Beyond local changes such as an increase in vascular permeability, extensive remodeling of the ECM and manipulation of liver-resident cell types, the hepatic PMN in PDAC is also shaped by systemic changes pertaining to the immune system. Like the initial acquisition of invasive properties, intravasation and extravasation, continuous immune surveillance forms yet another rate-limiting step in the metastatic cascade that needs to be overcome to ensure successful metastatic outgrowth of incipient pancreatic cancer cells. As such, pancreatic cancer cells have evolved a myriad of mechanisms by which they can evade host immune responses. This section will cover ways in which the primary tumor in PDAC can mold pre-metastatic sites into immunosuppressive environments.
It is widely recognized that macrophages are actively recruited to the microenvironment of the primary tumor where they foster disease progression by promoting angiogenesis, migration and intravasation, and repressing anti-tumor immune responses [
131,
132]. However, in recent years there has also been a growing appreciation for macrophages that populate distant (pre-)metastatic lesions and influence metastatic outcome at these sites. In a recent report, Nielsen et al. emphasized the significance of a distinct population of macrophages termed metastasis-associated macrophages (MAMs), which assist hepatic metastasis in PDAC by establishing a fibrotic microenvironment in the liver [
133].
MAMs originate from inflammatory monocytes which undergo differentiation following recruitment from the bone marrow to the liver. The chemotactic factors that initiate their recruitment were not evaluated in the study by Nielsen et al., but potential candidates include C-C motif chemokine ligang 2 (CCL2) and SDF-1 secreted by pancreatic cancer cells [
134,
135]. Once infiltrated in the liver, soluble factors derived from pancreatic cancer cells, for example colony stimulating factor-1 (CSF-1) [
136], trigger MAMs to produce and secrete granulin, a glycoprotein previously implicated in the wound healing response and breast cancer as a potent activator of stromal fibroblasts [
137,
138]. Consistent with this role, macrophage-derived granulin instigates the activation of HSCs into myofibroblasts during liver metastasis in PDAC [
133]. Unfortunately, our current mechanistic understanding of granulin-mediated HSC activation is limited due to conflicting reports with respect to the cognate cell-surface receptor for granulin. Nonetheless, granulin-activated HSCs are known to secrete high levels of ECM components, giving rise to a highly fibrotic environment. Besides collagen, one of the proteins particularly enriched in activated HSCs is periostin. This matricellular protein has previously been reported to encourage metastatic tumor development in colon cancer by augmenting cell survival via the Akt/PKB pathway [
139]. In another study, breast cancer stem cells were shown to induce periostin expression in fibroblasts to aid metastatic colonization, presumably because deposition of periostin creates a supportive microenvironment which resembles that of the primary niche. Furthermore, periostin was found to support stem cell maintenance by facilitating Wnt signaling, thereby allowing breast cancer stem cells to initiate metastatic growth [
140]. Nielsen et al. extended this work by demonstrating the requirement of periostin in the survival and growth of pancreatic cancer cells in vitro; a neutralizing antibody against periostin abolished the stimulatory effects of myofibroblast-conditioned medium on colony formation and proliferation of pancreatic cancer cells [
133]. Notably, induction of periostin expression in activated HSCs was found to be strictly regulated by MAM-derived granulin, and depletion of granulin in KPC mice significantly diminished metastatic growth of pancreatic cancer cells. These findings further illustrate the central role of granulin and, consequently, periostin in PDAC metastasis.
Though this newly-described mechanism supporting liver metastasis in PDAC may seem reminiscent of the role of Kupffer cells, Nielsen et al. propose that these liver-resident macrophages and MAMs exert distinct functions within the temporal sequence of events [
133]. While Kupffer cells are believed to predominantly facilitate initial seeding of pancreatic cancer cells, MAMs appear to be of greater relevance to subsequent cancer cell survival and growth. This may lead one to argue that MAMs are not involved in the early stages of PMN formation in the liver. In fact, it remains unclear from the study by Nielsen et al. whether MAM-induced activation of HSCs and concomitant fibrogenesis actually precede the arrival of pancreatic cancer cells in the liver. One methodological limitation of this study is the use of an experimental model of metastasis, which bypasses primary tumor development and hence fails to recapitulate the cardinal features of PMN formation [
141]. In this regard, spontaneous models of metastasis are considered the gold standard. Further studies employing these models are therefore anticipated in order to gain a more comprehensive and spatiotemporal understanding of how MAMs assist colonization of pancreatic cancer cells in the liver.
Either way, the pro-metastatic role of granulin-secreting MAMs in vitro and in vivo is consistent with clinical observations in PDAC patients correlating inflammatory monocyte density in the peripheral blood with decreased patient survival [
134]. Moreover, since granulin expression is already elevated in circulating inflammatory monocytes from KPC mice as well as from PDAC patients harboring liver metastases, it could have potential utility as a biomarker to predict metastasis in PDAC [
133]. In this respect, it would be worth assessing whether granulin is also upregulated in inflammatory monocytes from patients with premalignant or inflammatory lesions. If this is the case, it would greatly increase the potential of granulin as a predictive biomarker.
In a recent extension of this work, Schmid’s group uncovered an additional pro-tumorigenic function of MAMs that may causally link the abundance of inflammatory monocytes and decreased patient survival [
135]. Macrophages in general display a high degree of plasticity [
132]. In response to environmental stimuli, they can be polarized into a variety of phenotypes, with immune-stimulatory M1 and immunosuppressive M2 macrophages at either end of the spectrum. Initial seeding of pancreatic cancer cells and the development of micrometastases is accompanied by infiltration with tumoricidal cytotoxic T cells and M1-like MAMs, but upon progression to overt macrometastatic lesions, M2-like MAMs predominate, resulting in loss of T cell infiltration and effector function [
135]. Since the efficacy of immune checkpoint blockade is heavily dependent on the ability of cytotoxic T cells to infiltrate into tumors, this rendered metastatic lesions less responsive to treatment with an anti-PD-1 monoclonal antibody. Functional experiments in PDAC mice revealed that CSF-1, which is abundantly expressed by metastatic pancreatic cancer cells, is able to expand the population of MAMs while driving their differentiation toward an M2-like phenotype.
CSF-1 also induces granulin expression by MAMs. Conceivably, the resulting desmoplastic reaction forms a physical barrier that precludes T cell infiltration into the metastatic site, thereby protecting pancreatic cancer cells from tumor-targeting immune responses. In line with this theory, residual cytotoxic T cells were noted to cluster primarily in peripheral regions of anti-PD-1-resistant tumors, in close proximity to activated HSCs [
135]. Furthermore, the reduction in the number of cytotoxic T cells correlated with an accumulation of activated HSCs as well as enhanced collagen deposition. Interruption of metastasis-associated hepatic fibrosis by means of CSF-1 blockade or granulin depletion restored T cell infiltration and sensitized metastatic tumors to anti-PD-1 therapy in vivo. These data suggest that, via the release of CSF-1, metastatic pancreatic cancer cells are capable of undermining cytotoxic T cell-mediated immune surveillance by exploiting M2-polarized MAMs to construct a dense fibrotic stroma in the liver. Most importantly, they provide a strong rationale to explore the therapeutic benefit of immune checkpoint inhibitors in combination with compounds targeting granulin, and perhaps other pro-fibrotic factors, in advanced PDAC.
Myeloid-derived suppressor cells: Imposing immune tolerance
Our understanding of tumor immunology and the importance thereof has expanded rapidly over the years and has spawned a renewed surge of interest in another population of bone marrow-derived cells termed myeloid-derived suppressor cells (MDSCs) [
142]. MDSCs are a heterogeneous population of incompletely matured cells of myeloid origin endowed with potent immunosuppressive activity. In conditions of chronic inflammation and cancer, myelopoiesis is persistently stimulated which impairs the normal differentiation of granulocyte/monocyte precursors into mature granulocytes, monocytes or dendritic cells, leading to the accumulation of phenotypically similar but functionally distinct MDSCs in the peripheral blood, lymphoid organs and tumor-bearing tissues [
142,
143]. Indeed, numerous studies have reported increased prevalence of circulating MDSCs in patients with various types of cancer, for example breast cancer [
144,
145], non-small cell lung cancer [
144], head and neck carcinoma [
144], colorectal carcinoma [
146,
147], renal cell carcinoma [
148] and PDAC [
145,
149,
150].
It is likely that MDSCs originally evolved as safeguards against uncontrolled immune responses that may damage healthy tissues [
143]. Tumors, however, cunningly take advantage of this protective function and actively recruit MDSCs to permit tumor development and progression in the absence of tumor-specific immune attacks. Though a multitude of mechanisms have been described through which MDSCs perturb anti-tumor immunity, which immunosuppressive mechanism predominates seems to be partly influenced by the specific nature of the MDSC subset that is expanded. MDSCs can be broadly divided into granulocytic MDSCs and monocytic MDSCs. The prevalence of either of these two subsets differs depending on cancer type, though the currently available literature suggests that patients with PDAC accumulate both granulocytic and monocytic MDSCs in the peripheral blood and pancreatic tumor tissue [
142,
151‐
153].
Granulocytic MDSCs release high levels of arginase I, an enzyme that reduces the availability of
l-arginine in the microenvironment and in the circulation by hydrolyzing
l-arginine to
l-ornithine and urea [
151,
154].
l-arginine is required for T cell proliferation, cytokine production and expression of the ζ chain of the T cell co-receptor molecule CD3 (CD3ζ), so deficiency of this amino acid causes profound T cell dysfunction [
148]. Simultaneously, arginase I directly subserves tumor growth since the newly generated supply of
l-ornithine can be used for the synthesis of polyamines, compounds critically implicated in cell growth and survival [
155,
156]. The reactive oxygen species (ROS) hydrogen peroxide (H
2O
2) derived from granulocytic MDSCs systemically suppresses T cell function in a similar manner [
157]. In addition, ROS released by MDSCs mediate antigen-specific cytotoxic T cell tolerance by disrupting the integrity of the T cell receptor (TCR) complex [
158]. MDSCs have the ability to take up foreign antigens, process them and present the resulting peptides to T cells. By mediating the nitration of tyrosine residues within the TCR complex and causing its dissociation, MDSC-derived ROS interfere with the interaction between MDSCs and cytotoxic T cells and prevent the induction of a T cell response against the presented peptide [
158,
159]. In a genetically engineered mouse model that spontaneously develops PDAC, tumor infiltration with MDSCs was found to be almost mutually exclusive with the loss of intratumoral effector T cells, indicative of a strong inhibition of T cell function and proliferation [
160]. In agreement with this observation, targeted depletion of granulocytic MDSCs in KPC mice was shown to reestablish accumulation of activated cytotoxic T cells within the pancreatic tumor, accompanied by apoptosis of tumor cells and remodeling of the tumor stroma [
161].
While T cells are their prime targets [
143], MDSCs have also been reported to negatively regulate NK cell function by inducing NK cell anergy via membrane-bound TGF-β1 [
162,
163]. Other studies have implicated MDSCs in the recruitment and maintenance of regulatory T cells (T
reg), a subpopulation of T cells renowned for their role in sustaining tolerance to self-antigens and in restricting excessive immune responses [
160,
164]. Since many tumor antigens are in fact self-antigens, T
reg are also involved in limiting anti-tumor immune responses. Therefore it is not surprising that T
reg have been noted with increased prevalence in multiple forms of cancer, including PDAC [
165]. Moreover, MDSCs have been shown to support neovascularization, providing further evidence that they also directly favor tumor progression. To this end, MDSCs may either secrete pro-angiogenic factors such as bombina variegate peptide 8 (Bv8) [
166], VEGF or basic fibroblast growth factor (bFGF) [
167], or liberate VEGF from the tumor-associated stroma by means of MMP9 [
168,
169]. MDSCs may even actively participate in tumor angiogenesis by differentiating into endothelial-like cells capable of embedding in the growing vascular endothelium [
168].
While the association between MDSCs and the establishment of a PMN in the liver has not been thoroughly explored in PDAC, there is ample evidence to speculate that MDSCs indeed prepare the hepatic microenvironment for metastatic spread of pancreatic cancer cells by means of their pro-tumorigenic, immunosuppressive activities. Accumulation of MDSCs in the peripheral blood positively correlates with metastatic tumor burden in patients with PDAC and other types of cancer [
145], and ablation of MDSCs by means of a neutralizing antibody profoundly suppresses metastasis [
170]. Studies using a variety of tumor models, including a mouse model of PDAC, have revealed the capacity of MDSCs to home to and expand in the liver during pre-metastatic phases [
171,
172]. Importantly, depletion of liver MDSCs in mice harboring extrahepatic primary tumors was shown to dramatically reduce the frequency of liver metastases, indicating that these hepatic populations of MDSCs accelerate metastatic tumor growth in the liver [
172]. Here, MDSCs were found to inhibit cytotoxic T cell activation, proliferation and cytotoxicity, as well as induce the development of T
reg. Furthermore, MDSCs within the liver have been demonstrated to interact with Kupffer cells and augment their expression of the negative T cell costimulatory molecule PD-L1 [
171]. Notably, amplification of MDSCs can be detected in pancreatic tissue and in the peripheral blood of mice even before fully established primary tumors manifest in the pancreas [
173]. Whether MDSCs also have a direct effect on circulating pancreatic cancer cells, however, has not been investigated.
The above findings suggest that early tumor-mediated recruitment of MDSCs to the future metastatic site in the liver ensures prosperous metastatic outgrowth by shielding incoming pancreatic cancer cells from tumor-specific immunosurveillance. Accordingly, numerous reports have identified distinct mechanisms through which pancreatic cancer cells elicit the accumulation of MDSCs. Keratinocyte-derived chemokine (KC), the murine homolog of human C-X-C motif ligand 1 (CXCL1), is a soluble factor that is already secreted by precursor lesions in a murine model of PDAC [
172]. Ligation of KC to the KC receptor (CXCR2 in humans) is required for the previously described expansion of MDSCs in the livers of tumor-bearing mice. Mice deficient in the KC receptor showed a reduction in hepatic MDSC accumulation by more than 75%, while primary tumor growth remained unaffected. In the liver, MDSCs themselves secrete high levels of KC, among other regulatory and pro-inflammatory cytokines, so as to maintain the hepatic MDSC population. Consistent with these findings, Steele et al. [
170] noted greatly upregulated expression of CXCL1 in KPC mice. Inhibition or genetic deletion of its receptor CXCR2 nearly completely abrogated metastasis in this model. Depletion of MDSCs, which are naturally a prominent source of CXCR2, had a similarly profound effect on metastasis. These data indicate that KC/CXCL1-dependent recruitment of MDSCs to the liver is indeed a pivotal event in the formation of a hospitable PMN.
Other soluble factors that have been associated with the recruitment of MDSCs by pancreatic cancer cells include granulocyte/macrophage colony-stimulating factor (GM-CSF) [
174] and pancreatic adenocarcinoma upregulated factor (PAUF), a soluble protein previously implicated in pancreatic cancer metastasis [
175]. PAUF has been suggested to attract MDSCs to tumor sites in an SDF-1-dependent manner by upregulating the expression of CXCR4 on MDSCs. In addition, the tumor-associated stroma may also interact with MDSCs to encourage their expansion and enhance their function; a recent study demonstrated the capability of PSCs to secrete several cytokines and chemokines that promoted the differentiation of peripheral blood mononuclear cells into immunosuppressive MDSCs, including IL-6, VEGF and macrophage colony-stimulating factor (M-CSF), and SDF-1 and monocyte chemoattractant protein-1 (MCP-1), respectively [
176].
Interestingly, pancreatic cancer cell-derived exosomes have also been implicated in skewing the immune repertoire toward an immunosuppressive kind. Exosomes shed by pancreatic cancer cells can be internalized by CD14
+ monocytes and impart a monocytic MDSC phenotype by downregulating HLA-DR expression and activating signal transducer and activator of transcription 3 (STAT3) signaling, with increased arginase I and ROS production as a result [
153]. Another study reported that exosomes secreted by pancreatic cancer cells could expand both granulocytic and monocytic populations of MDSCs while lowering the number of dendritic cells, and proposed altered intracellular calcium fluxes as a potential mechanistic explanation [
177]. The specific contents of pancreatic cancer cell-derived exosomes that govern the observed reprogramming of monocytes is presently unclear. Still, because exosomes have previously been shown to be employed by the primary tumor in PDAC to distantly prepare the liver for metastasis, these exciting discoveries provide further evidence supporting a role for MDSCs in the formation of a hepatic PMN in PDAC.