Host Interactions with Engineered T-cell Micropharmacies

Adoptive T-cell therapies can localize to and kill antigen-positive cells in vivo (1–5). These characteristics have led to success against B cell–derived hematopoietic cancers. Chimeric antigen receptor (CAR) T-cell products have thus far been approved for B-cell lymphoma, B-cell leukemia, and multiple myeloma. Solid tumors, however, have remained refractory to such therapies and even most B-cell neoplasms ultimately relapse due to tumor heterogeneity and multiple immune escape mechanisms (6). We and many others are further exploring the therapeutic potential of adoptive T cells that can serve as delivery vehicles for drugs or biologic agents, also known as “targeted micropharmacies,” in addition to their intrinsic cytotoxic activity (7–14). More complex, and potentially more effective, engineered cells are in development, but the interactions of these highly modified cells with the host are poorly understood. A more complete characterization of targeted micropharmacies in immunocompetent hosts is critical to the safe clinical translation and further development of multifunctional adoptive T cells for the treatment of solid tumors.

Previously, our groups demonstrated that human T cells can be used in an enzyme–prodrug therapy approach to unmask highly toxic drugs selectively at tumors, thus improving antitumor activity (7). The Synthetic Enzyme-Armed KillER (SEAKER) cell platform was first demonstrated with CD19-targeted CAR T cells secreting the bacterial enzymes β-lactamase (β-Lac) or carboxypeptidase G2 (CPG2). We showed that these enzymes were delivered by CAR T cells to lymphomas in xenograft mouse models. Furthermore, systemic administration of prodrugs led to unmasking to form the corresponding parent drugs in the tumor microenvironment, leading to delayed tumor growth and improved survival, without systemic toxicity.

While compelling, these previous findings against human lymphoma in xenograft models may incompletely predict the activities of targeted micropharmacies in solid-tumor models, syngeneic systems, or when using T-cell receptor (TCR)–engineered T cells, rather than CAR T cells. Solid tumor microenvironments exploit multiple immune resistance mechanisms, such as antigen escape or downregulation, HLA downregulation, altered peptide presentation, upregulation of inhibitory immune checkpoint ligands, physical extracellular barriers and metabolic dysregulation to thwart T-cell persistence, cytotoxicity, and cytokine secretion (15–18). As a result, solid tumors present many challenges to the success of adoptive T-cell therapy. Therefore, understanding how T-cell drug-delivery vehicles localize and deliver therapeutic cargo to solid tumors and interact with the host immune system is of great interest.

T cells undergo homeostatic proliferation in empty T-cell niches, which complicates the study of their kinetics and biodistribution in xenograft mouse models (19, 20). How SEAKER cells perform in fully immunocompetent hosts has not been characterized. Moreover, the bacterial origin of the SEAKER enzymes used to date necessitates the evaluation of SEAKER enzyme functionality in immunocompetent hosts for further clinical development. In this study, we investigated (i) how the engineering of SEAKER cells would affect their pharmacokinetics and biodistribution in vivo; (ii) if a novel TCR T cell–based SEAKER platform (OT-1 SEAKER) could be used to allow us to target intracellular antigens; and (iii) whether the SEAKER platform could be therapeutically effective in immunocompetent hosts. We characterized these critical features and demonstrated that these cellular systems were functional, slowing tumor progression and resulting in significantly enhanced survival. Taken together, these results highlight the feasibility and effectiveness of adoptive T-cell micropharmacies with CAR T cells or TCR-engineered T cells in immunocompetent hosts and against solid tumors.

The active drug 5′-O-sulfamoyladenosine (AMS) was used in this study. Prodrugs AMS-Glu and Ceph-AMS were designed for the CPG2 and β-Lac systems, respectively. The prodrug AMS-Glu was designed with a glutamate masking group at the adenine 6-amino group of AMS. The prodrug Ceph-AMS was designed with a cephalothin masking group linked to the sulfamate nitrogen of AMS. AMS-Glu and Ceph-AMS were synthesized as described previously (7).

CPG2 and β-Lac proteins were produced and purified by GenScript as described previously (7). Constructs contain C-terminal hemagglutinin (HA) and His6 epitope tags and were purified by nickel affinity chromatography by GenScript.

To generate murine SEAKER cells, the genes encoding β-Lac and CPG2 were cloned into the SFG gamma retroviral vector [generously provided by R. Brentjens, Memorial Sloan Kettering Cancer Center (MSKCC)] alongside a P2A self-cleaving peptide and GFP or alongside the CAR constructs for the 1D3 anti-murine CD19 scFv or the 4H11 anti-MUC16 scFv with murine CD28 and CD3ζ genes, as described previously (4, 10, 21–23). Trackable β-Lac SEAKER cells were generated by cloning the β-Lac enzyme upstream of a Gaussia luciferase (gLuc) gene separated by a P2A site, and upstream of the mCherry gene via a T2A site (24).

Gibson assembly kits (New England Biolabs, E2611L) were used to generate all constructs according to manufacturer protocols. Retroviral producer cell lines were generated by transiently transfecting H29 cells with named constructs utilizing the ProFection Mammalian Transfection System according to the manufacturer's protocol (Promega, E1200). Viral supernatant from transfected H29 cells was collected on days 5, 6, and 7 posttransfection and filtered through a 45 μm filter (Corning, 431220) and incubated with Phoenix-ECO (pECO) to generate stable producer lines for murine cells (ATCC). Stable pECO producer cell lines were sorted for top 25% expressers, using a BD FACSAria 3 cell sorter. A list of constructs used to make pECO cell lines for transduction of primary murine T cells can be found in Supplementary Fig. S1.

All retroviral producer cell lines, human and murine tumor cells lines, and primary T cells were maintained in RPMI1640 medium (MSKCC Media Core Facility), at 37°C and 5% CO₂. Mouse T cells were maintained in RPMI1640 supplemented with sodium pyruvate and 2-mercaptoethanol (both from Life Technologies Corporation). All media was supplemented with 10% FBS, 2 mmol/L l-glutamine, 100 IU/mL penicillin, and 100 μg/mL streptomycin. All cells were routinely checked for mycoplasma via PCR and treated with Plasmocin (Invivogen) for 2 weeks, if positive. The EL-4 OVA (received 2018), B16F10 (received 2018), pECO (received 2018), and SET-2 (received 2017) cell lines were acquired from ATCC. B16F10 melanoma cells engineered to express the SIINFEKL peptide (B16-SIIN) were generously provided by A. Schietinger in 2020 (MSKCC). The H29 (received 2017), ID8 Muc16t (received 2018), EL-4 CD19 ffLuc (received 2018), and EL-4 Muc16t ffLuc (received 2018), pECO ffLuc/GFP (received 2018) cell lines were generously provided by the Brentjens lab. The EL-4 OVA and B16F10 cells lines were transduced to express ffLuc. Viral supernatant from pECO ffLuc/GFP cells was filtered through a 45 μm filter, incubated with EL-4 OVA cells, and spinoculated at 2,000 × g for 1 hour at 37°C for two consecutive days. Stable EL-4 OVA ffLuc⁺ cells were sorted for top 25% expressers, using a BD FACSAria 3 cell sorter. B16F10 cells were directly transduced with H29 supernatant from a transfection with a GFP/Luc construct, as described for pECO cells. Cell lines were frozen immediately after first passage and maintained in complete cell culture media supplemented with 5% DMSO at −180°C. Cell lines were maintained in culture for a maximum of 10 passages until new aliquots were thawed.

Nitrocefin assay was conducted as described previously (7). Spent culture media or tumor lysis samples in PBS, water, or phenol red-free RPMI1640 (Gibco, 11835030) were serially diluted (two-fold) and mixed 1:1 with 0.2 mmol/L nitrocefin (Abcam, ab145625). Samples were incubated 1 to 16 hours at room temperature and absorbance at 490 nm was read on a SpectraMax M2 plate reader (Molecular Devices). Data were analyzed with SoftMax Pro software 6.2.2. Recombinant β-Lac standard curves were run simultaneously to quantify β-Lac enzyme concentration.

A total of 1 × 10⁶/mL β-Lac GFP⁺ OT-1 were cultured for 48 hours in fresh media. Cells were spun at 500 × g for 5 minutes and supernatant fluid was collected. 2 × 10⁴ EL-4 OVA ffLuc⁺ cells were cultured in the presence or absence of SEAKER cell spent culture media. A total of 500 nmol/L ceph-AMS, 16 μg/mL tazobactam (Sigma-Aldrich, T2820–10MG), or both were added, and viability was assessed via luminescence.

A total of 1 × 10⁶/mL CPG-2 GFP+ OT-1 were cultured for 48 hours in fresh media. Cells were spun at 500 × g for 5 minutes and supernatant fluid was collected. A total of 1 × 10⁴ B16F10 cells were cultured in the presence or absence of SEAKER cell spent culture media. One hundred μmol/L AMS-glu was added and viability was assessed via CellTiter-Glo.

Samples with CPG2 enzyme were incubated with methotrexate (Accord Healthcare, 16729–277–30) at 450 μmol/L final concentration overnight. Absorbance at 320 nm was recorded on a NanoDrop spectrophotometer (Thermo Fisher Scientific), and decrease in UV signal signified substrate cleavage.

Mouse T cells were isolated from spleens of naïve C57/BL6 (Taconic), OT-1 C57BL/6-Tg(TcraTcrb)1100Mjb/J (Jackson), or Pmel C57BL/6-Tg(TcraTcrb)1100Mjb/J (Jackson) mice. OT-1 mice are transgenic for the TCR that recognizes chicken ovalbumin peptide SIINFEKL presented on murine cells. Pmel mice are transgenic for the TCR that recognizes the gp-100 antigen of melanocytes and melanoma. Splenocytes were put into a single-cell suspension by mechanical disruption using a 100 μm cell strainer. Splenocytes were collected and red blood cells were lysed using ACK lysis buffer to remove red blood cells (Thermo Fisher Scientific, A1049201). Splenocytes were activated overnight with CD3/CD28 Dynabeads (Thermo Fisher Scientific, 11452D) and 50 IU/mL human IL2 (MSKCC clinic). Activated T cells were transduced by centrifugation at 2,000 × g for 1 hour at 37°C with retroviral supernatant from transduced pECO cells on plates coated with RetroNectin (TakaraBio, T100B) for 2 consecutive days. Transduction efficiencies, as assessed by flow cytometry, ranged between 25% and 75%. A minimum transduction efficiency of 25% was used for all experiments, unless otherwise noted.

Cell samples were washed and stained in flow buffer (PBS, 1% FBS, 0.1% sodium azide) for 20 to 30 minutes. Antibodies used in the study are as follows: APC or FITC anti-mouse CD45.1 (clone: A20, Tonbo Biosciences), PE anti-myc (Clone: 71D10, Cell Signaling Technology), anti-Muc16t was generated and conjugated in house (9, 25), Alexa Fluor 700 anti-mouse CD45 (Clone: 30-F11, BioLegend), and PE anti-CD3 mouse (clone: UCHT1, BioLegend). Data were collected using a Guava easyCyte HT Flow Cytometer (Luminex), a LSRFortessa (BD Biosciences), or a Cytoflex LX (Beckman Coulter). Data were analyzed using Flowjo v10.4 software (Flowjo).

Sandwich ELISAs were performed on 96-well Immulon HBX plates (Thermo Fisher Scientific). A mouse IgG anti-β-Lac (clone: 3E11.G3, Thermo Fisher Scientific) was used to capture recombinant β-Lac and primary murine serum samples were used as primary antibody. A polyclonal anti-mouse IgG HRP antibody was used as a detection antibody (Novus Biologicals, NBP1–75130). Protein was detected using 3,3′,5,5′-tetramet hylbenzidine (TMB) substrate (Thermo Fisher Scientific, 34028) and H₂SO₄ acid quench, and then read on a SpectraMax M2 plate reader. Data were analyzed with SoftMax Pro software version 6.2.2.

T-cell and prodrug cytotoxicity assays with secreted enzymes were run for 24 to 48 hours until analysis using CellTiter-Glo (Promega, G7571) or luminescence using firefly (ffLuc) expressing tumor cells. For CellTiter-Glo based experiments, cells were analyzed in duplicate or triplicate in wells of a 96-well plate and equivalent volume of CellTiter-Glo reagent was added to each well. Following a 10-minute incubation at room temperature, samples were transferred to White 96-well Optiplates (Perkin Elmer) and luminescence was measured on a SpectraMax M2 plate reader. Data were analyzed with SoftMax Pro software version 6.2.2. The cytotoxicity of SEAKER cells was determined by luciferase-based assays. EL-4 OVA cells expressing ffLuc were used as target cells. Effector and tumor target cells were co-cultured in triplicate at the indicated effector-to-target (E:T) ratios using clear bottom, white 96-well assay plates (Corning 3903) with 2 × 10⁴ target cells in a total volume of 200 μL. Target cells alone were plated at the same cell density to determine maximum luciferase activity. Cells were cocultured for 4 to 18 hours, at which time d-luciferin substrate (Gold Biotech, LUCK-4G) was added at a final concentration of 0.5 μg/μL to each well. Emitted light was detected in a Wallac EnVision Multilabel reader (Perkin Elmer). Target lysis was determined as [1 − (RLUsample)/(RLUmax)] × 100.

All experiments were performed in compliance with all ethical regulations and in accordance with approved MSKCC Institutional Animal Care and Use Committee protocol 96–11–044. All mice were included in the analyses and no attrition was noted. Mice were 6 to 12 weeks old and weighed 18 to 30 g when treated. Both male and female mice were used. Female mice were randomized into groups to allow balance in groups for tumor growth before treatment. Male mice were maintained with their initial littermates to avoid fighting. Experiments were not blinded, but results were confirmed by blinded third parties. Experiments were replicated two or more times as indicated in the legends. For bioluminescence imaging (BLI) imaging, mice were dosed retro-orbitally with 2 mg of d-Luciferin in 100 μL PBS. All BLI was performed using a Xenogen IVIS Spectrum and analyzed using Living Image software 4.7.4 (Xenogen Biosciences) or Aura 4.0 (Spectral Instruments Imaging).

C57/BL6 mice (Taconic) were engrafted intraperitoneally with 2 × 10⁶ EL-4 OVA cells on day −6 or −7. Transduced OT-1 SEAKER cells were engrafted intraperitoneally on day 0. For β-Lac efficacy experiments, 4 mg/kg of Ceph-AMS was administered beginning on day 1, for 3 doses twice per day. For CPG2 efficacy experiments, 50 mg/kg AMS-glu was injected intraperitoneally from day 1 to day 5 twice every day.

C57/BL6 mice were engrafted subcutaneously with 1 × 10⁶ B16-SIIN on day –7. Tumors were engrafted using a 1:1 mixture with Matrigel (Corning, 354234). On day –1, mice were treated intraperitoneally with 100 mg/kg cyclophosphamide (Sigma-Aldrich, C0768–10G). On day 0, mice were engrafted intravenously via retro-orbitally injection with 1 to 3 ×10⁶ bulk engineered T cells.

Experimental parameters described in “B16F10 SIINFEKL melanoma models” above were used. Tumors were harvested at days 2, 5, 7, 10, 14, and then dissociated with Mouse Tumor Dissociation Kit (Miltenyi Biotec) in RPMI1640 and running on a gentleMACs (Miltenyi Biotec) for 10 to 20 minutes. Spleen samples were ground through a 100 μmol/L cell strainer into ACK lysis buffer and washed. Single-cell suspensions were washed and prepared for flow cytometry as described in “Flow Cytometry” above.

For BLI for T-cell tracking using gLuc⁺ T cells, mice were imaged on days 1, 2, 5, 7, 10, and 14 by injection of 100 μg of coelenterazine (Prolume, 3031–10) per mouse via retro-orbitally injection, with an exposure of 60 seconds.

For lymphoid organ SEAKER cell experiments, mice were harvested on day 6. Tumor and spleens were harvested as described above. Inguinal lymph nodes and lungs were removed and mechanically dissociated through a 100 μm strainer into RPMI-1640. Bone marrow was collected by removing femurs and crushing them with mortar and pestle in RPMI1640. Blood samples were collected through cheek bleeds into heparinized tubes. ACK lysis buffer was used to lyse blood cells for 5 minutes. Collected cells were washed in PBS and prepared for flow cytometry as described.

Experimental parameters described in “B16F10 SIINFEKL melanoma models” above were used. On day 5 after T-cell engraftment, tumor samples were harvested through a cell strainer into PBS to solubilize the cell-free proteins. Samples were centrifuged at 12,000 × g to remove all debris and clarified supernatant fluids were run in the nitrocefin assay.

Experimental parameters described in “B16F10 SIINFEKL melanoma models” above were used. Prodrug injections spanned days 4 to 8, with 4 mg/kg Ceph-AMS intraperitoneally twice daily. Tumor length and width was measured and tumor volume was determined using the formula: [(L × W)²]/2. Mice were euthanized once tumor volume exceeded 2 cm³ or if excessive distress or tumor ulceration was observed.

C57BL/6 mice received 3 × 10⁶ MUC16t⁺ ID8 cells intraperitoneally at day −21. On day 0, 3 × 10⁶ of anti-mCD19 or anti-mMuc16 gLuc CAR T cells were injected intraperitoneally with 100 μg of coelenterazine per mouse, with an exposure of 60 seconds, imaged immediately after injection. Images were taken serially at indicated days.

C57BL/6 mice were preconditioned with 100 mg/kg cyclophosphamide, followed by a subcutaneously engraftment of 1 × 10⁶ EL4 MUC16⁻ (left flank of mice) or MUC16⁺ (right flank of mice) tumors. At day 5, 3 × 10⁶ gLuc-mMUC16 cells were engrafted. On day 12 posttumor engraftment, 100 μg of coelenterazine was injected retro-orbitally, and mice were imaged prone.

C57BL/6 mice received subcutaneously engraftment of 1 × 10⁶ B16F10 cells at day −7. 100 mg/kg cyclophosphamide was injected intraperitoneally at day −1. Transgenic T cells from PMEL mice were transduced with gLuc and then engrafted retro-orbital T cells were imaged serially through retro-orbital injection of 100 μg coelentrazine.

C57BL/6 mice received subcutaneous engraftment of 1 × 10⁶ B16F10 SIINFEKL cells at day −7. 100 mg/kg cyclophosphamide was injected intraperitoneally at day −1. CPG2 OT-1 T cells were then engrafted, and tumors were harvested serially for flow cytometric analysis.

C57BL/6 mice were engrafted with 3 × 10⁶ ID8 ovarian tumor cells intraperitoneally at day −21. Mice were preconditioned with 100 mg/kg cyclophosphamide at day −1. On day 0, 3 × 10⁶ β-Lac-mMUC16 SEAKER cells were engrafted retro-orbital peritoneal lavages were performed with PBS on day 7 for flow cytometry and nitrocefin enzyme activity.

C57BL/6 mice were engrafted with 1 × 10⁶ B16F10 cells subcutaneously followed by twice every day dosing of AMS in PBS at dose levels ranging from 0 to 500 ug/kg, from day 9 to 13 posttumor engraftment

C57BL/6 mice were engrafted with 1 × 10⁶ B16F10 cells subcutaneously at day −7, 100 mg/kg of cyclophosphamide on day −1, and β-Lac PMEL T cells on day 0. Blood samples were collected via cheek bleed. Blood was clotted at room temperature for 10 minutes and spun at 500 × g for 10 minutes to harvest serum supernatant. Serum samples were tested for reactivity to recombinant untagged β-Lac at day 2, 5, and 7.

C57BL/6 mice were engrafted with 1 × 10⁶ B16F10 SIINFEKL cells subcutaneously at day −14, pretreated with 100 mg/kg cyclophosphamide at day −8 and treated with 3 × 10⁶ CPG-2 or Untransduced OT-1 cells retro-orbitally on day −7. Mice were retreated with 100 mg/kg cyclophosphamide on day −1 and reengrafted with trackable gLuc⁺ β-Lac SEAKER cells on day 0. Mice were imaged serially.

Data reported as mean ± SD unless otherwise noted. For in vitro experiments, technical replicates are displayed. For in vivo and ex vivo experiments, individual mice are displayed. Log-rank, unpaired or paired t tests were performed using Prism 8 software (GraphPad) when appropriate. Statistical significance was indicated as *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

The data generated in this study are available within the article and its supplementary data files or upon request from the corresponding author.

The SEAKER platform has been shown to be efficacious in human B-cell lymphoma models (7). However, these xenograft mouse models using NOD-SCID gamma (NSG) mice lack an endogenous adaptive immune system and natural killer (NK) cells, which may affect SEAKER-cell kinetics and persistence. Syngeneic models are useful in understanding the behavior of these complex cells for clinical translation. To understand key issues of kinetics, biodistribution, and immunogenicity of the SEAKER platform in immunocompetent hosts, we developed a syngeneic SEAKER cell system. Gene cassettes that included either the secreted β-Lac or CPG2 enzymes upstream of GFP via P2A cleavage sites were designed (Fig. 1A; Supplementary Fig. S2A). The best signal sequences for each enzyme were determined empirically previously (7) and used in this study. We transduced OT-1 T cells, which recognize the SIINFEKL OVA peptide presented on H-2Kb. Both β-Lac (Fig. 1B) and CPG2 (Supplementary Fig. S2B) constructs were successfully transduced into primary murine OT-1 T cells, to provide β-Lac OT-1 SEAKER cells and CPG2 OT-1 SEAKER cells for further evaluation herein. For β-Lac, transduction efficiencies ranged from 10% to 80%, with a mean transduction efficiency of 38% (Fig. 1A). For CPG2, transduction efficiencies ranged from 7% to 40%, with a mean transduction efficiency of 31% (Supplementary Fig. S2A). For all experiments, a lower threshold of 25% was used unless otherwise noted.

We have previously demonstrated that the β-Lac enzyme cleaves a Ceph-AMS prodrug to release the potent cytotoxic nucleoside analog AMS (7). Rapidly dividing cells are more susceptible to AMS (7, 26). To evaluate the ability of β-Lac OT-1 SEAKER cells to activate Ceph-AMS, spent culture media from these cells was collected and incubated with 180 nmol/L Ceph-AMS and EL-4 tumor cells. Although the Ceph-AMS prodrug alone was nontoxic, combination with preconditioned media from the murine OT-1 SEAKER cells led to substantial cytotoxicity against the EL-4 target cells, demonstrating that the enzyme was fully functional (Fig. 1C). Addition of the β-Lac inhibitor tazobactam abrogated the cytotoxicity, presumably by blocking conversion of the Ceph-AMS prodrug to the active AMS drug. In the SEAKER system, less than 1 ng/mL of β-Lac can convert nontoxic prodrug Ceph-AMS into highly toxic AMS (Supplementary Fig. S3; ref. 7). Using nitrocefin, we were able to quantitate β-Lac secretion at about 500 ng/mL of enzyme per 1 million SEAKER cells, per day (Fig. 1D). CPG2 cleaves the AMS-Glu prodrug substrate (7, 27). Conditioned media from CPG2 OT-1 SEAKER cells activated the cytotoxicity of an AMS-Glu prodrug (100 μmol/L) to induce toxicity on B16F10 cells (Supplementary Fig. S2C). Using methotrexate, we were able to quantitate CPG2 secretion at about 200 to 500 ng/mL per 1 million CPG2-secreting cells (Supplementary Fig. S2D and S2E).

Given the potent cytotoxicity of AMS, we reasoned that the anticancer activity of murine OT-1 SEAKER cells would be improved by the addition of the cognate prodrug. We performed a specific lysis experiment with OT-1 T cells expressing β-Lac cocultured with the EL-4 murine T-cell lymphoma cell line ectopically expressing full-length ovalbumin (EL-4 OVA) and ffLuc, in the presence or absence of the cognate prodrug (Fig. 1E). At each E:T tested, inclusion of the prodrug significantly enhanced cytotoxicity against EL-4 OVA cells. In addition to the OT-1 model, we also generated SEAKER cells based on established CAR T-cell constructs for antimurine CD19 and anti-MUC16t using wild-type (WT) murine T cells. Both anti-murine CD19 and anti-MUC16t SEAKER cells secreted enzymes, unmasked prodrug, and displayed enhanced cytotoxicity against antigen-positive and bystander cancer cells (Supplementary Figs. S4 and S5). These data demonstrated the generalizability of the murine SEAKER system to multiple target antigens using CAR (CD19, MUC16t) and TCR-modified (OT-1) T cells.

As a proof of concept, EL-4 OVA cells were engrafted intraperitoneally in C57BL/6 mice on day –7 to create a fully syngeneic tumor model system to test the SEAKER platform. Baseline tumor volume was measured on day –1, and β-Lac OT-1 SEAKER cells were engrafted intraperitoneally on day 0. Ceph-AMS prodrug administration began on day 1 at either 3 or 4 mg/kg twice every day for three doses (Fig. 1F). Mice treated with β-Lac OT-1 SEAKER cells and the Ceph-AMS prodrug demonstrated the lowest change in tumor progression across two replicate experiments (Fig. 1G, replicates shown in Supplementary Fig. S6). These data demonstrated that murine T cells can produce enough enzyme in vivo to unmask prodrug and delay tumor growth.

A similar experiment was performed using CPG2 OT-1 SEAKER cells and the corresponding AMS-Glu prodrug (50 mg/kg, BID for 12 doses; Supplementary Fig. S7A). In this experiment, 2 of 4 mice treated with the combination of CPG2 SEAKER cells and the AMS-Glu prodrug had a durable response showing no evidence of tumor for weeks (Supplementary Fig. S7B–S7D). The CPG2–AMS-Glu combination showed no overt systemic toxicity in this model (Supplementary Fig. S7E), as assessed by body weight change, whereas the β-Lac–Ceph-AMS combination did exhibit significant systemic toxicity and mouse weight loss (Supplementary Fig. S6C and S6H). Because β-Lac has fast enzyme kinetics for conversion of Ceph-AMS into AMS (7), we hypothesized that peritoneal β-Lac caused immediate conversion of the prodrug upon intraperitoneal injection, leading to systemic leakage of the active AMS drug. On the other hand, CPG2 has slower enzyme kinetics for conversion of its cognate prodrug AMS-Glu (7), and use of this slower system may be beneficial in scenarios where systemic drug leakage is of concern, thus mitigating drug toxicity. Nonetheless, these proof-of-concept experiments demonstrated that both enzyme–prodrug systems had antitumor efficacy in a syngeneic mouse model. However, these experiments utilized intraperitoneally injected SEAKER cells and prodrugs, which simplified the need for specific T-cell homing and localization to the tumor.

Models in which SEAKER cells must traffic and localize to solid tumor masses should better demonstrate the benefits of the SEAKER platform, as compared with intraperitoneally or in situ delivery models. Therefore, to better understand antigen-specific SEAKER-cell localization in solid tumors, we performed a time course of β-Lac OT-1 SEAKER cell localization to B16F10 melanoma tumors expressing the SIINFEKL ovalbumin peptide (B16-SIIN) in C57BL/6 mice. B16-SIIN tumors were engrafted subcutaneously on day –7. Cyclophosphamide was injected at 100 mg/kg intraperitoneally to precondition mice for adoptive T-cell transfer on day –1. β-Lac OT-1 SEAKER cells were engrafted intravenously on day 0. Mice were serially euthanized on days 2, 5, 7, 10, and 14, and tumors and spleens were harvested and analyzed by flow cytometry (Fig. 2A). OT-1 T cells have a CD45.1 congenic marker, allowing us to track the total OT-1 T cells as well as the transduced OT-1 SEAKER cells within the same animal. Both untransduced OT-1 T cells and OT-1 SEAKER cells exhibited a peak expansion at day 5 and began contracting by day 7 in the tumor (Fig. 2B–E). OT-1 T cells localized at 10-fold lower frequencies in the spleen than tumor, highlighting the specificity of antigen-specific T-cell localization. We observed a more pronounced contraction of OT-1 SEAKER cells as compared with untransduced T cells beginning at day 10 (Fig. 2F). Nonetheless, the kinetics of the OT-1 SEAKER cells matched that of the WT OT-1 T cells during the first week of the response, indicating functional fidelity. To further investigate the fitness of β-Lac OT-1 SEAKER cells, we compared the tumor control to mock transduced OT-1 T cells at a suboptimal dose of 3 × 10⁶ cells per mouse. Tumor growth kinetics were similar between β-Lac OT-1 SEAKER cells and mock transduced OT-1 T cells, demonstrating a transient control of tumor growth as compared with PBS treated mice (Fig. 2G and H).

T cells naturally localize to antigen-positive tissues, but also localize to secondary lymphoid organs (28), which might cause off-target toxicity with the SEAKER platform, due to prodrug unmasking. To understand whether SEAKER cells localize to off-target tissues, we treated B16-SIIN–bearing mice and harvested tumors, major secondary lymphoid organs (draining lymph nodes, bone marrow and spleen), blood, and the lungs at day 6 after β-Lac OT-1 SEAKER cell engraftment (Fig. 3A). The β-Lac OT-1 SEAKER cells localized at high percentages in the tumor, but they were present only at very low frequencies in off-target organs (Fig. 3B–E). Taken together, these findings demonstrated that SEAKER cells localized specifically to tumors and were rarely found in off-target secondary lymphoid organs during the peak expansion, which indicated a reduced risk of off-target toxicity.

To study the pharmacokinetics of SEAKER cells in vivo, we used Gaussia luciferase (gLuc), which enables BLI of antigen-specific T cells in both tumor and healthy tissues (Supplementary Figs. S8–S10; ref. 24). To track the exact localization of enzyme-secreting cells, we designed a β-Lac OT-1 SEAKER construct that includes a membrane-anchored gLuc and mCherry. These cells were used in the B16-SIIN model described above and mice were serially imaged by BLI (Fig. 4A and B). In accordance with the flow cytometric analysis of SEAKER cells (Fig. 2), we observed a peak in expansion at day 5 after T-cell engraftment followed by a contraction of detectable cells (Fig. 4C and D). Administration of rhIL-2 at 4.5 × 10⁵ IU/mouse/day for 6 days to stimulate T-cell growth further did not impact SEAKER cell expansion (Fig. 4D). Similar expansion kinetics were seen in the analogous CPG2 OT-1 SEAKER models (Supplementary Fig. S11). Taken together, these results demonstrated that SEAKER cell kinetics can be tracked longitudinally using BLI. Moreover, two separate methods of kinetic tracking (flow cytometry and BLI) both demonstrated the same kinetics, despite introduction of the foreign gLuc protein.

Although localization of SEAKER cells to antigen-positive tumors was robust, it was not yet evident whether these cells were delivering functional enzymes to tumors in the immunocompetent hosts. Thus, we harvested B16-SIIN tumors from mice treated with β-Lac OT-1 SEAKER cells and extracted the soluble protein fraction (Fig. 4E). When treated with the gLuc substrate coelenterazine, only tumor homogenates from the β-Lac-gLuc-mCherry OT-1 SEAKER-treated mice had any detectable BLI signal ex vivo (Fig. 4F). We then mixed tumor homogenate extracts with nitrocefin to assess β-Lac enzyme function directly. Both the tri-cistronic β-Lac-gLuc-mCherry OT-1 SEAKER and bi-cistronic β-Lac-GFP OT-1 SEAKER treated mice had significantly more enzyme activity than control mice in the tumor homogenates (Fig. 4G), and the bi-cistronic vector resulted in higher levels of expression.

These results showed that the level of secretion of β-Lac enzyme in vivo can be fine-tuned through the number of 2a elements. Each additional 2a site on the vector decreased the production efficiency of each genetic element. Levels of enzyme activity were sufficient to convert substrate ex vivo, indicating the feasibility of the SEAKER–prodrug approach in immunocompetent hosts. We also showed that targeted delivery of β-Lac enzyme was achievable in a peritoneal ovarian tumor model using the anti-MUC16 CAR model (Supplementary Fig. S12). This demonstrated that the SEAKER platform can deliver enzymes using both CAR and TCR-based antigen targeting mechanisms to a variety of anatomical locations in immunocompetent hosts.

Having established that SEAKER cells are functional in vivo and localize specifically to B16-SIIN tumors, we designed an efficacy model to test whether β-Lac OT-1 SEAKER cells can be potentiated with the nontoxic Ceph-AMS prodrug to delay tumor growth. Consistent with a previous report (29), direct treatment with the AMS parent drug induced pronounced systemic toxicity in mice, manifested by rapid weight loss, and had only a modest antitumor effect against B16 at the MTD of 0.1 mg/kg (Supplementary Fig. S13). To determine whether the SEAKER platform could improve antitumor efficacy and decrease toxicity, B16-SIIN tumor-bearing mice in our syngeneic model were engrafted with β-Lac OT-1 SEAKER cells on day 0, then treated with Ceph-AMS on days 4 to 8 (4 mg/kg, twice every day for 10 doses; Fig. 5A). Mice treated with β-Lac OT-1 SEAKER cells and prodrug demonstrated delayed tumor growth as measured by digital calipers (Fig. 5B). Furthermore, survival of mice given SEAKER cells and prodrug was significantly enhanced compared with mice treated with SEAKER cells or prodrug alone (Fig. 5C). In addition, we observed no overt toxicity, as evidenced by no decrease in weight of mice treated with SEAKER plus prodrug (Fig. 5D). These results showed that the SEAKER platform is efficacious against melanoma solid tumors without the characteristic toxicity associated with direct administration of the AMS parent drug, due to local generation at the tumor.

One potential limitation of the SEAKER platform is that the SEAKER cells may themselves be subject to the cytotoxic effects of the prodrug they unmask. To assess this issue, we used our trackable SEAKER cells in the subcutaneously efficacy model established above (Fig. 5) and imaged for the presence of the T cells every day during the prodrug dosing regimen (Fig. 6A). We found that groups of mice given the SEAKER cells alone or SEAKER cells plus prodrug demonstrated identical T-cell pharmacokinetics during the dosing schedule (Fig. 6B and C). We speculate that the SEAKER cells may survive because the prodrug is administered at the end of the T-cell expansion period, when the T cells are presumably slowing proliferation. This may explain the resistance of SEAKER cells to the active cytotoxic drug unmasked in vivo.

The bacterial origin of SEAKER enzymes raises the possibility that they will be immunogenic in immunocompetent hosts, potentially preventing the cellular or enzymatic functions. Mice treated with β-Lac OT-1 SEAKERs mounted a humoral response to the enzyme that peaked at day 5 after T-cell injection, which followed a similar kinetic time course to that of the SEAKER cells themselves (Supplementary Fig. S14A–S14E). When those same serum samples were incubated with recombinant β-Lac and nitrocefin substrate, equivalent levels of substrate cleavage resulted from enzymes treated with naive mouse serum or β-Lac OT-1 SEAKER-treated mouse serum (Supplementary Fig. S14F and S14G). This indicated that, although the antibodies that bind β-Lac were induced, they did not block the function of the enzyme.

Multiple rounds of adoptive T-cell transfer might increase antitumor efficacy and T-cell localization in the tumor. However, immune responses against adoptive T cells have been documented after single injections, preventing additional rounds of cell administration (8). We assessed whether SEAKER cells could be readministered in previously-treated mice to explore the impacts of retreatment upon the pharmacokinetics of the cells. B16-SIIN tumor-bearing mice received a preconditioning dose of cyclophosphamide at 100 mg/kg intraperitoneally at day –8 then a dose of either PBS or β-Lac-GFP OT-1 SEAKER cells at day –7. Mice were then stratified into two further groups: one cyclophosphamide dose or three consecutive cyclophosphamide doses at 100 mg/kg i.p. gLuc⁺ SEAKER cells were then engrafted into all groups on day 0 and imaged day 5 after engraftment (Fig. 6D). Mice that had not been pretreated engrafted the SEAKER cells by day 5. In contrast, four of five mice treated with two successive SEAKER doses and one cyclophosphamide dose rejected the second SEAKER cell engraftment (Fig. 6E and F). However, mice who received three doses of cyclophosphamide prior to the second adoptive transfer showed enhanced engraftment of the second SEAKER cell administration (Fig. 6E and F). Furthermore, primary engraftment of wild-type OT-1 T cells or CPG2 OT-1 SEAKER cells had no impact on subsequent engraftment of β-Lac OT-1 SEAKER cells (Supplementary Fig. S14). These results demonstrated that the host immune responses to SEAKER cells can be overcome through simple pharmacologic manipulation. Taken together, these studies show that the SEAKER-produced enzymes are still functional despite being bound by antibodies, and that multiple rounds of SEAKER cell treatment can be achieved with stringent preconditioning, similar to what is used in humans clinically.

We have examined the kinetics, biodistribution, and efficacy of complex, cargo-delivering, adoptive T cells in immunocompetent hosts bearing solid tumors, revealing both opportunities and potential challenges. The results presented demonstrate the feasibility and efficacy of adoptive T-cell micropharmacies in syngeneic systems. Traditionally optimized for their cytotoxic capacity, T cells can now be used for consistent, large-scale, highly-localized cargo delivery. This will allow for effective therapeutic killing, even for poorly expressed, downregulated, or heterogenous antigens within a tumor with limited access. This study presents four conceptual advances for adoptive T-cell micropharmacies: (i) cargo delivery and efficacy in an immunocompetent system, (ii) cargo delivery and efficacy in solid-tumor models, (iii) expansion of the target repertoire of SEAKER cells to intracellular antigens using TCRs, and (iv) circumvention of the SEAKER and enzyme immunogenicity issues.

Syngeneic tumor models enable the study of adoptive T cells in immunologically competent hosts. In contrast, xenograft models in NSG mice lack common gamma chain–dependent immune cells, such as T cells, B cells, and NK cells (30, 31). Furthermore, innate immune function in these models is altered. As a consequence, secondary lymphoid organ formation is also altered. Competition for prosurvival cytokines is largely nonexistent, which allows adoptively transferred human cells to maintain homeostatic proliferation in the absence of antigen, confounding pharmacokinetic and biodistribution studies (19, 20). Critically, antimouse reactive human cells selectively expand in xenograft models and induce graft-versus-host disease at late timepoints (32). Investigation of true T-cell pharmacokinetics is not feasible in these models. Syngeneic mouse models circumvent all of the caveats of xenograft models. The OT-1 immunocompetent SEAKER models developed in this study provided systems in which to test important pharmacokinetic parameters of SEAKER cells in advance of clinical translation. In the context of a complete immune system, SEAKER cells are still able to localize at high concentrations to antigen-positive tumors and to deliver cargo with minimal off-tumor localization.

Solid tumors present additional challenges to effective adoptive T-cell therapy in comparison to hematopoietic cancers. T-cell penetration in solid tumors is often aberrant, due to altered tumor extracellular matrix, trapping by suppressive immune cells, and necrotic acellular cores (33, 34). Solid tumors also express antigens that are found on vital healthy tissues. This leads to toxicities that limit the effectiveness of many solid tumor therapies. In addition, solid tumors are notoriously heterogenous. Many tumor-specific antigens are only present in a subset of cells within the tumor, leading to inevitable escape after adoptive T-cell therapy. The B16-SIIN tumor model is highly aggressive, with high resistance to T cell–mediated killing in vivo. Addition of the SEAKER enzyme/prodrug system led to delayed solid tumor growth and improved survival without the characteristic systemic toxicity of chemotherapeutics, even in the absence of substantial T-cell killing.

Immunogenicity is a barrier for synthetic or living therapeutics. However, adenovirus and microbial-derived therapies have seen clinical success despite their potential immunogenicity (35, 36). Synthetic proteins developed in a given species may still elicit an immune response in the same species, due to small changes in the amino acid sequences, new sequences at fusion junctions of chimeric proteins, or antigen presentation of the synthetic protein. For example, both mice and humans mount humoral immune responses against species-matched CAR proteins (8). These immune responses may explain why some patients do not achieve complete responses when treated with CAR T cells or have poor persistence or responses to later infusions of the same product. Despite their immunogenicity, CAR T cells and other synthetic therapeutics are efficacious for some B-cell cancers. Persistence of adoptive T cells, even if immunogenic, may be prolonged by use of effective lymphodepleting regimens. In syngeneic mice, lymphodepleting preconditioning is a requirement for adoptive T-cell engraftment. In this study, both the SEAKER enzymes and cells elicited an immune response in the immunocompetent model, even within the short timeframe of the aggressive B16F10-SIINFEKL model. Nonetheless, the SEAKER enzymes remained functional despite antibody binding, and secondary engraftment of SEAKER cells could be achieved through stringent preconditioning regimens.

The SEAKER platform is a modular small-molecule delivery platform that takes advantage of intrinsic T-cell function to deliver therapeutic cargo to tumors. Antigen stimulation increases and localizes therapeutic payloads to tumors via two mechanisms. Firstly, stimulated T cells are retained in their target tissues and undergo proliferation. This leads to high tumor concentrations of T cells, whereas trace numbers of cells localize to off-target tumors. Second, activated T cells express more retroviral transgenes (37). T-cell activation leads to activation of retroviral promoter LTRs and expression of downstream genes (38). Although we used constitutively expressed enzymes, these two mechanisms localize the enzyme to its target tissue. When tumor, SEAKER cells, and prodrug are all injected intraperitoneally, toxicity in the mice is observed, due to systemic leakage of unmasked AMS from the peritoneal cavity. However, in a more clinically relevant adoptive transfer model using intravenously administered SEAKER cells and a solid subcutaneous tumor, antitumor benefit was observed without the characteristic toxicity of AMS. The SEAKER platform has the highest therapeutic window when T cells must traffic into the tumor from circulation. Once in the target tissue, any cell in the vicinity of unmasked drug will be susceptible, offering the possibility to kill antigen negative cells and stromal cells. Future iterations of the SEAKER platform may also include T-cell activation inducible promoters, such as NFAT, to further control the delivery of payloads.

The potential for targeted delivery of synthetic cargo to tissues has applications outside of cancer. Enzyme–prodrug therapy could be used in a variety of diseases to deliver drugs directly to the affected tissues. In this study, we expanded the application of SEAKER cells to include both CAR T cells and TCR T cells, demonstrating the wider translational potential of targeted micropharmacies. Cellular micropharmacy–directed therapy could be used in a variety of disease states, such as inflammatory or autoimmune conditions, to deliver potent immunosuppressives to target tissues, sparing systemic immunosuppression. Critically, a barrier to further application of the SEAKER platform in inflammatory disease states is the need for cells that do not kill their target. This will require further engineering and exploration of other cell types as targeted micropharmacies.

C.M. Bourne reports grants from NIH NCI F31 and MSK MERIT Sawyers Fellowship during the conduct of the study. T.J. Gardner reports a patent for SEAKER cell technology patent pending to CoImmune; and reports employment with ArsenalBio. D.S. Tan reports grants, personal fees, and nonfinancial support from CoImmune, Inc. during the conduct of the study; grants, personal fees, and nonfinancial support from CoImmune, Inc. outside the submitted work; also has a patent for PCT/US2018/040629 pending to CoImmune, Inc., a patent for PCT/US2018/040633 pending to CoImmune, Inc., a patent for PCT/US2018/040640 pending to CoImmune, Inc., a patent for PCT/US2018/040639 pending to CoImmune, Inc., and a patent for PCT/US2021/015424 pending. D.A. Scheinberg reports grants from NIH and CFCR during the conduct of the study; grants, personal fees, and other support from Coimmune; personal fees and other support from Sellas and Actinium; personal fees from Eureka, Repertoire, Atengen, Oncopep, Sapience, Pfizer; and other support from Clade outside the submitted work; also has a patent for SEAKER cells pending to Coimmune. No disclosures were reported by the other authors.

C.M. Bourne: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. P. Wallisch: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M.M. Dacek: Conceptualization, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. T.J. Gardner: Conceptualization, resources, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. S. Pierre: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. K. Vogt: Conceptualization, resources, supervision, validation, investigation, methodology, writing–original draft, writing–review and editing. B.C. Corless: Conceptualization, resources, investigation, methodology, writing–review and editing. M.A. Bah: Resources, investigation, methodology. J.E. Romero-Pichardo: Validation, investigation, methodology. A. Charles: Conceptualization, validation, investigation, methodology. K.G. Kurtz: Conceptualization, investigation, methodology. D.S. Tan: Conceptualization, resources, funding acquisition, methodology, project administration, writing–review and editing. D.A. Scheinberg: Conceptualization, resources, supervision, funding acquisition, methodology, writing–original draft, project administration, writing–review and editing.

We thank the Renier Brentjens lab (MSK) for providing reagents and expertise throughout this project. Financial support for this work was provided by the Leukemia and Lymphoma Society (to D.A. Scheinberg), grants from the NIH (P30 CA008748 to C. B. Thompson, R01 CA055349 to D.A. Scheinberg, P01 CA023766 to D.A. Scheinberg and D.S. Tan, R35 CA241894 to D.A. Scheinberg and D.S. Tan, F31 CA254311 to C.M. Bourne, T32 CA062948–Gudas to B.C. Corless, and F31 CA261179 to B.C. Corless), Tudor Funds (to D.A. Scheinberg), Commonwealth Foundation and Experimental Therapeutics Center of MSKCC (to D.S. Tan and D.A. Scheinberg), and the Memorial Sloan Kettering Maximizing Excellence in Research, Innovation and Technology Sawyers Fellowship (to C.M. Bourne).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.