IL-12-engineered human PSMA-CAR T cells for the treatment of advanced prostate cancer

Abstract

Adoptive cell therapies used to treat advanced prostate cancer are being developed to target several tumor-associated antigens, including prostate-specific membrane antigen (PSMA). Chimeric antigen receptor (CAR) T cell therapy using the single-chain variable fragment (scFv) derived from the humanized murine mAb clone J591 as the antigen-binding domain has shown promising anti-tumor activity. However, it has also been associated with macrophage activation syndrome and other unwanted toxicities, highlighting the need for more specific and human-derived antigen-binders with optimized construct designs for improved safety and efficacy. Here, we optimize a human scFv-based PSMA-targeted CAR (hPSMA-CAR) with highly selective PSMA targeting. We further introduce a membrane-bound IL-12 (mbIL12) molecule, which enhances potency with increased T cell expansion, IFN-γ production, and anti-tumor cell activity in vitro. Using two clinically relevant bone-metastatic prostate cancer models, we show that mbIL12-engineered hPSMA-CAR T cells drive potent in vivo anti-tumor responses. In summary, we have developed a promising therapeutic that has potential to promote safe and effective treatment of advanced PSMA+ prostate cancer.

Graphical abstract

PSMA is used as a diagnostic marker for advanced prostate cancer, but attempts to use it as a therapeutic target for immunotherapy approaches have been challenging. Here, the authors outline systematic optimization of a fully human PSMA CAR T cell with membrane-bound IL-12 engineering for the effective treatment of prostate cancer.

Introduction

Metastatic castration-resistant prostate cancer (mCRPC) represents a challenging stage of prostate cancer with a 5-year survival rate around 30% and limited effective treatment options.¹ Although immune checkpoint blockade (ICB) has had striking clinical successes in select solid and hematological malignancies, these therapies have not shown similar therapeutic responses in mCRPC.² Therefore, bispecific T cell engagers (BITEs) and chimeric antigen receptor (CAR) T cell therapies have been an active area of clinical investigation.^3,4,5,6,7 Several tumor targets are attractive in mCRPC, including prostate-specific membrane antigen (PSMA).⁸ Although expressed at low levels on some normal tissues (salivary glands, duodenal mucosa, proximal tubules, and brain),^{9,10,11,12,13} its overexpression in advanced prostate cancer makes it a favored target for therapeutic interventions. Several groups have reported clinical experiences with CAR T cells and BITEs targeting PSMA, which have, to date, been challenged by severe adverse events.^6,14,15 The therapies of note are the PSMAxCD3 bispecifics, which have reported high titers of anti-drug antibodies following repeat infusions, and armored PSMA-CAR T cells with treatment-related adverse events (TRAEs).^16,17,18 As an example, the monoclonal antibody J591 binds PSMA with picomolar affinity and has been routinely used to target PSMA. Humanized J591 and related anti-PSMA single-chain variable fragments (scFvs), which are being clinically tested in CAR T cell therapies, still have the potential to be immunogenic. As such, there is a critical need for potent and safe PSMA-targeted CAR T cell therapies.

Here, we report the systematic optimization of a human PSMA (hPSMA) CAR T cell for the treatment of mCRPC. Using Eureka Therapeutics’ E-ALPHA phage display library, we identified a lead human anti-PSMA scFv and optimized the CAR extracellular spacer and transmembrane domains for selective PSMA targeting. We found that a dCH2 spacer together with a CD28 transmembrane domain on a 4-1BBz-containing CAR showed the greatest specificity for PSMA-expressing tumor cells. To address suboptimal potency of the hPSMA-CARs, scFv affinity maturation (AM) was performed but did not significantly improve the potency of hPSMA-CAR T cells. We then evaluated engineering hPSMA-CAR T cells with a membrane-bound IL-12 (mbIL12) molecule to enhance anti-tumor activity. hPSMA-CAR/mbIL12 T cells showed improved tumor cell killing, expansion, and cytokine secretion while maintaining highly specific PSMA targeting. Using an in vitro macrophage:tumor:T cell co-culture model, we show an improved cytokine safety profile of hPSMA-CAR/mbIL12 T cells compared with J591-based CAR T cells. Finally, we showed that our hPSMA-CAR/mbIL12 T cells resulted in robust anti-tumor activity in in vivo and bone-metastatic prostate cancer models. These data highlight PSMA-CAR T cells engineered with mbIL12 as a promising therapy for the treatment of mCRPC, with broader implications for the development of safe and effective CAR T cells for other solid tumors.

Results

Development of a human-derived PSMA-CAR T cell

We first attempted to develop a PSMA-CAR using a human scFv to reduce toxicity and T cell clearance due to anti-scFv immunogenicity. Eureka Therapeutics’ E-ALPHA phage display platform was used to screen anti-PSMA antibodies in scFv format using PSMA-overexpressing Jurkat cells. After three rounds of panning, 1,080 clones were screened, and of these, two unique clones were obtained, ET260-1 and ET260-2. These clones were confirmed by DNA sequencing. Both clones stained PSMA-positive cells but not PSMA-negative cells (Figure S1). Binding kinetic studies of the ET260 clones showed that ET260-2 had lower binding (on-rate) and dissociation (off-rate) compared with ET260-1 (Figure S1; Table S1).

The ET260-1 and ET260-2 scFvs were cloned into a CAR construct comprising a CH3- IgG extracellular spacer (CH2-deleted, termed dCH2), a CD4 transmembrane domain (CD4tm), an intracellular 4-1BB costimulatory domain (41BB), and a CD3z cytolytic domain (CD3z), along with a truncated CD19 gene (CD19t) separated by a T2A skip sequence (referred to as hPSMA-CARs) (Figure 1A). The J591 scFv was cloned into the same construct to serve as a reference for the activity of our hPSMA-CARs. T cells were transduced with lentivirus encoding the CAR constructs and enriched for CD19t (Figure S2A). We detected the CAR directly through the dCH2 extracellular spacer by anti-Fc staining, and expression was comparable across the three PSMA-CARs (Figure 1B). We then tested these CARs against PC-3 cells in a coculture over 72 h at an effector-to-tumor cell (E:T) ratio of 1:4. These cells were engineered to express varying levels of PSMA (Figure S3). The ET260-1 CAR had greater tumor cell killing, activation, and IFN-γ secretion compared with the ET260-2 CAR (Figures 1C–1E). While J591-based CAR T cells showed superior tumor cell killing, activation, and IFN-γ release against PSMA-expressing PC-3 cells compared with the hPSMA-CAR T cells, J591 CAR T cells also showed significantly higher tumor cell killing and activation against the PSMA-negative PC-3 parental cell line (WT) compared with the hPSMA-CARs (Figures 1C and 1D). For the purpose of our studies, we use activity against PC-3 WT as a proxy for on-target, off-tumor toxicities due to the low levels of PSMA^10,13,19,20 (Figure S3).

Figure 1.

Development of a human PSMA-CAR T cell

(A) Schema of hPSMA-CAR scFv variants with a dCH2 Fc extracellular linker, CD4tm, 4-1BB costimulatory domain, and CD3ζ signaling domain construct. (B) Flow cytometric analysis of CARs on the surface of primary T cells via detection of the Fc linker domain. (C Tumor cell killing of hPSMA-CARs over 72 h when co-cultured with PSMA-expressing cell lines at an effector-to-tumor ratio (E:T) of 1:4. Tumor cell killing was calculated relative to the UTD T cell condition. (D Expression of 4-1BB on hPSMA-CARs compared to the J591-CAR after 72 h. (E) Production of IFN-γ at 72 h as determined by ELISA. (F and G) Tumor cell killing and fold expansion of hPSMA CARs after 8-day LTK at an E:T of 1:20. Data are presented as mean values ± SEM. p values indicate differences between ET260-1-dCH2(4tm)BBz and ET260-2-dCH2(4tm)BBz and between ET260-1-dCH2(4tm)BBz and J591-dCH2(4tm)BBz using a two-tailed Student’s t test.

To further assess the functionality of these CARs, we performed extended long-term killing (LTK) assays over 8 days at an E:T of 1:20. The hPSMA-CARs did not perform equivalently to J591-CARs, with lower tumor cell killing and fold expansion (Figures 1F and 1G). Importantly, the ET260-1 CAR showed promising effector function against PC-3 PSMA cell lines and limited activity against PC-3 WT, and based on these results, the ET260-1 scFv was selected for further optimization. These studies suggest that while human scFv-based PSMA-CARs can target PSMA+ tumor cells in vitro, their activity was diminished compared with the humanized J591-CAR. This may be explained by the binding kinetics of J591, which showed higher binding (on-rate) and lower dissociation (off-rate) compared with the ET260 scFvs (Table S1).

Optimization of hPSMA-CAR extracellular spacer and transmembrane domains

We next evaluated whether optimization of the CAR backbone with the ET260-1 scFv could improve the potency of hPSMA-CARs. We evaluated hPSMA-CAR constructs using three different extracellular spacer lengths (dCH2, CD8h, and HL) and three different transmembrane domains (CD4tm, CD8tm, and CD28tm), all using the BBz backbone (Figure 2A). T cells showed comparable transduction efficiency and were subsequently enriched for CD19t for further studies (Figure S2B). We observed higher Fc staining in the dCH2(28tm)BBz-containing hPSMA-CAR compared with the original CD4tm construct and the J591-CAR (Figure 2B). To evaluate the in vitro functionality of these CAR T cells, we performed tumor cell killing assays against PC-3 cells engineered with varying levels of PSMA over 72 h. ET260-1-dCH2(4tm)BBz and ET260-1-dCH2(28tm)BBz CAR T cells performed comparably to J591-CAR T cells against PC-3 PSMA^mid and PSMA^hi cells (Figure 2C). The dCH2(28tm) containing hPSMA-CAR T cells showed higher 4-1BB expression and IFN-γ secretion compared with the dCH2(4tm), CD8h(8tm), and HL(28tm)-containing hPSMA-CAR T cells. These versions of hPSMA-CAR T cells showed minimal increases in non-specific activity against PC-3 WT cells, compared with J591-CAR T cells (Figures 2C and 2D). hPSMA-CAR T cells containing dCH2(CD28tm) showed improved activity against our PC-3 PSMA cell line compared with dCH2(4tm) but were still suboptimal compared with J591-CAR T cells (Figures 2F and 2G). Although we observed suboptimal activity of our ET260-1-based CAR T cells compared with J591-CAR T cells, it is important to highlight their specificity and sensitivity to PSMA combined with limited activity against the WT cell line. In summary, we improved the potency of the hPSMA-CAR T cells with dCH2(28tm)BBz, but it remained suboptimal in long-term tumor cell killing potential compared with J591-CAR T cells.

Figure 2.

Optimization of hPSMA-CAR extracellular spacer and transmembrane domains

(A) Schema of hPSMA-CAR variants with different extracellular linkers (dCH2, CD8h, HL) and transmembrane domains (CD4tm, CD28tm, CD8tm), a 4-1BB costimulatory domain, and a CD3ζ signaling domain. (B) Flow cytometric analysis of ET260-1 CAR variants on the surface of primary T cells via detection of Fc. (C) Tumor cell killing of hPSMA-CARs over 72 h when co-cultured with PSMA-expressing cell lines at an E:T of 1:4. (D) Expression of 4-1BB on ET260-1 CAR variants compared to J591 after 72 h. (E) Production of IFN-γ at 72 h as determined by ELISA. (F and G) Tumor cell killing and fold expansion of ET260-1 CAR variants after 8-day LTK at an E:T of 1:20. Data are presented as mean values ± SEM. p values indicate differences between ET260-1-dCH2(4tm)BBz and ET260-1-dCH2(28tm)BBz and between ET260-1-dCH2(28tm)BBz and J591-dCH2(4tm)BBz using a two-tailed Student’s t test.

AM of hPSMA scFv impacts CAR stability and activity in vitro

We hypothesized that the potent activity of J591-CAR T cells was, in part, due to its high affinity for PSMA. We therefore sought to improve the binding kinetics of our ET260-1 scFv through AM. Indeed, we observed that while J591 had a high on-rate, it also had a particularly slow off-rate (Table S1). In contrast, our ET260-1 scFv had a lower on-rate and a higher off-rate than J591. To address this, AM using error-prone PCR was performed, identifying 10 unique affinity-matured scFv clones carrying different mutations, which were shown to bind Jurkat-PSMA cells but not Jurkat parental cells. These clones were subjected to surface plasmon resonance (SPR) to determine affinity for the PSMA extracellular domain protein (Figure S4). The parental scFv, ET260-1(WT), served as a positive control (Table 1). In addition to ET260–1, J591 served as a reference of kinetic activity, and based on the binding kinetics of the clones relative to both ET260–1 and J591, five clones were selected and cloned into our dCH2(CD28tm)BBz CAR construct (Figure 3A). Specifically, clones were selected based on their improved KD compared with the parental ET260 and kinetics that resembled J591, i.e., higher on-rate and slow off-rate (Table S3). T cells were engineered to express CARs containing the respective scFv and enriched for CD19t. Interestingly, of the five clones, clones 56 and 69 did not show stable CAR expression (by Fc staining), while clone 39 showed two distinct populations. Clones 58 and 67 most closely resembled the parental CAR expression pattern (Figure 3B).

Table 1.

Affinities and binding kinetics of affinity-matured hPSMA scFvs

	on-rate ka1 (1/Ms)	off-rate kd1 (1/s)	KD1 (M)
ET260-1	1.97E+05	2.20E−03	1.120E−8
AMET260-1-39	6.76E+05	5.42E−03	8.01E−09
AMET260-1-69	1.37E+04	1.26E−04	9.19E−09
AMET260-1-67	2.38E+04	4.99E−05	2.10E−09
AMET260-1-58	3.61E+04	5.75E−05	1.60E−09
AMET260-1-56	1.97E+05	4.01E−04	2.03E−09
J591	5.84E+05	2.04E−4	3.50E−10

Binding to PSMA was determined by surface plasmon resonance spectroscopy (SPR). Kinetic constants from SPR measurements were determined, and KD was calculated from the values of the rate constants.

Figure 3.

Affinity maturation of ET260-1 scFv impacts CAR stability and activity in vitro

(A) Schema of hPSMA-CAR scFv variants. (B) Flow cytometric analysis of affinity-matured scFvs in primary T cells as determined by CD19t and Fc expression. (C) Tumor cell killing of AM-CARs over 72 h at an E:T 1:4. (D) Expression of 4-1BB on CAR T cells after 72 h co-culture. (E) Production of IFN-γ by AM-CAR T cells after 72 h by ELISA. (F) Activity of AM-CARs against endogenous expressers of PSMA as determined by IFN-γ ELISA after a 48 h co-culture at an E:T of 1:1. Data are presented as mean values ± SEM. p values indicate differences between ET260-1-dCH2(28tm)BBz and ET260-1(58)-dCH2(28tm)BBz and between ET260-1(58)-dCH2(4tm)BBz and J591-dCH2(4tm)BBz using a two-tailed Student’s t test.

Functional testing of our AM-CAR T cells showed that clone 58 performed most comparably to the parental scFv with respect to tumor cell killing, activation, and IFN-γ production after 72 h (Figures 3C–3E). While we observed high tumor cell killing and IFN-γ secretion against PC-3 WT by J591-CAR T cells, clone 58 did not show increased activity against the WT cell line (Figures 3C–3E). Over the course of the 8-day assay, AM-CAR T cells continued to show suboptimal activity compared with J591-CAR T cells (Figures S5A and S5B). We also assessed the activity of our hPSMA-CAR T cells against endogenous PSMA expressers by co-culturing T cells and quantifying IFN-γ production (Figure 3F). We observed cytokine secretion with both ET260-1 parental and AM-CAR T cells against the endogenous PSMA expressers, CWR22Rv1, LNCaP, and C4-2. Additionally, we found that this activity was antigen-dependent with IFN-γ secretion increasing with antigen density and peaking against PC-3 PIP, which has higher PSMA expression than our PC-3 PSMA^hi cell line (Table S2). These studies collectively suggest that, in this system, there were limited improvements in CAR T cell functionality following AM of the hPSMA scFv.

Engineering hPSMA-CAR T cells with mbIL12 rescues potent CAR T cell functionality

Previous work has demonstrated the impact of IFN-γ signaling on CAR T cell responses in solid tumors, including driving their direct cytolytic activity.^21,22,23 Since we consistently showed dampened IFN-γ production by hPSMA-CAR T cells compared with J591-CAR T cells, associated with reduced tumor cell killing and T cell proliferation, we hypothesized that addition of IL-12 signaling, which is known to enhance IFN-γ secretion, may improve these activities in our hPSMA-CAR T cells. Our group recently reported a novel membrane-bound form of IL-12 (mbIL12), which was safe and robustly improved CAR T cell functionality in solid tumors.²³ Here, we manufactured hPSMA-CAR T cells with mbIL12 to determine whether this engineering strategy could rescue potent and specific PSMA-directed CAR T cell functionality (Figure 4A). We transduced human T cells with both CARs (with CD19 control, ET260-1, and ET260-1[58] scFvs) and mbIL12 (Figure 4B). Our short-term killing assay showed comparable tumor cell killing by our hPSMA-CAR T cells with mbIL12, with no increases in non-specific targeting of PC-3 WT cells (Figure 4C). 4-1BB expression did not show any striking differences between the hPSMA-CARs alone and their mbIL12-expressing counterparts (Figure 4D). When we investigated CD25, another marker of activation, we observed higher expression in the mbIL12-containing CAR T cells compared with ET260-1-CAR T cells alone against low- and high-PSMA-expressing tumor cells (Figure 4E). Similarly, IFN-γ secretion improved with the addition of mbIL12 to hPSMA-CAR T cells, reaching levels comparable to J591-CAR T cells (Figure 4F). We also performed the 8-day LTK assay with the AM hPSMA-CAR with mbIL12 against WT and PC-3 PIP cells, which again showed similar overall activity between ET260-1 and ET260-1 (58), albeit improved fold expansion and IFN-γ production by the parental CAR engineered with mbIL12 (Figures S6A–S6C). We therefore moved forward with interrogating the hPSMA-CAR/mbIL12 T cells incorporating the parental ET260-1 scFv.

Figure 4.

Engineering hPSMA-CARs with mbIL12 rescues CAR T cell functionality

(A) Schema of hPSMA-CAR variants with mbIL12 expression. (B) Flow cytometric analysis of ET260-1 CAR and ET260-1(58) CAR with or without mbIL12, as detected by extracellular Fc and intracellular IL-12. (C) Tumor cell killing of hPSMA-CARs at an E:T of 1:4 over 72 h. Tumor cell killing was calculated relative to the CD19-CAR/mbIL12 T cell condition. (D) Expression of 4-1BB over 72 h on hPSMA-CARs. (E) Expression of CD25 on hPSMA-CARs after 72 h. (F) Secretion of IFN-γ by hPSMA-CARs with mbIL12 after 72 h as determined by ELISA. (G) Tumor cell killing of hPSMA-CARs with mbIL12 and J591 against PC-3 PIP at an E:T of 1:10 over 5 days at varying concentrations of anti-IFN-γR1 blocking antibody and isotype control. Data are presented as mean values ± SEM. p values indicate differences between ET260-1-dCH2(28tm)BBz and ET260-1-dCH2(28tm)BBz/mbIL12 and between ET260-1-dCH2(28tm)BBz/mbIL12 and J591-dCH2(4tm)BBz using a two-tailed Student’s t test.

The improved activity in the mbIL12-engineered CAR T cells to comparable or greater levels than J591-CAR T cells supported the hypothesis that this activity was, at least in part, IFN-γ-mediated. To test this further, we blocked IFN-γ signaling in our co-culture system and found that tumor cell killing by J591-CAR T cells was inhibited in a dose-dependent manner with IFN-γR blockade (Figure 4G). While tumor cell killing by hPSMA-CAR/mbIL12 T cells was also inhibited by IFN-γR blockade, the effect was less pronounced compared with J591-CAR T cells. These results indicate that J591-CAR T cell anti-tumor activity is dependent on IFN-γ signaling and that the benefits of mbIL12 on hPSMA-CAR T cells are only partially dependent on IFN-γ signaling, highlighting IFN-γ-independent functionality of IL-12 signaling in our system.

The functional benefits of mbIL12 were further demonstrated in LTK assays (8 days), with ET260-1-CAR/mbIL12 T cells showing improved tumor cell killing compared with the parental scFv-based CAR T cells (Figure 5A). Importantly, T cell expansion and IFN-γ secretion were greater with mbIL12 compared with the parental scFv-based CAR T cells and J591-CAR T cells (Figures 5B and 5C). Additionally, although our ET260-1-CAR/mbIL12 T cells showed improved killing against the low expresser, activity did not increase against PC-3 WT cells (Figure 5A). Furthermore, ET260-1-CAR/mbIL12 T cells showed improved expansion compared with the parental scFv against the low expresser, but expansion did not reach the levels seen in J591(Figure 5B). We repeated the LTK assay with additional PSMA-positive and PSMA-negative tumor cells and again demonstrated optimal specificity for PSMA-expressing tumors using ET260-1-CAR/mbIL12 T cells (Figures S7A–S7C).

Figure 5.

mbIL12-engineered hPSMA-CAR T cells demonstrate improved activity in a recursive tumor cell assay in vitro

(A) Tumor cell killing of hPSMA-CARs over 8 days when co-cultured with PSMA-expressing cell lines at an E:T of 1:20. Tumor cell killing was calculated relative to the CD19-CAR/mbIL12 T cell condition. (B) Fold expansion of hPSMA-CARs compared to J591-CAR after 8-day co-culture. (C) Production of IFN-γ at 8 days as determined by ELISA. (D) Schema of rechallenge with PSMA-CARs co-cultured with PC-3 PSMA^lo, PC-3 PSMA^hi, or PC-3 PIP and rechallenged with tumor cells every 3 days. (E) PC-3 PSMA^lo (left), PC-3 PSMA^hi (center), or PC-3 PIP (right) cell counts were quantified by flow cytometry. (F) Fold expansion of T cells after each rechallenge was quantified by flow cytometry. (G) Production of IFN-γ by T cells following each rechallenge as determined by ELISA. Data are presented as mean values ±SEM. p values indicate differences between ET260-1-dCH2(28tm)BBz and ET260-1-dCH2(28tm)BBz/mbIL12 and between ET260-1-dCH2(28tm)BBz/mbIL12 and J591-dCH2(4tm)BBz using a two-tailed Student’s t test.

To stress test our hPSMA-CAR T cells further, we recursively challenged CAR T cells with three rechallenges of PSMA tumor cells over the course of 2 weeks (Figure 5D). Supporting our previous in vitro results, ET260-1-CAR/mbIL12 T cells showed superior tumor cell killing compared with T cells expressing ET260-1-CAR alone and were comparable to J591-CAR T cells (Figure 5E). Additionally, ET260-1-CAR/mbIL12 T cells exhibited higher T cell expansion against PSMA-high expressers and PC-3 PIP compared with both J591-CAR and ET260-1-CAR T cells alone (Figure 5F). Lastly, ET260-1-CAR/mbIL12 T cells secreted higher and more sustained IFN-γ levels compared with J591-CAR or ET260-1-CAR T cells alone (Figure 5G). ET260-1-CAR/mbIL12 T cells were less sensitive to PSMA-low expressing tumor cells. In contrast, J591-CAR T cells showed higher T cell expansion and IFN-γ secretion against PSMA-low expressers. Evaluation of T cell phenotypes (expression of CD62L and CCR7) showed little differences between the CAR T cells throughout the tumor rechallenge assay (Figure S8). Collectively, our data support mbIL12 engineering as a strategy for improving hPSMA-CAR T cell anti-tumor activity against PSMA tumor cells while maintaining their selectivity and specificity.

hPSMA-CAR/mbIL12 T cells demonstrate a dampened cytokine profile in the presence of monocyte-derived suppressive M1 and M2 macrophages

Clinically evaluated PSMA-targeting strategies to date have been marred by severe adverse events attributed to cytokine-release syndrome and macrophage-activation syndrome. To evaluate the possibility of our hPSMA-CAR T cells potentiating similar toxicities, CAR T cells were cocultured in the presence of M1-polarized or M2-polarized macrophages using our previously described in vitro system.²⁴ Macrophages were differentiated and polarized into M1-like (CD80^high, CD163⁻, CD206^low) or M2-like (CD80^low, CD163⁺, CD206^high) macrophages (Figures 6A and S9). CAR T cells, macrophages, and our PC-3 parental or PSMA^Hi tumor cells were cocultured at an effector:macrophage:tumor cell ratio of 1:2:4 (Figure 6B). As measures of activity, we quantified activation of CAR T cells via CD25 after 72 h and secretion of IL-6 and IFN-γ at 24 h against PC-3 WT (Figures 6C–6E) and PC-3 PSMA^Hi (Figures 6F–6H). While there was no significant difference in IL-6 secretion between our hPSMA-CAR and hPSMA CAR/mbIL12 T cells against PC-3 WT in the tumor cell-only and macrophage conditions, J591-CAR T cells showed significantly higher IL-6 secretion (Figure 6D). We again observed significant increases in CD25 expression with the hPSMA-CAR/mbIL12 T cells compared with the hPSMA-CAR T cells alone across the different macrophage conditions (Figure 6F). Importantly, although there was increased activation with the hPSMA-CAR/mbIL12 T cells in the presence of M1 or M2 macrophages, there was no increases in IL-6 secretion (Figure 6G). In contrast, the J591-CAR T cells consistently showed higher CD25 expression and IL-6 secretion in the tumor cell-only and M2 macrophage conditions compared with hPSMA-CAR T cells (Figure 6G). Our hPSMA-CAR/mbIL12 T cells secreted higher IFN-γ in the tumor cell-only condition, while the J591-CAR T cells secreted higher IFN-γ in the presence of M1 and M2 macrophages (Figure 6H). These data indicate that while hPSMA-CAR/mbIL12 T cells improve activation and IFN-γ production compared with hPSMA-CAR T cells alone in the presence of M1 or M2 macrophages, the presence of mbIL12 does not increase IL-6 production, suggesting a favorable safety profile.

Figure 6.

mbIL12-engineered hPSMA-CAR T cells demonstrate improved activity and cytokine profile in the presence of macrophages

(A) Confirmation of differentiation of M1-like and M2-like macrophages by detection of CD80 (left) and CD163 (right). To differentiate M1-like macrophages, cells were cultured with GM-CSF, IFN-γ, LPS, and IL-6. To differentiate M2-like macrophages, cells were cultured with M-CSF, IL-4, IL-13, and IL-6. All cytokines and LPS were used at 20 ng/mL. (B) Schema of the T cell suppression assay (TSA), cells were cultured at an effector:macrophage:tumor cell (E:M:T) of 1:2:4, supernatant was collected at 24 h, and cells were processed for flow cytometry at 72 h. (C and F) Expression of CD25 on T cells cocultured with PC-3 WT (C) or PC-3-PSMA^hi (F) at 72 h. (D and G) IL-6 levels in supernatant by ELISA following coculture with PC-3 WT (D) or PC-3-PSMA^hi (G) for 24 h. (E and H) IFN-γ levels in supernatant by ELISA following coculture with PC-3 WT (E) or PC-3-PSMA^hi (h) for 24 h.

hPSMA-CAR/mbIL12 T cells demonstrate robust efficacy in a clinically relevant bone metastatic prostate cancer model

We next evaluated the therapeutic efficacy of hPSMA-CAR and hPSMA-CAR/mbIL12 T cells using an intratibial (i.ti.) C4-2 prostate cancer bone metastasis model (Figure 7A). C4-2-ffluc was engrafted i.ti., and tumor burden was monitored using bioluminescence flux imaging. Mice were treated with a single dose of CAR T cells (CD19-CAR and CD19-CAR/mbIL12 n = 5; ET260-1-CAR n = 6; ET260-1-CAR/mbIL12 n = 6; or J591-CAR n = 6) by retro-orbital (r.o.) intravenous (i.v.) injection. While mice treated with ET260-1-CAR T cells showed heterogeneous responses (50% complete response rate) in this model, ET260-1-CAR/mbIL12 and J591-CAR T cells showed robust and durable anti-tumor responses (Figure 7B). While not statistically significant, mice treated with J591-CAR T cells showed 5/6 complete responses, while ET260-1-CAR/mbIL12 T cells showed 100% complete responses (Figures 7C and 7D). We observed a greater persistence of T cells in the peripheral blood of mice treated with ET260-1/mbIL12 T cells compared with J591-CAR and ET260-1-CAR T cells alone (Figure 7E). We validated the improved therapeutic efficacy of ET260-1-CAR/mbIL12 T cells using LAPC-9, a patient-derived xenograft from a patient with bone metastatic prostate cancer that endogenously expresses PSMA (Figure S10). Mice treated with ET260-1 CAR/mbIL12 had improved survival (50% complete response) compared with ET260-1 CAR (20% complete response) and showed no changes in body weight following treatment (Figures S9B and S9C). These results provide strong in vivo evidence that our ET260-1/mbIL12 CAR T cells promote robust and durable anti-tumor responses against PSMA+ prostate cancer bone metastasis.

Figure 7.

hPSMA-CAR T cells engineered with mbIL12 demonstrate robust efficacy in a clinically relevant bone metastatic prostate cancer model

(A) Schema of the intratibial (i.ti.) model of C4-2_ffluc. Tumor cells (150k) were engrafted intratibially (D0), imaged and sorted into treatment groups on D20, and treated on D21 with 0.5e6 CAR T cells intravenously. Tumor burden was quantified using bioluminescent imaging (BLI). (B) Average tumor flux of mice. (C) Kaplan-Meier survival for CD19, CD19/mbIL12, ET260-1, ET260-1/mbIL12, and J591-CAR T cells. (D) Quantification of flux from individual mice in each treatment group (CD19 n = 5, CD19/mbIL12 n = 5, ET260-1 n = 6, ET260-1/mbIL12 n = 6, and J591 n = 6). (E) Quantification of human CD45+ cells in the peripheral blood of mice treated with T cells at day 7 post-treatment. Data are presented as mean values ± SEM. p values indicate differences between ET260-1-dCH2(28tm)BBz/mbIL12 and J591-dCH2(4tm)BBz using a two-tailed Student’s t test.

Discussion

Herein, we systematically optimized hPSMA-CAR T cells, which are anticipated to reduce anti-CAR immunogenicity and improve potent and selective targeting of PSMA. We first compared two human scFvs (ET260-1 and ET260-2) with the humanized J591 scFv in CAR T cells. Our results highlight ET260-1 as the lead scFv over ET260-2 but with suboptimal activity compared with J591, which led us to further optimize the CAR extracellular spacer, transmembrane domains, as well as affinity-mature the scFv. The combination of dCH2 and CD28tm in BBz-containing CAR T cells improved tumor cell killing, T cell activation, and IFN-γ secretion. Although we improved upon the original CAR, we continued to observe suboptimal T cell expansion and cytokine secretion compared with J591-based CAR T cells, even following AM of the human scFv to improve the kinetics of on- and off-rates. These results demonstrate that higher scFv affinity of CARs does not necessarily result in greater functionality but may, in some instances, increase the risk of non-specific targeting.

The proposed safety benefits of our human scFv are 2-fold: 1) a human CAR avoids the potential for development of neutralizing human anti-mouse antibodies (HAMA). HAMA has resulted in reduced persistence of CAR T cells and, in some cases, led to severe allergic reactions.^25,26,27,28 We have shown our ET260-1 scFv to be selective and specific to higher expression of PSMA, which we expect to mitigate potential on-target, off-tumor toxicities associated with lower expression of PSMA in normal tissues, including the salivary gland²⁹ and in the brain.^11,12

Studies using J591 in combination with drug conjugates, toxic payloads, or as the antigen-binding domain of CAR T cells have demonstrated activity but have also been associated with unwanted toxicities.^6,30 These studies have also highlighted the tight binding of J591 to PSMA,^31,32,33,34 which our binding kinetic studies also confirmed. We postulate that J591’s activity against our PC-3 parental cell line, which has low but detectable levels of PSMA, leads to potent activation and is a likely explanation for some of the observed toxicities in PSMA-CAR T cell clinical investigations. Even when coupled with mbIL12 engineering, our hPSMA-CAR T cells demonstrated potent targeting of PSMA+ tumor cells without increasing activation or targeting of PC-3 cells with very low PSMA expression. The safety of our hPSMA-CAR T cells was further supported by lower IL-6 production in the presence of macrophages, even with mbIL12 engineering, compared with J591-based CAR T cells. The lower affinity of our hPSMA-CAR T cells showed little to no activity against negligible PSMA-expressing PC-3 WT cells, whereas J591-CAR T cells were consistently active against these cells. This result is a potential advantage of our ET260-1-CAR compared with the J591-CAR in terms of selectivity in targeting higher levels of PSMA in cancer cells.

Increasing evidence points to the essential role of IFN-γ signaling for CAR T cell activation and early killing potential against solid tumors.^21,22 Our rationale was that by introducing an immunostimulatory cytokine to our CAR T cells, we could bridge this gap in functional activity while retaining the safety and specificity of our human ET260-1 scFv-based CAR T cells. Based on recent published work from our laboratory, we combined our hPSMA-CARs with an mbIL12 cytokine.²³ Indeed, mbIL12-engineered hPSMA-CAR T cells completely restored functional activity to a similar and even greater extent than J591-CAR T cells. This activity was not entirely IFN-γ-dependent, since our hPSMA-CARs still retained activity even following IFN-γR blockade. The activity of the hPSMA-CAR indicates that there is a mechanism of action intrinsic to our hPSMA construct that synergizes with mbIL12 for potent tumor cell killing in the absence of IFN-γ signaling. This also suggests that there are other roles of IL-12 signaling that may benefit our hPSMA-CARs. Future studies are warranted to understand the mechanisms regulating this IFN-γ-independent signaling by mbIL12, which may include ICAM1, FasL, or other adhesion-related molecules. Importantly, this may be accomplished through similar beneficial activation using other cytokine modulators, including IL-7, IL-15, and IL-21.

We recognize that our hPSMA-CAR is limited in its single-antigen targeting ability, as antigen heterogeneity and low antigen density may limit durable response rates. As exemplified by other PSMA-targeting strategies, which have shown transient responses,^6,35,36 there is likely a need to recruit the endogenous immune system to mount a robust anti-tumor response. By incorporating mbIL12, we aim to improve the therapeutic efficacy of our hPSMA-CAR T cells while also eliciting antigen spread and endogenous anti-tumor immunity. While the current models limited our ability to investigate these phenomena, future studies using syngeneic hPSMA-knock-in (KI) mice will be a high priority.

In summary, we have optimized an hPSMA-CAR T cell therapy, which was further engineered with mbIL12 signaling, representing an innovative approach to driving specific and potent targeting of PSMA+ cancers. These studies warrant further translational work and potential clinical testing.

Materials and methods

Cell lines

Human metastatic prostate cancer cell line PC-3 (ATCC CRL-1435) was cultured in RPMI-1640 (Corning) containing 10% fetal bovine serum (FBS, Hyclone), and 1× antibiotic-antimycotic (AA, Gibco) (complete RPMI). PC-3 PIP (a kind gift from Dr. Katharina Leuckerath, UCLA) was cultured in complete RPMI. C4-2 (CRL-3314) was purchased from ATCC and cultured in Dulbecco’s modified Eagle’s medium (DMEM):F12 (Life Technologies) with 10% FBS. C4-2 cells were transduced with lentivirus containing firefly luciferase to generate C4-2_ffluc. CWR22Rv1 (CRL-2505) (a kind gift from Dr. Robert Reiter, UCLA) and LNCaP were cultured in complete RPMI.

The human fibrosarcoma cell line HT1080 (ATCC CCL-12) and the human embryonic kidney cell line 293T (ATCC CRL-3216) were cultured in DMEM (Life Technologies) containing 10% FBS, 1× AA, 25 mM HEPES (Irvine Scientific), and 2 mM L-glutamine (Corning (complete DMEM).

The human prostate cancer-derived xenograft LAPC-9 (a kind gift from Dr. Robert Reiter, UCLA) was serially passaged in male NOD.Cg-Prkdcscid IL2rgtm1Wjl/SzJ (NSG) mice, and single-cell suspensions were prepared using the Miltenyi Biotec human tumor dissociation kit (cat: 130-095-929) per the manufacturer’s protocol. LAPC-9 cells were transduced with lentivirus to express enhanced green fluorescent protein (eGFP)/firefly luciferase (LAPC-9-eGFP-ffLuc).

hPSMA antibody panning

Jurkat cells engineered to overexpress PSMA were used for antibody panning against Eureka Therapeutics’ E-ALPHA phage library, a collection of human scFv antibody phage display libraries containing over 10 × 10¹⁰ unique clones. The phage libraries included naive and semi-synthetic human scFv antibodies. The E-ALPHA library was used to screen for the selection of human antibody constructs (e.g., scFv) specific for PSMA. In a first step, the library was negatively screened using Jurkat parental cells to eliminate clones that may recognize cell surface molecules other than PSMA. The remaining clones were screened in three rounds using Jurkat-PSMA engineered cells to select PSMA+ clones. The positive clones were confirmed by flow cytometry for specifically binding to hPSMA on Jurkat-PSMA cells and not Jurkat parental cells. Finally, unique selected clones were further confirmed by flow cytometry using the PSMA-positive cancer cell lines, LnCAP and Jurkat-PSMA cells and, as negative controls, PSMA-negative cell lines Jurkat and PC-3.

DNA constructs and lentivirus production

PC-3 tumor cells were engineered to express PSMA by transduction with MSCV retrovirus carrying the hPSMA gene (accession #:NC_000011.10). PSMA⁺ cells underwent fluorescence-activated cell sorting (FACS) using the BD FACSAria cell sorter. Single-cell clones were generated and selected based on antigen density for low, mid, and high PSMA expression.

The ET260 scFv sequences were generated and provided by Eureka Therapeutics and described above. The J591 scFv was based on the PSMA CAR in patent US20110268656A1. The two distinct ET260 scFv chain variants with differing CDR regions were cloned into a CH2-deleted version (ΔCH2, 129-amino acid middle-length) of the IgG4 Fc spacer, CD4 transmembrane, 4-1BB intracellular costimulatory domain, and CD3ζ cytolytic signaling domain CAR construct. The 4-1BB and CD3ζ cytolytic domains were consistent throughout all future construct iterations. A T2A ribosomal skip sequence separated the CAR construct from the truncated CD19 (CD19t) tag. All CAR constructs were cloned into the epHIV7 lentiviral vector under the EF1α promoter and a GMCSFRα signal sequence. The J591 CAR was cloned into the ΔCH2 linker and CD4tm construct.

For transmembrane and linker combination comparison, a short hinge linker (HL, 22-amino acid length) was combined with a CD28 transmembrane domain (CD28tm), the ΔCH2 linker was combined with the CD28tm, and lastly a CD8 hinge (C8h, 49 amino acids) was combined with a CD8tm. All constructs contained the 4-1BB intracellular costimulatory domain and the CD3ζ signaling domain.

AM and clone selection

An AM library of ET260-1 scFv clones was constructed by error-prone PCR using a random mutagenesis kit (GeneMorph II, Agilent) according to the manufacturer’s instructions. Briefly, 100 ng of ET260-1 phagemids were used as a template in a 20-cycle PCR reaction performed using the manufacturer’s guidelines to attain the desired mutation rate. The PCR products were purified and re-converted into phagemids. These phagemids were transduced into TG1 competent cells by electroporation; the phage library was generated by precipitation of TG1 cell culture supernatants containing phage. TG1 cells are an electrocompetent E. coli amber suppressor strain (supE) suitable for protein expression and preparation of antibody or peptide phage display libraries. AMET260-1 was obtained from a phage library constructed with a medium mutation rate and an effective size of 7 × 10⁶ unique clones.

Plate panning against the AMET260-1 phage library was performed using His-tagged PSMA ECD (Lys44-Ala750) protein-coated Corning 96-well clear, flat-bottom, polystyrene high-binding plates at a concentration of 2 μg/mL in a volume of 100 μL/well at room temperature overnight. An irrelevant His-tagged protein (MCT4) was used as a negative control. After blocking with 3% non-fat milk, three rounds of selection were carried out under increasingly stringent conditions by increasing the incubation and washing times. The washing buffer contained 5 μg/mL of free PSMA ECD protein to compete with plate-coated antigen.

AM scFvs were cloned into the ΔCH2/CD28tm vector. The mbIL12 (mbIL12) construct was generated using the p35 and p40 genes (p35, NC_000003.12; p40, NC_000005.10) separated by a G4S spacer and linked to the CD28 transmembrane domain, as described previously.²³

Lentivirus was generated as previously described.³⁷ Lentiviral titers, as determined by CD19t, EGFRt, or IL-12 expression, were quantified using HT1080 cells.

SPR

The binding affinity of the scFv clones to the extracellular domain of human PMSA (Lys44-Ala750) was measured by SPR using a BiaCore X-100 instrument (Cytiva). The binding parameters between the anti-PMSA scFvs and PSMA protein were measured using a Protein G sensor chip (Cytiva) according to the manufacturer’s instructions for multi-cycle kinetics (MCK) analysis. Briefly, the recombinant N-terminal Fc-tagged PSMA protein (hFc-PMSA) was loaded onto the protein G Sensor Chip at 10 μg/mL. The scFv proteins were passed over the sensor chip at varying concentrations of 10, 5, 2.5, 1.25, 0.625, 0.3125, 0 μg/mL in consecutive runs. The data was analyzed using a 1:1 binding site model and BiaCoreTM X-100 Evaluation Software. The binding parameters—association (on-rate) constant ka, dissociation (off-rate) constant kd, and equilibrium dissociation constant KD—were calculated.

T cell isolation, lentiviral transduction, and ex vivo expansion

Leukapheresis products were obtained from consented research participants (healthy donors) under protocols approved by the City of Hope (COH) Internal Review Board (IRB). On the day of leukapheresis, peripheral blood mononuclear cells (PBMCs) were isolated using SepMate PBMC Isolation Tubes (StemCell Technologies), followed by multiple washes in PBS with 2% FBS. To obtain depleted PBMCs (dPBMC), cells were depleted of CD25 and CD14 using CD25 and CD14 beads (Miltenyi Biotec) and sorted using the autoMACS Pro Separator. Depletion was confirmed by flow cytometry, and cells were frozen for future use in CryoStor CS5 cryopreservation media (BioLife Solutions).

T cell activation and transduction were performed as previously described.³⁷ Where indicated, cells underwent a second lentiviral transduction 24 h after the first transduction. Cells were replenished with fresh cytokine X-VIVO mixture every 2–3 days. Cells were ex vivo manufactured, enriched for CD19t or EGFRt, and cryopreserved as previously described. CAR T cells were characterized pre- and post-enrichment for purity and phenotype by flow cytometry.

For in vitro assays, cells were thawed 1 day prior and rested overnight. Cells were resuspended in 1× cytokine X-VIVO at 1 million cells per mL. For in vivo studies, cells were thawed on the day of treatment and resuspended at the desired treatment concentration in 1× PBS.

Flow cytometry

For flow cytometric analysis, cells were processed in FACS staining solution (FSS) (Hank’s balanced salt solution without Ca²⁺, Mg2+, or phenol red (HBSS−/−, Life Technologies) containing 2% FBS and 0.5 g sodium azide in 500 mL). Cells were washed once, followed by incubation with antibodies for 30 min at 4°C in the dark. Antibodies were conjugated with Brilliant Violet 510 (BV510), Brilliant Violet 570 (BV570), SuperBright 600 (SB600), Brilliant Violet 650 (BV650), fluorescein isothiocyanate (FITC), phycoerythrin (PE), peridinin chlorophyll protein complex (PerCP), PE-CF594, PE-Cy5, PerCP-Cy5.5, PerCP-eFlour710, PECy7, allophycocyanin (APC), or APC-Cy7 (or APC-eFluor780), and Alexa Fluor 700 (AF700).

Antibodies against human antigens used include: CD3 (BD Biosciences, Cat: 563109, Clone: SK7), CD4 (BD BioLegend, Cat: 300534, Clone: RPA-T4), CD8 (Invitrogen, Cat: 56-008-742, Clone: SK1), CD19 (BD Biosciences, Cat: 557835, Clone: SJ25C1), CD45 (BD Biosciences, Cat: 563204, Clone: HI30), CD69 (BD Biosciences, Cat: 341652, Clone: L78), CD137 (Invitrogen, Cat: 63-137-942, Clone: 4B4-1), Tim-3 (BioLegend, Cat: 345028, Clone: F38-2E2), IL-12 (BD Bioscience, Cat: 554575, Clone: C11.5), Lag-3 (BD Biosciences, Cat: 565718, Clone: T47-530), PD-L1 (Invitrogen, Cat: 15-598-342, Clone: MIH1), CD25 (Invitrogen, Cat: 46-0259-42, Clone: BC96), PSMA (BioLegend, Cat: 342508, Clone: LNI-17), PD-1 (eBiosciences, Cat: 47-2799-42, Clone: J105), CD62L (BioLegend, Cat: 304832, Clone: DREG-56), CD45RO (Biolegend, Cat: 304246, Clone: UCHL1), and CCR7 (Biolegend, Cat: 353212, Clone: G043H7).

Cell viability was determined using 4′, 6-diamidino-2-phenylindole (DAPI, Sigma, Cat: D8417). Unless otherwise stated, antibodies were used at a dilution of 1:100. Flow cytometry was performed on a MACSQuant Analyzer 10 or MACSQuant Analyzer 16 (Miltenyi Biotec), and the data were analyzed with FlowJo software (v10.8.1, TreeStar).

For intracellular flow cytometry, cells were processed using the BD Cytofix/Cytoperm fixation permeabilization kit (Cat: 554714) according to the manufacturer’s protocol.

PSMA quantification was done using the Bangs Laboratories Quantum Simply Cellular mouse IgG1 kit, per the manufacturer’s instructions.

In vitro macrophage differentiation

Primary human M1-like and M2-like macrophages were differentiated as previously described.²⁴ Briefly, frozen human monocytes were thawed and cultured in cytokine-containing RPMI +10% FBS and differentiated for 7–10 days. To differentiate M1-like macrophages, cells were cultured with GM-CSF (BioLegend, 572903). The media was changed once after 3–5 days to media containing GM-CSF, IFN-γ (BioLegend, 570202), LPS (Sigma-Aldrich, L3012-5MG), and IL-6 (BioLegend, 570804). To differentiate M2-like macrophages, cells were cultured with M-CSF (BioLegend, Cat: 574804). The media was changed once after 3–5 days to media containing M-CSF, IL-4 (BioLegend, 574004), IL-13 (BioLegend, 571102), and IL-6. All cytokines and LPS were used at 20 ng/mL. After differentiation, macrophages were lifted using PBS-EDTA, and phenotype was assessed by flow cytometry to confirm successful polarization. Cells were counted and used for further studies.

In vitro cell killing

For tumor cell killing assays, T cells and tumor cells were co-cultured at indicated E:T ratios in X-VIVO with 10% FBS in 96-well plates. For co-cultures set up for 72 h processing time points, an E:T of 1:4 was used, and for 8-day cocultures, an E:T of 1:20 was used. Cells were processed for flow cytometry at the indicated time points. CAR T cell killing was calculated by comparing cell counts relative to counts of target cells co-cultured with untransduced T cells (UTD) or non-targeting CARs. Tumor cells were gated as DAPI-negative, single cells, CD45-negative.

For recursive tumor cell challenge assays, T cells were initially co-cultured with PSMA-expressing cell lines at an E:T of 1:4 for PC-3 PSMA^hi/lo and 1:5 for PC-3 PIP. CAR T cells co-cultured with PC-3 PSMA^hi/lo were rechallenged every 3 days up to 3 times with twenty-thousand tumor cells. CAR T cells co-cultured with PC-3 PIP were rechallenged every 3 days, first with 75k tumor cells, then 100k tumor cells, and the final rechallenge with 150k tumors cells. This was due to the high proliferation of the CAR T cells and to ensure an adequate rechallenge. The remaining viable cells were quantified by flow cytometry to determine tumor cell and T cell counts prior to every rechallenge and 3 days after the last rechallenge.

In vitro T cell suppression assay

CAR T cells, macrophages, and target tumor cells were co-cultured in X-Vivo +10% FBS in the absence of exogenous cytokines in 96-well plates. Cells were plated at an effector:macrophage:target ratio of 1:2:4 to model prostate cancer with PC-3 WT or PC-3-PSMA^hi cells. Supernatant was collected after 24 and 72 h for ELISA, and cells were trypsinized and collected for flow cytometry after 72 h. Tumor cell killing and T cell activation were evaluated after 72 h by flow cytometry.

ELISA cytokine assays

Supernatant collected from tumor cell killing assays were collected at the indicated time points and stored at −20°C for future use. Supernatants were used to analyze secretion of IFN-γ with the ELISA Ready-SET-GO! kit (Cat: 88-7316-88) according to the manufacturer’s protocol. Plates were read at 450 nm and 650 nm using a Cytation5 imaging reader with Gen5 microplate software V3.08 (BioTek).

In vivo animal studies

All animal experiments were performed under protocols approved by the COH Institutional Animal Care and Use Committee. All mice were co-housed in a maximum barrier, pathogen- and opportunist-free animal facility. For i.ti. tumor studies, C4-2 (0.15 × 10⁶) and LAPC-9 (0.75 × 10⁵) were prepared in a final volume of 30 μL HBSS−/− and injected into the tibia. Tumor growth was monitored at 1–2× per week via non-invasive bioluminescence imaging (Xenogen, LagoX), and flux signals were analyzed with Aura software (Spectral Instruments Imaging). To image mice, 150 μL of D-luciferin potassium salt (PerkinElmer) in PBS at 4.29 mg/mouse was injected intraperitoneally (i.p.). Mice were treated with T cells (1 × 10⁶ or 0.5 × 10⁶) in 100 μL PBS r.o. i.v. on day 15 or day 21. Humane endpoints were used in determining survival. At pre-determined time points or at moribund status, mice were euthanized. Mice were euthanized by CO2 and cervical dislocation upon signs of distress, such as labored or difficulty breathing, apparent weight loss, impaired mobility, or evidence of being moribund.

Statistical analysis and reproducibility

Data are presented as mean ± standard error of the mean (SEM), unless otherwise stated. Statistical comparisons between groups were performed using the unpaired, two-tailed Student’s t test to calculate the p value, unless otherwise stated.

Data and code availability

All data associated with this study are present in the paper or the supplemental information. Materials are available from S.J.P. under a material transfer agreement.

Acknowledgments

We thank the staff members of the following cores at the Beckman Research Institute at City of Hope Comprehensive Cancer Center: Animal Facility and Small Animal Imaging for their excellent technical assistance. Work performed in the Small Animal Imaging Core was supported by the National Cancer Institute of the National Institutes of Health under grant no. P30CA033572. Research reported in this publication was supported by the Eureka Therapeutics Sponsored Research Agreement (PI: Priceman), the Fiterman Family Foundation fund, and the Doug and Rhonda Collier Foundation fund. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author contributions

S.J.P., along with L.S.L., provided the conception and construction of the study and wrote the manuscript. L.S.L., Z.C., Y.Y., J.P.M., Z.Y., K.Z., J.Y., W-C.C., and S.J.P. provided the design of experimental procedures, data analysis, and/or interpretation. L.S.L., Z.C., Y.Y., J.P.M., Z.Y., K.Z., J.Y., and W-C.C. performed experimental procedures. Z.C., S.J.F., and V.W.C. assisted in writing/editing the manuscript. S.J.P. supervised the study. All authors reviewed the manuscript.

Declaration of interests

S.J.P. is a scientific advisor to and receives royalties from Imugene Ltd, Adicet Bio, Port Therapeutics, and Celularity. L.S.L., S.J.P., and Eureka Therapeutics employees are listed as co-inventors on a patent related to the development of human constructs targeting PSMA, which is co-owned by City of Hope and Eureka Therapeutics. S.J.P. and J.P.M. are listed as co-inventors on a patent on the development of membrane-bound IL-12 engineered CAR T cells for the treatment of cancer, which is owned by City of Hope.