Targeting the IGF-Axis Potentiates Immunotherapy for Pancreatic Ductal Adenocarcinoma Liver Metastases by Altering the Immunosuppressive Microenvironment

Pancreatic ductal adenocarcinoma (PDAC) is currently the fourth leading cause of cancer-related deaths in the industrialized world (1, 2). Despite remarkable advances in current diagnostic techniques (3), it has been widely reported that PDAC may surpass colorectal cancer to become the second leading cause of cancer-related death in the United States by 2030 (4). There is a dearth of effective therapies for pancreatic cancer and the 5-year survival still stands at 5% to 9%, the lowest of any common malignancy, identifying this disease as an obvious “unmet need” (5, 6). The most notable clinical features of PDAC are its propensity for aggressive local invasion, metastasis (mainly to the liver) and inherent resistance to conventional therapies. While approximately 50% of patients with pancreatic cancer present with evidence of distant disease, particularly in the liver, the remaining patients have localized disease without detectable metastases. Of these, 15% to 20% of patients are operable and therefore potentially amenable to curative therapy, while approximately 30% have locally advanced disease. The poor prognosis associated with pancreatic cancer is related mainly to the high incidence of synchronous liver metastases and postoperative recurrence (7). The liver is the major site of recurrent disease (8–10). Chemotherapy with gemcitabine or 5-fluorouracil, irinotecan and oxaliplatin (FOLFIRINOX) is still the standard of care for metastatic pancreatic cancer, but it is associated with high toxicity and limited survival advantage (10, 11). Thus, therapies that can effectively reduce the incidence of pancreatic cancer liver metastases and/or metastatic outgrowth would have tremendous potential to significantly improve survival.

Recent advances in targeting immune checkpoints such as programmed cell death-1 (PD-1) and CTL antigen-4 (CTLA-4) have yielded promising therapeutic results in several aggressive and treatment-refractory cancers such as malignant melanoma, small cell lung cancer and renal cell carcinoma (12–14). However, to date, immunotherapy has failed to show promise in the treatment of PDAC (15, 16). This, despite compelling clinical evidence for an effector T-cell infiltrate in PDAC, that, when present, has positive prognostic implications and can become reactive to the autologous tumors when further potentiated ex vivo (17). This failure of immunotherapy to alter the course of PDAC may be due, at least in part, to the presence of immunosuppressive cells such as myeloid-derived suppressor cells (MDSC) and tumor-associated macrophages (TAM) that polarize to an immunosuppressive M2 phenotype in the tumor microenvironment (TME) and impede T cell–mediated cytotoxicity, as well as promote PDAC progression through the release of proangiogenic and proinvasive factors such as VEGF and matrix metalloproteinases (MMP). Thus, therapeutic approaches that can co-target the immunosuppressive TME of PDAC, while also reversing T-cell exhaustion could enhance the efficacy of immunotherapy and are currently being sought (16, 18) (reviewed in ref. 19).

There is compelling experimental and clinicopathologic evidence that the insulin-like growth factor (IGF) axis plays an important role in PDAC progression (20–22). A recent analysis of surgical PDAC resections revealed type I IGF receptor (IGF-IR) overexpression in 53% of the tumors and this correlated with higher tumor grade. Moreover, high IGF-IR expression levels were also observed in the tumor-associated stroma and high IGF-IR expression in both compartments was associated with shorter overall survival and resistance to chemotherapy (22–24).

The major immune cell subtypes of both innate and adaptive immunity including the myeloid-derived mononuclear cells, natural killer (NK) cells and T and B lymphocytes express the IGF-IR and are responsive to IGF ligands (25). Although the role of IGF-IR in the development and function of these cells is complex, there is compelling evidence that within the TME, the IGF axis promotes an anti-inflammatory, immunosuppressive response that enables cancer expansion. IGF-I was shown to negatively regulate dendritic cell (DC) activation, thereby impairing antigen presentation (26) and it was also shown to stimulate the proliferation of immunosuppressive regulatory T cells (Treg; refs. 27–29). IGF-IR activation was also linked to polarization of macrophages to the protumorigenic M2 phenotype (30). IGF targeting was shown to alter the tumor–immune microenvironment (TIME) in colon cancer, reducing anti-inflammatory cytokines (31). IGF-I may also play a role in the tumor-promoting effect of myeloid-derived suppressor cells (MDSC) by regulating TGFβ expression (32). Collectively, these studies identify the IGF axis as a contributor to an anti-inflammatory, protumorigenic TME.

We reported on the bioengineering and characterization of a novel IGF-inhibitor, the IGF-Trap—a fusion protein consisting of the entire extracellular domain of the human IGF-IR fused to the Fc portion of human IgG₁. Similar to the cognate receptor, the IGF-Trap binds the ligands IGF-I and IGF-II (but not insulin) with high affinity, reducing ligand bioavailability in the circulation (ref. 33; and reviewed extensively in ref. 34). We have shown that the IGF-Trap inhibits liver metastases of several highly aggressive carcinomas (33, 35). Moreover, we recently reported that IGF-I could directly induce the expression of N2-associated transcripts in vitro in bone marrow–derived neutrophils and moreover, treatment of colon carcinoma MC-38 injected mice with the IGF-Trap, markedly reduced the proportion of immunosuppressive N2-polarized neutrophils infiltrating hepatic metastases (36). These data suggested that inhibition of IGF signaling could reverse the protumorigenic effects of a polarized TME.

Here, we sought to determine whether the IGF-Trap could enhance the efficacy of an immune checkpoint blocker by altering the TIME within hepatic PDAC metastases, thereby potentiating CTL-mediated tumor cell kill to reduce metastatic expansion. We found that IGF axis targeting profoundly altered the immune landscape of hepatic PDAC metastases and enhanced the effect of a PD-1 blocking antibody, causing a marked reduction in metastatic outgrowth.

The pancreatic ductal adenocarcinoma (PDAC) LMP cell line originated from a tumor that arose spontaneously in the genetically engineered Kras G12D/+; p53R172H/+; Pdx1 Cre (KPC) mouse model, and developed in the Lowy laboratory as described in detail elsewhere (37). In syngeneic B6.129 F1 mice implanted in the pancreas with LMP cells, tumor growth and metastasis mimic the aggressive clinical behavior of PDAC. The murine pancreatic cancer line KPC FC1199, referred to here as FC1199, was generated in the Tuveson laboratory (Cold Spring Harbor Laboratory, New York, NY) from PDA tumor tissue obtained from KPC mice of a pure C57BL/6 background, as described previously (38), and was a generous gift from the Tuveson laboratory. Both cell lines were authenticated by the ATCC using Short Tandem Repeat (STR) analysis as described in the National Institute of Standards and Technology (NIST) granted U.S. patent (No. 9,556,482). The cells were routinely tested for common murine pathogens, as per the McGill University Animal Care Committee and the McGill University Biohazard Committee guidelines. Mycoplasma testing was carried out routinely using the Applied Biological Materials Mycoplasma PCR Detection Kit (ABM, last relevant test and verification carried out in June 2020, most recently confirmed on August 20, 2021). The cells were routinely grown in a humidified incubator at 37°C with 5% CO₂ in Dulbecco modified Eagle medium (Thermo Fisher Scientific) supplemented with 100 U/mL penicillin and 100 μg/mL streptomycin solution (Sigma), and 10% FBS (Thermo Fisher Scientific). To minimize genetic drifts and phenotype changes, a frozen stock of these cells was maintained and cells thawed and passaged in vitro for up to 2–3 weeks only, prior to use in the experiments.

All mouse experiments were carried out in strict accordance with the guidelines of the Canadian Council on Animal Care (CCAC) “Guide to the Care and Use of Experimental Animals” and under the conditions and procedures approved by the Animal Care Committee of McGill University (AUP number: 5260). Mouse experiments were performed in male and female B6.129 F1 mice that are syngeneic to the LMP cells or in C57Bl/6 male mice that are syngeneic to the FC1199 cells. BL6-Col-GFP mice in which type I collagen is genetically tagged with GFP were a kind gift from Dr. Tatiana Kisseleva (Department of Surgery, University of California, San Diego, La Jolla, CA). They were backcrossed for one generation onto the 129S1/Svlmj (Jackson laboratories) background to obtain heterozygous Col-GFP mice that were crossbred with BL6 mice to generate first generation Bl6.129-Col-GFP F1 mice used for the analysis of activated HSC. All mice were bred in the animal facility of the Research Institute of the McGill University Health Center and used for the experiments at the ages of 7–12 weeks old.

Spontaneous PDAC liver metastases were observed following the intra-pancreatic implantation of 5 × 10⁵ LMP cells in 25 μL Matrigel (Corning) mixed with 25 μL PBS, as described (39). Animals were euthanized 21 days post tumor implantation, at which time metastases were visible on the surface of the liver and were enumerated and sized without prior fixation.

Experimental liver metastases were generated by intrasplenic/portal injections of 1 × 10⁵ or 5 × 10⁵ tumor cells (as indicated), followed by splenectomy as we previously described (40). Animals were euthanized 21 days later, and visible metastases on the surface of the liver were enumerated and sized without prior fixation. Where indicated, fragments of the livers were also fixed in 10% phosphate-buffered formalin, paraffin-embedded, and 5-μm sections stained with hematoxylin and eosin to detect micrometastases and quantify the metastatic burden, as shown. The bioengineering of a third-generation IGF-Trap with improved physicochemical and pharmacodynamic properties was recently described (35). This Trap (IGF-Trap 3.3) was used in all the in vivo experiments, as indicated. For immunotherapy the anti PD-1 antibody—Clone RMP1-14 and isotype control rat IgG2a from BioXcell (West Lebanon, NH) were used, as indicated.

The NanoString's GeoMx digital spatial profiling was used to spatially profile formalin-fixed paraffin-embedded (FFPE) sections derived from mice bearing liver metastases. The analysis was performed by the NanoString Pilot Lab. Initially, 12 regions of interest (ROI) (200 μm–600 μm in diameter) were selected within the tumor and IME enriched tissue areas using the following morphology markers: PanCK (Invitrogen/Thermo Fisher clone AE1/AE3), syto83 (Thermo Fisher), CD8 (Abcam, clone EPR21769) and CD45 (Novus, clone EM-05) with additional staining for CD3, αSMA, and F4/80, as required. A cocktail of 46 primary antibodies including IgG isotype controls (rat IgG2a, rat IgG2b, and rabbit IgG) and three housekeeping proteins (Histone H3, Ribosomal protein S6, GAPDH) conjugated to unique oligonucleotide tags with a UV photo-cleavable linker were incubated with 5-μm–thick FFPE sections. The photo-cleaved oligos were then quantified using the optical barcodes in the NanoString's nCounter platform (nCounter Analysis System, NanoString). For analysis, digital counts were first normalized with internal spike-in controls, then with housekeeping genes and the background was subtracted using the IgG controls.

B6129F1 mice were injected via the intrasplenic/portal route with 1 × 10⁵ or 5 × 10⁵ LMP cells as indicated, and the livers perfused at the time intervals indicated, first with PBS and then with 4 mL of a 4% paraformaldehyde solution. The perfused livers were placed in 4% paraformaldehyde for 48 hours and then in 30% sucrose for an additional 48 hours before they were stored at -80°C. For immunostaining, 10-μm cryostat sections were prepared, incubated first in a blocking solution (1% bovine serum albumin (BSA) and 1% FBS in PBS) and then for 1 hour each with the primary antibodies, used at the indicated dilutions, followed by the appropriate Alexa Fluor-conjugated secondary antibodies (the antibodies used in this study are listed in Supplementary Table S1), all at room temperature. Sections stained with the secondary antibodies only were used as controls in all the experiments. After washing with PBS, an autofluorescence quenching kit (VectorTrueVIEW, Burlingame) was used to reduce tissue autofluorescence and sections counterstained with 1 mg/mL 4,6-Diamidino-2-Phenylindole, dihydrochloride (DAPI, Invitrogen). The sections were mounted in the Prolong Gold anti-fade reagent (Molecular Probes) and confocal images were captured with a Zeiss LSM-880 microscope with a spectrum detection capability. The immunostained cells were quantified blindly in at least 8 images acquired per section, per group.

To analyze early changes in the TIME, mice were injected with 1 × 10⁵ tumor cells via the intrasplenic/portal route and the livers removed 14 days later (or as indicated). Liver homogenates were prepared in cold PBS and filtered through a stainless steel mesh, using a plunger. The filtrates were centrifuged at 500 rpm to separate the hepatocytes, the supernatants containing the non-parenchymal cell fraction centrifuged at 1,400 rpm and the pellets resuspended in 10 mL of a 37.5% Percoll solution in Hank balanced salt solution containing 100 U/mL heparin and centrifuged at 1,910 rpm for 30 minutes to obtain the immune cell–rich fraction. Prior to flow cytometry, red blood cells were removed using the ACK (ammonium chloride-potassium) solution and 1 × 10⁶ cells were immunostained with the indicated antibodies. Data acquisition was with a BD LSRFortessa and FACS Diva software and the data analyzed using the FlowJo software. For FC on hepatic leukocytes, single cells were gated on the basis of size (FSC), granularity (SSC), viability using an eFluor 780 fixable dye (eBioscience, Thermo Fisher) and the expression of CD45.

Experimental liver metastases were generated by injecting 1 × 10⁵ LMP cells via the intrasplenic/portal route. Livers were resected 14 days later and immune cell isolated and stimulated for 4 hours with PMA (5 ng/mL; Sigma) and isonomycin (500 ng/mL; Sigma) in the presence of a protein transport inhibitor (BD GolgiStop). The activated T cells were first immunostained for extracellular markers and then fixed and permeabilized for IFNγ staining prior to analysis by FC.

RNA was extracted from G-MDSCs and CD3⁺CD8⁺ T cells using TRIzol (Ambion, Life Technologies). cDNA was synthesized from isolated RNA using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems), as per the manufacturer's protocol. qPCR was performed in a Bio-Rad Light Cycler (Bio-Rad), using SYBR (Roche). Two micrograms of total RNA were reverse transcribed and the cDNA analyzed using the primer sets listed in Supplementary Table S1. Changes in expression levels were calculated using the ΔΔC_t values and GAPDH was used to normalize for loading.

Splenocytes from naïve mice were isolated and red blood cells lysed as described above. Splenic CD3⁺ T cells were sorted by FACS, stained with CellTrace Carboxyfluorescein succinimidyl ester (CFSE; Thermo Fisher Scientific) and incubated for 48 hours in RPMI with Dynabeads Mouse T-Activator CD3/CD28 (Thermo Fisher Scientific) in a 96-well plate at 37ºC. Liver-derived MDSC from mice which were pretreated, or not, with the IGF-Trap were isolated 14 days post tumor injection and sorted as described above. MDSCs were then added to the preactivated splenic T cells at a ratio of 1:1. In the control condition, no MDSC were added to the activated T cells. After 48 hours of coincubation, the cells were harvested and analysis of CFSE intensity was performed using the BD LSRFortessa.

The nonparametric Mann–Whitney test was used to analyze all metastasis data and a two-tailed Student t test was used to analyze ex vivo and in vitro data and the IF results. Box-and-whiskers plots were used to show individual values, where applicable. Where indicated, the middle bar denotes the median value, the box limits extend from the 25th to 75th percentiles and the whiskers denote the lowest and highest values. Survival data was analyzed by the Gehan–Breslow–Wilcoxon Test.

We have previously shown that treatment with the IGF-Trap altered the TME in the liver (36, 41) and significantly reduced the outgrowth of liver metastases in several preclinical models of aggressive carcinomas (33, 35). The murine LMP cells originated from the LSL-Kras^G12D/+;LSL-Trp53^R172H/+;Pdx-1-Cre (KPC) transgenic model of spontaneous PDAC (42). When implanted orthotopically into the pancreas of syngeneic Bl6.129 F1 mice, these cells mimic the pathology of the human disease and metastasize spontaneously to the liver, eventually causing accumulation of ascites and morbidity within 3–4 weeks following implantation (37, 43). We tested the effect of the IGF-Trap on the growth of LMP cells in vivo and found that while this treatment had a marked effect on liver metastases, significantly reducing their numbers and sizes (Fig. 1A–C), it did not alter local tumor growth in the pancreas (Fig. 1D). This suggested that in this model, the IGF-Trap may have exerted its effect on hepatic metastases by altering the TME in the liver. To further investigate this possibility, we profiled the immune microenvironment in the livers of LMP-injected and IGF-Trap treated (or non-treated) mice, using the NanoString GeoMx digital spatial profiler (44), to identify global changes in the expression of immune cell surface markers in the TIME that resulted from IGF-Trap treatment. We analyzed the TIME associated with both micrometastatic lesions and large diffuse liver metastases to distinguish early and late events in immune cell recruitment, and within each liver section, compared the intensity of each signal to that associated with “tumor-free” areas within the same section, to identify changes specific to tumor-infiltrating immune cells (Supplementary Fig. S1A). When relative signal intensities in IGF-Trap and vehicle (PBS)-treated controls were compared, we observed significant differences in the accumulation of several immune cell subtypes in the TME of IGF-Trap treated, as compared to control mice (Fig. 2A and B). Notable among them were a significant increase in the expression of CD11c and MHCII in both small and large metastases, consistent with increased recruitment and activation of DCs, decreased accumulation of CD11b⁺ and Ly6G⁺ cells, consistent with a reduction in bone marrow–derived granulocytic cells (neutrophils or granulocytic (G)-MDSC), increased presence of CD4⁺ T cells that was accompanied by increased expression of several immune checkpoints [PD-1, T-cell immunoglobulin and mucin domain-containing (TiM)-3 and lymphocyte activation gene (LAG)-3)] in early-stage metastases, and an increase in CD68⁺ cells accompanied by reduced CD163 expression in the larger metastases, indicative of a reduction in M2 tumor–associated macrophages. In addition, this analysis also revealed a marked reduction in TGFβ1 levels in and around the metastases, consistent with an overall reduction in immunosuppressive signals (Fig 2B; Supplementary Fig. S1B).

Having identified global changes in the liver TIME in IGF-Trap–treated mice, we sought to identify specific immune cell subtypes whose recruitment and/or phenotype were affected by this treatment. We therefore used flow cytometry (FC) and immunofluorescence microscopy (IF) to compare the immune cell infiltrates in mice treated, or not, for 2 weeks with the IGF-Trap following the intrasplenic/portal inoculation of LMP cells. We confirmed that IGF-IR activation levels were reduced in CD11b⁺ myeloid cells by immunohistochemistry (IHC) performed on liver sections from the treated mice (Supplementary Fig. S2). Consistent with the data obtained with the digital profiler, we found that the accumulation of CD11b⁺Ly6C⁺Ly6G⁺ myeloid cells, particularly (CD11b⁺Ly6C^midLy6G^high) G-MDSC was markedly reduced in the livers of IGF-Trap–treated mice, as compared with controls (Fig. 3A and see also Supplementary Fig. S3). We also found in the IGF-Trap–treated mice an increased accumulation of (CD11b⁺CD11c⁺) DC that was accompanied by a 35% increase in MHCII expression in these cells (Fig. 3B), indicative of their increased activation levels in the treated mice. Consistent with previous findings in a colon carcinoma metastasis model (36), we also found a significant increase in the proportion of Intercellular adhesion molecule (ICAM)-1⁺ (CD11b⁺Ly6G^highLy6C^low) neutrophils in the treated mice, indicative of reduced N2 polarization (Fig. 3C). Taken together, these results suggested that IGF-Trap-mediated blockade of IGF-IR signaling resulted in an altered and less immunosuppressive TIME in the liver, with an enhanced potential to mount an antitumor immune response. This conclusion was further supported by IHC performed on liver cryostat sections obtained from tumor-bearing mice. Namely, we observed a marked increase in CD11c⁺MHCII⁺ DC around metastatic foci in IGF-Trap–treated mice (Fig. 3D, top), as compared with controls. Moreover, in both control and IGF-Trap–treated mice, we detected F4/80⁺CD206⁺ macrophages around the metastases, indicative of the accumulation of M2-polarized TAM in the livers. However, the number of these TAM was significantly reduced (47.3%) in mice treated with the IGF-Trap (Fig. 3D, bottom). When G-MDSCs derived from IGF-Trap treated mice were incubated with CD3⁺ T cells, we found that suppression of T cell proliferation by the MDSC was reduced in comparison to control MDSCs (supplementary Fig. S4), although the difference did not reach statistical significance (P = 0.08), likely due to inter-sample variability. Finally, when CD11b⁺LY6G^high cells were isolated using FACS sorting and their RNA analyzed by qPCR, we found in cells isolated from the IGF-Trap–treated mice a significant reduction in Vegf and arginase (Arg1) expression (Fig. 3E), consistent with reduced immunosuppressive potency.

Taken together, these results identified in IGF-Trap–treated mice a significant shift in the TIME of PDAC liver metastases from a predominantly immunosuppressive and prometastatic TME to one that is more conducive to antitumor immune responses. To determine whether this reduced accumulation and functional potency of immunosuppressive cells was, in fact, associated with increased accumulation in the liver of CD4⁺ and CD8⁺ T cells, we performed an IHC analysis on cryostat sections derived from tumor-inoculated mice and found increased accumulation of these T cells in the IGF-Trap–treated mice (Fig. 4A).

To determine whether IGF-Trap treatment affected T cell activation and cytokine production in the liver, we isolated immune cells from the livers of LMP-inoculated mice and measured IFNγ production levels following stimulation of the cells with phorbol-12- myristate-13-acetate (PMA) and ionomycin. Flow cytometric analyses revealed increased IFNγ levels in CD3⁺ CD8⁺ T cells derived from the IGF-Trap–treated mice (Fig. 4B), suggesting an increased functional cytotoxic activity. This was also confirmed when granzyme (Grz)B expression levels were measured by qPCR in CD3⁺CD8⁺ T cells isolated by FACS from similarly treated mice, revealing increased GrzB expression levels in T cells isolated from IGF-Trap–treated mice (Fig. 4C). Finally, this decrease in the accumulation and function of immunosuppressive cells, coupled with increased T cell activation was associated with a reduction in Ki67⁺ tumor cells within metastatic foci, suggesting that the number of proliferating tumor cells was markedly reduced (Fig. 4D). Collectively, these data strongly suggested that IGF-Trap treatment caused a profound change in the otherwise, immune-tolerant and prometastatic microenvironment associated with metastatic expansion in the liver (45, 46).

HSCs are activated during the early stages of liver metastasis and play a key role in the induction of a prometastatic microenvironment in this organ (45–47). Normally quiescent within the space of Disse, they are activated in response to factors released by innate immune cells that are recruited to sites of tumor invasion, differentiate into myofibroblast-like cells and characteristically express the myofibroblastic cell surface markers α-smooth muscle actin (α-SMA) and desmin and produce type I collagen (48, 49), as well as immunoregulatory chemokines (48). Previously, we have shown that IGF-I plays a role in HSC activation in the proinflammatory microenvironment induced by tumor cell entry into the liver (41). To investigate the effect of the IGF-Trap on HSC activation following injection of LMP cells, we used the Bl6.129-Col-GFP mice, in which type I collagen (ColI) is genetically GFP-tagged and activated HSC can be identified on the basis of coexpression of GFP-ColI and α-SMA. Consistent with our previous findings (41), we found here that treatment of these mice with the IGF-Trap significantly reduced HSC activation in response to the metastatic cells (Fig. 4E).

Having observed the marked differences in the immunosuppressive landscape within the TIME of liver metastases in mice treated with the IGF-Trap, we next analyzed how these changes affected metastatic expansion following injection of LMP cells via the intrasplenic/portal route to generate experimental liver metastases. Because we have previously observed a sexual dimorphism in the control of the TIME of liver metastases (as reported in ref. 43), we performed the experiments in age-matched male and female mice to rule out sex-specific effects. In both sexes, we observed a significant reduction in the numbers and sizes of liver metastases (Fig. 5A,–H), as compared with vehicle-treated controls, when using the protocol depicted in Fig. 5I. This suggested that the changes in the TIME impacted metastatic expansion in the livers of the treated mice, regardless of their sex.

The failure of PDAC patients to respond to immunotherapy is thought to be due, at least in part, to immunosuppressive cells such as polarized TAMs and MDSCs that infiltrate the primary tumors (16, 18). While the TIME associated with PDAC liver metastases has not been extensively explored (partially because surgical resections of PDAC liver metastases are rare), we have recently documented the accumulation of immunosuppressive cells such as G-MDSCs and monocytic (Mo)-MDSC in the livers of mice bearing spontaneous or experimental LMP metastases (43). The data above have clearly shown that IGF-Trap treatment caused a reduction in the accumulation and/or polarization of several immunosuppressive cell types in the livers. Of significance, however, the increased accumulation of T cells in the treated mice was also associated with increased expression of PD-1 (Fig. 2B; Supplementary Fig. S1). We also confirmed the expression of PD ligand (PD-L)1 in LMP cells, as well as in a second PDAC cell line the FC1199 cells syngeneic to C57Bl/6 mice (Supplementary Fig. S5). We asked therefore whether combining a PD-1 inhibitor with the IGF-Trap will have a reciprocal enhancing effect, potentiating antitumor immunity in this organ, and further reducing the metastatic burden. Mice inoculated with LMP cells via the intrasplenic/portal route were treated on alternate days with 5 mg/kg IGF-Trap or 10 mg/kg of an anti-murine PD-1 antibody (or a nonimmune IgG isotype as control) for a total of 5 injections each over a period of 15 days (see time line in Supplementary Fig. S6). Liver metastases were then enumerated 21 days post tumor injection and compared with mice injected with each of these inhibitors alone. As expected, treatment with the IGF-Trap reduced liver metastasis when used alone and the addition of the IgG isotype control to the IGF-Trap had no effect on the metastatic burden. Treatment with the anti–PD-1 antibody alone also reduced the number of liver metastases, suggesting that blockade of this immune checkpoint enabled an antitumor immune response in these mice. These tumor-inhibitory effects were enhanced, when the two treatments were combined and this was observed in both male (Fig. 6A,–E) and female (Fig. 6F,–J) mice, and reflected in further reductions in both the median number and mean size of the metastases, revealing an additive inhibitory effect on the growth of liver metastases when the two drugs were combined. This marked effect of the combinatorial immunotherapy on liver metastases was not specific to the strain or the LMP model, as similar effects were also observed when C57Bl/6 mice were injected with the syngeneic FC1199 cells and treated in a similar manner (Supplementary Fig. S7). Furthermore, mice that were continuously treated with the combinatorial immunotherapy had a significantly prolonged survival, as compared with untreated mice, whereas treatment with the IGF-Trap or anti–PD-1 antibodies alone had a more minor (and not statistically significant) effect on survival (Fig. 6K).

To elucidate the mechanism of action of the combinatorial treatment, we compared the immune cell infiltrate in mice injected with LMP cells treated with each of the inhibitors alone or with the combination, using FC and IF. Tumor-injected mice were treated as indicated (Fig. 7) and their livers analyzed 14 days post tumor inoculation. As expected, we observed a significant decrease in MDSC accumulation in IGF-Trap–treated mice, as well as in mice that received the combination therapy, as compared to control mice or mice treated with the anti–PD-1 antibody, and this was particularly evident for the accumulation Ly6G^high G-MDSCs (Fig. 7A). In mice treated with anti-PD-1 antibodies (alone or in combination with IGF-Trap), we observed a significant increase in the accumulation of both CD4⁺ and CD8⁺ T cells and a corresponding decrease in PD-1 expression on CD8⁺ cells, and this was confirmed by both FC and IF. However, no change was observed in the expression of CTLA-4 or LAG3 on these cells (Fig. 7B,–D). Finally, FC revealed a marked increase in IFNγ-producing CD8⁺ T cells in mice treated with the combinatorial therapy as compared to single agent–treated mice (Fig. 7E) and in addition, GrzB mRNA levels in T cells isolated from mice treated with IGF-Trap and combinatorial therapy were also significantly increased relative to controls (Fig. 7F). Together, these data showed a potentiation of a T-cell response that was optimal in mice treated with the combination therapy.

Despite recent successes of single-agent immunotherapy in the treatment of several highly aggressive malignancies, the majority of cancers, including PDAC remain unresponsive. This, despite evidence of T cell–activating tumor antigens on PDAC cells (reviewed in refs. 50, 51). Among several factors that may contribute to this failure to respond are the production of immunosuppressive IL10 and TGFβ by PDAC cells and the recruitment to the tumor site of immunosuppressive cells such as MDSC and M2 TAM (ref. 52; reviewed in ref. 19). This may be particularly true in the liver, where recruitment of immunosuppressive bone marrow–derived myeloid cells (BMDC) was documented prior to (53, 54), and following tumor cell entry into the liver (45, 46). Thus, combination therapies that can activate cytotoxic T cells, while also targeting immunosuppressive cells in the TME hold much promise in the treatment of this disease. In this study, we show for the first time, that targeting the IGF axis alters the TIME of PDAC liver metastases, reducing the recruitment and activity of several immunosuppressive cell types and resulting in increased T-cell accumulation in the liver. Treatment with anti–PD-1 antibodies, on the other hand, while not directly increasing T-cell recruitment, decreased the proportion of PD-1⁺ T cells. Our results suggest that the combined effect of increased T-cell numbers (caused mainly by IGF-Trap treatment) and decreased T-cell exhaustion (due to anti PD-1 antibody) was an enhancement of T cell activity, and this resulted in the more potent antimetastatic effect of the combinatorial therapy (see diagram in Fig. 8).

Intriguingly, although the growth of liver metastases in animals treated with the IGF-Trap was inhibited, we did not observe a difference in the growth of the local, orthotopically implanted pancreatic tumors in the same mice. We have previously documented differences in the TME of local pancreatic LMP tumors and their hepatic metastases. While a high proportion of MDSC identified in the liver were of the Mo-MDSC subset, this population was lacking in the pancreatic tumors (43). Mo-MDSC were among immune cell types affected by IGF-Trap treatment and their absence in the pancreatic tumors may be one factor affecting the differential response to the treatment. In addition, we observed that the liver is the main site of IGF-Trap accumulation (35), and a lower bioavailability in the pancreatic tumors may have contributed to the divergent effect.

We have shown here that IGF-Trap treatment profoundly altered the immune landscape of liver metastases, affecting a multiplicity of immune cell types. IGF-IR is widely expressed on innate and adaptive immune cells (55, 56), and IGF-IR blockade could, therefore, affect the recruitment and/or function of each of these cell types directly. Alternatively (or in addition to), this broad effect may be due to transcriptional regulation of a central factor regulating immunosuppression in the liver. We have previously shown that treatment of bone marrow-derived CD11b⁺Ly6G⁺ cells with IGF-I upregulated the expression of both TGFβ1 and VEGF (36), and as we have shown here (Fig. 2; Supplementary Fig. S1), IGF-Trap treatment significantly reduced TGFβ1 levels in the TME of liver metastases, at both early and late stages of expansion. This could account for the overall reduction in the accumulation of tumor-associated immunosuppressive cells such as MDSC, M2 macrophages, and N2 neutrophils that are commonly regulated by TGFβ (46). Furthermore, because the recruitment of neutrophils and monocytes are early events in the mobilization of innate immune cells into the ME of liver metastases, triggering later events such as activation of HSC that are an additional source of TGFβ (45), a blockade of IGF-IR signaling in these cells could result in a general decrease in the recruitment of immunosuppressive cells and thereby, increased accumulation and activation of T cells. Using a tumor-free model, we have also previously shown that reduced IGF-I bioavailability impairs hepatic stellate cell activation directly (41). Thus, direct and indirect effects on HSC could contribute to a state of enhanced antitumor immunity. We have previously reported that although the IGF-Trap consists of the extracellular domain of the human receptor, no humoral response to this protein could be detected in injected mice (33). However, we cannot entirely rule out the possibility that an immune response to the IGF-Trap played some role or that a direct effect on tumor cell proliferation in the liver also contributed.

Our results add to the growing body of evidence that immunotherapy can be rendered more effective when combined with strategies that target the TME in PDAC and other advanced solid and treatment-refractory tumors. Some combinations with targeted therapy (e.g., ref. 57 and www.clinicaltrials.org - NCT02758587) and chemotherapy (NCT03977272, NCT03983057) have, in fact, already advanced to clinical trials (reviewed in ref. 19). Diana and colleagues, analyzing the prognostic value of PD-1/PD-L1 expression together with infiltration of CD8⁺ lymphocytes and Treg in 145 surgical PDAC resections found that both PD-1⁺ and CD8⁺ T-cell infiltrates were independent prognostic markers in patients treated with adjuvant chemotherapy, predicting a better outcome (58). Moreover, a genomic analysis recently identified a PDAC subtype with increased activation of CD8⁺ T cells and overexpression of CTLA-4 and PD-1 within surgical resections that corresponded to higher frequency of somatic mutations and tumor-specific neoantigens (59), suggesting that patient stratification based on mutational signatures may identify a patient subpopulation particularly sensitive to immunotherapy. Collectively, these data identify PDAC as a malignancy that may be highly responsive to immune-based therapy under optimizing conditions. In our model, we found a significant effect on liver metastases when mice were treated with a single checkpoint inhibitor, namely anti–PD-1 antibodies, despite an unaltered expression of both CTLA-4 and LAG3 on CD8⁺ T cells. This suggests that the effect of immunotherapy, in this model, may be even further optimized by combining several immune checkpoint blockers with IGF-Trap treatment, and this may also be required in the clinical management of this disease.

Liver metastases were identified as one of several factors predictive of poor response to immunotherapy (60, 61). This could be a contributing factor to the resistance of PDAC to immunotherapy, as a large proportion of patients with PDAC already harbor hepatic metastases at the time of diagnosis or relapse with liver metastases following surgical excision of the primary tumor. Our data, taken together with other findings, suggest that for patients with resectable primary PDAC tumors that are free of liver metastases at the time of diagnosis, combining immunotherapy with targeted anti-IGF-IR therapy could improve treatment outcome.

A.M. Lowy reports personal fees from Tanabe, Corcept, Autobahn, and personal fees from Kinnate outside the submitted work. P. Brodt reports grants from MITACS, McGill University Health Center Foundation, and grants from Canadian Institute for Health Research during the conduct of the study; in addition, P. Brodt has a patent for Soluble IGFR Fc Fusion Proteins and Uses Thereof Application No. 12857915.8 Case Ref. 05001770-467EP European Patent Office issued, a patent for Novel Soluble IGF Receptor Fc Fusion Proteins For The Treatment Of Cancer And Metastasis - US - International (PCT) Patent Application No. PCT/CA2012/050899 United States issued, and a patent for Therapeutic Applications Of Type 1 Insulin-Like (IGF-1) Receptor Antagonists - United States Provisional Patent Application No. 63/120,442. pending. No disclosures were reported by the other authors.

M. Hashimoto: Data curation, formal analysis, investigation, methodology, writing–original draft. J.D. Konda: Data curation, methodology. S. Perrino: Data curation, methodology. M.C. Fernandez: Data curation, methodology. A.M. Lowy: Resources, writing–review and editing. P. Brodt: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing.

The authors acknowledge with gratitude the staff of the Immunophenotyping Facility of the Research Institute of the McGill University Health Center and in particular, Marie-Helene Lacombe and Ekaterina Iourtchenko, for their help with flow cytometry and cell sorting, Simon Milette for his assistance with flow cytometry analyses, the NanoString Technologies scientific staff, in particular Drs. Liang Zhang and Yan Liang, for excellent support in the analysis of the GeoMax immune profiling data and Dr. Matthew Leibovitch (Research Institute of the Mcgill University Health Center) for his help with data analysis. Dr. Masakazu Hashimoto was the recipient of the Oversea Medical Research Grant from Sumitomo Life welfare and Culture Foundation, a Research Fellowship from the Uehara memorial Foundation and a fellowship from the Research Institute of the McGill University Health Center. This work was made possible by grant PP2-141724 from the Canadian Institute for Health Research, and grant IT04483 from MITACS (to P. Brodt). Research support from the MUHC Foundation and the MI4 initiative of McGill University (to P. Brodt) and NIH award CA155620 (to A.M. Lowy) are also gratefully acknowledged.

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.