Pancreatic ductal adenocarcinoma (PDAC) is a lethal form of pancreatic cancer with an average 5-year survival rate of 11.5% from 2012 to 2018, according to recent data provided by Surveillance, Epidemiology, and End Results (SEER). In 2022, pancreatic cancer accounts for 3% of all cases and 8% of all deaths across cancer types, making it one of the top-five most life-threatening cancers. PDAC is positively correlated with age, having a median diagnosis age of 68 years; however, no sex preference has been determined [1]. Due to the lack of cancer-specific symptoms and prognosis biomarkers, most patients have non-resectable spread tumors at the time of diagnosis. Though PDAC typically forms at the head of the pancreas, where the stomach and the duodenum join and the site of chronic pancreatitis, the underlying mechanisms are still not fully understood.

Risk factors of PDAC include chronic pancreatitis, obesity, tobacco use, type 2 diabetes, and inherited genetic alternations such as mutations in tumor suppressor genes STK11, BRCA1, BRCA2, CDKN2A, and genes regulating DNA damage response and DNA repair [2]. More than 90% of PDAC patients have oncogenic KRAS mutations. Specifically, KRAS^G12D mutation is the most dominant oncogenic mutation and is present in approximately 40% of PDAC cases [3], promoting pancreatic tumorigenesis and maintaining tumor growth [4]. KRAS regulates almost all hallmarks of pancreatic cancer, especially activating essential signaling pathways for proliferation and survival, rewiring anabolism, and suppressing immune response in the tumor microenvironment (TME) [5]. Mouse model studies suggest that mutant KRAS alone can induce replication stress in pancreatic epithelial cells and is not sufficient to drive malignancy. Other factors are required to promote PDAC development, including chronic inflammation (pancreatitis) and loss of tumor suppressor genes, among which CDKN2A, CDKN2B, TP53, and SMAD4 are frequently detected in PDAC accompanied by KRAS mutations [6]. According to integrated genomic analysis, PDAC can be generally classified into four different subtypes: squamous/quasi-mesenchymal/basal-like, pancreas progenitor/classical, immunogenic, and aberrantly differentiated endocrine exocrine (ADEX)/exocrine-like subtypes [7]. The squamous subtype is associated with the poorest overall outcomes and is least dependent on KRAS signaling.

2. Hallmarks of the Pancreatic Tumor Microenvironment

Malignant cells rewire the microenvironment. The main cell populations of the PDAC TME consist of pancreatic stellate cells (PSCs), cancer-associated fibroblasts (CAFs), and myeloid cells, as well as regulatory T cells, B cells, and neuronal cells ^[8][9]. They can either support or block tumor development and maintenance depending on their received signals, and the sum of their effects results in unique hallmarks of the pancreatic TME: dense desmoplasia, high tissue stiffness, severe hypoxia, abnormal angiogenesis, nutrient deprivation, marked neuropathy, extensive immune suppression, and symbiosis. In detail, PSCs and CAFs produce abundant extracellular matrix (ECM) molecules such as proteoglycans, collagens, and fibronectin ^[9][10], leading to fibrosis and tissue rigidity and thus contributing to hypoxia and supporting tumor progression and metastasis ^[10][11][11,12]. Hypoxia rewires tumor metabolism, promotes tumor proliferation, survival, and epithelial-to-mesenchymal transition (EMT), and contributes to immune suppression ^[12][13]. Increased tissue stiffness drives PDAC invasion ^[13][14]. However, distinct subtypes of CAFs may exhibit opposite roles, such as that α-smooth muscle actin-positive (αSMA+) myofibroblasts restrain tumor growth partially by preventing the infiltration of immune suppressive cells ^[14][15][15,16]. Immune cells are another major component in the pancreatic TME. The major myeloid cell populations include macrophages, neutrophils, monocytes, and dendritic cells (DCs). In response to inflammation, pancreas infiltrated macrophages drive the acinar-to-ductal metaplasia transdifferentiation by producing inflammatory cytokines C-C motif chemokine ligand 5 (CCL5) and tumor necrosis factor α (TNFα) to activate nuclear factor kappa B (NFκB) signaling pathway, resulting in ECM remodeling and epithelial cell transformation ^[16][17]. In PDAC, tumor-associated macrophages (TAMs), tumor-associated neutrophils (TANs), and myeloid-derived suppressor cells (MDSCs) are dominant while DCs are sparse, preventing cytotoxic T cells and natural killer (NK) cells from penetrating the tumor ^[17][18][18,19]. In addition, regulatory CD4+ Foxp3+ T cells (Tregs) and regulatory B cells accumulate in advanced PDAC to promote immune tolerance by secreting inhibitory cytokines such as interleukin (IL)-10, transforming growth factor β (TGFβ) and IL-35 ^[9][19][20][10,20,21]. Tregs also interact with DCs and suppress their costimulatory ligand expression to restrain CD8+ T cell activation ^[21][22]. Surprisingly, ablation of Tregs in a PDAC mouse model accelerates tumor progression ^[22][23]. Mechanistic analysis reveals that Treg depletion causes CAF reprogramming by loss of tumor-restraining αSMA+ fibroblasts and gain of C-C motif chemokine receptor 1 (CCR1) ligand expression, recruiting myeloid cells to restore the immune suppression. Moreover, γδT cells constitute about 40% tumor infiltrating T cells in human PDAC, which are considered major sources of immune suppressive checkpoint ligands ^[23][24]. KRAS can increase granulocyte-macrophage colony-stimulating factor (GM-CSF) expression in mouse pancreatic ductal epithelial cells, and GM-CSF upregulation is also observed in human pancreatic neoplasia lesions ^[24][25]. GM-CSF recruits Gr1+ myeloid cell infiltration, and their pro-tumor activity is mediated by CD8+ T cell suppression ^[24][25][25,26]. In addition, cancer-associated mesenchymal stem cells (MSCs), rather than normal pancreas MSCs, stimulate alternative polarization of macrophages ^[26][27]. A recent study reveals that lysine demethylase 3A (KDM3A) is an epigenetic regulator of the immunotherapy response whose effect is mediated by transcription factors KLF5 and SMAD4 in PDAC ^[27][28]. Epidermal growth factor receptor (EGFR) is their downstream factor, and its inhibition facilitates CD8+ T cell infiltration, reduces myeloid cells, and sensitizes pancreatic tumors to combination immunotherapy (CD40 agonist, programmed cell death protein 1 (PD-1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) blockade). The pancreatic TME is highly innervated ^[28][29]. In a clinical study, 100% of PDAC patients (132/132) had neural invasion ^[29][30]. Tumor cells secrete neurotrophins to promote neuron infiltration and stimulate neuron growth ^[30][31]. On the other hand, premalignant pancreatic cells are prone to invade the spinal cord along sensory neurons. Ablation of sensory neurons in PDAC mouse models blocks the inflammatory signal transduction from pancreatic neoplasia to the central nervous system and hinders disease progression ^[31][32][32,33]. Sensory neurons produce stress molecules such as catecholamines that bind to β-adrenergic receptors on PDAC cells to promote tumorigenesis and tumor growth ^[33][34][34,35]. Upon pathway activation, PDAC cells increase nerve growth factor (NGF) expression, leading to perineural invasion and enlarged intratumoral nerves ^[34][35]. Blockage of the NGF/neurotrophic receptor tyrosine kinase 1 (Trk) pathway impairs tumor growth, prolongs mouse survival, and enhances tumoricidal effect of gemcitabine in spontaneous PDAC mouse models ^[34][35]. While subdiaphragmatic vagotomy accelerates PDAC progression, systemic administration of bethanechol, a muscarinic agonist, impairs tumor growth and prolongs mouse survival ^[35][36]. Mechanism dissection reveals that the cholinergic receptor muscarinic 1 (CHRM1) receptor expressed by tumor cells is responsible for the cholinergic suppressive effect via modulating mitogen-activated protein kinases (MAPK) and phosphatidylinositol-3-kinase (PI3K)-AKT pathways. Moreover, a recent study discovered that neurons nourish PDAC cells with serine to facilitate protein translation. Increased NGF production, in turn, exacerbates tumor innervation ^[36][37]. The symbiotic relationship between cancer cells and cells of the TME supports PDAC growth. KRAS promotes the secretion of sonic hedgehog protein (SHH) by cancer cells, which can induce extensive proteomic changes in PSCs ^[37][38]. The changes include the upregulation of ECM components such as matrix metalloproteinases (MMPs) and collagens, suggesting that KRAS is a driver of tumor desmoplasia. SHH also elevates growth factors insulin-like growth factor 1 (IGF1) and growth arrest-specific 6 (GAS6) that reciprocally activate IGF1R/AXL-AKT signaling pathway and increase spare mitochondrial capability in PDAC cells. In addition, fibroblast growth factor 1 (FGF1) secreted by CAFs is essential for paracrine MYC activation and protein stability coordinately with pancreatic tumor cell-autonomous signals ^[38][39]. Moreover, KRAS upregulates type I cytokine receptor complexes (IL2rγ-IL4rα and IL2rγ-IL13rα1) in pancreatic neoplasia. Tumor-infiltrated T helper 2 (Th2) cells produce IL-4 and IL-13 to activate Janus kinase 1 (JAK1)- signal transducer and activator of transcription 6 (STAT6)-MYC axis and enhance glycolysis, thus promoting tumorigenesis ^[39][40]. The accumulation of Th2 cells in TME needs intercellular cooperation. TNFα and IL-1β from tumor cells enable activation of CAFs to produce thymic stromal lymphopoietin (TSLP), which promotes Th2 polarization via DC conditioning ^[40][41]. Monocyte-recruited basophils stabilize the Th2 phenotype in pancreatic tumor-draining lymph nodes by releasing IL-4 ^[41][42]. Finally, Th2 cells are recruited into the TME in response to Th2-attracting chemokines secreted by tumor cells ^[40][41]. Rapid tumor growth causes nutrient deprivation in the TME. Thus, the reciprocal intercellular interaction is critical for nutrient exchange ^[42][43]. PDAC cells are addict to glucose and glutamine ^[4][43][4,44], and circulating lactate is a primary carbon source for the tricarboxylic acid (TCA) cycle in fasted mice ^[44][45]. To fulfill the high demand of amino acids, PDAC elevates macropinocytosis via KRAS to scavenge macromolecules from surroundings ^[45][46][46,47]. In a biological process called reverse Warburg effect ^[47][48], tumor cells stimulate CAFs to secrete metabolic intermediates such as pyruvate and lactate, which are reciprocally taken in by tumor cells for ATP production. Moreover, alanine secreted by activated PSCs serves as an alternative carbon source for PDAC to fuel TCA cycle and biosynthesis of non-essential amino acids (NEAA) and lipids ^[48][49]. Notably, the secretion of alanine by PSCs requires cancer cell stimulated autophagy activity. Despite PDAC cells’ ability to exploit stroma cells to adapt to the nutrient-deprived condition, CD8+ T cells usually exhibit impaired function and proliferation ^[49][50]. Specifically, MDSCs and TAMs express high levels of arginase and nitric oxide synthase, which consume arginine in the pancreatic TME, a critical amino acid for T cell activation ^[50][51]. The physicochemical features and intercellular crosstalk in the TME not only regulate pancreatic tumorigenesis, tumor maintenance and metastasis, but they are also critical elements determining tumor responses to cancer therapies. Hypoxia, a nonnegligible resistance inducer, has been extensively studreviewed elsewhere ^[51][52][52,53]. Thus, reswearchers focus on the interplay between cancer cell and non-cancer cell counterparts that prevents PDAC from apoptosis upon therapeutic treatments. Accumulated evidence suggests that PSCs, CAFs, TAMs and TANs/MDSCs are major players of therapy resistance, and their multifaced roles are discussed in the following sections.

3. Tumor-Associated Macrophages and Therapy Resistance

Macrophages regulate tissue development and maintain tissue homeostasis ^[53][124]. They are tissue-resident or infiltrated from circulating bone marrow-derived monocytes. Due to their hyperplastic nature, macrophages are polarized distinctly depending on stimuli and are mainly classified into five subtypes: pro-inflammatory M1 and immune-tolerant M2 (M2a, M2b, M2c, and M2d). Macrophages are phagocytes that can engulf and digest foreign pathogens and apoptotic cells regardless of polarization status. In addition, macrophages are professional APCs that process and present antigens for T cell recognition. Macrophages are abundant in the pancreatic TME, and M2 TAMs are correlated with poor overall survival ^[54][125]. However, M1 and M2 definitions could not accurately describe the heterogeneity of TAMs, which is fully reflected by single-cell transcriptional analysis. For example, Tie2+ M2 TAMs are a predictive marker of poor prognosis in multiple cancers, including PDAC, which may regulate angiogenesis via the ang2-Tie2 axis ^[55][56][126,127]. TAMs in PDAC are composed of pancreas resident macrophages orientated from the yolk sac and circulating monocytes ^[57][128]. Notably, embryonic progenitor-derived TAMs express more ECM molecules, regulating collagen deposition and fibrosis. In contrast, monocyte-derived TAMs have high cytokine expression and antigen presentation molecules, suggesting a role in modulating cancer immunity. CSF1 signaling is crucial for macrophage differentiation, infiltration, local expansion, and survival ^[58][129]. C-C motif chemokine ligand 2 (CCL2)-C-C Motif Chemokine Receptor 2 (CCR2) axis is the major chemoattractant signaling of macrophages. Blockage of either pathway decreases TAM population and impedes tumor growth ^[59][130], making them both promising therapeutic targets to limit TAM pro-tumor activities. TAMs exacerbate desmoplasia, angiogenesis, nutrient deprivation, and immune suppression to promote tumor growth by producing cytokines, chemokines, growth factors, and ECM components ^[55][60][126,131]. Meanwhile, TAMs are a key player in regulating therapy resistance. TAMs accumulate in TME after therapies ^[61][62][63][95,99,132], contributing to pancreatic tumor recurrence. To confer chemoresistance, TAMs release deoxycytidine via the transcription factor C/EBPδ to interfere with the uptake and metabolism of gemcitabine ^[64][65][93,133]. TAMs stimulate PDAC cells to upregulate cytidine deaminase expression, which eliminates gemcitabine ^[66][94]. By accumulating around blood vessels, TAMs promote tumor revascularization via secreting VEGF-A after chemotherapy to support tumor relapse in mouse models ^[63][132]. TAMs are also involved in radiotherapy resistance. Whereas the enrichment of CCR2+ macrophages has been observed in PDAC after radiotherapy, neutralizing or genetic depletion of CCL2 improves radiotherapy responses and attenuates tumor growth in mouse models ^[62][99]. Upon KRAS targeted therapy, macrophage infiltration dramatically increases in therapy-resistant PDAC tumors in pre-clinical models ^[61][95], which are essential and sufficient to drive KRAS bypass. Mechanistically, PADC cells elevate the production of CCL2 chemokine to attract CCR2+ M2-like macrophage infiltration, which reciprocally provides tumor cells with abundant TGFβ to promote KRAS-independent tumor growth. TGFβ is a robust driver of EMT, which is associated with KRAS inhibition resistance in PDAC and lung cancer cells ^[67][68][69][134,135,136]. TAMs may also induce EMT in PDAC via the secretion of MMP9 ^[70][137]. PDAC is irresponsive to immunotherapy. TAMs in peri-tumor regions form a barrier against T cells ^[71][100]. Although chemotherapy induces immune suppressive TAMs, these p21^high TAMs respond to CD40 agonists ^[72][138]. TAMs secrete granulin upon M-CSF stimulation in the TME, and granulin is essential for local fibrosis and exclusion of T cells at the metastatic site in mouse models ^[73][96]. Depletion of granulin allows T cell entry in the liver and sensitizes metastatic PDAC tumors to PD-1 blockade. In addition, PDAC upregulates necroptosis complex component receptor-interacting serine/threonine-protein 1 (RIP1) in both tumor epithelial cells and stroma to induce chemokine CXCL1 expression, which attracts macrophage infiltration ^[74][97]. Interaction of TAMs and tumor cells via ligation of Mincle and Sin3A-associated protein 130 (SAP130) polarizes TAMs to immune suppressive M2 phenotype, resulting in adaptive immune suppression and tumor progression. Depletion of RIP1 in epithelial cells or TME cells is protective against PDAC. In addition, chemical inhibition of RIP1 reprograms TAMs toward immunogenic M1 via STAT1 activation ^[75][98]. Educated M1 TAMs elicit the cytotoxicity of CD8+ T cells and promote Th1 and Th17 differentiation of T helper cells. Combination therapy of RIP1 inhibitor and PD-1 blockade synergistically induces tumor immunity and suppresses PDAC tumor growth in vivo. In conclusion, the targeting or reprogramming of TAMs is likely to enhance the tumoricidal effect of multiple therapies through disruption of intercellular crosstalk between TAMs and other cell counterparts, leading to remodeling of the pro-tumor to anti-tumor TME. It is noticed that macrophage status is a highly dynamic spectrum and determined by external stimuli; thus, in vivo reprogramming of TAMs may be more challenging than previously expected.

4. Tumor-Associated Neutrophils and Therapy Resistance

Neutrophils are the most abundant and short-lived innate immune cells, and they are responsible for mediating the rapid innate host defense against pathogens ^[76][139]. They infiltrate solid tumors and have attracted much attention in recent years ^{[77][78][79][80]}[140,141,142,143]. It is still challenging to distinguish between TANs and polymorphonuclear MDSCs (PMN-MDSCs) (a.k.a. granulocytic MDSCs (G-MDSCs)) because of the lack of unique markers, while both are considered immune suppressive in most cancer cases. Similar to macrophages, TANs are classified into two major polarization states, anti-tumor N1 and pro-tumor N2 ^[81][144]. TGFβ renders the compromised cytotoxicity of TANs via a superoxide-dependent mechanism and promotes the anti-inflammatory N2 phenotype. On the other hand, IFNβ, IFNγ, and GM-CSF have been shown to polarize TANs into a pro-inflammatory, APC-like N1 state ^[82][83][145,146]. In contrast to N1 TANs, which have lobulated and hyper-segmented nuclei and express CD101 marker ^[84][147], N2 TANs appear immature with circular and less lobulated nuclei and high levels of CD170 ^[85][148]. To note, the phenotype of TANs is more dynamic and heterogeneous than the dichotomized classification, so a mixed state is usually observed. In PDAC, TAN presence is an independent prognosis factor for tumor recurrence and overall survival ^[86][149]. By producing ROS and arginase 1 (ARG1), modulating multiple signaling pathways, and forming neutrophil extracellular traps (NETs), TANs not only promote tumor metastasis and angiogenesis but also suppress NK and T cell cytotoxicity and induce immune tolerance ^[76][78][139,141]. Neutrophils express CXCR1 and CXCR2 receptors whose ligands such as CXCL1, 2, 5, and 8 are dramatically upregulated by PDAC cells, and high CXCL5 is positively correlated with poor patient overall survival ^[87][102]. In addition, cytokines G-CSF and GM-CSF, rather than M-CSF, regulate neutrophil recruitment, survival, and differentiation, and they are significantly upregulated in PDAC versus a normal pancreas. Targeting TANs improves conventional therapy response in PDAC. IL6 receptor blockade suppresses STAT3 phosphorylation in both myeloid cells and tumor cells, thus sensitizing tumor cells to gemcitabine ^[88][80]. In addition, CXCR2 inhibition prevents compensatory infiltration of CXCR2+ TANs upon CCR2+ TAM depletion in PDAC models, resulting in improved tumoricidal immunity and better response to chemotherapy regimens ^[87][102]. In a KRAS-driven sarcoma model, depletion of TANs by anti-Ly6G neutralizing antibody enhances radiotherapy responses ^[89][104]. Furthermore, TANs (or G-MDSCs) drive immunotherapy resistance. By multidimensional imaging, G-MDSCs expressing high lectin-like oxidized low-density lipoprotein receptor 1 (LOX-1) and ARG1 are shown to reduce the expression of granzyme B and Ki67 in colocalized T cells ^[90][103]. Another recent study reveals that the immunosuppressive role of IL-17 in PDAC is mediated by TANs ^[91][101]. Specifically, IL-17 recruits neutrophils in TME and promotes NETs formation. Blockade of IL-17 enables PDAC cells to respond to checkpoint blockade, and it synergizes with PD-1 to impair tumor growth in a CD8+ T cell-dependent manner in pre-clinical models. In addition, there is a positive correlation between NETosis and poor overall survival in PDAC patients. Depletion of CXCR2+ TANs attenuate PDAC metastasis, promotes T cell entry, and sensitizes tumor cells to PD-1 blockade in mice ^[92][107]. In a later study dissecting immune heterogeneity of various PDAC subclones, immunotherapy-resistant ones lack T cell infiltration but enrich G-MDSCs ^[93][106]. Tumor cells secrete CXCL1 to recruit CXCR2+ G-MDSCs. Depletion of CXCL1 overcomes resistance to combination immunotherapy (anti-CD40 agonist, anti-PD-1 antagonist, and anti-CTLA-4 antagonist) in syngeneic mouse models. Similarly, in a p53R172H mutant PDAC model, neutrophils are recruited by tumor cells via the CXCL2/5–CXCR2 axis. Depletion of TANs increases T cell infiltration and enhances the tumoricidal activity of CD40 agonist and gemcitabine/nab-paclitaxel combination ^[94][105]. The lack of T cell activation indicates that the addition of T cell immune checkpoint blockade may further impair tumor growth. Additionally, STAT5 inhibition by lorlatinib blocks tumor-induced granulopoiesis and suppresses neutrophil migration, leading to enhanced immunotherapeutic responses and PDAC regression in vivo ^[95][150]. In summary, immune suppression is the dominant role of TANs/G-MDSCs, and they can rewire the TME by autocrine or paracrine mechanisms. Depletion of TANs or MDSCs by CXCR2 inhibitors augments immunotherapy response in several cancer models ^[96][97][151,152]. Whether CXCR2 inhibition can also sensitize PDAC patients to checkpoint blockade needs further clinical investigation. Besides targeting neutrophil recruitment, modalities to repolarize N2 TANs into N1 pro-inflammatory phenotype are in development as well, such as combinations of TGFβ signaling inhibitors and immune checkpoint blockade in clinical trials. Survival pathways for TANs, including PI3K gamma/delta (PI3Kγ/δ), are other promising targets to reduce the TAN population.