7, Biolegend, 1:200), NK1.1 PerCP-Cy5.5 (PK136, Biolegend, 1:200), PD-1 BV605 (29F.1A12, Biolegend, 1:200), CTLA-4 APC (UC10-4B9, Biolegend, 1:200), CD11b APC-Cy7 (M1/70, Invitrogen, 1:200), Ly6G BUV737 (1A8, BD, 1:200), Ly6C Alexa Fluor 488 (HK1.4, Biolegend, 1:200), CD19 BV785 (6D5, Biolegend, 1:200), CD11c PE-Cy7 (HL3, BD, 1:200), MHC-II (I-A/I-E) BV711 (M5/114.15.2, Biolegend, 1:200) and CD103 BV605 (2E7, Biolegend, 1:200). Gating strategies are shown in Supplementary Figs. 5 and 6.

GeoMx High-plex transcriptome analysis

NanoString GeoMx Digital Spatial Profiler (DSP) whole transcriptome analysis (WTA) was performed on tumour sample collected at day 14 after tumour inoculation. Parafin-embedded tissue blocks were sectioned by the UCLA Translational Pathology Core Laboratory (TPCL). NanoString GeoMx DSP was performed at the Technology Center for Genomics and Bioinformatics (TCGB) core at UCLA. Twelve regions of interest (ROI) were chosen across four different tumour sections in each treated group, followed by library preparation and sequencing. A tumour sample in the unarmoured CAR-T treatment group was not used because the tissue section was damaged.

Transcriptome data analysis was performed using NanoString Spatial Data Analysis software (GeoMx DSP Software v.3.0.0.109). Raw files were analysed for quality control, followed by sequence alignment. Count matrix files were used to perform differential gene expression analysis using DESeq2 (ref. ⁵²). For PCA analysis, fragments per kilobase of transcript per million mapped reads values were normalized by log₂ + 1 transformation and PCA was plotted on the basis of a correlation matrix using the prcomp package v.3.6.2. PCA plots were visualized using the factoextra package v.1.0.7 and ggpubr package v.0.6.0. S1-12 was excluded as an outlier in GSEA on the basis of the PCA analysis. Pathway analysis was performed using GSEA (v.4.3.3)⁵³. All computational analyses were carried out in RStudio v.4.1.0. Heat maps were generated using the package pheatmap v.1.0.12. One-sided Fisher’s exact test was performed using the fisher.test() function in R.

Splenic T cell response to hSTEAP1-positive and negative RM9 cells after CAR-T treatment

RM9-hSTEAP1-Fluc tumour-bearing mice received 10 million CAR-T cells on day 4 after tumour inoculation. Spleens were collected on day 12 and splenic T cells were purified. The T cell enrichment cocktail was supplemented with 1 µg ml⁻¹ Biotin-SP (long spacer) AffiniPure F(ab’)₂ Fragment Donkey Anti-Human IgG (H + L) (Jackson ImmunoResearch, 709-066-149) to remove CAR-T cells. Isolated splenic T cells were maintained in the mouse T cell culture medium until use. A day before co-culture with splenic T cells, 3 × 10⁶ tumour cells were seeded in a T75 flask, supplemented with 20 ng ml⁻¹ recombinant mouse IFNγ (Biolegend) and incubated for 24 h at 37 °C in the presence of 5% CO₂. On the next day, tumour cells were collected using TryPLE express Enzyme (Gibco) and washed twice with the mouse T cell culture medium without cytokines. Tumour cells (2 × 10⁵) and the same number of splenic T cells were mixed in 200 µl of mouse T cell culture medium without cytokines, seeded in a 96-well V-bottom plate (Corning) and incubated for 48 h at 37 °C in the presence of 5% CO₂. Supernatant was collected and mouse IFNγ was quantified by ELISA.

Manufacture and functional testing of CAR Jurkat cells

Jurkat cells were transduced with 10× concentrated lentiviral vector using a Retronectin-coated plate as described above. Surface expression of the CAR was detected using Biotin-SP-conjugated AffiniPure F(ab’)2 Fragment Goat Anti-Human IgG, F(ab’)2 Fragment Specific (Jackson ImmunoResearch, 109-066-006, 1:200) and PE-streptavidin (Biolegend, 1:200). CAR⁺ Jurkat T cells (50,000) were stimulated with eBioscience cell stimulation cocktail (Invitrogen). Secreted IL-12 variants were quantified by ELISA. Human scIL-12-His and CBD-human scIL-12-His were used as standards.

Manufacture and functional testing of human CAR-T cells (A)

The following procedures were performed to generate data in Fig. 8a–c. Healthy donor human peripheral blood mononuclear cells (PBMCs) (STEMCELL Technologies) were subjected to EasySep CD8 negative selection (STEMCELL Technologies). CD8⁺ T cells (1 × 10⁶ per ml) were stimulated with T cell TransAct (1:100, Miltenyi Biotec) in RPMI media supplemented with 5% human serum (Bloodworks Northwest) and 50 U ml⁻¹ human IL-2 (Peprotech). Next day, 5× concentrated lentiviral vector was added to the culture with 10 µg ml⁻¹ protamine sulfate (MP Biomedicals). After 24 h, media were replaced with fresh IL-2 to remove viruses and TransAct. Seven days after the addition of viruses, cells were stained with biotin-anti-Fab (Jackson Immunoresearch, 109-066-006, 1:100) followed by PE-streptavidin (Thermo Fisher, 1:200), and CAR⁺ cells were sorted (except untransduced cells from which all live cells were sorted) on a BD FACSymphony S6 system. Sorted cells (5 × 10⁴) were expanded with irradiated feeder cells (2 × 10⁶ mixed PBMCs, three individual donor PBMCs from STEMCELL Technologies) and 4 × 10⁵ Epstein–Barr virus-transformed B-lymphoblastoid cells (Fred Hutch Research Cell Bank), 50 U ml⁻¹ IL-2, 10 ng ml⁻¹ IL-15 and 0.3 ml anti-CD3 antibodies (OKT3, Miltenyi Biotec) in 1 ml culture volume. After 10 days, T cells were used for functional assays. For the killing assay, 1 × 10⁵ T cells were co-cultured with 1 × 10⁴ target cells (22Rv1 hSTEAP1 KO + rescue or 22Rv1 hSTEAP1 KO)²⁴ lentivirally transduced to express mCherry in 96-well flat-bottom plates (Corning). Target killing was monitored in IncuCyte (Sartorius), detecting fluorescence from mCherry every 4 h using whole-well scanning. The integrated intensity per well was normalized to values from the first timepoint. For cytokine production, 1 × 10⁵ each of T cells and target cells were co-cultured in 96-well round-bottom plates (Corning) with 1:1,000-diluted Golgi Stop (monensin, BD) and Golgi Plug (brefeldin A, BD) for 18 h. Cells were then surface stained with Live/Dead Aqua (Thermo Fisher, 1:1,000) and FITC-CD8 (SK1, Biolegend, 1:200) for 15 min at 4 °C, fixed and permeabilized with BD Cytofix/Cytoperm (BD), intracellularly stained with APC-IFNγ (4S.B3, Biolegend, 1:100) and PE-IL-12 (20C2, BD, 1:100) for 15 min at 4 °C, and analysed on a BD FACSymphony A5 system.

Manufacture and functional testing of human CAR-T cells (B)

The following procedures were performed to generate data in Fig. 8d–h and Extended Data Fig. 10d–f. PBMCs were isolated from fresh whole human male blood in acid citrate dextrose (Research Donors) using Ficoll Paque Plus (Cytiva) and cryopreserved using CryoStor CS5 (STEMCELL Technologies) until use. CD3⁺ T cells were isolated using a Magnisort Human T cell Enrichment kit (Invitrogen) and activated using Dynabeads Human T-Activator CD3/CD28 (Thermo Fisher) in CTS OpTmizer T Cell Expansion SFM (Gibco) supplemented with CTS OpTmizer T Cell Expansion Supplement, 2 mM l-glutamine, 5% CTS Immune Cell SR (Gibco), 1% penicillin/streptomycin, 10 ng ml⁻¹ recombinant human IL-2 and 0.5 ng ml⁻¹ recombinant human IL-15. At 24 h after activation, T cells were transduced with a RetroNectin-coated plate as described above. After 2 days, the activation beads were magnetically removed, and T cells were expanded until use. %CAR⁺ was assessed using biotinylated anti-human IgG and streptavidin-PE as described above.

Manufacture and functional testing of human CAR-T cells (C)

The following procedures were performed to generate data in Fig. 8i–l and Extended Data Fig. 10g–i. Human CAR-T cells were generated as previously described²⁴. Briefly, PBMCs were isolated from a de-identified healthy donor. Cryopreserved PBMCs were thawed, washed and resuspended in T cell media (TCM base) consisting of AIM-V medium (Gibco) supplemented with 55 mM β-mercaptoethanol, human male AB plasma (Sigma) and GlutaMAX. Following cell counting, CD4⁺ and CD8⁺ T cells were isolated using human CD4 and CD8 microbeads (Miltenyi Biotec) according to manufacturer protocol. CD4⁺ T cells were cultured in CD4 media (TCM base supplemented with 0.5 ng ml⁻¹ IL-15 and 5 ng ml⁻¹ IL-7), and CD8⁺ T cells were cultured in CD8 media (TCM base supplemented with 50 U ml⁻¹ IL-2 and 0.5 ng ml⁻¹ IL-15). T cells were activated on Day 0 using Dynabeads Human T-Activator CD3/CD28 following manufacturer instructions. On Day 2, CD4⁺ and CD8⁺ T cells were transduced with high-titre lentivirus at a multiplicity of infection (MOI) of 12.5, based on titration on HEK293T cells. On Day 4, activation beads were removed, cells were counted, and transduction efficiency was assessed by flow cytometry using R-PE-conjugated Protein L (Sino Biological, 2 µl/2.5 × 10⁵ cells). CAR-modified T cells were expanded until day 14 before experiments. The CAR-T cells and 22Rv1-Fluc-RFP cells were co-cultured as reported previously to evaluate target cell killing and secretion of IL-12 and IFNγ²⁴.

In vivo functional assays of primary human CAR-T cells

For the subcutaneous 22Rv1 model, a total of 2 × 10⁶ 22Rv1-Fluc-RFP cells resuspended in 50 µl of 50% PBS (−) and 50% Phenol Red-free Matrigel were subcutaneously injected on the left side of the back of each male NSG mouse on day 0. The mice received 5 million CAR-T cells intravenously on day 13, followed by blood samplings on days 20, 24 and 28 for serum IL-12 quantification. Serum samples that showed values below the detection limit were plotted at the detection limit value in the logarithmic graph. Tumour burden was tracked using a digital caliper as described above. For the 22Rv1 disseminated tumour model, 1 × 10⁶ 22Rv1-Fluc-RFP cells were intravenously injected on day −14. Five million human UTD, hSTEAP1-BBζ and armoured hSTEAP1-BBζ CAR-T cells at a defined CD4:CD8 ratio of 1:1 suspended in 100 µl of PBS were injected intravenously on day 0. Serial BLI images were recorded to assess the progression of tumours, and retro-orbital bleeds were collected at day 12 to quantify serum IL-12.

Statistical analysis

Data and statistical analyses were performed using MicroSoft Excel v.16.88 and GraphPad Prism 10 except for transcriptomics data, which were analysed using R (as described above). P < 0.05 was considered statistically significant. Tests used are indicated in figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

1. Maude, S. L. et al. Chimeric antigen receptor T cells for sustained remissions in leukemia. N. Engl. J. Med. 371, 1507–1517 (2014).

   Article  PubMed  PubMed Central  Google Scholar 

2. Brudno, J. N. et al. T cells genetically modified to express an anti-B-cell maturation antigen chimeric antigen receptor cause remissions of poor-prognosis relapsed multiple myeloma. J. Clin. Oncol. 36, 2267–2280 (2018).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

3. Raje, N. et al. Anti-BCMA CAR T-Cell Therapy bb2121 in relapsed or refractory multiple myeloma. N. Engl. J. Med. 380, 1726–1737 (2019).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

4. Rafiq, S., Hackett, C. S. & Brentjens, R. J. Engineering strategies to overcome the current roadblocks in CAR T cell therapy. Nat. Rev. Clin. Oncol. 17, 147–167 (2020).

   Article  PubMed  Google Scholar 

5. Chen, N., Li, X., Chintala, N. K., Tano, Z. E. & Adusumilli, P. S. Driving CARs on the uneven road of antigen heterogeneity in solid tumors. Curr. Opin. Immunol. 51, 103–110 (2018).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

6. Tugues, S. et al. New insights into IL-12-mediated tumor suppression. Cell Death Differ. 22, 237–246 (2015).

   Article  CAS  PubMed  Google Scholar 

7. Rogge, L. et al. Selective expression of an interleukin-12 receptor component by human T helper 1 cells. J. Exp. Med. 185, 825–831 (1997).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

8. Melero, I., Rouzaut, A., Motz, G. T. & Coukos, G. T-cell and NK-cell infiltration into solid tumors: a key limiting factor for efficacious cancer immunotherapy. Cancer Discov. 4, 522–526 (2014).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

9. Gordon-Alonso, M., Hirsch, T., Wildmann, C. & Bruggen, P. V. Galectin-3 captures interferon-gamma in the tumor matrix reducing chemokine gradient production and T-cell tumor infiltration. Nat. Commun. 8, 793 (2017).

   Article  PubMed  PubMed Central  Google Scholar 

10. Trinchieri, G. Interleukin-12 and the regulation of innate resistance and adaptive immunity. Nat. Rev. Immunol. 3, 133–146 (2003).

    Article  CAS  PubMed  Google Scholar 

11. Zhou, F. Molecular mechanisms of IFN-γ to up-regulate MHC class I antigen processing and presentation. Int. Rev. Immunol. 28, 239–260 (2009).

    Article  CAS  PubMed  Google Scholar 

12. Sathe, P. et al. The acquisition of antigen cross-presentation function by newly formed dendritic cells. J. Immunol. 186, 5184–5192 (2011).

    Article  CAS  PubMed  Google Scholar 

13. Steding, C. E. et al. The role of interleukin-12 on modulating myeloid-derived suppressor cells, increasing overall survival and reducing metastasis. Immunology 133, 221–238 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

14. Leonard, J. P. et al. Effects of single-dose interleukin-12 exposure on interleukin-12-associated toxicity and interferon-gamma production. Blood 90, 2541–2548 (1997).

    CAS  PubMed  Google Scholar 

15. Atkins, M. B. et al. Phase I evaluation of intravenous recombinant human interleukin 12 in patients with advanced malignancies. Clin. Cancer Res. 3, 409–417 (1997).

    CAS  PubMed  Google Scholar 

16. Kerkar, S. P. et al. Tumor-specific CD8⁺ T cells expressing interleukin-12 eradicate established cancers in lymphodepleted hosts. Cancer Res. 70, 6725–6734 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

17. Pegram, H. J. et al. Tumor-targeted T cells modified to secrete IL-12 eradicate systemic tumors without need for prior conditioning. Blood 119, 4133–4141 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

18. Zhang, L. et al. Evaluation of γ-retroviral vectors that mediate the inducible expression of IL-12 for clinical application. J. Immunother. 35, 430–439 (2012).

    Article  PubMed  PubMed Central  Google Scholar 

19. Zhang, L. et al. Improving adoptive T cell therapy by targeting and controlling IL-12 expression to the tumor environment. Mol. Ther. 19, 751–759 (2011).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

20. Zhang, L. et al. Tumor-infiltrating lymphocytes genetically engineered with an inducible gene encoding interleukin-12 for the immunotherapy of metastatic melanoma. Clin. Cancer Res. 21, 2278–2288 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

21. Ishihara, J. et al. Targeted antibody and cytokine cancer immunotherapies through collagen affinity. Sci. Transl. Med. 11, eaau3259 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

22. Mansurov, A. et al. Collagen-binding IL-12 enhances tumour inflammation and drives the complete remission of established immunologically cold mouse tumours. Nat. Biomed. Eng. 4, 531–543 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

23. Armstrong, A. J. et al. Five-year survival prediction and safety outcomes with enzalutamide in men with chemotherapy-naive metastatic castration-resistant prostate cancer from the PREVAIL trial. Eur. Urol. 78, 347–357 (2020).

    Article  CAS  PubMed  Google Scholar 

24. Bhatia, V. et al. Targeting advanced prostate cancer with STEAP1 chimeric antigen receptor T cell and tumor-localized IL-12 immunotherapy. Nat. Commun. 14, 2041 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

25. Jia, Z. et al. IL12 immune therapy clinical trial review: novel strategies for avoiding CRS-associated cytokines. Front. Immunol. 13, 952231 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

26. Kohli, K., Pillarisetty, V. G. & Kim, T. S. Key chemokines direct migration of immune cells in solid tumors. Cancer Gene Ther. 29, 10–21 (2022).

    Article  CAS  PubMed  Google Scholar 

27. del, Rio, M. L., Bernhardt, G., Rodriguez-Barbosa, J. I. & Förster, R. Development and functional specialization of CD103⁺ dendritic cells. Immunol. Rev. 234, 268–281 (2010).

    Article  Google Scholar 

28. Hildner, K. et al. Batf3 deficiency reveals a critical role for CD8α⁺ dendritic cells in cytotoxic T cell immunity. Science 322, 1097–1100 (2008).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

29. Hirschhorn, D. et al. T cell immunotherapies engage neutrophils to eliminate tumor antigen escape variants. Cell 186, 1432–1447 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

30. Youn, J. I. & Gabrilovich, D. I. The biology of myeloid-derived suppressor cells: the blessing and the curse of morphological and functional heterogeneity. Eur. J. Immunol. 40, 2969–2975 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

31. Johnston, R. J. et al. Bcl6 and Blimp-1 are reciprocal and antagonistic regulators of T follicular helper cell differentiation. Science 325, 1006–1010 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

32. Nurieva, R. I. et al. Bcl6 mediates the development of T follicular helper cells. Science 325, 1001–1005 (2009).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

33. Yu, D. et al. The transcriptional repressor Bcl-6 directs T follicular helper cell lineage commitment. Immunity 31, 457–468 (2009).

    Article  CAS  PubMed  Google Scholar 

34. Muramatsu, M. et al. Class switch recombination and hypermutation require activation-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell 102, 553–563 (2000).

    Article  CAS  PubMed  Google Scholar 

35. Revy, P. et al. Activation-induced cytidine deaminase (AID) deficiency causes the autosomal recessive form of the Hyper-IgM syndrome (HIGM2). Cell 102, 565–575 (2000).

    Article  CAS  PubMed  Google Scholar 

36. Shaffer, A. L. et al. Blimp-1 orchestrates plasma cell differentiation by extinguishing the mature B cell gene expression program. Immunity 17, 51–62 (2002).

    Article  CAS  PubMed  Google Scholar 

37. Ryffel, B. Interleukin-12: role of interferon-gamma in IL-12 adverse effects. Clin. Immunol. Immunopathol. 83, 18–20 (1997).

    Article  CAS  PubMed  Google Scholar 

38. Williford, J. M. et al. Recruitment of CD103 dendritic cells via tumor-targeted chemokine delivery enhances efficacy of checkpoint inhibitor immunotherapy. Sci. Adv. 5, eaay1357 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

39. Sasaki, K. et al. Engineered collagen-binding serum albumin as a drug conjugate carrier for cancer therapy. Sci. Adv. 5, eaaw6081 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

40. Gattinoni, L. et al. Removal of homeostatic cytokine sinks by lymphodepletion enhances the efficacy of adoptively transferred tumor-specific CD8 T cells. J. Exp. Med. 202, 907–912 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

41. Brudno, J. N. & Kochenderfer, J. N. Recent advances in CAR T-cell toxicity: mechanisms, manifestations and management. Blood Rev. 34, 45–55 (2019).

    Article  CAS  PubMed  Google Scholar 

42. Lee, E. H. J. et al. Antigen-dependent IL-12 signaling in CAR T cells promotes regional to systemic disease targeting. Nat. Commun. 14, 4737 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

43. Hu, J. et al. Cell membrane-anchored and tumor-targeted IL-12 (attIL12)-T cell therapy for eliminating large and heterogeneous solid tumors. J. Immunother. Cancer 10, e003633 (2022).

    Article  PubMed  PubMed Central  Google Scholar 

44. Zhang, L. et al. Enhanced efficacy and limited systemic cytokine exposure with membrane-anchored interleukin-12 T-cell therapy in murine tumor models. J. Immunother. Cancer 8, e000210 (2020).

    Article  PubMed  PubMed Central  Google Scholar 

45. Jones, D. S. et al. Cell surface-tethered IL-12 repolarizes the tumor immune microenvironment to enhance the efficacy of adoptive T cell therapy. Sci. Adv. 8, eabi8075 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

46. Luo, Y. et al. IL-12 nanochaperone-engineered CAR T cell for robust tumor-immunotherapy. Biomaterials 281, 121341 (2022).

    Article  CAS  PubMed  Google Scholar 

47. Ribba, A. S. et al. Ser968Thr mutation within the A3 domain of von Willebrand Factor (VWF) in two related patients leads to a defective binding of VWF to collagen. Thromb. Haemost. 86, 848–854 (2001).

    Article  CAS  PubMed  Google Scholar 

48. Zufferey, R. et al. Self-inactivating lentivirus vector for safe and efficient in vivo gene delivery. J. Virol. 72, 9873–9880 (1998).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

49. Lanitis, E. et al. Optimized gene engineering of murine CAR-T cells reveals the beneficial effects of IL-15 coexpression. J. Exp. Med. 218, e20192203 (2021).

    Article  CAS  PubMed  Google Scholar 

50. Tanito, K. et al. Comparative evaluation of natural killer cell-mediated cell killing assay based on the leakage of an endogenous enzyme or a pre-loaded fluorophore. Anal. Sci. 37, 1571–1575 (2021).

    Article  CAS  PubMed  Google Scholar 

51. Bankhead, P. et al. QuPath: open source software for digital pathology image analysis. Sci. Rep. 7, 16878 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

52. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).

    Article  PubMed  PubMed Central  Google Scholar 

53. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA 102, 15545–15550 (2005).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

54. Sasaki, K. et al. Spatial transcriptomics data of RM9-hSTEAP1 tumor tissues treated with CAR-T cells (Datasets). NCBI GEO https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE300750 (2025).

Download references

Authors and Affiliations

1. Department of Bioengineering, Imperial College London, London, UK

   Koichi Sasaki, Pooja Kaur, Chuyi Wang, Takumi Matsuo, Olivier Dubois, Po-Chuan Chiu, Charanjit Singh, Ioanna Panagi, Laurine Noblecourt, Maria Nikolaidi & Jun Ishihara

2. Human Biology Division, Fred Hutchinson Cancer Center, Seattle, WA, USA

   Vipul Bhatia, Gerardo Javier & John K. Lee

3. Division of Hematology/Oncology, Department of Medicine, David Geffen School of Medicine at UCLA, Los Angeles, CA, USA

   Vipul Bhatia, Truman Chong, Gerardo Javier & John K. Lee

4. Translational Science and Therapeutics Division, Fred Hutchinson Cancer Center, Seattle, WA, USA

   Yuta Asano, Jakob Bakhtiari & Aude G. Chapuis

5. Department of Microbiology, Immunology and Molecular Genetics, UCLA, Duarte, CA, USA

   Donny Gun

6. Department of Hematology and Hematopoietic Cell Transplantation, City of Hope, Duarte, CA, USA

   Saul J. Priceman

7. Department of Immuno-Oncology, Beckman Research Institute of City of Hope, Duarte, CA, USA

   Saul J. Priceman

8. Division of Medical Oncology, University of Washington, Seattle, WA, USA

   Aude G. Chapuis

9. Parker Institute for Cancer Immunotherapy at UCLA, Los Angeles, CA, USA

   John K. Lee

10. Exploratory Oncology Research and Clinical Trial Center (EPOC), National Cancer Center, Chiba, Japan

    Jun Ishihara

Contributions

K.S., J.K.L. and J.I. conceived the project. K.S., V.B., Y.A., I.P., L.N., S.J.P., A.G.C., J.K.L. and J.I. designed experiments. K.S., V.B., Y.A. and J.B. performed experiments. P.K., C.W., T.M., O.D., P.-C.C., D.G., C.S., L.N., M.N. and T.C. assisted with tumour experiments. V.B. and G.J. analysed spatial genomics data. K.S. and J.I. wrote the original draft. K.S., V.B., Y.A., S.J.P., A.G.C., J.K.L. and J.I. edited the manuscript with input from all other authors. J.K.L. and J.I. supervised the research.

Corresponding authors

Correspondence to John K. Lee or Jun Ishihara.

Competing interests

K.S., Y.A., A.G.C., J.K.L. and J.I. are inventors on a US provisional patent application (No. 63/650,382) covering the technology described in this work. K.S. and J.I. are inventors on international patent applications (PCT/US2019/020685) covering CBD-IL-12 protein therapy. J.K.L. is an inventor on international patent applications (PCT/US2023/062428) related to STEAP1 CAR-T cells. J.K.L. holds equity in, serves on the scientific advisory board of, and receives research funding from PromiCell Therapeutics. J.K.L. is a consultant for Lyell Immunopharma. J.I. is a founder and shareholder in Arrow Immune Inc. J.I. is a scientific advisor of EsoBiotec. The remaining authors declare no competing interests.

Peer review information

Nature Biomedical Engineering thanks Daniel Goodman, Lisa Volpatti and Di Yu for their contribution to the peer review of this work.

Source Data

Source Data for Figs. 1–5, 7 and 8 and Extended Data Figs. 2–4 and 6–10.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions