Differential Effects of Depleting versus Programming Tumor-Associated Macrophages on Engineered T Cells in Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDA) has a dismal 5-year survival rate of 9% (1, 2). The disease is resectable in 10% to 20% of patients, and aggressive and toxic treatments, including chemotherapy and radiotherapy, are rarely curative. Immunotherapy that augments endogenous immune system activity, such as immune-checkpoint blockade, can be effective in other advanced malignancies, yet has not been successful in most PDA patients (3, 4). Resistance to immune-checkpoint blockade may reflect, in part, too few neoantigens (5). Infusion of T cells engineered to express tumor-specific T-cell receptors (TCR) or chimeric antigen receptors has yielded some therapeutic responses in patients with cancer, and may be particularly useful for tumors with few neoantigens. We previously showed (6) that CD8⁺ T cells engineered to express a high-affinity TCR specific to mesothelin (TCR_Msln) infiltrate pancreatic tumors and prolong survival in an autochthonous PDA mouse model (Kras^G12D/+;Trp53^R172H;p48-Cre, denoted as KPC; ref. 7). The KPC model has been predictive of therapeutic responses in patients (reviewed in refs. 8, 9).

Mesothelin (Msln) is a self-antigen that has low expression in mesothelial cells that line vital organs, including the lung and the heart (10), and in fibroblasts during inflammation (11). Msln has high expression in pancreatic tumor cells (6, 12), and therapy with TCR_Msln CD8⁺ T cells targets the tumor specifically, without overt toxicities to normal tissues (6). We have also isolated corresponding human TCRs for clinical translation (6). However, because infused TCR_Msln cells in the KPC model become progressively dysfunctional in the tumor and contract in number over time (6), repeated T-cell infusions are administered to achieve therapeutic benefit (6) and strategies to modulate the tumor microenvironment (TME) could enhance potency.

PDAs are notorious for robust desmoplasia, orchestrated largely by activating mutations in the Kras proto-oncogene. Myeloid cells, particularly tumor-associated macrophages (TAM), predominate in the tumor stroma (13–15). TAMs often express immunosuppressive factors and inhibitory ligands, support tumor angiogenesis, and inhibit endogenous T cells (16). Nevertheless, we have found that T cells colocalize with TAMs in situ in human PDA, and the presence of T-cell infiltrates correlates positively with TAM numbers (15). Thus, modulating TAMs could potentially be leveraged to enhance T-cell–based therapies.

In healthy tissues, macrophage homeostasis is maintained by macrophage colony-stimulating factor (Csf1), which promotes differentiation of hematopoietic stem cells toward the myeloid lineage during development and inflammation (17). Csf1 binds the receptor Csf1R, inducing phosphorylation and activation of several signaling pathways, including Mapk and Stat3, to promote myeloid cell survival and proliferation. Csf1R signaling can also promote immune tolerance to transplantation antigens (18) and Csf1R blockade depletes TAMs and enhances endogenous T-cell antitumor activity in several mouse cancer models (19, 20). Targeting this pathway is in early-stage clinical trials and has exhibited antitumor activity in diffuse-type tenosynovial giant cell tumors (21).

Changing the functionality of TAMs in tumors from a suppressive state to an antitumor state (TAM programming) could be a promising alternative to TAM depletion for cancer therapy. Beatty and colleagues showed that anti-CD40 can increase recruitment of TAM precursors with a direct tumoricidal function into PDA (13). Anti-CD40 did not reprogram suppressive TAMs within the tumor, but instead recruited tumoricidal TAM progenitors. Unexpectedly, the antitumor effects of anti-CD40 were independent of endogenous T cells. Agonistic anti-CD40 also improves response to immune-checkpoint blockade in pancreatic cancer models (22), potentially reflecting recruitment and activation by anti-CD40 of Batf3⁺ dendritic cells (DCs; ref. 23).

Here, we investigated the impact of TAM depletion using Csf1R blockade or TAM programming with agonistic anti-CD40 in combination with TCR_Msln-engineered cell therapy in KPC mice. The results demonstrated that TAM depletion diminished the antitumor activity of infused effector CD8⁺ T cells, whereas TAM programming enhanced the accumulation and longevity of TCR_Msln-engineered cells but still failed to overcome engineered T-cell dysfunction in the TME. The results support both the safety and clinical potential of anti-CD40 and engineered T-cell therapy for PDA patient treatment, yet, also highlight the potential for immune modulation that impact endogenous vs. adoptively transferred T cells distinctly, as well as the need for further investigation into fundamental mechanism(s) governing antigen-specific T-cell dysfunction in pancreatic cancer.

The Fred Hutchinson Cancer Research Center (FHCRC), University of Washington, and the University of Minnesota Institutional Animal Care and Use Committees approved all animal studies. Kras^LSL-G12D/+;Trp53^LSL-R172H/+;p48Cre (KPC) mice (Fred Hutchinson Cancer Research Center) mice initially generated on a mixed background (7) and were serially backcrossed to C57Bl/6J mice (The Jackson Laboratory; Stock No 00664) and determined to be >99.6% pure by SNP analyses as previously described (6). The animals were housed at the Fred Hutchinson Cancer Research Center and University of Minnesota. P14 mice (24), which are of C57Bl/6J origin, were bred to the Thy1.1⁺ congenic strain (The Jackson Laboratory; Stock 000406) to generate P14 Thy1.1^+/+ or P14 Thy1.1^+/− T cells as a source for engineering and housed at the University of Washington and University of Minnesota.

Surgically resected PDA specimens were obtained from patients (n = 11) who provided written informed consent by NWBioTrust (Department of Pathology, University of Washington School of Medicine). PDA samples were selected from patients who underwent surgery without neoadjuvant therapy, and this study cohort has been previously described (15). Resected tumor and normal adjacent (Nadj) pancreas were collected immediately following surgery and placed into ice-cold complete media (DMEM containing 10% FBS + 1% penicillin/streptomycin). Tissues were subdivided for flash frozen (RNA/DNA) or were further processed to generate single-cell suspensions of mononuclear cells. In the latter case, tumors were weighed, minced with a razor blade, incubated for 30 minutes in collagenase (1 mg/mL), mechanically filtered (70 μm), and washed twice in complete media. Single-cell suspensions were cryopreserved in 10% DMSO and stored at −80°C for subsequent analysis. Tissues for gene-expression analyses were flash frozen in liquid nitrogen and stored at −80°C for further processing. All studies using human specimens were approved by the FHCRC Institutional Review Board and conducted according to the principles expressed in the Declaration of Helsinki.

A retroviral vector containing the high-affinity mesothelin-specific (1045) TCR was generated as previously described (6). Briefly, 2.2 × 10⁶ Platinum-E (Plat-E, ATCC) retroviral packaging cells were plated on 10-cm tissue culture–treated plates in Plat-E media [DMEM (Gibco), 10% FBS (Gibco), 2% L-glutamine (Sigma), 1% penicillin/streptomycin (Sigma), blasticidin (10 μg/mL, Sigma), puromycin (1 μg/mL, Sigma)] for 24 hours at 37°C, 5% CO₂. On day 2, Plat-E cells were transfected with the MIGRI-TCR1045α−P2A-TCR1045β plasmid using Effectene (Qiagen). The MIGRI plasmid has been previously described (25). The cloning of the high-affinity 1045 TCR has been previously described (6). On day 3, Plat-E media were replaced with T-cell media (DMEM, 10% FBS, 2% L-glutamine, 1% penicillin/streptomycin, β-mercaptoethanol), and cells were further incubated at 32°C, 5% CO₂. On days 4 and 5, viral supernatants were harvested and passed through a 0.45-μm filter (Sigma) for immediate use.

Single-cell suspensions of mononuclear cells from female P14 Thy1.1⁺ spleens were generated by mechanical disruption and red blood cell lysis (ACK). Splenic mononuclear cells were stimulated in vitro with anti-CD3 (1 μg/mL; clone 145-2C11, BD Biosciences) and anti-CD28 (1 μg/mL; clone 37.51, BD Biosciences) in 10 mL of complete T-cell media containing recombinant human IL2 (rIL2, 50 U/mL) upright in T25 flasks at 37°C, 5% CO₂. On days 1 and 2 after stimulation, bulk splenocytes containing activated T cells were transduced with the MIGRI-TCR1045α−P2A-TCR1045β retrovirus by spinfection in 12-well plates containing polybrene (10 μg/mL) and rIL2 (50 U/mL) for 90 minutes at 1,000 × g at 30°C as described (6). On day 5, T cells were screened for transduction efficiency by flow-cytometric staining with CD8-e450 (clone 53-6.7; BD Biosciences), Thy1.1-PerCP (clone OX-7; BD Biosciences), Vβ9-PE (clone MR10-2; BD Biosciences), and/or an Msln_406-414-H-2D^b-APC tetramer generated by the FHCRC Immune monitoring core. On day 7 after in vitro stimulation, transduced T cells were restimulated in vitro with peptide-pulsed, irradiated splenocytes from B6 mice (3500 R) at a 5:1 APC to T-cell ratio in the presence of rIL2 (50 U/mL). All T-cell cultures were supplemented with rIL2 (50 U/mL) every other day for the duration of in vitro culture. On day 5 after the second stimulation in vitro, >90% of the CD8⁺Thy1.1⁺ T cells expressed TCR_MSLN. T cells were resuspended in sterile saline and infused into mice as described below.

We previously described a protocol optimized to promote the expansion of mesothelin TCR-engineered cells in mice (6). Briefly, KPC mice are enrolled in treatment studies when they achieve 3–6 mm pancreatic tumors as determined by serial monitoring with high-resolution ultrasound (Vevo 2100). Enrolled mice received cyclophosphamide (Cy, 180 mg/kg, IP, UW Pharmacy), and 6 hours later received IP 1 × 10⁷ congenic (Thy1.1⁺) CD8⁺ transduced to express the high-affinity TCR_Msln (1045(6)) + 1 × 10⁷ irradiated and peptide (Msln_406-414, GQKMNAQAI)-pulsed syngeneic splenocytes. Engineered T cells had been stimulated 2× in vitro prior to transfer (described above). Recipients also received rIL2 (2 × 10⁴ U/IP every other day for 8 days) to promote T-cell expansion after transfer.

KPC mice with 3–6 mm tumors were injected with purified anti-CSF1R (anti-CD115, clone AFS98; ref. 18) derived from a hybridoma generously provided by Dr. Miriam Merad (Icahn School of Medicine at Mount Sinai, New York, NY). Mice received 1 mg/mouse IP on day −3 and day 0 (e.g., day of T-cell therapy), followed by biweekly injections of 0.5 mg/mouse IP for 8 or 28 days. A separate cohort of KPC mice was treated once with 100 μg agonistic anti-CD40 (FGK145 hybridoma kindly provided by Dr. Stephen Schoenberger, La Jolla Institute for Allergy and Immunology, San Diego, CA) or isotype control (Rat IgG2a, 2A3; Bio X Cell). Anti-CSF-1R and anti-CD40 were purified from culture supernatants of hybridomas grown in a CELLine Flask (BD) in serum-free hybridoma medium (Sigma) or purchased from Bio X Cell.

Peripheral blood mononuclear cells (PBMC) were collected prior to organ harvest. Blood was collected in PBS containing 20 mmol/L EDTA, then lysed with 1× ACK Lysis buffer (Thermo Fisher) at room temperature and centrifuged at 10,000 rpm for 1 minute. PBMCs were then resuspended in 1 mL complete media (DMEM containing 10% FBS + 1% penicillin/streptomycin), stored on ice for <2 hours, and subsequently stained for flow-cytometric analyses. Tumor, pancreas, spleen, metastases, or lung were also collected and placed into ice-cold media (DMEM containing 10% FBS + 1% pen/strep). Tissues were subsequently divided for analyses for RNA (flash frozen), IHC (formalin-fixed), and immunofluorescence (OCT compound). For flow-cytometric analyses of single cells, tissues were weighed, minced, and then incubated for 20 minutes in collagenase (1 mg/mL; Sigma) at 37°C, filtered (70-μm filters, Sigma), and washed 2× in complete media. Live cells were counted by Trypan blue exclusion.

Single-cell suspensions of mononuclear cells isolated from blood, spleen, and tumor were stained with the following fluorophore-conjugated monoclonal antibodies: CD45 (Ly5; BioLegend), Thy1.1 (OX-7; BD Biosciences), CD8α (53-6.7; BD Biosciences), CD11b (M1/70; BD Biosciences), Ly6C (HK1.4; BioLegend), Gr-1 (RB6-8C5; BD Biosciences), CD115 (AFS98; BioLegend), PD-1 (J43; Thermo Fisher), granzyme B (NGZB; BioLegend), IFNγ (XMG1.2; BD Biosciences), TNFα (MP6-XT22; BD Biosciences), CD11c (N418; BioLegend), CD64 (X54-5/7.1; BioLegend), SIRPa (P84; BioLegend), Ly6G (1A8; BioLegend), CD19 (1D3; BD), and NK1.1 (PK136; eBioscience). Human antibodies used included CD11b (ICRF-44; eBioscience), CD115 (12-3A3-1B10; eBioscience), and CD163 (GH1/61; BioLegend). To measure intracellular cytokine production, engineered T cells ± Msln_406-414 peptide or gp33 peptide (KAVYNFATM, GenScript) were incubated in vitro for 5 hours in the presence of GolgiPlug (5 μg/mL; BD Biosciences), stained for surface antigens, fixed and permeabilized (BD Biosciences Fixation/Permealization kit, according to protocol), and then stained with appropriate antibodies. To quantify the frequency of cells expressing intranuclear proteins T-bet (4B10), Foxp3 (FJK-16s), Eomes (Dan11mag), and Ki67 (SolA15), cells were surface stained, as indicated above, followed by eBioscience Foxp3 Fixation/Permeabilization solutions prior to nuclear staining. Data were acquired on an LSRII or FacsCanto (BD Biosciences) and analyzed using FlowJo v10.3 software. Cell numbers infiltrating tissues were normalized to tumor or spleen weight.

RNA was generated from flash frozen tumor samples using the Qiagen RNAeasy kit according to the manufacturer's protocol. RNA content was confirmed using a Nanodrop, and 100 ng of RNA was used for each reaction. The standard NanoString Technologies hybridization protocol was followed at the FHCRC Genomics facility. Gene expression was determined using the nCounter PanCancer Immune Profiling Panels (both mouse and human specific kits, NanoString Technologies). Reporter gene-specific nucleic acid probes appended with fluorophore barcodes were detected and used to quantify a specific subset of immune genes (26). The data were quality controlled according to NanoString's protocols. Gene expression was analyzed using nSolver v4.0 (NanoString Technologies).

Tumors were fixed in 10% formalin for 48 hours and paraffin-embedded. Sections (4 μm) were stained with hematoxylin and eosin, Masson's trichrome, or Picrosirius red at the University of Washington Histology Core. Primary antibodies used recognized CD163 (Novocastra, 10D6, 1:200), cleaved caspase-3 (Cell Signaling Technology, D175, 1:200), and Ki67 (Thermo Fisher, clone SolA15, 1:200). Slides were scanned using the Nanozoomer Digital Pathology slide scanner (Hamamatsu), and Visiopharm software was used to identify regions of interest (ROI, i.e., tumor tissue, excluding normal tissue) sampled at 100%. The software was trained to detect immunoreactivity using a project-specific configuration based on a threshold of pixel values as we previously reported (6). The number of positively stained cells was measured in 3–5 nonoverlapping 20× fields using NIS-Element imaging software (Nikon's universal software platform, n = 3–5 mice per group). Collagen was quantified from 2 tumor sections stained with Picrosirius red, and intensity of the red staining was assessed in a blinded manner across 3–5 20× fields as follows: 0, no staining detected; 1, light staining; 2, moderate; 3, moderate–high staining intensity; 4, high staining, as we previously described (ref. 6; n = 3–6 animals per cohort).

Normal pancreas and tumors were embedded in OCT (Tissue-Tek), frozen, and stored at −80°C. Sections (7 μm) were fixed in acetone at −20°C for 10 minutes. Sections were rehydrated with PBS + 2.5% bovine serum albumin (BSA) and incubated for 1 hour at room temperature with primary antibodies to CD11b-PE/Dazzle (BioLegend, M1/70; 1:200), CD8α (BD, 53-6.7; 1:25), and panCK-FITC (Sigma-Aldrich, F3418; 1:200) diluted in PBS + 2% BSA. Slides were washed three times in PBS + 2.5% BSA and incubated with anti-mouse 546 (1:1,000, Invitrogen) for 1 hour at room temperature in the dark to detect CD8⁺ T cells. Slides were then washed 3× with PBS + 2% BSA, washed 3× with PBS, and mounted in DAPI Prolong Gold (Life Technologies). Images were acquired on a Leica DM6000 epifluorescent microscope at the University of Minnesota Center for Immunology and analyzed using Imaris 9.1.0 (Bitplane).

Correlation between myeloid gene and cytokines gene expression in Fig. 1 is in whole based upon data generated by the TCGA Research Network: https://www.cancer.gov/tcga. The data include a total of 165 primary PDAs available from TCGA and analyzed using excel and GraphPad software. A large subset of these has been described (27).

Statistical analyses were performed using GraphPad software. All mouse experiments reflect n = 3–5 mice per group. Student t test was used to compare two-group data. One-way ANOVA and Tukey after test were used for multiple comparisons. Pearson r was used to determine correlational significance. Data were expressed as mean ± SEM. P < 0.05 was considered significant. *, P < 0.05; **, P < 0.005; ***, P < 0.0005.

We previously showed a positive correlation between CD163⁺ macrophage abundance and localization with endogenous CD8⁺ T cells in human PDA (hPDA), with a fraction of CD163⁺ macrophages expressing PD-L1 (15). CD163 is often used to identify immunosuppressive, protumorigenic M2 macrophages (28), but it is also expressed by some tissue-resident macrophages, including a subpopulation of perivascular macrophages in the central nervous system (29). In further analyses of these tissues described in our prior study (15), we found CSF1 (M-CSF) to be significantly increased in invasive PDA (n = 11) compared with normal adjacent pancreas tissue (n = 3; Fig. 1A). In contrast, CSF2 (GM-CSF) and CSF3 (G-CSF) were not significantly different between normal adjacent pancreas and tumors in patient samples. Gene-expression analyses in pancreatic tumors in the TCGA database (30) showed a significant correlation between CD163 and CSF1, between CSF1 and its receptor CSF1R, and between CD163 and CSF1R (Fig. 1B). In contrast to CD11b⁺CD163⁻ cells, most of the CD11b⁺CD163⁺ macrophages in hPDA expressed CSF1R (CD115; Fig. 1C and D). Consistent with prior studies (13, 14), CD11b⁺ myeloid cells were abundant in both preinvasive lesions (pancreatic intraepithelial neoplasms, PanINs), primary pancreatic tumors, and spontaneous metastases in KPC mice (Fig. 1E). CSF1 was similarly increased in human and mouse PDA relative to normal pancreas tissue, whereas CSF2 and CSF3 were significantly higher in mouse compared with human PDA (Fig. 1F). In contrast to normal mouse pancreas, normal adjacent human pancreas is not not true normal, which can contribute to differences when comparing mouse and human gene expression data. Tissue macrophages can be M1-like and express high MHC II, costimulatory molecules and IL12 (ref. 13; in mice, also express high Ly6C and contribute to antitumor responses, ref. 31) or M2-like, which are characterized by low Ly6C and MHC II expression and produce the immunosuppressive cytokine IL10 (32). Granulocytes express high Gr-1 (and Ly6G) and have intermediate expression of Ly6C (Ly6C^int), whereas mononuclear phagocytes do not express Ly6G (33). We differentiated between granulocytes (CD45⁺CD11b⁺Gr1^highLy6C^int), Ly6C^high TAMs (CD45⁺CD11b⁺Ly6C^high), and Ly6C^low TAMs (CD45⁺CD11b⁺Gr-1⁻Ly6C⁻; Fig. 1G). The most frequent intratumoral CD11b⁺ subset in KPC mice was Ly6C^low TAMs. The phenotypes of cells identified using this gating strategy were further assessed and included CD19⁺ B cells and intratumoral NK cells (NK1.1⁺) as control cell populations. Both Ly6C^high and Ly6C^low TAMs expressed common mononuclear myeloid markers CD11c, CD64, and SIRPα and had variable expression of MHC II (Fig. 1H). Csf1R (CD115) expression, albeit dim, was expressed the highest on intratumoral Ly6C^low TAMs (Fig. 1H).

To investigate the direct role of Csf1R signaling on myeloid subsets and indirect impact on antigen-specific T cells, we administered a monoclonal antibody against Csf1R (anti-Csf1R/CD115; ref. 18) in combination with Msln-specific T cells according to a therapy protocol previously shown to have activity in KPC mice with invasive disease (ref. 6; Fig. 2A), which includes administration of cyclophosphamide (Cy) to achieve transient lymphodepletion, followed by (6 hours later) infusion of 1 × 10⁷ P14 Thy1.1⁺ T cells retrovirally transduced to express a high-affinity TCR specific for Msln_406-414/H-2D^b. The transduced T cells were stimulated twice in vitro with antigen and expanded in recombinant human IL2 prior to transfer. To promote expansion of transferred engineered T cells, recipient animals also received as a vaccine 1 × 10⁷ irradiated syngeneic splenocytes pulsed with Msln peptide (6). As seen in healthy animals (34), Csf1R had high expression on circulating Ly6C^high inflammatory monocytes (IM) and Ly6C^low patrolling monocytes (PM) in untreated KPC mice (Fig. 2B). At day 8 after T-cell infusion, anti-Csf1R did not significantly modify circulating CD45⁺ or CD11b⁺ cell frequencies, but did significantly reduce both circulating monocyte populations, while concomitantly increasing granulocyte frequency (Fig. 2C). By day 28, PM precursor frequency remained significantly decreased in the blood, whereas IM frequency rebounded to near baseline levels (Fig. 2C). PMs are relatively long-lived in circulation (>1 week) compared with IMs, which have a circulating half-life <4 hours and rapidly differentiate into PMs in wild-type mice (34). The half-life of IMs in tumor-bearing mice may be slightly prolonged because of myeloid cytokines such as Csf1 secreted by pancreatic tumor cells (14). However, the fact that Csf1R blockade caused sustained reduction of PMs but not IMs suggests either that Csf1R signaling promotes IM differentiation into PMs or that the kinetics of conversion and/or replenishment of IMs are altered in cancer.

We next investigated whether intratumoral myeloid subsets were altered by Csf1R blockade. At day 8 after T-cell transfer, anti-Csf1R treatment significantly decreased Ly6C^low TAM frequency, resulting in a 2-log reduction in number (Fig. 2D). As seen with circulating granulocytes, both the frequency and number of intratumoral granulocytes were also significantly increased with Csf1R blockade at day 8 (Fig. 2D), a finding also observed in some transplantable tumor models (35). In contrast with day 8 findings, granulocytes were instead significantly reduced by day 28 after Csf1R blockade (Fig. 2E). Ly6C^low TAM frequency remained significantly reduced compared with control tumors 28 days after treatment with engineered T cells and anti-Csf1R (Fig. 2E). Thus, Csf1R blockade had differential temporal effects on intratumoral granulocytes during the treatment course.

Histologic analyses were performed to determine if Csf1R blockade and TAM depletion affected the stroma. We previously showed that administration of TCR_Msln significantly altered the stroma, as characterized by collagen reduction in the extracellular matrix (ECM), which was evident at day 8 but not sustained to day 28 after transfer when the T cells have largely lost function (6). However, concomitant Csf1R blockade maintained ECM loss at 28 days after T-cell transfer (Supplementary Fig. S1A and S1B). Thus, Csf1R signaling may promote ECM deposition and/or stability, potentially by promoting survival and/or proliferation of a subset of fibro-inflammatory TAMs (36, 37). Anti-Csf1R treatment significantly decreased tumor apoptosis at both 8 and 28 days following infusion of TCR_Msln-engineered T cells (Supplementary Fig. S1C). We also observed a trend for reduced tumor cell proliferation in anti-Csf1R–treated mice compared with controls, suggesting Csf1R signaling may support tumor cell proliferation (Supplementary Fig. S1D). Quantification of collagen showed a significant decrease due to anti-Csf1R treatment (Supplementary Fig. S1E). ECM remodeling may influence tumor cell survival and/or proliferation, as well as have effects on engineered T-cell migration and accumulation.

We next investigated how Csf1R blockade influenced polyclonal endogenous CD8⁺ T cells compared with engineered tumor antigen–specific T cells. Anti-Csf1R caused a significant increase in frequencies of both endogenous (CD8⁺Thy1.1⁻) and TCR_Msln-engineered (CD8⁺Thy1.1⁺) T cells in both circulation and in tumors of KPC mice (Fig. 3A–C). Endogenous CD8⁺ T-cell number increased 5- to 10-fold in tumors at both early and later time points (Fig. 3D). In contrast, intratumoral engineered T-cell numbers were unchanged at day 8 and only modestly increased (∼2-fold) at day 28 (Fig. 3E). Thus, despite significantly lowering the TAM:engineered T-cell ratio by 6- to 8-fold (Fig. 3F) and enhancing accumulation of endogenous T cells and frequency of engineered T cells, Csf1R blockade had limited impact on engineered T-cell expansion and accumulation. We previously showed that TCR_Msln-engineered T cells become progressively dysfunctional in PDA, as characterized by reduction in tumor cell killing and loss of ability to produce IFNγ (6). The addition of Csf1R blockade resulted in a higher frequency of dysfunctional engineered T cells in the tumor at day 8 after transfer (Fig. 3G and H), which correlated with a higher frequency of cells expressing PD-1 (Fig. 3I).

The unanticipated finding that TAM depletion did not enhance engineered T-cell function in PDA prompted us to quantify the impact of T-cell therapy alone on myeloid cell composition in PDA. Serial (biweekly) TCR_Msln cell infusions into KPC mice modified the frequency and number of intratumoral myeloid subsets. A significant increase in Ly6C^high CD11b⁺ frequency and number was observed after treatment with TCR_Msln cells but not control T cells (Fig. 3J), and intratumoral Ly6C^high CD11b⁺ cell frequency significantly correlated with mouse survival (Fig. 3K). Thus, T-cell therapy promoted the recruitment, survival, and/or expansion of Ly6C^high TAMs in PDA. Although it remains unclear what factors activated T cells produce to influence the recruitment and/or expansion of myeloid cells, possibilities include CCL2, CCL5, or GM-CSF (38). Additional investigation into the cross-talk between tumor-reactive T cells and myeloid cells in the TME should help improve immunotherapy design.

Because TAM depletion failed to augment TCR_Msln cell function and TCR_Msln cell therapy efficacy correlated with TAM accumulation (Fig. 3J), we next tested if modulation of TAM function rather than depletion could enhance engineered T-cell antitumor effects. Agonistic anti-CD40 promotes the influx of monocytes into PDA, which can differentiate into tumoricidal macrophages or dendritic cells (13, 23). However, it is not known if anti-CD40 can enhance self/tumor-reactive engineered effector T cells. Therefore, we administered a single dose of anti-CD40 at the time of TCR_Msln-engineered cell therapy in KPC mice (Fig. 4A). Anti-CD40 dramatically enhanced both the acute expansion and persistence of infused CD8⁺Thy1.1⁺ TCR_Msln-engineered T cells in circulation (Fig. 4B and C). PD-1 on circulating TCR_Msln cells was significantly lower in mice that received anti-CD40 compared with isotype control for at least 14 days after transfer (Fig. 4C), suggesting that anti-CD40 may enhance the quality of infused T cells. Anti-CD40 resulted in a >10-fold increase in intratumoral engineered T-cell numbers at days 8 (Fig. 4D) and 28 (Fig. 4E). Anti-CD40 also significantly increased the frequency and number of infused T cells in the lung at day 8, a site where mesothelin had low expression (Fig. 4D) without evidence of respiratory distress or toxicity to normal tissues, suggesting this may be nonspecific (Supplementary Fig. S2). By day 28, the number of infused T cells in the lung was no longer significantly different (Fig. 4E). In situ staining showed that anti-CD40 increased CD8⁺ T-cell accumulation in PDA (Fig. 4F). Anti-CD40 also significantly increased the frequency of infused T cells that express Ki67 and/or the cytolytic molecule granzyme B at day 28 (Fig. 4G–I). However, anti-CD40 did not rescue engineered T-cell IFNγ production in PDA (Fig. 4J and K). This was not due to specific deficiencies in the introduced TCR signaling because stimulation of the engineered T cells via the endogenous TCR (P14 which recognizes gp33/H-2D^b) also did not rescue function. IFNγ production by splenic TCR_Msln cells was significantly diminished after a single dose of anti-CD40 compared with TCR_Msln cells isolated from control- or isotype-treated mice (Fig. 4K), which may reflect changes in the local inflammatory environment in the spleen following anti-CD40. Taken together, these data indicate that anti-CD40 promotes engineered tumor-reactive T-cell longevity but fails to fully rescue the function of engineered T cells in pancreatic cancer.

Previously, we showed that a single infusion of engineered T cells had only transient antitumor activity (6). To test if anti-CD40 prolonged antitumor activity, we assessed stroma remodeling and tumor cell apoptosis. Anti-CD40 sustained collagen loss and tumor cell killing following engineered T-cell infusion through day 28 after infusion (Fig. 5A–C). Histologic analyses of the heart and lung, sites of normal mesothelin expression, showed no evidence of increased apoptosis or pathologic damage (Supplementary Fig. S2). To gain insight into how anti-CD40 might promote T-cell persistence and tumor cell killing, we evaluated the immune composition in the tumors following treatment. Anti-CD40 in combination with TCR_Msln T cells significantly increased the frequency of both CD45⁺ hematopoietic cells (Fig. 5D) and CD11b⁺ myeloid cells (Fig. 5E) at day 8. Anti-CD40 plus T-cell therapy increased granulocyte frequency at day 8 but decreased it by day 28 (Fig. 5D). Anti-CD40 also significantly increased the frequency and number of Ly6C^high TAMs in PDA at day 8 after T-cell transfer (Fig. 5D and E), while leaving the number of Ly6C^low TAMs unchanged. PD-1, an inhibitory receptor that has been identified on immunosuppressive M2-like TAMs that interfere with tumor immunity (39), was highly expressed on the Ly6C^low TAMs (Fig. 5F). However, anti-CD40 reduced the frequency of these PD-1⁺Ly6C^low TAMs with a corresponding increase in Ly6C^highPD-1^low TAM at day 8 (Fig. 5F). Taken together, these data support potential benefits of myeloid programming rather than ablation to enhance antitumor activity of engineered T-cell therapy, and are supported by the fact that CSF1 expression trends with prolonged PDA patient overall survival (Supplementary Fig. S3). However, there are clearly additional barriers that suppress sustained engineered T-cell cytokine production that are not overcome by these specific myeloid-targeted therapies. The enhanced tumor killing may reflect the activity of anti-CD40 in promoting tumoricidal macrophages, as has been reported in PDA (13). The cyclophosphamide preconditioning regimen prior to T-cell therapy transiently induced death of all circulating myeloid cell subsets (Supplementary Fig. S4A and S4B), which may indirectly affect how myeloid-targeted therapies influence engineered T-cell responses. The results highlight the complexity of activities by myeloid subsets and suggest that the most effective strategies for enhancing adoptive cellular therapies involving transfer of activated effector T cells may differ from the requirements for amplifying evolving endogenous T-cell responses.

Most PDA patients succumb rapidly to their disease because of resistance to even the most aggressive therapies. Because current immunotherapies such as anti–PD-1/PD-L1 or anti–CTLA-4 provide minimal therapeutic benefit for the majority of patients, we were pursuing an engineered T-cell therapy that redirects T-cell specificity to target a commonly overexpressed self/tumor antigen mesothelin (40). Despite significant antitumor activity of these TCR_Msln cells in autochthonous murine PDA, intratumoral engineered T cells decrease in both quantity and quality over a relatively short time period (6), suggesting the TME may limit treatment efficacy. TAMs are abundant in mouse and human PDA and are immunosuppressive to endogenous T cells (13, 14, 36, 41). Nevertheless, we found that TAM depletion, including the most abundant Ly6C^low subset, following Csf1R blockade had unfavorable effects on engineered T cells. Alternatively, anti-CD40 increased Ly6C^high TAM accumulation without altering Ly6C^low TAM numbers and increased the longevity of engineered T cells. However, the changes in TAMs and the myeloid infiltrate in PDA by anti-CD40 were insufficient to maintain the functions of persisting engineered T cells, including cytokine production.

Genetic fate mapping studies have demonstrated that macrophages in most adult tissues arise from yolk-sac–derived hematopoietic progenitors during embryonic development (42). Such tissue-resident macrophages perpetuate indefinitely throughout adulthood via stromal cell production of Csf1 that promotes macrophage survival and proliferation (42). In settings of inflammation, such as cancer (43) or atherosclerosis (44), circulating monocytes are recruited into the tissue parenchyma and differentiate into macrophages, contributing to the overall tissue macrophage pool. It has been suggested that Csf1-dependent tissue-resident TAMs form the primary macrophage subset that contributes to PDA growth (36). Our study corroborates previous work demonstrating Csf1R-dependent maintenance of Ly6C^low TAMs in PDA (36). These tissue-resident TAMs are a major source of ECM components and angiogenic factors (36). Our observation that Ly6C^low TAM depletion following Csf1R blockade decreased collagen and tumor cell proliferation is consistent with tissue-resident TAMs promoting tumor growth, perhaps indirectly by depositing ECM and/or directly via production of factors that can induce EGF/MAPK signaling in tumor cells (37, 45).

Our findings suggest that tissue-resident TAMs have differential effects on engineered versus endogenous T cells. Ablation of Ly6C^low TAMs using Csf1R blockade was not sufficient to enhance acute intratumoral accumulation of infused engineered effector T cells, despite increasing endogenous CD8⁺ T-cell accumulation. Similarly, Csf1R blockade has been shown to deplete the normal pancreas of resident macrophages and prevent diabetes onset mediated by the adoptive transfer of T cells specific to islet antigens (46). By contrast, the impact on the accumulation of endogenous T cells is consistent with a study that showed Ly6C^low TAMs inhibit endogenous T-cell infiltration and/or prevent their retention in PDA (20). The antigen specificity of such endogenous CD8 T cells is unknown, however, and the enhanced infiltration might be largely nonspecific. In contrast to naïve or quiescent memory T cells, engineered T cells are activated and differentiated effector T cells that express a distinct set of adhesion molecules and chemokine receptors. Thus, it is possible Ly6C^low TAM depletion enhances the expression of chemokines or factors that support endogenous naïve/memory T-cell recruitment, but that such factors are irrelevant for effector T-cell recruitment. Alternatively, the influx of endogenous CD8⁺ T cells may successfully compete for critical homeostatic cytokines at the tumor site that would otherwise be available for engineered T cells.

The fact that CSF1 expression trended with prolonged overall survival in patients with pancreatic cancer and that TAMs correlated with CD8⁺ T-cell numbers and location in human PDA (15) suggest a TAM subset, or specific TAM qualities, in PDA may support effector T-cell responses. Because both tissue-resident– and monocyte-derived TAMs can take up tumor antigens (36), MHC I–restricted antigen presentation by TAMs may be critical for sustaining effector functions of engineered T cells. iNOS production by TAMs promotes adoptively transferred CD8⁺ T-cell antitumor activity in a mouse syngeneic lymphoma model (47), whereas iNOS production suppresses naïve T-cell activation (48), suggesting loss of iNOS production might preferentially interfere with activity of the infused effector cells.

In contrast to Ly6C^low TAM depletion, a single dose of agonistic anti-CD40 enhanced the expansion, intratumoral accumulation, and persistence of adoptively transferred TCR-engineered T cells. CD40, a TNFR superfamily member, has broad cell distribution including macrophages, dendritic cells, B cells, and endothelial cells. In the current report, we focused on the effects of agonistic anti-CD40 on TAMs, which has been identified as a target (13). Agonistic anti-CD40 can promote upregulation of adhesion molecules on endothelial cells and macrophage proinflammatory cytokine production, which could enhance the accumulation and activation of adoptively transferred tumor antigen–specific CD8⁺ T cells. Because we observed that infusion of TCR_Msln cells increased the accumulation of a similar Ly6C^high TAM population without anti-CD40, our data suggest that one potential mechanism of antitumor activity by engineered T-cell therapy is via engaging the myeloid lineage. Thus, the significant increase in tumor cell apoptosis we observed in mice treated with anti-CD40 and TCR_Msln cells could in part reflect further enhancement of tumoricidal macrophages, as reported (13).

Combinatorial targeting of TAMs is a potential therapeutic strategy. However, agonistic anti-CD40 administered <3 days prior to chemotherapy has been shown to induce lethal hepatotoxicity in mice via Csf1R-dependent macrophages (49), indicating a sensitivity to timing and sequencing when incorporating multiple treatment modalities. Therefore, given the inherent complexity of our treatment protocol, which includes cyclophosphamide preconditioning, tumor antigen–specific CD8⁺ T cells, a vaccine, and exogenous IL2 (6), we focused on either Csf1R blockade or an anti-CD40 agonist in combination with T-cell therapy. In mouse immunogenic implantable tumor models, the combination of anti-CD40 and Csf1R blockade synergize to significantly slow tumor growth, whereas Csf1R blockade had no antitumor effect alone (50, 51). Using a potentially less immunogenic and genetically engineered melanoma model, similar synergistic antitumor effects of dual anti-CD40 and Csf1R blockade was shown (52). However, these studies required frequent, prolonged, and high dosing of both anti-CD40 and Csf1R blockade. Our results demonstrated no benefit to persistence or function of transferred TCR_Msln T cells with Csf1R blockade and anti-CD40 failed to rescue T-cell cytokine production. Our T-cell therapy protocol uses cyclophosphamide preconditioning, which transiently diminishes circulating hematopoietic cells (including all myeloid cell subsets) in the blood, and we did not pursue experiments combining Csf1R blockade and agonist anti-CD40. Because both TAM-modulating strategies failed to rescue IFNγ production by engineered T cells in the tumor, future studies will need to elucidate how best to design combinatorial regimens modulating the myeloid compartment with engineered T cells to maximize therapeutic benefit and minimize toxicity. On the basis of our observation of differential responses by engineered T cells compared with endogenous T cells following TAM depletion, studies examining the impact of immune modulating reagents on endogenous anticancer immune responses appear insufficient for predicting appropriate combinations for enhancing engineered T-cell therapies.

R.H. Pierce reports receiving commercial research grants from Juno Therapeutics, Immunomics Therapeutics, Pulse Biosciences, and X4 Pharmaceuticals, has ownership interest (including stock, patents, etc.) in Oncosec Medical, Celgene, and Five Prime Therapeutics, and is a consultant/advisory board member for AbbVie, Calithera Biosciences, Sensei Biotherapeutics, Pulse Biosciences, and Immunomics Therapeutics. P.D. Greenberg is a consultant for Juno Therapeutics, FLX Bio, Elpiscience, and Nextech, reports receiving a commercial research grant from Juno Therapeutics, has ownership interest (including stock, patents, etc.) in FLX Bio, Elpiscience, and Nextech, is a consultant/advisory board member for Juno Therapeutics, FLX Bio, Elpiscience, and Nextech. S.R. Hingorani reports receiving a commercial research grant from Halozyme Therapeutics to Fred Hutchinson Cancer Research Center and is a consultant/advisory board member for Halozyme Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: I.M. Stromnes, S.R. Hingorani

Development of methodology: I.M. Stromnes, A. Hulbert, S.R. Hingorani

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I.M. Stromnes, A.L. Burrack, P. Bonson, C. Black, J.S. Brockenbrough, J.F. Raynor, E.J. Spartz, R.H. Pierce, S.R. Hingorani

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I.M. Stromnes, A.L. Burrack, P. Bonson, R.H. Pierce, P.D. Greenberg, S.R. Hingorani

Writing, review, and/or revision of the manuscript: I.M. Stromnes, A.L. Burrack, J.F. Raynor, P.D. Greenberg, S.R. Hingorani

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases):J.S. Brockenbrough

Study supervision: I.M. Stromnes, S.R. Hingorani

I.M. Stromnes is supported by an AACR Pancreatic Cancer Action Network Career Development Award (17-20-25-STRO) and an Institutional Research Grant 124166-IRG-58-001-55-IRG65 from the American Cancer Society. Support was provided by the National Cancer Institute, CA018029 and CA033084 (P.D. Greenberg) and CA161112 (S.R. Hingorani), Giles W. and Elise G. Mead Foundation (to S.R. Hingorani), a Cancer Center Support Grant P30CA015704 Supplement (S.R. Hingorani and P.D. Greenberg) and a Pancreatic Cancer Action Network Grant 16-65-GREE (P.D. Greenberg). P.D. Greenberg receives support from Juno Therapeutics, a Celgene company.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.