Promoting tumor antigen presentation unleashes therapeutic efficacy of Tgfbr2-deficient TCR-engineered T cells in pancreatic ductal adenocarcinoma

Abstract

Background

Pancreatic ductal adenocarcinoma (PDA) is a lethal malignancy with limited therapeutic options. Adoptive cell therapy (ACT) using tumor-reactive T cell receptor-engineered T cells (TCR-T) is a promising therapeutic modality for targeting PDA. However, additional strategies are needed to overcome the suppressive tumor microenvironment (TME) that impairs T cell persistence and function. Tumor growth factor-β (TGFβ) is a prominent immunosuppressive component of the TME, but is also believed to be critical for T cell memory.

Methods

We develop aggressive immunocompetent orthotopic Kras^LSL-G12D/+;Trp53^LSL-R172H/+;p48-Cre (KPC) PDA mouse models to identify obstacles to sustained activity of TCR-T cell therapy targeting the native tumor antigen, mesothelin (Msln). During ACT, we track TCR-T cell fate, function, antitumor effects, and interactions with type I dendritic cells using flow cytometry, two-photon microscopy, and live-cell imaging. Based on an abundance of TGFβ and a halt in TCR-T cell differentiation in orthotopic tumors, TCR-T cells with defective Tgfbr2 are tested in a new model, where peptide-major histocompatibility complex-I (pMHC-I) is incrementally augmented on KPC tumor cells.

Results

TCR-T cells accumulated in orthotopic PDA but were rendered dysfunctional despite lacking features of exhaustion. KPC tumor cells express minimal constitutive MHC-I, resulting in suboptimal recognition by TCR-T cells. Modestly increasing pMHC-I on KPC tumor cells restored canonical differentiation of TCR-T cells into PD-1⁺ progenitor exhausted (T_PEX) in PDA, but failed to rescue interferon-γ (IFNγ) production. TGFβ was enriched in tumor nests, and deletion of Tgfbr2 in TCR-T cells enhanced dendritic cell interactions, reduced intratumoral TGFβ deposition, restored IFNγ production, and improved therapeutic efficacy without compromising TCR-T cell longevity or tissue residency.

Conclusions

Modestly enhancing tumor-intrinsic pMHC-I unleashes therapeutic efficacy and functionality of Tgfbr2-disrupted TCR-T cells in PDA.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Pancreatic ductal adenocarcinoma (PDA) is a lethal malignancy that lacks durable therapies. T cell receptor-engineered T cell (TCR-T) therapy targeting mesothelin (Msln) is a promising approach based on preclinical studies in autochthonous mouse models but is limited by loss of donor T cell functional persistence. Tumor growth factor-β (TGFβ) is a highly suppressive cytokine that is elevated in the pancreatic tumor microenvironment but can also promote T cell memory.

WHAT THIS STUDY ADDS

• We demonstrate that immunocompetent PDA mouse models lack sufficient peptide-major histocompatibility complex-I (pMHC-I) to evoke differentiation, accumulation, and antitumor function of engineered TCR-T cells targeting Msln. Augmenting tumor pMHC-I permits antitumor T cell differentiation, interactions with dendritic cells (DCs), and therapeutic efficacy mediated by TCR-T cells deficient in Tgfbr2. At a late time point following adoptive cell therapy (ACT), TCR-T cells lacking Tgfbr2 outcompete wild-type counterparts in non-gastrointestinal tissues, including healthy pancreas.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Enhancing tumor antigen presentation and deleting Tgfbr2 in TCR-T cells can overcome T cell dysfunction and enhance ACT efficacy without compromising immunological memory, attributes that correlate with altered T cell:DC dynamics in PDA.

Introduction

Pancreatic ductal adenocarcinoma (PDA) is the third deadliest malignancy in the USA.¹ Patients with metastatic PDA have a median overall survival of less than 1 year.² Chemotherapy regimens provide limited survival benefits,^{2 3} and while recently approved rat sarcoma viral oncogene homolog (RAS) inhibitors improve outcomes,⁴ secondary resistance develops rapidly.⁵ PDA also exhibits primary resistance to immune checkpoint blockade strategies that are effective in other cancers. The highly suppressive and fibroinflammatory tumor microenvironment (TME) of PDA is believed to be a major barrier to both cytotoxic and immunological treatment strategies.⁶ However, tumor cell-intrinsic mechanisms of resistance also likely contribute. Indeed, constitutive major histocompatibility complex (MHC)-I is frequently low in PDA,^7–10 which may impair cytotoxic CD8⁺ T cell-mediated tumor recognition and antitumor effects. Further, tumor cell MHC-I loss can cooperate with CD4⁺ regulatory T cells to facilitate PDA metastasis.¹¹

Adoptive cell therapy (ACT) is a promising therapy for solid malignancies that involves infusion of tumor-reactive T cells. Transfer of T cells engineered to express a tumor-reactive T cell receptor (TCR) or neoantigen-specific tumor-infiltrating lymphocytes have shown encouraging activity in gastrointestinal malignancies, including PDA.^12–14 We previously developed a TCR-engineered T cell therapy targeting the self/tumor antigen, mesothelin (Msln),¹⁵ which is overexpressed in nearly all PDAs.¹⁶ Serial infusions of murine T cells virally transduced to express the high-affinity Msln_406-414:H-2D^b-specific TCR (TCR1045) in combination with vaccine^{15 17} showed compelling antitumor effects in the autochthonous Kras^LSL-G12D/+;Trp53^LSL-R172H/+;p48-Cre (KPC) genetically-engineered mouse model of PDA that recapitulates the genetics, progression, and suppressive TME characteristic of human PDA.¹⁸ However, TCR1045 T cells diminished over time, and remaining cells were rendered dysfunctional.¹⁵ Blockade of multiple immune checkpoints (anti-PD-1, anti-Lag3, anti-Tim3) failed to rescue TCR1045 T cell function in autochthonous tumors.¹⁹ Further, reprogramming myeloid cells using agonistic anti-CD40 or depleting suppressive myeloid cells through targeting of Csf1r was insufficient to rescue engineered T cell dysfunction.²⁰ However, anti-CD40 combined with a vaccine did markedly improve TCR1045 T cell expansion and persistence in KPC tumors.²⁰

Here, we employ an immunocompetent orthotopic KPC model to define and overcome fundamental mechanisms of resistance to ACT. Unexpectedly, TCR1045 T cells infiltrating orthotopic tumors arrest in a dysfunctional early precursor state that is distinct from classical T cell exhaustion. Modestly increasing tumor peptide-MHC-I (pMHC-I) restores TCR-T cell differentiation toward a classical exhausted state. We further identify that tumor growth factor-β (TGFβ)-associated latency-associated peptide (LAP) is elevated within orthotopic tumor cell nests. Although TGFβ has well-established roles in suppressing T cells,^21–33 its contribution to limiting TCR-T cell-based therapy of PDA has not been deeply interrogated. We demonstrate that incrementally enhancing pMHC-I in KPC tumor cells confers an advantage to TCR-T cells lacking Tgfbr2, which remain polyfunctional in the TME and prolong survival. Collectively, our findings identify two independent mechanisms that constrain ACT efficacy, insufficient pMHC-I and suppressive TGFβ, and provide a conceptual framework for designing more effective T cell-based therapies for PDA patient treatment.

Methods

Animals

The University of Minnesota Institutional Animal Care and Use Committee approved all studies (Protocol ID: 2305–41080A). Male and female C57Bl/6J (000664), P14 (037394-JAX), Thy1.1 (000406), Tgfbr2^Flox/Flox (012603), and dLck-Cre (012837) mice aged 6–8 weeks were obtained from The Jackson Laboratory. Xcr1-Venus mice³⁴ were kindly provided by Dr. Tsuneyasu Kaisho (Wakayama Medical University) via Dr. Brian Fife (University of Minnesota). 1045 TCR exchange (1045^TREX) mice that express the Msln_406-414:H-2D^b-specific TCR within the Trac locus¹⁷ were crossed to Thy1.1 mice. For two-photon studies, 1045^TREX mice were crossed to Actb-eGFP mice (003291, The Jackson Laboratory). Previously described Kras^LSL-G12D/+;Trp53^LSL-R172H/+;p48-Cre (KPC) mice speed-backcrossed to the C57Bl/6J strain¹⁵ were employed. All strains were genotyped using in-house or Transnetyx protocols and maintained in specific pathogen-free facilities.

Tumor cell culture

KPC2 tumor cells were derived from a primary KPC tumor and previously described.³⁵ KPC271 and KPC451 lines were derived from independent B6 KPC mice. For two-photon imaging, KPC2 cells were transduced with mScarlet in the MIGR1 plasmid (kindly provided by Dr. Marc Jenkins, University of Minnesota) and sorted to >95% purity. For cytotoxicity assays, KPC2 cells were transduced with Incucyte Nuclight NIR lentivirus (Sartorius) and maintained under antibiotic selection using 1 µg/mL puromycin (Gibco). The KPC2^MslnER cell line was generated by transducing KPC2 cells to express Msln_406-414 (GQKMNAQAI) using “PresentER” plasmid³⁶ and sorted to >95% purity.

Orthotopic tumor implantation

Orthotopic tumor implantation was performed as previously described.^{35 37 38} Briefly, intrapancreatic injection of 1×10⁵ tumor cells in 60% Matrigel (Corning) was completed following surgical peritoneal access in anesthetized mice. Buprenorphine and carprofen were delivered preoperatively and postoperatively for analgesia.

ACT protocols

Tumor-bearing mice on day 5–6 post-tumor or healthy mice (online supplemental figures S6 and S7) received 120 mg/kg or 180 mg/kg cyclophosphamide (Cy) intraperitoneally (i.p.) followed 8 hours later by 5×10⁶ 1045^TREX T cells activated in vitro with 1–10 μg/mL of Msln_406-414 peptide (Genscript) and 10 ng/mL of recombinant human interleukin-2 (rhIL-2, Peptrotech) 72 hours prior to i.p. injection. Cryopreservation, which is used in most human ACT treatments, was omitted to streamline the protocol. Where indicated, BiVax (100 µg Msln_406-414 + 50 µg Poly I:C (Tocris))³⁹ or TriVax (100 µg Msln_406-414 + 50 µg Poly I:C+50 µg agonistic anti-CD40 (Clone FGK4.5, BioXcell))⁴⁰ was delivered i.p. with ACT. 5 µg of rhIL-2 was administered i.p. with ACT and at 48 hours and 96 hours thereafter. For experiments in KPC mice (online supplemental figure S5), congenically distinct P14 dLck-Cre⁺ × Tgfbr2^WT/WT or P14 dLck-Cre⁺ × Tgfbr2^Flox/Flox T cells were retrovirally transduced with TCR1045, expanded, and co-transferred at a 1:1 ratio with 1×10⁷ Msln_406-414-pulsed, irradiated splenocytes as described.¹⁵

High-resolution ultrasound

Tumors were identified in anesthetized mice based on anatomic landmarks, hypoechoic features, and clear tumor boundaries using ultrasound (Vevo F2, VisualSonics).^{15 38} Tumor volume was calculated as follows: 0.5 (length × width²). Mice were euthanized when tumors reached ≥500 mm³ in volume or when moribund.

Intravenous labeling and single cell suspension generation

Intravenous labeling was performed, and single cell suspensions were generated as described previously.³⁸ Briefly, minced tumor and pancreas were incubated with 1 mg/mL collagenase XI (Sigma-Aldrich) for 15 min in an Incu-Shaker (Carolina) at 37 °C. Spleens were mechanically dissociated, followed by RBC lysis using tris-ammonium chloride buffer (Gibco). Mechanically dissociated livers and minced lungs were digested for 1 hour using collagenase type I (Worthington Biochemicals) and dissociated using a gentleMACS tissue dissociator (Miltenyi), followed by mononuclear cell enrichment via Percoll (Sigma Aldrich) gradient. Intraepithelial lymphocytes (IEL) within small intestines (SI) were isolated using a method established in the Masopust laboratory.⁴¹ Briefly, SI Peyer’s patches and contents were removed, followed by incubation for 30 min at 37 °C with 0.154 mg/mL dithioerythritol (Sigma Aldrich) in 10% Hank’s Balanced Salt Solution (Corning). IEL were enriched via Percoll gradient. All single cell suspensions were passed over 40 µm or 70 µm filters and washed twice prior to further processing.

Ex vivo T cell stimulation for measuring cytokine production

Isolated splenic and intratumoral cells were incubated at 37 °C in 5% CO₂ for 4.5 hours ± 1–5 µg/mL of Msln_406-414 or a 1:500 dilution of Cell Stimulation Cocktail (CSC, Tonbo). Cells were cultured with GolgiPlug (1:1000; BD) and GolgiStop (1:1500; BD).

Flow cytometric analyses

Flow cytometry antibodies are listed in online supplemental table 1. Msln_406-414: H-2D^b-BV421 tetramer was obtained from the NIH Tetramer Core. Cultured cells or single cell suspensions were washed twice in fluorescence-activated cell sorting (FACS) buffer: phosphate buffer solution (Gibco), 2% fetal bovine serum, 0.01% sodium azide (Ricca Chemical Company). Cells were then incubated with a mixture of antibodies diluted at 1:100–500 in FACS buffer for 1 hour at room temperature (single cell suspensions) or 45 min at 4 °C (cultured cells). Intracellular staining of cytokines and transcriptional factors was performed using Cytofix/Cytoperm fixation/permeabilization (BD) or Foxp3 fixation/permeabilization (Tonbo) kits, respectively. Intracellular staining samples were incubated with antibodies diluted at 1:100–300 in perm/wash buffer overnight at 4 °C and washed twice with perm/wash buffer before resuspending in FACS buffer. Samples were analyzed within 24 hours using a 5-laser Cytek Aurora spectral cytometer with data acquisition via SpectroFlo software (Cytek). Some samples were analyzed using LSR-II or LSRFortessa conventional cytometers (BD) with data acquisition via FACSDiva (BD, V.9). All data analysis was performed using FlowJo software (V.10).

In vitro MHC-I induction, proliferation, cytokine production, cytotoxicity, and co-culture assays

For MHC-I measurements, KPC-derived tumor cells were incubated with or without titrating amounts of murine recombinant interferon-γ (IFNγ, R&D Systems) for 48 hours. For proliferation assays, CellTrace Violet (Invitrogen) labeled 1045^TREX splenocytes were stimulated with titrating amounts of Msln_406-414 and 10 ng/mL rhIL-2 for 72 hours as described.⁴² For cytokine production, in vitro-activated 1045^TREX effector T cells were restimulated with 1 µg/mL plate-bound anti-CD3 (Clone 17A2, BD) ± 20 ng/mL Tgfβ1 (R&D Systems) for 24 hours. For cytotoxicity assays, in vitro-activated 1045^TREX T cells were enriched using negative selection (Miltenyi) and co-cultured with previously plated KPC2 NIR⁺ tumor cells at a 5:1 effector:target (E:T) ratio. Tumor cell growth was monitored using an Incucyte SX5 instrument (Sartorius). For co-culture assays, in vitro-activated 1045^TREX T cells were ‘rested’ for 24 hours followed by co-culture with tumor cells pretreated ±100 ng/mL IFNγ for 24 hours ±5 µg/mL of Msln_406-414 for 1 hour. Co-cultures were performed at a 5:1 E:T ratio for 24 hours.

CRISPR/Cas9 mediated deletion of Tgfbr2

Two guide RNAs (gRNA), 5’-G*A*C*CGCACCGCCAUUGU and 5’-C*U*C*AGUUAACAGUGAUG (Synthego), targeting exon 1 of Tgfbr2 were complexed at a 3:1 ratio with TrueCut Cas9 V2 (Invitrogen). In vitro-activated 1045^TREX T cells were suspended in RT P4 nucleofection buffer (Lonza), 4 µM Alt-R Cas9 electroporation enhancer (Lonza), and gRNA/Cas9 in nucleofection cuvettes (Lonza). Electroporation (program CM137) was performed using a 4D Nucleofector (Lonza). Editing efficiency was determined using Inference of CRISPR Edits (ICE) online software (Synthego) following DNA isolation (DNeasy Blood and Tissue Kit, Qiagen) and Sanger sequencing (Eurofins) with primer 5’-TGTCGCGGCTGCATATC. Edited cells were used in downstream assays within 24 hours of electroporation.

Bone marrow-derived dendritic cell generation, live-cell imaging, and RNA sequencing

Bone marrow cells from femurs and tibias of C57Bl/6J mice were collected and cultured in media supplemented with 10 ng/mL murine GM-CSF (R&D Systems) and 100 ng/mL murine FLT3-L (R&D Systems) for 9 days. Resultant bone marrow-derived dendritic cells (BMDCs) were labeled with 0.5 µM CellTracker Deep Red (Invitrogen), subsequently pulsed with 0.1 µg/mL of Msln_406-414 for 1 hour, and then plated on chamber slides (Ibidi) coated with 1% bovine serum albumin (BSA; Sigma Aldrich) for 90 min. T cells with Tgfbr2 edited via CRISPR/Cas9 and WT T cells were labeled with 10 µM CellTracker Green CMFDA (Invitrogen) and 2 µM Rhod-2 AM (Invitrogen). For live-cell imaging, co-cultures were prepared at 5:1 T cell to DC ratios and immediately imaged using a Nikon Ti-E confocal microscope. Calcium flux, immune synapse formation, and cell interactions were quantified using FIJI (V.2.18.0) in ImageJ (V.1.54 p). For RNA sequencing, BMDCs pulsed with 0.1 µg/mL of Msln_406-414 and T cells were co-cultured at a 1:1 E:T for 24 hours. BMDCs were then isolated using CD11b selection kits (Miltenyi), and RNA sequencing was performed (Plasmidsaurus). Data were analyzed using web-based software (Plasmidsaurus).

Two-photon imaging and analysis

KPC2 mScarlet⁺ tumors were orthotopically implanted in Xcr1-Venus recipients and subsequently treated with ACT using 5×10⁶ Tgfbr2^WT or Tgfbr2^CRISPR-edited 1045^TREX × Actb-eGFP⁺ T cells plus TriVax. Tumors obtained 7 days after ACT were embedded in 0.5% agarose, cut into 200 µm slices, and mounted on coverslips. Movies were acquired using an MP SP5 two-photon microscope TCS (Leica) equipped with Mai Tai HP Deep See lasers (SpectraPhysics), an 8000 Hz resonant scanner, a 25×3/0.95 NA objective, and two non-descanned and two hybrid detectors. Tissue was excited at 860 or 890 nm with the following collections: second harmonic generation (SHG) <440 nm, mScarlet 569–593 nm, GFP 500–520 nm, Venus 515–528 nm. 60-frame videos were collected over 30 min in 3–5 geographic locations per tumor separated by at least 1 mm.

TrackMate in FIJI was used for two-photon analysis. cDC1s and T cells were identified using the Laplacian of Gaussian detector, and cell movement was tracked using the linear assignment problem tracker. Cell motility was calculated by fitting the mean square displacement with the persistent random walk model as follows, where $S$ is the migration speed and $P$ is the persistence time:

$${\displaystyle {\displaystyle \mathrm{M}\mathrm{S}\mathrm{D}\left(t\right)=2S^{2}P\left(t-P\left(1-e^{-\frac{t}{P}}\right)\right)}}$$

The motility coefficient is as follows, where $n_d$ is of unity dimension:

$$\dot{\mu }=\frac{S^{2}P}{n_d}$$

The total motility coefficient is the sum of motility components:

$$\mu ={\dot{\mu }}_x+{\dot{\mu }}_y$$

Statistical outliers were removed by ROUT (Q=1%). Individual cell track features (displacement, total distance traveled, confinement ratio, mean track speed, and mean directional change rate) were used to further characterize cell movement.

Immunofluorescent microscopy

Tissues were preserved in OCT, rapidly frozen, and stored at −80 °C. 7 µm sections were mounted on glass slides and fixed with acetone for 10 min. Following blocking with 5% BSA, samples were incubated with primary antibodies (listed in online supplemental table 1) at 1:100–200 dilutions in 1% BSA for 1 hour at room temperature. ProLong Gold antifade with 4′,6-diamidino-2-phenylindole (DAPI, Invitrogen) was applied, and samples were stored at 4 °C prior to imaging on DM4-B upright or Thunder Imager widefield microscopes (Leica). MHC-I and latency-associated peptide (LAP) within cytokeratin⁺ (CK) tumors were quantified using QuPath (V.0.6.0). CD8⁺ and CD11c⁺ cells were manually quantified by two independent scorers.

Statistical analyses

For in vivo experiments, male and female mice were randomized prior to treatment and monitored in an unblinded fashion without prespecified inclusion or exclusion criteria, with group size dictated by historic power calculations. Post-hoc exclusions were not performed. Unpaired two-tailed Student’s t-tests or Welch’s t-tests were used when comparing two groups. One-way analysis of variance (ANOVA) with Tukey’s or Dunnett’s posttests were used when comparing more than two groups. Comparisons were performed with statistical significance groups of *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001 using GraphPad Prism (V.10).

Results

Tumor-infiltrating TCR1045 T cells arrest in a dysfunctional progenitor state in orthotopic KPC PDA

We previously identified a high-affinity murine TCR (TCR1045) specific for Msln_406-414:H-2D^b and established that serial TCR1045-engineered T cell infusions combined with a cell-based vaccine prolong survival of KPC mice with autochthonous PDA.¹⁵ ACT with TCR1045 T cells is also effective in models of ovarian cancer that naturally overexpress Msln.^{43 44} We subsequently generated TCR exchange (1045^TREX) mice, in which TCR1045 is integrated within Trac during T cell development.¹⁷ The arrangement permits physiologic T cell development and TCR regulation, and provides a reproducible source of Msln-specific CD8⁺ T cells.^{17 42}

While invaluable, autochthonous PDA models are restricted by challenges in development of multifocal tumors, resulting in difficulties enrolling animals with similar disease burdens and practical breeding limitations. Consequently, we developed an orthotopic KPC model to more readily identify strategies to enhance ACT efficacy. The KPC2 cell line is derived from primary tumors from a female B6 KPC mouse with metastatic PDA, expresses Msln, upregulates the restricting element for Msln_406-414, H-2D^b, in response to IFNγ, and is refractory to PD-1/PD-L1 blockade.³⁵ We orthotopically implanted KPC2 cells into the pancreas of syngeneic mice and enrolled recipients to receive in vitro activated 1045^TREX T cell ACT alone or in combination with vaccine (figure 1A). As most human protocols use pre-activated T cells followed by lymphodepleting chemotherapies, including cyclophosphamide (Cy), to create a niche for transferred T cells,⁴⁵ we administered Cy prior to transfer of 1045^TREX T cells. To promote donor T cell expansion, some cohorts received vaccination with Poly I:C and Msln_406-414 peptide (BiVax)³⁹ or TriVax,⁴⁰ which adds agonistic anti-CD40 antibody.³⁵ ACT with TriVax reduced orthotopic tumor weights twofold compared with control mice on day 14 (figure 1B). Expectedly, donor T cell engraftment and expansion was significantly enhanced by vaccination (figure 1C,D). Infused T cells were enriched within tumors, supporting accumulation in response to Msln antigen.

Figure 1. Tumor-infiltrating TCR1045 T cells arrest in a dysfunctional progenitor state in orthotopic KPC PDA. (A) Schematic illustrating cyclophosphamide (Cy, 180 mg/kg) followed by ACT with 5×10⁶ 1045^TREX T cells and BiVax or TriVax 5 days after KPC2 tumor implantation. Animals were evaluated on day 14 (9 days after ACT). (B) KPC2 tumor weights. (C–D) Thy1.1⁺ donor frequency of CD8⁺ cells (C) or number per LN or mg of tissue (D). (E) Diagram depicting T cell differentiation states, including Tcf1⁺Slamf6⁺ T_STEM, Tcf1⁺Slamf6⁺PD-1⁺ T_PEX, and PD-1⁺Lag3⁺ T_EX. (F) Representative histograms (upper) of Tcf1⁺Slamf6⁺ (T_STEM) frequency (lower left) or Tcf1^– frequency (lower right) of donor CD8⁺ cells. (G) Representative histograms (upper) of PD-1⁺ (T_PEX) frequency of donor T_STEM (lower). (H) Representative histograms (upper) of Klrg1⁺ (T_EFF) frequency of donor CD8⁺ cells (lower). (I) Representative histograms (upper) of PD-1⁺ frequency (lower left) or PD-1⁺Lag3⁺ (T_EX) frequency (lower right) of donor CD8⁺ cells. (J) Representative plots (upper) depicting IFNγ⁺ frequency (lower left) or IFNγ⁺TNFα⁺ frequency (lower right) of donor CD8⁺ cells within spleen or tumor following restimulation with Msln_406-414. Dots represent individual animals, and n=3–4/group in all panels with mean±SEM shown. ANOVA with Tukey post-test was used for comparisons in all panels. ACT, adoptive cell therapy; Cy, cyclophosphamide; dLN, draining lymph node; IFNγ, interferon-γ; KPC, Kras^LSL-G12D/+;Trp53^LSL-R172H/+;p48-Cre; PDA, pancreatic ductal adenocarcinoma; T_EX, terminally exhausted T cells; T_PEX, progenitor exhausted T cells; T_STEM, self-renewing stem-like progenitor T cells; TNFα, tumor necrosis factor-α.

To examine T cell differentiation patterns, we first evaluated CD62L to quantify central memory (T_CM) formation. Approximately 70% of donor T cells within pancreatic tumor-draining lymph nodes (dLNs) adopted a CD44⁺CD62L⁺ central memory (T_CM) phenotype, whereas only ~5% of tumor-infiltrating donor T cells maintained a T_CM phenotype (online supplemental figure S1A), consistent with differentiation within tumors. We then enumerated T cell subsets differentiating along a trajectory, including self-renewing stem-like progenitor cells (T_STEM), which are thought to upregulate PD-1 and form a progenitor exhausted (T_PEX) subset, followed by differentiation into terminally exhausted T cells (T_EX, figure 1E, gating scheme in online supplemental figure S1B).^{38 46} Donor 1045^TREX T cells residing in dLNs nearly exclusively formed Tcf1⁺Slamf6⁺ T_STEM (figure 1F). Approximately one-third of intratumoral donor cells also formed these precursors, markedly higher than our prior studies of endogenous tumor-specific T cells.³⁸ Although Tcf1 was maintained in donor T cells in dLNs, around 60% downregulated Tcf1 in tumors, again consistent with differentiation in tumors. However, less than 10% of T_STEM formed PD-1⁺ T_PEX in both tissues (figure 1G). Terminally differentiated Klrg1⁺ effectors (T_EFF) were also rare and comprised less than 10% of all donor T cells (figure 1H). While approximately 20% of intratumoral donor T cells expressed PD-1, supportive of some degree of antigen recognition in vivo, PD-1 and Lag3 co-expressing T_EX were not observed (figure 1I).

To examine function, donor 1045^TREX T cells were restimulated with Msln_406-414 cognate peptide. While around 60% of splenic donor T cells produced IFNγ, less than 20% of tumor-infiltrating donor T cells produced IFNγ and/or tumor necrosis factor-α (TNFα, figure 1J). Collectively, these data support that 1045^TREX T cells acquire a dysfunctional yet non-exhausted, precursor state in orthotopic KPC2 tumors.

Genetic deletion of Tgfbr2 in engineered T cells enhances function in vitro

We hypothesized that TGFβ, which is abundant in both murine and human PDA, might prevent 1045^TREX T cell differentiation due to its broad immunosuppressive effects.⁴⁷ Latency-associated peptide (LAP), a pro-domain which can serve as a surrogate for TGFβ,⁴⁸ was significantly enriched within KPC2 orthotopic tumor parenchyma and nearly absent in stromal and peritumoral regions in untreated mice (figure 2A,B). Therefore, tumor cells and/or other cells within tumor nests produce abundant TGFβ.

Figure 2. Genetic deletion of Tgfbr2 in engineered T cells enhances function in vitro. (A) Representative images at 10× or 40× magnification from KPC2 tumors isolated from untreated control mice evaluating CK, LAP, F4/80, and DAPI. The upper right representative image depicts tumor parenchyma and non-tumor areas determined by CK. (B) Box (IQR and median) and whisker (minimum and maximum) plot depicting the frequency of LAP⁺ areas within CK⁺ parenchymal and CK^– non-tumor areas quantified from 2 to 4 KPC2 tumor regions per animal with n=3/group. (C) Representative histograms of proliferation of 1045^TREX WT or KO CD8⁺ T cells stimulated with 1 µg/mL Msln_406-414 for 72 hours as assessed by CTV dilution. (D) Frequency of 1045^TREX WT or KO CD8⁺ T cells stimulated with 1 µg/mL Msln_406-414 by number of divisions. (E) Frequency of 1045^TREX WT or KO CD8⁺ T cells stimulated with 0.1–10 μg/mL Msln_406-414 reaching ≥6 divisions. (F) Number of 1045^TREX WT or KO CD8⁺ T cells stimulated with 1 µg/mL Msln_406-414 by number of divisions. (G) Schematic illustrating activation followed by restimulation with anti-CD3 antibody ±20 ng/mL Tgfβ1 of 1045^TREX WT or KO CD8⁺ T cells to assess for cytokine production. (H–K) IFNγ⁺ (H), TNFα⁺ (I), IFNγ⁺TNFα⁺ (J), or GzmB⁺ (K) frequency of 1045^TREX WT or KO CD8⁺ T cells following restimulation as in (G) and compared with control conditions without restimulation. Representative histograms are shown in lower panels. (L) gMFI of IFNγ⁺ in 1045^TREX WT or KO CD8⁺ T cells. (M) Fold proliferation of KPC2 NIR⁺ tumor cells during co-culture with 1045^TREX WT or KO CD8⁺ T cells at a 5:1 E:T ratio. (C–M) Mean±SD from technical triplicates and are representative of at least three independent experiments. Student’s t-test was used in (D–F) and ANOVA with Tukey post-test was used in (H–M) for comparisons. CK, cytokeratin; CTV, CellTrace Violet; E:T, effector:target ratio; gMFI, geometric mean fluorescent intensity; IFNγ, interferon-γ; LAP, latency-associated peptide; TNFα, tumor necrosis factor-α.

To test the hypothesis that TGFβ constrains engineered T cell differentiation, 1045^TREX mice were crossed with Tgfbr2^Flox/Flox dLck-Cre mice to delete Tgfbr2 in T cells. To validate the knockout, we compared in vitro proliferation of 1045^TREX Tgfbr2^Flox/Flox dLck-Cre^– (WT) to dLck-Cre⁺ (KO) T cells. Expectedly, loss of Tgfbr2 enhanced naïve 1045^TREX CD8⁺ T cell proliferation in response to titrating Msln_406-414 peptide (figure 2C–E). This translated to a significant increase in KO T cell number with ≥7 divisions (figure 2F).

To assess the role of Tgfbr2 in effector cytokine production, we restimulated previously in vitro-activated T cells with anti-CD3 to ensure that triggering of TCR was equivalent and measured cytokine production (figure 2G). Loss of Tgfbr2 in 1045^TREX T cells increased IFNγ, TNFα, and granzyme B (GzmB) production ≥twofold (figure 2H–K). The addition of Tgfβ1 during restimulation reduced cytokine production only in WT T cells. 1045^TREX KO T cells also exhibited a threefold increase in IFNγ production on a per-cell basis (figure 2L). Deletion of Tgfbr2 enhanced cytotoxicity against KPC2 cells, whereas cytotoxicity by WT T cells was modest (figure 2M). These data are consistent with prior studies demonstrating that TGFβ limits CD8⁺ T cell proliferation, cytokine production, and cytotoxicity in other settings.^{49 50}

Abrogating Tgfbr2 fails to overcome the differentiation arrest and dysfunction of engineered T cells infiltrating orthotopic PDA

We next tested the hypothesis that Tgfbr2 is a barrier to T cell differentiation in PDA. ACT was performed in orthotopic KPC2-bearing mice using 1045^TREX WT or KO T cells and BiVax (figure 3A). Therapies similarly reduced KPC2 tumor weights by ~30% compared with control animals (figure 3B). Although WT and KO ACT resulted in similar antitumor effects, frequencies and numbers of 1045^TREX KO T cells were enriched twofold compared with WT counterparts in PDA (figure 3C,D). Approximately 90% of WT and KO donor T cells within dLNs adopted a T_STEM phenotype, and these precursors comprised ~30% of intratumoral donor T cells (figure 3E,F), similar to prior observations (figure 1F). Again, while most T cells infiltrating tumors downregulated Tcf1 (figure 3G), indicative of downstream differentiation from T_STEM, PD-1⁺ T_PEX cells were rare in both dLNs and tumors (figure 3H,I). Klrg1⁺ T_EFF cells were increased in KO compared with WT cells but comprised <10% of KO donor cells in PDA (figure 3J,K), indicative of poor T_EFF development regardless of TGFβ responsiveness. Less than 7% of donor T cells acquired a T_EX phenotype, and T_EX cells were modestly decreased in cells that lacked Tgfbr2 (figure 3L,M). Although low PD-1, Lag3, and Tox were detected in both WT and KO donor cells, supporting some degree of Msln recognition, frequencies were significantly reduced compared with endogenous, antigen-experienced CD44⁺ T cells in untreated mice (figure 3N–S).

Figure 3. Abrogating Tgfbr2 fails to overcome the differentiation arrest and dysfunction of engineered T cells infiltrating orthotopic PDA. (A) Schematic illustrating cyclophosphamide (Cy, 120 mg/kg) followed by 1045^TREX WT or KO ACT (5×10⁶ T cells) plus BiVax delivered 5 days after KPC2 tumor implantation. Mice were evaluated on day 15 (10 days after ACT). (B) KPC2 tumor weights. (C–D) Thy1.1⁺ donor frequency of CD8⁺ cells (C), or number per LN or mg of tissue (D). (E–G) Representative plots (E) of Tcf1⁺Slamf6⁺ (T_STEM) frequency (F) or Tcf1^– frequency (G) of donor CD8⁺ cells. (H–I) Representative plots (H) depicting PD-1⁺ (T_PEX) frequency of T_STEM (I). (J–K) Representative plots (J) depicting Klrg1⁺ (T_EFF) frequency of donor CD8⁺ cells (K). (L–M) Representative plots (L) depicting PD-1⁺Lag3⁺ (T_EX) frequency of donor CD8⁺ cells (M). (N–O) Representative histograms (N) of PD-1⁺ frequency of endogenous CD44⁺CD8⁺ T cells in untreated (Untx) mice or Thy1.1⁺ 1045^TREX donor CD8⁺ cells (O) in KPC2-bearing animals treated with ACT. (P–Q) Representative histograms (P) of Lag3⁺ frequency of endogenous CD44⁺CD8⁺ T cells or donor CD8⁺ cells (Q). (R–S) Representative histograms (R) of Tox⁺ gMFI of endogenous CD44⁺CD8⁺ T cells or donor CD8⁺ cells (S). (T–V) Representative plots (T) depicting IFNγ⁺ (U) or IFNγ⁺TNFα⁺ (V) frequency of intratumoral donor CD8⁺ cells following restimulation with Msln_406-414 or CSC. All results are compiled from two independent experiments with dots representing individual animals, n=4–8/group, and mean±SEM shown. ANOVA with Tukey post-test was used in (B,O,Q,S) and Student’s t-test was used for comparisons in all remaining panels. ACT, adoptive cell therapy; CSC, Cell Stimulation Cocktail; Cy, cyclophosphamide; dLN, draining lymph node; gMFI, geometric mean fluorescent intensity; IFNγ, interferon-γ; TNFα, tumor necrosis factor-α.

While most donor T cells residing in spleens produced IFNγ and/or TNFα in response to Msln_406-414 (figure 3T–V), less than 20% of tumor-infiltrating 1045^TREX WT or KO donor cells produced IFNγ and/or TNFα in response to restimulation. This was modestly augmented in KO cells and overcome by non-specific activation. Thus, although abrogating Tgfbr2 signaling enhances in vitro T cell functionality and expansion, it fails to promote T cell differentiation and effector function in KPC2 orthotopic tumors.

Tgfbr2 deletion promotes engineered T cell interactions with intratumoral cDC1

Type I DCs (cDC1s) are essential for cross-presenting tumor-associated antigens to prime and reactivate T cells.^{37 51} To determine if TCR1045 T cells could associate with cDC1s in orthotopic PDA, we used CRISPR/Cas9 to delete Tgfbr2 in activated 1045^TREX-GFP T cells (figure 4A,B). KPC2 mScarlet⁺ orthotopic tumors were established in Xcr1-Venus mice, which permitted concurrent visualization of donor T cells, tumor cells, and cDC1s. ACT was performed with Cy followed by 1045^TREX-eGFP Tgfbr2^WT or Tgfbr2^CRISPR T cells. All recipients were treated with TriVax, which includes agonistic αCD40 to promote antigen-presenting cell (APC) activation. Therefore, any differences detected would be attributable to loss of Tgfbr2 in T cells rather than the vaccine regimen itself. Tumors were evaluated by two-photon and SHG imaging on day 14 (figure 4C,D, online supplemental video). The overall motility of intratumoral 1045^TREX-eGFP Tgfbr2^CRISPR T cells was reduced compared with Tgfbr2^WT counterparts, as measured by mean distance traveled and displacement (figure 4E). Characteristics of Tgfbr2-deleted T cell dynamics included modestly lower movement speed, increased directional changes, and reduced confinement ratios (ie, a measure of how efficiently a migrating cell travels away from its starting point; figure 4F). These data suggest that TGFβ may interfere with T cell interactions with other cells in the TME. Conversely, cDC1s in recipients of Tgfbr2^CRISPR T cells exhibited increased motility, distance traveled, displacement, and movement speed (figure 4G,H), suggesting loss of Tgfbr2 in TCR-T cells increases intratumoral cDC1 migration and sampling of the TME. Thus, T cell-intrinsic deletion of Tgfbr2 results in altered T cell and DC motility, potentially sustaining interactions between the two cell types.

Figure 4. Tgfbr2 deletion promotes engineered T cell interactions with intratumoral cDC1. (A) Inference of CRISPR Editing analysis from representative CRISPR/Cas9-mediated deletion of Tgfbr2 in 1045^TREX T cells. (B) Tgfbr2 histograms from representative preparation of 1045^TREX Tgfbr2^CRISPR cells with secondary antibody only (upper histogram) and Cas9 only edited cells (lower histogram) as controls. (C) Schematic illustrating cyclophosphamide (Cy, 180 mg/kg) followed by 5×10⁶ 1045^TREX-eGFP Tgfbr2^WT (WT) or Tgfbr2^CRISPR (CRISPR) T cell ACT with TriVax in Xcr1-Venus mice 5 days after implantation of KPC2 mScarlet⁺ tumors. Two-photon imaging was performed on day 12 (7 days after ACT). (D) Representative still frames of two-photon (see online supplemental video) from animals treated with 1045^TREX-eGFP Tgfbr2^WT or Tgfbr2^CRISPR ACT. White arrowheads show cDC1s contacting donor T cells. (E) Motility coefficients (left), total distance traveled (middle), and displacement (right) of donor 1045^TREX-eGFP Tgfbr2^WT or Tgfbr2^CRISPR CD8⁺ cells. (F) Mean speed (left), mean directional change, and confinement ratio of donor CD8⁺ cells. (G) Motility coefficients (left), total distance traveled (middle), and displacement (right) of cDC1s in mice treated with 1045^TREX-eGFP Tgfbr2^WT or Tgfbr2^CRISPR ACT. (H) Mean speed (left), mean directional change, and confinement ratio of cDC1s. (I) Representative images of live-cell imaging of BMDCs co-cultured with 1045^TREX Tgfbr2^WT or Tgfbr2^CRISPR T cells. Upper panels show T cells contacting DCs with insets shown at higher magnification. Lower panels show representative Ca²⁺ flux assessed by Rhod-2 AM over time with cells of interest indicated by arrowheads. (J) Fold change in Ca²⁺ in T cells contacting DCs from live-cell imaging. (K) T cell:DC contact duration from live-cell imaging. (L) Gene set enrichment analysis from RNA sequencing of DCs. Dots represent individual cells in all panels with mean±SEM shown. Two-photon measurements are derived from n=3/group. Live-cell imaging shows measurements from one of two representative experiments. Student’s t-test was used for comparisons in (E–H) and Welch’s t-test was used in (J–K). BMDC, bone marrow-derived dendritic cell; Cy, cyclophosphamide; cDC1, type I dendritic cell; SHG, second harmonic generation.

Given cDC1s in the two-photon in vivo studies may be induced and/or activated by αCD40, we performed in vitro live-cell imaging of Msln_406-414-pulsed BMDCs co-cultured with 1045^TREX Tgfbr2^WT or Tgfbr2^CRISPR T cells. Tgfbr2^CRISPR T cells exhibited increased calcium flux and prolonged contact duration with DCs (figure 4I–K), supporting more pronounced immune synapse formation. Additionally, RNA sequencing was performed on BMDCs following 24 hours of co-culture to evaluate DC-T cell crosstalk. Functional gene set enrichment analysis revealed that BMDCs exposed to Tgfbr2^CRISPR T cells exhibited significantly increased inflammatory signatures (IFNγ, interferon-α, inflammatory response, TNFα signaling via NFκB), mitosis, protein secretion, and apical junctions, while demonstrating reduced oxidative phosphorylation (figure 4L). Overall, these data support that Tgfbr2 deletion in TCR-T cells increases cDC1 environmental sampling while slowing donor T cell motility, resulting in prolonged cDC1-T cell interactions and improving T cell and DC activation.

Strategies to enhance TCR1045 recognition of KPC cells in vitro

KPC tumor cells express low constitutive MHC-I that can be augmented by IFNγ.^{8 9 35} Although TCRs can be sensitive to few pMHC molecules on APCs, we hypothesized that low pMHC-I by KPC cells limits TCR1045 cell antigen recognition. We first validated our prior findings³⁵ and confirmed that KPC2 cells express abundant Msln (figure 5A). Next, 1045^TREX WT or KO T cells were co-cultured with KPC2 tumor cells and assessed for upregulation of TCR signaling markers CD25 and CD69 (figure 5B). Although native antigen recognition was modest, loss of Tgfbr2 significantly increased recognition from 11% to 39% (figure 5C), indicating that KPC2 cells endogenously process and present Msln_406-414 and targeting Tgfbr2 can amplify antigen recognition. Enhancing MHC-I via IFNγ significantly increased 1045^TREX WT target recognition, further reinforcing that Msln is endogenously processed. GzmB production was similar between WT and KO T cells (figure 5D).

Figure 5. Strategies to enhance TCR1045 recognition of KPC cells in vitro. (A) Intracellular (IC) and membrane-bound (MB) Msln protein staining in KPC2 PDA cells by flow cytometry with secondary antibody only staining as a negative control. Plots are representative of at least three independent experiments. (B) Schematic illustrating preparation of 1045^TREX WT and KO T cells along with KPC tumor cell lines for co-culture. T cells were activated in vitro with Msln_406-414 for 48 hours followed by a 24-hour rest period. Tumor cells were plated ±100 ng/mL IFNγ for 24 hours and pulsed ±5 µg/mL Msln_406-414 for 1 hour prior to co-culture. Prepared T cells were plated at a 5:1 E:T ratio and co-cultured for 24 hours. (C–D) Representative plots (upper) from co-cultures with KPC2 cells depicting CD25⁺CD69⁺ (C) or GzmB⁺ (D) frequency of 1045^TREX WT and KO CD8⁺ cells (lower). (E) Schematic showing generation of KPC2^MslnER, which is engineered to produce Msln_406-414 peptide targeted to the endoplasmic reticulum (ER). (F–G) Representative plots (upper) from co-cultures with KPC2^MslnER cells depicting CD25⁺CD69⁺ (F) or GzmB⁺ (G) frequency of 1045^TREX WT and KO CD8⁺ cells (lower). (C–D, F–G) show mean±SD from technical triplicates and are representative of at least three independent experiments. ANOVA with Tukey post-test was used for all comparisons. E:T, effector:tEffector:Target ratio; GzmB, granzyme B; IFNγ, interferon-γ.

To determine if pMHC-I was limiting, KPC2 cells were engineered to express Msln_406-414 targeted to the endoplasmic reticulum (ER) to create ‘KPC2^MslnER’ cells (figure 5E). Due to low constitutive MHC-I in KPC cells, it remained uncertain if the ‘PresentER’ system, which results in physiologic levels of antigen presentation,³⁶ could generate sufficient peptide to stabilize cell surface MHC-I. KPC2^MslnER cells were co-cultured with 1045^TREX WT or KO T cells as in figure 5B. This modestly increased CD25 and CD69 in WT T cells from 11% to 25% and KO T cells from 39% to 45% (p=0.003 and 0.034, respectively, figure 5C, F). GzmB production was similarly elicited in T cells between KPC2 and KPC2^MslnER, with or without IFNγ pretreatment (figure 5D, G). The increased KPC recognition granted by increased Msln_406-414 epitope remained significantly inferior to maximal responses provoked by pulsing tumor cells with supraphysiologic levels of soluble peptide with or without IFNγ, which resulted in 60–90% of T cells co-expressing CD25 and CD69 and producing GzmB. Similar findings were observed with KPC271 and KPC451 derived from independent KPC mice (online supplemental figure S2). Thus, pMHC-I is limiting in KPC tumor cells and only incrementally restored by the KPC2^MslnER system.

Msln_406-414 overexpression increases PDA-intrinsic MHC-I

To assess how the PresentER system influenced MHC-I in KPC2 cells, we measured H-2D^b and H-2K^b following titrating IFNγ. Both H-2D^b and H-2K^b were nearly undetectable in untreated KPC2 cells by antibody staining (figure 6A,B). Poor constitutive MHC-I was also observed in KPC271 and KPC451 cells (online supplemental figure S3). While modest, constitutive H-2D^b and H-2K^b were significantly increased in KPC2^MslnER cells. Both MHC-I alleles were further increased in KPC2^MslnER cells as compared with KPC2 cells after exposure to low IFNγ concentrations (≥0.1 ng/mL, figure 6C). Thus, increasing Msln_406-414 availability stabilizes cell surface pMHC-I.⁵² Moreover, low constitutive MHC-I appears to be a conserved feature of KPC cells, consistent with other reports.^{35 53 54}

Figure 6. Msln_406-414 overexpression increases PDA-intrinsic MHC-I in vivo. (A) Representative histograms of H-2D^b (left) or H-2K^b (right) in untreated or IFNγ-treated KPC2 or KPC2^MslnER cells. Fluorescence minus one (FMO) controls are shown. (B) H-2D^b (left) or H-2K^b (right) gMFI on untreated KPC2 or KPC2^MslnER cells. (C) H-2D^b (left) or H-2K^b (right) gMFI on KPC2 or KPC2^MslnER cells pretreated with IFNγ. (D) Representative images of cytokeratin (CK) and MHC-I within autochthonous KPC pre-invasive or invasive lesions, and orthotopic KPC2 or KPC2^MslnER tumors from untreated mice. Higher magnifications are shown in right panels from insets (white boxes). White arrowheads show MHC-I colocalized on CK⁺ tumor cells. (E) Box (IQR and median) and whisker (minimum and maximum) plot depicting frequency of MHC-I⁺ areas within CK⁺ tumors from n=3/group with MHC-I quantified in 1–4 tumor areas per animal. Dots are replicates in (B–C) with mean±SD shown and are representative of at least three independent experiments. Student’s t-tests were used in (B–C) and ANOVA with Dunnett’s post-test was used in (E) for comparisons. gMFI, geometric mean fluorescent intensity; IFNγ, interferon-γ; MHC-I, major histocompatibility complex-I; PDA, pancreatic ductal adenocarcinoma.

In light of our prior work demonstrating that TCR1045 cells are effective in autochthonous PDA,¹⁵ we then interrogated MHC-I levels in orthotopic and autochthonous tumors. Orthotopic KPC2, KPC2^MslnER, and autochthonous KPC tumors were co-stained for H-2D^b and H-2K^b in a single fluorescent channel and assessed by immunofluorescent (IF) microscopy using cytokeratin (CK) to define tumor cells (figure 6D,E). In autochthonous tumors, low but detectable MHC-I was observed in pre-invasive and invasive regions and confined to non-tumor (CK^–) cells residing within tumor stroma. In contrast, stromal cells were sparse in KPC2 tumors and MHC-I was only detectable on ~5% of CK⁺ areas. MHC-I was significantly increased in KPC2^MslnER tumors in both the stroma and within CK⁺ tumor cells. Thus, while constitutive tumor cell MHC-I is low in KPC models, MHC-I can be augmented by modest epitope overexpression in vivo. These models therefore allow for examination of PDA with disparate pMHC-I levels, reflecting variable MHC-I observed in human disease.^{7–11 53}

Augmenting pMHC-I in tumor cells enables Tgfbr2-deficient TCR-T cell therapeutic efficacy

To test the hypothesis that pMHC-I is a barrier to the efficacy of Tgfbr2-deleted T cells in vivo, we performed 1045^TREX WT and KO T cell ACT in mice with orthotopic KPC2^MslnER tumors. Tumor-bearing mice were treated with Cy followed by ACT without vaccination (figure 7A). Tumor growth was reduced in mice treated with ACT and further suppressed when 1045^TREX KO cells were employed (figure 7B,C). Survival was also significantly prolonged by 1045^TREX KO ACT (figure 7D). Nine days after transfer, 1045^TREX KO cell frequency (twofold) and number (fivefold) in PDA were increased compared with WT T cells (figure 7E). Tumor-infiltrating CD8⁺ T cells were also augmented >twofold in recipients of 1045^TREX KO cells by orthogonal IF imaging (figure 7F,G). Additionally, the frequency of CD8⁺ T cells within 20 µm of CD11c⁺ APCs was doubled in recipients of KO versus WT 1045^TREX T cells (figure 7H), reinforcing that abrogating TGFβ signaling may increase and/or prolong DC-T cell interactions. Intratumoral CD11c⁺ APCs were significantly increased >twofold in KPC2^MslnER tumors compared with KPC2 tumors derived from untreated mice (online supplemental figure S4A,B). Therefore, modestly increasing antigenicity through the PresentER system may activate endogenous T cells that produce factors driving APC accumulation. ACT with 1045^TREX KO T cells also decreased intratumoral LAP by nearly threefold (online supplemental figure S4C,D), supporting decreased TGFβ production within the TME.

Figure 7. Augmenting pMHC-I in tumor cells enables Tgfbr2-deficient TCR-T cell therapeutic efficacy. (A) Schematic illustrating cyclophosphamide (Cy, 120 mg/kg) followed by 1045^TREX WT or KO ACT (5×10⁶ T cells) delivered without vaccine 6 days after KPC2^MslnER tumor implantation. Tumor growth and survival were monitored, and additional animals were evaluated on day 15 (9 days after ACT). (B) Representative ultrasound images of KPC2^MslnER tumors from mice treated with 1045^TREX WT or KO ACT on day 16. (C) Tumor volumes from untreated (Untx) control or treated mice. (D) Survival of control and treated mice. (E) Representative plots (left) depicting Thy1.1⁺ donor frequency of CD8⁺ cells (middle) or number per LN or mg of tissue (right). (F) Representative images from tumors of mice treated with 1045^TREX WT or KO ACT on day 15 evaluating LAP, CD8, and CD11c. Right panels show higher magnification of insets (white boxes) with arrowheads showing CD8⁺ T cells in close proximity to CD11c⁺ cells. (G) CD8⁺ cells per mm². (H) Frequency of CD8⁺ cells within 20 µm of a CD11c⁺ cell. (I) Representative plots (left) of PD-1⁺ (T_PEX) frequency of 1045^TREX WT or KO CD8⁺ T_STEM (right). (J) Representative plots (left) of Klrg1⁺ (T_EFF) frequency of 1045^TREX WT or KO CD8⁺ donor cells (right). (K) Representative plots (left) depicting PD-1⁺Lag3⁺ (T_EX) frequency of donor CD8⁺ cells (right). (L) Representative plots (left) depicting IFNγ⁺ frequency of donor CD8⁺ cells restimulated with Msln_406-414 or Cell Stimulation Cocktail (CSC, right). (C–D) Two independent experiments and contain n=4/group. (E, I–L) Compiled from two independent experiments with dots representing individual animals, n=9/group, and mean±SEM shown. (G–H) Box (median and IQR) and whisker (minimum and maximum) plots include n=3/group with cells quantified from 9–18 high-power fields per tumor by two independent scorers. For comparisons, ANOVA with Tukey post-test was used in (C), Cox regression was used in (D), Welch’s t-test was used in (G–H), and Student’s t-test was used in all remaining panels. ACT, adoptive cell therapy; dLN, draining lymph node; IFNγ, interferon-γ; LAP, latency-associated peptide; pMHC-I, peptide-major histocompatibility complex-I; TCR-T, T cell receptor-engineered T cell.

1045^TREX WT and KO T_STEM were enriched in dLNs and comprised <5% of intratumoral transferred T cells (online supplemental figure S4E). Tcf1 downregulation was modestly increased in KO T cells residing in dLNs, similar to other reports,⁵⁵ and Tcf1 was rarely expressed by tumor-infiltrating T cells. T_PEX readily formed in both tissues within KPC2^MslnER-bearing mice (figure 7I). While just 10% of intratumoral WT donor T cells differentiated into Klrg1⁺ T_EFF, this population increased fourfold when Tgfbr2 was disrupted (figure 7J). Additionally, PD-1⁺Lag3⁺ T_EX constituted 40% of WT T cells and were modestly reduced in KO T cells (figure 7K), unlike prior observations where T_EX failed to form (figures 1 and 3). T_EX co-expressing PD-1, Lag3, and Tox were also detected in tumors (online supplemental figure S4F). The three exhaustion-associated markers were prevalent on intratumoral WT and KO donor T cells and only slightly reduced compared with endogenous CD44⁺ T cells in untreated mice (online supplemental figure S4G–I). Functionally, loss of Tgfbr2 markedly augmented IFNγ production in response to restimulation with Msln_406-414 in both splenic and intratumoral donor T cells (figure 7L). More than 70% of splenic 1045^TREX WT and KO T cells maintained function in response to peptide restimulation, which was further enhanced by Tgfbr2 deletion. Whereas only 25% of intratumoral WT cells produced IFNγ on restimulation, nearly 80% of intratumoral KO T cells produced IFNγ. Thus, canonical TCR-T cell differentiation was restored by incrementally enhancing pMHC-I, and loss of Tgfbr2 promoted accumulation of functional T_EFF within the TME.

To ensure that our observations were not unique to KPC2^MslnER, we employed autochthonous KPC PDA, where we previously established efficacy of TCR1045 T cells.¹⁵ When co-transferred into the same KPC hosts to control for any differences in tumor size (online supplemental figure S5A–C), Tgfbr2 KO T cells more readily differentiated into effector cells and exhibited superior function as compared with the WT TCR1045 counterparts (online supplemental figure S5D,E). Thus, T cell-intrinsic Tgfbr2 deletion enhances the function of T cells targeting a native antigen, and improving antigen presentation in KPC cells can recapitulate findings in autochthonous PDA, thereby providing a valuable tool for the field.

Having identified that modest improvements to antigen presentation can unveil the potency of Tgfbr2 KO TCR1045 cells, we turned to addressing a potential limitation of targeting TGFβ signaling in T cells. TGFβ can promote T cell memory, including central memory and tissue-resident memory T cells (T_RM).⁵⁶ Moreover, T_RM signatures correlate with immunotherapy outcomes in cancer patients.⁵⁷ As the efficacy of ACT is associated with T cell persistence and acquisition of memory traits,⁵⁸ use of T cells lacking Tgfbr2 could interfere with memory establishment and undermine ACT durability. Such characteristics would be best studied by rechallenging previously cured animals at a memory timepoint. However, neither scenario was feasible due to the aggressive and incurable nature of orthotopic KPC models when employing TCR-T cell therapy. Thus, we compared the kinetics and persistence of 1045^TREX WT and KO T cells co-transferred into identical non-tumor-bearing hosts in the experimental setting of ACT that included vaccination with BiVax (online supplemental figure S6A). During the 6 weeks following ACT, circulating 1045^TREX KO T cells were enriched 10-fold to 20-fold over WT T cells and comprised ~80% of CD8⁺ T cells (online supplemental figure S6B–D). Frequencies of circulating 1045^TREX KO memory phenotype T_CM (CD127⁺CD62L⁺) were modestly reduced compared with WT donor cells, and replaced by effector memory (T_EM, CD127⁺CD62L^–) and T_EFF (CD127^–CD62L^–) phenotypes (online supplemental figure S6E–H). However, numbers of circulating 1045^TREX KO cells adopting T_CM, T_EM, and T_EFF phenotypes were all increased 10- to 100-fold. Thus, while loss of Tgfbr2 on TCR-T cells mildly biases away from T_CM by proportion, this is compensated by the substantial proliferative and/or survival advantage conferred by abrogation of TGFβ signaling.

We next enumerated tissue-resident 1045^TREX T cells within inguinal LN, spleen, pancreas, lung, liver, and intraepithelial SI (SI-IEL) 8 weeks after co-transfer (online supplemental figure S7A). Parenchymal 1045^TREX KO T cells, identified by a lack of CD45⁺ intravenous labeling, were ~10-fold more abundant than WT donor T cells in all tissues apart from SI-IEL (online supplemental figure S7B). Most (>98%) resident CD8⁺ T cells in SI-IEL were host-derived, indicating that TCR-T cells failed to infiltrate and/or persist in the SI-IEL, irrespective of Tgfbr2. We then compared T_RM-associated markers CD69, CD49a, Cxcr6, and CD103 on donor T cells. Endogenous CD8⁺ T cells within SI-IEL served as a positive control given nearly all CD8⁺ T cells within SI-IEL are T_RM.⁵⁹ CD69⁺, CD49a⁺, and Cxcr6⁺ frequencies were similar or modestly reduced in KO compared with WT donor cells across all tissues apart from SI-IEL (online supplemental figure S7C–E). All three T_RM-associated markers were reduced 30%–50% in 1045^TREX KO cells within the SI-IEL, consistent with studies demonstrating that TGFβ contributes to intestinal T_RM.⁶⁰ Expectedly, CD103 was undetectable on tissue-residing 1045^TREX KO T cells (online supplemental figure S7F) given TGFβ induces CD103.⁵⁶ Collectively, these data support that TGFβ is not required for the development of phenotypic T_CM and non-intestinal T_RM during ACT that includes vaccination. However, TGFβ signaling does sustain intestinal T_RM. Additionally, T_RM can form in the absence of CD103, despite its known role to promote tissue retention through binding of E-cadherin⁵⁶ and consistent with some prior studies.^{61 62} Therefore, protecting TCR-T cells from TGFβ promotes effector differentiation and function while largely preserving T cell memory phenotypes, at least in tissues apart from the digestive tract.

Discussion

Here, we extend our prior evaluations of TCR1045 T cells targeting Msln to orthotopic KPC models to improve efficiency in identifying strategies to enhance T cell-based therapies for PDA. We demonstrate that KPC cells display insufficient pMHC-I to promote canonical T cell differentiation in vivo, despite eliciting TCR1045 T cell activation in vitro. The in vivo differentiation trajectory of TCR1045 T cells toward T_EFF and T_EX states is restored by modestly increasing tumor cell production of the Msln_406-414 epitope. Increasing Msln_406-414 using the PresentER system modestly but significantly increases MHC-I on KPC cells. These data support that limited pMHC-I by KPC cells is a major barrier to CD8⁺ T cell therapeutics and likely extend to most MHC-I targets, reinforcing rationale for enhancing MHC-I in human PDA where expression is heterogeneous.^{7–11 53}

In lymph nodes, TCR1045 T cells adopt a PD-1^–Tcf1⁺Slamf6⁺ stem-like state and fail to form PD-1⁺ T_PEX. In KPC orthotopic tumors, TCR1045 T cells readily downregulate CD62L and Tcf1 but fail to further differentiate into T_EX. Despite lacking a canonical T_EX phenotype, TCR1045 T cells exhibit poor functionality in response to cognate antigen. Thus, additional mechanisms beyond the canonical T cell exhaustion paradigm driven by persistent TCR signaling promote T cell dysfunction in PDA. Given modestly augmenting Msln_406-414 in tumor cells restored canonical TCR1045 T_PEX and T_EX formation, insufficient TCR signaling within the suppressive TME of PDA may also drive T cells into a dysfunctional state. As T_PEX have garnered substantial interest following their identification as a primary target of PD-1/PD-L1 pathway blockade,^{38 46} a lack of sufficient T_PEX formation due to insufficient tumor antigenicity may promote resistance to such immune checkpoint blockade strategies in patients.

While increasing tumor Msln_406-414 epitope was sufficient to rescue T_EX formation, it failed to promote T_EFF differentiation. Here, we show that TGFβ is a major barrier to T_EFF differentiation of TCR1045 T cells when pMHC-I is not limiting. These data broadly align with prior studies targeting TGFβ in various cancers.^21–33 Tgfbr2-deficient TCR1045 T cells decreased intratumoral LAP, a surrogate for TGFβ production, consistent with enhanced T cell inflammatory cytokine production and potential suppression of TGFβ production. Disrupting Tgfbr2 also decreased TCR1045 T cell motility while markedly increasing cDC1 motility and augmenting inflammatory gene signatures in BMDCs. Thus, T cell-intrinsic deletion of Tgfbr2 results in bidirectional crosstalk with DCs that enhances both cell subsets and could partially underlie improved T cell functionality. Increased numbers of T cells lacking Tgfbr2 adopted memory and tissue residency phenotypes during an ACT protocol, further supporting the value of protecting T cells from TGFβ. Moreover, targeting Tgfbr2 in TCR1045 cells did not result in overt adverse events during experimental windows, including animals observed for 8 weeks in which transferred Tgfbr2 KO cells comprised over 80% of the CD8⁺ T cell compartment.

Although abrogating TGFβ signaling in TCR-engineered T cells can amplify donor T cell expansion, infiltration, function, and interactions with DCs without compromising T cell memory, our findings call for additional strategies that augment pMHC-I to maximize the therapeutic efficacy of TCR-engineered T cells. Targeting mitogen-activated protein kinase-dependent autophagy,⁹ Wnt,⁶³ transcription enhancer factor 1,⁶⁴ or epigenetic regulators^{11 65} are all promising approaches to increase Msln and/or MHC-I. Alternatively, focal adhesion kinase inhibitors or cGAS/STING pathway agonists could be used to nonspecifically promote antigen processing and cross-presentation.^{66 67} RAS inhibitors, which have antitumor effects in metastatic PDA, can remodel the TME, and partially depend on T cells for efficacy,^{4 68–71} represent an additional translational approach to enhance immune-based therapeutics for PDA patients.

Data availability statement

Data sharing not applicable as no datasets were generated and/or analyzed for this study.