Combination systemic therapies with immune checkpoint inhibitors in pancreatic cancer: overcoming resistance to single-agent checkpoint blockade

Immune checkpoint inhibitors have demonstrated broad single-agent antitumor activity and a favorable safety profile that render them attractive agents to combine with other systemic anticancer therapies. Pancreatic cancer has been fairly resistant to monotherapy blockade of programmed cell death protein 1 receptor, programmed death ligand 1, and cytotoxic T-lymphocyte associated protein 4. However, there is a growing body of preclinical evidence to support the rational combination of checkpoint inhibitors and various systemic therapies in pancreatic cancer. Furthermore, early clinical evidence has begun to support the feasibility and efficacy of checkpoint inhibitor-based combination therapy in advanced pancreatic cancer. Despite accumulating preclinical and clinical data, there remains several questions as to the optimal dosing and timing of administration of respective agents, toxicity of combination strategies, and mechanisms by which immune resistance to single-agent checkpoint blockade are overcome. Further development of biomarkers is also important in the advancement of combination systemic therapies incorporating checkpoint blockade in pancreatic cancer. Results from an impressive number of ongoing prospective clinical trials are eagerly anticipated and will seek to validate the viability of combination immuno-oncology strategies in pancreatic cancer.


Introduction
Monoclonal antibodies targeting the programmed cell death protein 1 receptor (PD-1) and programmed death ligand 1 (PD-L1) are approved as cancer immunotherapy for a number of solid tumors and hematologic malignancies [1]. Early studies demonstrated expression of PD-L1 in human pancreatic cancer tissues associated with poor prognosis and evidence of antitumor activity with PD-1/ PD-L1 blockade in pancreatic cancer mouse models in vivo [2][3][4][5]. However, pancreatic cancer has been fairly resistant to single-agent checkpoint blockade in the clinical setting as initial phase I trials enrolling advanced pancreatic cancer patients produced overall response rates (ORRs) of 0% with anti-PD-1 and anti-PD-L1 therapy [6][7][8]. Similarly, no objective responses were seen in locally advanced or metastatic pancreatic cancer patients treated with cytotoxic T-lymphocyte associated protein 4 (CTLA-4) inhibitor monotherapy [9]. These negative but important clinical studies underscored the primary or innate resistance of pancreatic cancer to checkpoint inhibitors rather than acquired resistance, which would otherwise be seen in those that initially respond but eventually develop resistance to these agents [10].
There is increasing evidence to suggest that both tumor cell-intrinsic and tumor cell-extrinsic factors contribute to the primary resistance of pancreatic cancer to checkpoint blockade. Immune active tumors that are sensitive to checkpoint inhibitors such as melanoma, lung squamous cell carcinoma, or lung adenocarcinoma are characterized by an abundance of CD8+ tumor-infiltrating lymphocytes (TILs), while pancreatic cancer (except those with defects in mismatch repair) represents an immune quiescent tumor characterized by lack of infiltration by effector T-cells that is otherwise critical in driving the antitumor response to checkpoint blockade [11]. This property of pancreatic adenocarcinoma (PDAC) has been attributed, in part, to the prominent PDAC desmoplastic stroma that may impede access to the tumor by TILs [12]. Additionally, PDAC is intrinsically a low mutational burden or low neoantigen-expressing tumor, which is important given that higher mutational load correlates to higher levels of neoantigens capable of inducing antitumor responses to immune checkpoint blockade [12].
Other tumor cell-intrinsic and tumor cell-extrinsic mechanisms of resistance to immune checkpoint inhibitors in pancreatic cancer include (1) the ability of pancreatic cancer cells to evade the host antitumor immune response (immunoediting) through expression of immune checkpoints such as PD-L1 and indoleamine-2,3-dioxygenase (IDO), secretion of granulocyte-macrophage colony-stimulating factor (GM-CSF) resulting in a myeloid cell-inflamed phenotype, and upregulation of regulatory T-cells (Tregs) and (2) induction of immune tolerance by direct interaction between cancer cells and tumor antigen-specific T-cells (immune privilege) through downregulation of antigen presenting major histocompatibility complex (MHC) molecules, expression of Fas ligand and decreased Fas receptor signaling, and expression of Foxp3 [13,14].
Evidence is emerging to support combination systemic therapies on a backbone of immune checkpoint inhibition to overcome resistance to single-agent PD-1/PD-L1/ CTLA-4 blockade in pancreatic cancer. PD-1/PD-L1 inhibitors, in particular, have shown broad single-agent activity across a spectrum of cancers with safety and tolerability profiles that render them attractive agents for combination with other anticancer therapies [1]. In this review, we highlight the current developmental landscape of combination regimens incorporating systemic therapies and checkpoint blockade in pancreatic cancer. In particular, we review the preclinical evidence supporting the rational combination and transition to prospective clinical trials that have thus far reported on the safety and efficacy of combination systemic therapies with checkpoint inhibitors in pancreatic cancer. We end with a discussion on future considerations that are important to optimizing the antitumor efficacy of these combinations in this lethal malignancy.

Search criteria
A literature search up to April 17, 2018 using the keywords "pancreatic cancer, " "PD-1, " "PD-L1, " "checkpoint, " and "immunotherapy" was conducted in MEDLINE and generated a total of 1836 hits. Preclinical or prospective clinical studies investigating combination regimens involving PD-1, PD-L1, or CTLA-4 inhibitors and ≥ 1 systemic therapies in the treatment of pancreatic cancer were included. An additional manual search was performed to include preliminary results from relevant abstracts investigating combination therapies. Only studies published in English language were included. Using these criteria, the list of studies was narrowed to a total of 30 preclinical studies (22 fully published and 8 abstracts) and 14 prospective clinical studies (5 fully published and 9 abstracts) that were included in this review.

Chemotherapy
Among the earliest preclinical studies demonstrating synergistic antitumor effects with combination checkpoint inhibition and chemotherapy in pancreatic cancer involved the administration of anti-PD-L1 therapy and gemcitabine in mice models in vivo without overt toxicity (Table 1). Gemcitabine with delayed anti-PD-L1 therapy (≥ 14 days after gemcitabine) and gemcitabine with simultaneous PD-L1 blockade showed enhanced tumor suppression compared to either modality alone; however, only simultaneous combination therapy resulted in complete responses (CRs) in treated mice [4]. In a transgenic mouse model of resectable PDAC, neoadjuvant PD-1 inhibition and gemcitabine significantly reduced local recurrence and improved survival compared to either modality alone, while promotion of natural killer (NK) cell activation with the addition of anti-CD96 antibody to adjuvant gemcitabine enhanced control of distant metastases [15]. This preclinical model therefore highlighted the potential of combination strategies with PD-1 blockade to target acquired resistance (preventing recurrence) in addition to the more commonly investigated goal of enhancing response to PD-1 inhibition (targeting primary resistance). Administration of cisplatin using nanoparticle carriers along with anti-PD-L1 therapy produced preliminary evidence of enhanced tumor suppression in an orthotopic mouse model of pancreatic cancer [16].
Targeting of either the H3K4 methylation-specific histone methyltransferase, mixed-lineage leukemia 1 (MLL1), with the epigenetic agent verticillin A, JAK/ STAT pathway with ruxolitinib, mitogen-activated protein kinase (MAPK) pathway with a MAPK kinase (MEK inhibitor, cholecystokinin (CCK) receptor, DNA methyltransferase with decitabine, or Bcl-2 in combination with checkpoint blockade significantly enhanced tumor growth suppression, when compared to controls, in pancreatic cancer-carrying mouse models [19][20][21][22][23][24]. MLL1 normally catalyzes the trimethylation of H3K4 to activate immune inhibitory PD-L1 transcription in tumor cells [19]. The JAK/STAT pathway putatively upregulates PD-L1 expression and immunosuppressive cytokine production by tumor cells that altogether decrease effector T-cell function [21]. Myeloid cells have been shown to support immune evasion in PDAC by upregulating PD-L1 expression in a MAPK-dependent manner [22]. CCK has been implicated in increasing fibrosis and reducing the influx of TILs in pancreatic cancer [23]. Decitabine, a DNA hypomethylating agent, was shown to increase the amount of CD8+ TILs as monotherapy. This prompted the addition of PD-1 antibody following decitabine treatment that showed greater suppression of tumor growth compared to either agent alone in PDAC mice [24]. Lastly, blocking of Bcl-2-Associated athano-Gene 3 (BAG3) has been shown to decrease the number of immunosuppressive TAMs in PDAC [20].

Tumor microenvironment
Preclinical evidence posits that the acidic pH of the TME has immunosuppressive effects by inhibiting T-cell activation and abrogating interferon-γ (IFN-γ) and tumor necrosis factor alpha (TNF-α) secretion [25]. Buffer therapy with sodium bicarbonate in drinking water and anti-PD-1 therapy significantly diminished tumor growth in a PDAC mouse model compared to either therapy alone (Table 1). Given the concerns for translating such high doses of sodium bicarbonate in the clinical setting, the same group has preliminarily demonstrated the ability to increase sensitivity to anti-PD-1 monotherapy by combining the carbonic anhydrase IX (CAIX) inhibitor DH348 or lactate dehydrogenase A inhibitor FX11 with anti-PD-1 therapy in PDAC-carrying mice [26].
Hyaluronan has been shown to contribute to tumor promotion and depletion by PEGylated recombinant human hyaluronidase PH20 (PEGPH20) 24 h prior to anti-PD-1 or anti-PD-L1 therapy significantly suppressed tumor growth, when compared to either modality alone, in high hyaluronan-expressing pancreatic tumor mice models [27]. Gene expression of immunosuppressive markers such as interleukin 10 (IL-10) and Foxp3 was higher in hyaluronan-high tumors and suggested a relationship between hyaluronan level and immune suppression.

Vaccines
Combining cyclophosphamide and a GM-CSF cell-based vaccine (GVAX) with anti-PD-1 or anti-PD-L1 antibodies in mouse models of pancreatic cancer showed significantly increased survival compared to anti-PD-1 monotherapy (Table 1) [28]. Notably, combination systemic therapy was associated with significantly increased IFNγ-producing CD8+ TILs in the metastatic PDAC TME, decreased Tregs, and decreased CTLA-4 expression on CD4+ and CD8+ T-cells when compared to controls. The same group later demonstrated preliminary antitumor efficacy with the combination of Annexin A2-specific Listeria monocytogenes vaccine and anti-PD-1 therapy in a PDAC mouse model [29]. The pairing of checkpoint blockade to local radiation therapy (RT) and a SIY antigen vaccine enhanced tumor regression in otherwise immune quiescent pancreatic cancer mouse models (Table 1) [30].

Cytokines and chemokines
Production of chemokines such as CCL2, CXCL12, and CXCR4 contribute to the immunosuppressive PDAC TME by facilitating T-cell trapping in the stroma and effector T-cell exclusion [31]. Regulation of cytokines such as interferons and TNF-α also contribute to the immunosuppressive PDAC TME through upregulation of PD-L1 [31]. Targeting the chemokine (C-X-C motif ) ligand 12 (CXCL12) through the chemokine (C-X-C motif ) receptor 4 (CXCR4) inhibitor, AMD3100, in PDAC-bearing mice with anti-PD-L1 therapy significantly reduced tumor volume by 48 h, when compared to controls, with no further decreases in tumor volume over the following 4 days [32]. Interestingly, CTLA-4 blockade did not augment the antitumor effect of AMD3100. A subsequent study established the efficacy of a bispecific PD-L1 and CXCL12 fusion protein or trap with > 1000× higher affinity for mouse PD-L1 than that between endogenous PD-1 and PD-L1 [33].
In preclinical PDAC mouse models, combined targeting of PD-L1 and IL-6 correlated with increased intratumoral effector T-cells and increased T-cells with a Th1 phenotype, while inhibiting pancreatic cancer growth compared to either modality alone [34]. Targeting of IL-18, CCR2, or CXCR2 in combination with PD-1/ PD-L1 blockade has shown preclinical anticancer efficacy as well (Table 1) [35][36][37]. Delivery of an antitumor cytokine payload demonstrated feasibility in PDAC mice treated with single-domain antibodies against PD-L1 fused with IFNγ or IL-2 in vivo; reduced tumor burden seen from targeting IFNγ was associated with decreased numbers of CD11b + cells and transition of intratumoral macrophages towards an M1-like phenotype [38].

Adoptive T-cell therapy
In initial investigations of fusion receptor constructs comprised of PD-1 and the costimulatory protein CD28 transduced into transgenic murine CD8+ T-cells specific for ovalbumin (OT-1), complete tumor regressions in PDAC-carrying mice were observed with 300-fold increases in IL-2 and IFN-γ production and increased T-cell proliferation [39]. Reimplanted tumors were rejected in 9/11 treated mice vs. 0/6 naïve mice, which was indicative of a memory response. The same group has recently presented preliminary findings of synergistic T-cell-induced tumor cell cytotoxicity in mouse pancreatic cancer cell lines cocultured with OVA-specific CD4+ and CD8+ T-cells transduced with a PD-1-CD28 fusion receptor [40].

Immune costimulatory proteins and immunostimulants
In studies demonstrating resistance to single-agent checkpoint blockade in PDAC-bearing mice, addition of CD40 (antigen presenting cell costimulatory protein) agonistic antibody and/or chemotherapy reversed refractoriness to checkpoint blockade by priming T-cell responses (Table 1) [41,42]. Combination therapy significantly improved survival (targeting primary resistance) and conferred immunologic memory as demonstrated by curative protection from multiple tumor rechallenges, when compared to controls (targeting acquired resistance). Triple therapy with mAb-AR20.5 (anti-MUC1), anti-PD-L1 therapy, and PolyICLC (immunostimulant) cured 50% of mice subcutaneously injected with PDAC cells by 70 days and retained immunologic memory as evidenced by tumor antigen-specific rejection of tumors reimplanted in treated mice but not in control mice [43]. Preliminary antitumor efficacy has also been shown with triple therapy involving the immunomodulatory antifungal ciclopirox olamine, anti-PD-1, and anti-CTLA-4 therapy in PDAC-carrying mice [44].

Vaccines
In an open-label phase Ib trial, patients with previously gemcitabine-treated, advanced PDAC were randomized to receive ipilimumab (arm 1) or ipilimumab with GVAX (arm 2) and demonstrated grade 3-4 immune-related adverse events (AEs) in 20% of patients in both arms (colitis, Guillain-Barre syndrome (GBS), and nephritis in arm 1 and colitis, rash, and pneumonitis in arm 2) that resolved to grade 1 with steroids (except for the cases of nephritis and GBS) [45]. There was a trend towards significant improvement in overall survival (OS) in favor of the combination arm despite the small sample size of 30 in this study (Table 2).
A German group isolated antigen-primed monocyte derived dendritic cells (DCs) from 44 patients with stage IV pancreatic cancer who failed first-line chemotherapy and demonstrated a median OS of 8 months with DC vaccine alone; however combination DC vaccine and PD-L1 blockade was able to induce secondary stabilization of disease of 4-8 months in 5/10 patients who failed to respond to previous DC therapy [46]. Preliminary results including 2 partial responses (PRs) were observed from a pilot study by the same group investigating lower dose nivolumab in combination with DC vaccine therapy in 7 patients with stage IV pancreatic cancer (Table 2) [47].

Chemotherapy
The CTLA-4 inhibitor tremelimumab in combination with weekly gemcitabine demonstrated preliminary efficacy and tolerability with the most common grade 3-4 toxicities being asthenia (11.8%) and nausea (8.8%) in a phase Ib trial enrolling treatment-naïve metastatic pancreatic cancer patients [48]. A dose-finding, multi-arm phase Ib trial enrolled patients with advanced solid tumors to 6 different treatment arms; in the metastatic PDAC cohort, a final 17 patients with treatment-naïve or previously-treated disease received combination pembrolizumab, gemcitabine, and nab-paclitaxel with responses observed only in those who were previously untreated (Table 2) [49,50]. Notably, gemcitabine and nab-paclitaxel were reduced to 800 mg/m 2 and 100 mg/ m 2 in the pretreated cohort, which was ultimately closed due to futility [50]. Immune-related AEs (all grades) were seen in 47.1%. The most common grade 3-4 AEs were neutropenia (46.7%) and thrombocytopenia (20%) in the treatment-naïve cohort. The combination of ipilimumab and gemcitabine was well-tolerated but produced an ORR of 12.5% and lower than that of gemcitabine alone, historically [51]. Alternatively, the feasibility and preliminary efficacy of nivolumab + nab-paclitaxel (arm A) and nivolumab + gemcitabine + nab-paclitaxel (arm B) has been demonstrated in locally advanced or metastatic PDAC where the most common grade 3-4 AEs in arm A were pulmonary embolism, neutropenia, and anemia in 2/11 patients (18%) and anemia in 2/6 (33%) in arm B [52]. Neoadjuvant pembrolizumab with concurrent capecitabine and RT was relatively tolerated in resectable or borderline resectable PDAC with no grade 4 toxicities reported and no major surgical complications reported within 30 days post-surgery (Table 2) [53].

Targeted therapies
Preliminary findings from a phase Ib dose-finding trial of a PD-1 inhibitor plus PARP inhibitor in patients with treatment-refractory advanced solid tumors and demonstrated the feasibility of this pairing (Table 2) [55]. Furthermore, a PR and prolonged stable disease (SD) of up to 281 days were observed in pancreatic cancer patients treated with this combination.

Other immune checkpoint inhibitors
Preliminary results from a phase I Japanese trial investigating the oral IDO inhibitor epacadostat and pembrolizumab showed a PR thus far in1 pancreatic cancer patient out of 15 patients with treatment-refractory advanced solid tumors (Table 2) [56]. In the overall cohort, there was 1 dose-limiting toxicity (DLT) of grade 4 rhabdomyolysis in the epacadostat 100 mg + pembrolizumab arm and 12 patients experienced all-grade AEs (80.0%) while 2 patients (13.3%) had grade 3 liver disorder and grade 4 rhabdomyolysis (1 each) with the combination.
A recent phase II trial reported preliminary findings of modest efficacy with durvalumab and tremelimumab in the second-line treatment of metastatic pancreatic cancer (Table 2) [57]. There were more grade ≥ 3 treatment-related AEs with combination durvalumab and tremelimumab (22%) than single-agent durvalumab (6%). Grade ≥ 3 treatment-related AEs seen in the combination arm included diarrhea (9.4%) and fatigue (6.3%), while ascites (3.1%), hepatitis (3.1%), and increased lipase (3.1%) were among those observed in the monotherapy arm. Discontinuation of therapy occurred in 9.4% and 3.1% of patients in the combination and monotherapy arms, respectively.

Fusion proteins
A pilot phase I study demonstrated a PR in 1 patient with mismatch repair (MMR) deficiency out of 5 patients with pretreated advanced pancreatic cancer using a bifunctional fusion protein of anti-PD-L1 antibody fused to the extracellular domain of TGFβ receptor II (TGFβ trap) [58]. A maximum-tolerated dose (MTD) was not reached at the highest dose level of 20 mg/kg every 2 weeks (Table 2).

Mechanisms of immune resistance and rational combination strategies
The mechanisms underlying the resistance of pancreatic cancer to the antitumor immune response have been extensively reviewed and are largely attributed to perturbations in immune surveillance, the process of immunoediting, and immune privilege [13,[59][60][61]. Although pancreatic cancer has been relatively resistant to singleagent checkpoint blockade when compared to more immune sensitive tumors, strategies for overcoming primary resistance to immunotherapy in pancreatic cancer are readily available from investigations in other solid tumor types focused on improving the antitumor efficacy of checkpoint blockade through combining various therapeutic modalities on a checkpoint inhibitor backbone [62][63][64]. Attractive agents to combine with checkpoint inhibitors should, in principle, directly stimulate CTLs, inhibit tumor-induced immunosuppressive factors, inhibit Tregs, and/or activate NK cell activity ideally through nonredundant pathways [63]. Here, for example, immunochemotherapy has been widely established across several tumor types where: (1) gemcitabine can increase class I human leukocyte antigen (HLA) expression, tumor antigen cross-presentation, and selectively eliminate MDSCs, (2) docetaxel can decrease immunosuppressive MDSCs, (3) paclitaxel can stimulate antigenpresenting cells and improve cancer cell permeability to granzyme B, (4) irinotecan can decrease Tregs and MDSCs, and (5) doxorubicin can promote immunogenic cell death, increase cancer cell permeability to granzyme B, and enhance antigen presentation by dendritic cells [49]. It has been shown, however, that vinorelbine may result in a bystander effect or the inadvertent death of neighboring immune cells that may interfere with the ability to mount an antitumor immune response [65].
Targeting the TME to attenuate immunosuppression, induce immunogenic tumor cell death, enhance antigen presentation, and/or prolong survival of immune-effector cells represents another approach to boost the anticancer immune response given the well-established immunosuppressive features of the PDAC TME [66]. Vitamin D priming may also serve as a potential adjunct to PDAC therapy given that the vitamin D receptor is expressed in the stroma of human pancreatic tumors and is involved in stromal remodeling and increased intratumoral drug delivery that may result in tumor volume reduction in combination strategies [67]. Furthermore, although checkpoint blockade alone is effective in removing immune suppression, it potentially does not provide a sustaining means for immune activation; strategies on priming antitumor immune-effector cells or rescuing dysfunctional immune-effector cells are also becoming increasingly recognized concepts to further improve the efficacy of checkpoint inhibitors in pancreatic cancer [11,62,68].
Indeed, a growing body of preclinical evidence supports the incorporation of checkpoint inhibitors in combination strategies with systemic therapies that address several pathways contributing to the immune resistance of pancreatic cancer (Fig. 1). For example, combination regimens incorporating checkpoint blockade have been shown to: (1) decrease TAMs and reprogram TAMs to promote antigen presentation and antitumor T-cell activity, (2) increase CD8+ TILs, decrease CD4+Foxp3+  Tregs, and improve effector T-cell/Treg ratios in tumors often associated with increased IL-2 and IFN-γ production and a Th1 phenotype, (3) upregulate PD-L1 and CTLA-4 on PDAC cells, (4) decrease numbers of CD11b + cells with transition of intratumoral macrophages towards an M1-like phenotype, and (5) overcome the fibrotic and immunosuppressive TME of PDAC, while altogether producing enhanced antitumor activity when compared to controls [17,18,28,34,38,39]. Notably, these studies fall into the same category as the overwhelming number of preclinical and clinical studies to date that have investigated the ability of combination regimens incorporating to PD-1/PD-L1/CTLA-4 inhibitors to target primary resistance or improve response to checkpoint blockade in pancreatic cancer (Tables 1 and  2). There is a small but relevant number of studies that have highlighted the potential for combination strategies with checkpoint inhibitors to target acquired resistance or prevent recurrence through the ability to induce immunologic memory to subsequent tumor rechallenge and reverse resistance to single-agent checkpoint blockade across several preclinical PDAC mouse models [39,[41][42][43].

Candidates for combinations with immune checkpoint inhibitors
Based on the above rationale, a roadmap for partnering of therapeutic modalities with PD-1, PD-L1, and CTLA-4 inhibitors can be broadly conceptualized to strategies that can: (1) convert non-T-cell inflamed or immunologically "cold" tumors to T-cell inflamed or immunologically "hot" tumors and (2) enhance or rescue responses achieved with single-agent checkpoint blockade [69]. For the first approach that is essentially targeting primary resistance, evidence exists for vaccines, oncolytic viruses, immune cell co-stimulatory agonists (CD137 (4-1BB), CD134 (OX40), glucocorticoid-induced TNF receptor (GITR), and CD40), adoptive T-cell therapies and chimeric antigen receptor (CAR) T-cells, chemotherapy, RT, and targeted therapies as candidate agents for combination in this category [69]. The combination of RT and checkpoint blockade, in particular, is being explored extensively in clinical trials across numerous tumor types given the ability of RT to prime antitumor T-cell responses; the development of this combination as cancer therapy has been extensively reviewed elsewhere [70]. For the second approach that can target acquired resistance, inhibitors of other immunosuppressive molecules or immune checkpoints such as IDO, TGFβ inhibitors, angiogenesis inhibitors, and Treg depletion represent potential candidates for combination with anti-PD-1/ PD-L1/CTLA-4 antibodies [69].

Translation into clinical settings: dosing, timing, toxicities, and treatment setting
To date, only phase I-II trials have reported out with various combination regimens incorporating PD-1, PD-L1, and CTLA-4 inhibitors in PDAC patients and overall, they have shown modest efficacy that certainly represent improvements over the dismal response rates seen in clinical trials of single-agent checkpoint blockade in advanced PDAC (Table 2) [6][7][8][9]. However, combination therapies with checkpoint inhibitors in pancreatic cancer are still in early phases of clinical development and there are no large, randomized phase III trials for comparison to current treatment standards. Nevertheless, there are several points to consider from current clinical trials investigating combination regimens integrating checkpoint blockade in pancreatic cancer that will likely inform future studies in this area.
In an early phase I trial investigating the combination of first-line tremelimumab and gemcitabine in metastatic pancreatic cancer, the pharmacokinetics (PKs) of gemcitabine was similar in the absence or presence of tremelimumab at the highest dose level of 15 mg/kg [48]. The MTD was full-dose gemcitabine 1000 mg/m 2 (weekly for 3 weeks with 1 week off ) and tremelimumab 15 mg/kg on day 1 every 84-day cycles. The most common grade 3-4 treatment-related AE at the MTD was neutropenia (18%) and toxicity associated with combination therapy was essentially similar to that of tremelimumab monotherapy. Several phase I-II trials have since demonstrated the feasibility of adding standard-dose checkpoint inhibitors to full-dose gemcitabine and/or nab-paclitaxel in the firstand second-line treatment settings in advanced pancreatic cancer (Table 2) [49][50][51][52]54]. Immune-related AEs have ranged from 5.9% to 47.1% (all grades) with combination chemotherapy and anti-PD-1 or anti-CTLA-4 therapy [48,50]. Premedication with intravenous dexamethasone 12 mg on the day of chemotherapy infusion was shown to decrease grade 3-4 AEs (2.2 events/ patient) compared to those without premedication (1.1 events/patient) [50].
Efficacy to combination chemotherapy and checkpoint blockade appears to be better in the treatment-naïve setting compared to the second-line setting in advanced PDAC, which posits that combination regimens with checkpoint inhibitors may be better served when introduced earlier in the treatment sequencing of metastatic PDAC regimens [50,52]. There appears to be increased toxicities with greater numbers of agents or therapeutic modalities included with checkpoint inhibition [52][53][54]. Recently, the most promising efficacy seen to date occurred in a phase II trial investigating the quadruplet regimen of gemcitabine + nab-paclitaxel + durvalumab + tremelimumab in treatment-naïve metastatic PDAC with the most common grade ≥ 3 AEs being hypoalbuminemia (45%), abnormal lipase (45%), anemia (36%), fatigue (27%), abnormal white blood cells (27%), and hyponatremia (27%) [54]. In phase I-II trials to date, combination chemotherapy with PD-1/PD-L1/CTLA-4 inhibitors have employed concurrent administration of chemotherapy and checkpoint blockade [48][49][50][51][52]54]. In the neoadjuvant treatment of borderline resectable PDAC, chemoradiation was also initiated concurrently with pembrolizumab in a phase Ib/II trial [53]. There is preclinical data to support simultaneous administration given that concurrent chemotherapy and PD-L1 blockade resulted in CRs that were not seen with delayed anti-PD-L1 therapy (≥ 14 days after gemcitabine) [4]. Notably, in a phase I dose-escalation trial of first-line gemcitabine and tremelimumab in metastatic pancreatic cancer, addition of tremelimumab 1 month after initiation with gemcitabine showed 0 grade 3-4 AEs of diarrhea compared to 2 grade 3-4 AEs of diarrhea in the concurrent therapy group though there was no significant difference in events of diarrhea over all cycles between arms (45% vs. 27%, respectively, p = 0.505) [48].
Phase I-II trials of dual checkpoint blockade or combination immunotherapy with checkpoint blockade in previously-treated advanced PDAC have similarly shown toxicity profiles typical of the respective agents [45,57]. Specifically, the safety profile of combination GVAX + ipilimumab in PDAC patients was similar to that of ipilimumab alone in melanoma patients [45]. Combination durvalumab and tremelimumab, however, had higher grade ≥ 3 AEs (22%) than durvalumab alone (6%) leading to more treatment discontinuations (9.4% vs. 3.1%) [57]. Although limited by a small sample size of 5 patients with pretreated advanced pancreatic cancer, treatment with a bifunctional anti-PD-L1 and TGFβ receptor II fusion protein was efficacious and well-tolerated with a MTD not reached at the highest dose, altogether highlighting the feasibility of multitargeted fusion constructs involving checkpoint blockade [58].
A phase Ib investigating ipilimumab vs. ipilimumab + GVAX vaccine in gemcitabine-treated locally advanced or metastatic PDAC was among the first to demonstrate an interesting immuno-oncology concept in pancreatic cancer: only in the combination immunotherapy arm were delayed SD and radiographic responses associated with declines in tumor markers seen, and 2 of such cases had localized disease while 1 patient had lungonly metastases [45]. Furthermore, most patients demonstrating a response required ≥ 12 weeks of therapy. The authors proposed that immunotherapy should be initiated earlier in the treatment course, i.e., locally advanced or resected disease, to allow more time to induce immune responses as well as produce less immune tolerance given a smaller disease burden. Selecting PDAC patients in this manner will also allow patients who have better reserve more time to recover from immune-related AEs and be eligible for retreatment. Indeed, this concept of earlier integration of checkpoint blockade and immunotherapy in patients with likely lower disease burden has shown promising efficacy in the neoadjuvant, consolidative (after definitive therapy), and adjuvant settings in other tumor types [71][72][73][74][75]. There are several ongoing clinical trials investigating combination systemic therapies with checkpoint blockade in the perioperative treatment of pancreatic cancer to see if such an immuno-oncology approach is beneficial in the non-metastatic setting ( Table 3).

Biomarkers
Regardless of the treatment setting in the pancreatic cancer, biomarkers to guide the optimal selection of candidates for immune checkpoint blockade-based therapies are desperately needed for this malignancy. Several metaanalyses have corroborated the prognostic value of PD-L1 expression in PDAC patients, however, its utility as a predictive biomarker for checkpoint inhibitors has yet to be as validated as it has been in other tumor types [76,77]. Tumors with mismatch repair deficiency (dMMR)/ microsatellite instability (MSI) or high tumor mutation burden (TMB) have been shown to respond to checkpoint inhibitors though dMMR is relatively rare in PDAC (frequency of 0.8%) and pancreatic cancer has among the lowest TMB across tumor subtypes [78,79]. However, investigations are ongoing and greater efforts are being undertaken to identify novel immuno-oncology biomarkers in pancreatic cancer. Recent correlation of cytolytic immune activity with mutational, structural, and neoepitope features in human PDAC samples has identified potential genomic signatures predictive of low cytolytic T-cell activity and expression signatures for multiple immune checkpoints other than PD-1/PD-L1 predictive of high immune cytolytic activity that altogether provide impetus for further investigation into therapeutic strategies that target multiple other immune checkpoints in pancreatic cancer [80]. An abundant microbiome within the pancreatic tumor that is distinct from that of the gut has interestingly been shown to promote immunosuppression that is characteristic of PDAC; targeting of this microbiome reverses tumor immune tolerance and enables efficacy for checkpoint inhibition-based therapy [81]. Continued advances in immuno-oncology biomarker development will ideally allow for selection of PDAC patients most likely to benefit from checkpoint blockade-based therapies based on a comprehensive immune profile-an individualized approach that remains a cornerstone to achieving precision immuno-oncology [82].

Conclusion
Despite the dismal activity of single-agent checkpoint blockade in pancreatic cancer, PD-1, PD-L1, and CTLA-4 inhibitors have shown broad antitumor activity as single agents in other tumor types and a relatively tolerable toxicity profile rendering them attractive agents to combine with systemic therapy. Indeed, there is a growing body of early clinical evidence to suggest the feasibility and efficacy in combining checkpoint blockade with other forms of systemic therapy in pancreatic cancer. Although phase I-II data support the concurrent administration of standard-dose checkpoint blockade with full-dose systemic therapies including chemotherapy, there remains several questions on dosing, timing, toxicity, and patient selection for checkpoint inhibitor-based combination therapies in pancreatic cancer that warrant further prospective validation. Results from ongoing clinical trials investigating combination strategies with checkpoint blockade in pancreatic cancer are eagerly awaited and will hopefully provide answers to many looming questions in this arena (Table 3). PD-1 programmed cell death protein 1 receptor, PD-L1 programmed death ligand 1, CTLA-4 cytotoxic T-lymphocyte associated protein 4, CY cyclophosphamide, ORR objective response rate, DLT dose-limiting toxicities, FOLFIRINOX folinic acid, 5-fluorouracil, irinotecan, and oxaliplatin, OS overall survival, GVAX GVAX, allogeneic pancreatic tumor cells transfected with granulocyte-macrophage colony-stimulating factor (GM-CSF) gene, SBRT stereotactic body radiation therapy, TGF transforming growth factor, FOLFOX 5-fluorouracil, folinic acid, and oxaliplatin, PFS progression-free survival, MSI/MSS microsatellite instability/microsatellite stable, RT radiotherapy