Tumorigenesis is commonly due to a combination of somatic mutations, chromosomal rearrangements, copy number alterations, and epigenetic changes that damage the natural cell cycle and regulatory pathways. Sequencing of tumor samples has changed the understanding of disease biology and has revealed inter-tumoral and intra-tumoral heterogeneity. In turn, there has been a shift in therapeutic strategy to identify potential targets and deliver more personalized care. Based on these data, two different models of PDAC carcinogenesis have been proposed. The first is a stepwise progression through the accumulation of genetic alterations in the following order: oncogene KRAS (90% of patients), followed by tumor suppressors CDKN2A (60%), then TP53 (80%) and SMAD4 (40%) [9, 10, 11]. Accumulated mutations in these four genes leads to cell cycle proliferation in both human and mouse studies [12, 13]. Additional studies have suggested that the early activation of KRAS and subsequent RAS signaling pathway is the main driver of PDAC development and concomitant mutations in the three tumor suppressor genes accelerates the development and dissemination of cancer [14, 15, 16, 17]. This theory suggests a slow, gradual development process with a late clinical presentation of the disease.

The second theory proposes that PDAC occurs through the simultaneous knockout of genes, as opposed to a stepwise fashion, in the setting of complex chromosomal rearrangements. Essentially, several structural alterations occur in a single cell cycle on a few chromosomes that leads to rearrangement of multiple driver genes. These changes result in rapid carcinogenesis and dissemination [18]. The main difference between the two theories is whether PDAC occurs through a stepwise or punctuated progression, sometimes referred to as chromothripsis (Fig. 1, Ref. [19]). The punctuated progression theory is compatible with patients who present with sudden onset of advanced disease.

Fig. 1.

The hypothetical accumulation of somatic events over time as a pancreatic cell undergoes dysplasia, including structural (chromothripsis) and simple somatic mutations under a punctuated model, stepwise model, or hybrid model. This figure was edited and reproduced with copyright permission from reference [19].

Collisson et al. [21] was the first study to categorize PDAC into subtypes based on the genomic landscape. At the time, there was a paucity of PDAC samples available so the authors combined data from primary PDAC samples into two different studies [21, 27]. Based on this analysis, the authors defined three subtypes: classical, quasimesenchymal, and exocrine-like. The classical subtype had high expression of epithelial and adhesion-associated genes and was sensitive to erlotinib in vitro. The quasi-mesenchymal subtype was defined by high expression of mesenchymal-associated genes. These patients had worse survival outcomes compared with the other two subtypes and were more sensitive to gemcitabine on in vitro analysis. The exocrine-like subtype demonstrated high expression of tumor cell derived digestive enzyme genes. The exocrine-like subtype was not identified in any of the mouse or human cell lines [21]. These subtypes have been validated with additional published PDAC datasets.

Research and treatment of PDAC is complicated by the overwhelming stromal component within the tumor. The dense stroma tissue and paucity of malignant cells makes it difficult to perform molecular analysis. Moffitt et al. [22] overcame this obstacle by including normal pancreas, primary PDAC, and metastatic PDAC samples in their analysis. The authors employed blind source separation and digitally separated tumor from stroma and normal tissue gene expression. Using this method, two subtypes were identified: classical and basal-like. The classical subtype was characterized by high adhesion-associated gene expression, ribosomal and epithelial gene expression, and increased GATA6 expression. This subtype overlapped with Collisson’s “classical subtype”. The basal-like subtype was comprised of tissue with high laminins and keratins like the basal-like subtype of bladder and breast cancers. This subtype was associated with worse outcomes.

Moffitt et al. [22] classified the stroma gene expression into two subtypes: normal and activated. Activated stroma was characterized by genes associated with a strong immune response and normal stoma contained markers for pancreatic stellate cells. The stroma subtypes were independently predictive of outcomes. Patients with activated stroma had worse long-term outcomes versus patients with normal stoma (median overall survival (OS): 15 months versus 24 months, respectively). Both basal and classical subtypes were within normal and activated stroma. The normal stoma, classical subtype had the lowest hazard ratio on Cox regression, while the basal-like, activated subtype had the highest. These data suggested a cumulative effect of PDAC subtype and stoma subtype on survival.

Bailey et al. [23] performed an analysis of 456 PDAC samples that used a combination of whole-genome and RNA sequencing to determine mutational mechanisms, potential genomic events, and expression profiles. The samples were initially enriched and selected for high epithelial content $>$40% to balance stromal gene expression. Four subtypes were defined. The squamous subtype had upregulation of TP63$\mathrm{\Delta }$N transcriptional network, hypermethylation of pancreatic endodermal cell-fate determining genes, and enrichment for TP53 and KDM6A mutations. This subtype overlaps with the quasimesenchymal subtype from Collisson et al. [21]. Pancreatic progenitor tumors were defined by expression of genes involved in early pancreatic development (FOXA2/3, PDX1, and MNX1) and metabolism [23]. This subtype overlapped with the classical subtype as defined by Collisson et al. [21]. The aberrantly differentiated endocrine exocrine (ADEX) subtype featured upregulation of genes that regulated networks involved in KRAS activation and genes associated with exocrine and endocrine differentiation [23]. This subtype corresponded to Collisson’s exocrine-like subtype [21]. In addition, immunogenic tumors were classified as an enrichment of genes associated with infiltrating B and T cells [23].

Raphael et al. [20] compared the three molecular classification systems using The Cancer Genome Atlas. This study performed genomic, transcriptomic, and proteomic profiling of 150 PDAC samples [20]. The authors applied a clustering technique to reproduce the subtype classifications of Collisson et al. [21], Moffitt et al. [22], and Bailey et al. [23]. Given that the low neoplastic cellularity of PDAC made analysis challenging, Raphael et al. [20] divided tumors into high and low purity samples based on the malignant cellularity of the tissue. High-purity tumors were classified into two groups: the basal-like/squamous which correlated to the previously described quasi-mesenchymal (Collisson et al. [21]), squamous (Bailey et al. [23]), or basal-like group (Moffitt et al. [22]) and the classical group that corresponded to the previously described classical (Collisson et al. [21]), progenitor (Bailey et al. [23]), or classical group (Moffitt et al. [22]). The immunogenic and ADEX groups described by Bailey et al. [23] and the exocrine-like group reported by Collisson et al. [21] were only found in low tumor cellularity samples, which may have represented contaminating gene expression from non-neoplastic cells. However, these subtypes should not be entirely disregarded since there has been variation among these landmark studies, as well as the subsequent studies, in the way samples were collected, stored, and analyzed. The validity of the immunogenic, ADEX, and exocrine-like subtypes still requires further investigation.

Raphael et al. [20] proposed that based on these data, the two consensus subtypes should be basal-like/squamous and classical/progenitor. This proposal was further validated with subsequent long non-coding RNA sequencing, methylation analysis, and proteomics [20]. The two subtypes were predictive of overall survival following surgery in two different analyses [13, 20]. These data suggest that these subtypes may provide clinically relevant information to assist with strategies related to treatment.

The underlying mechanism for the evolution of PDAC into either of these two subtypes is still unclear. Currently, basal-like tumors are associated with poor differentiation, worse survival, and resistance to therapy while classical tumors are more differentiated, have better survival, and improved response to chemotherapy [24]. However, these subtypes were primarily evaluated in patients with localized primary tumors who underwent resection and not among individuals with metastatic disease. Building on this work, Chan-Seng-Yue et al. [24] created a dataset of whole genomes and transcriptomes generated from purified epithelium of primary and metastatic PDAC tumors. Based on these data, a more granular classification system was proposed: basal-like A, basal-like B, hybrid, classical A, and classical B. Even within the two accepted molecular subtypes of PDAC, there was transcriptomic heterogeneity that was correlated with clinical disease manifestation. When divided into basal-like A versus B and classical A versus B signatures, metastatic disease from localized disease could be differentiated, respectively. The hybrid subtype was inconsistently classified due to multiple expression profiles. This stratification system still requires validation, but it does demonstrate the importance of including metastatic sites in PDAC molecular studies.

Somatic structural rearrangement of chromosomes is a common mutation that causes gene disruption, activation, and/or fusions; these mutations drive carcinogenesis. Small genomic sequencing studies of PDAC demonstrated that PDAC genomes include extensive and complex chromosomal rearrangements [28, 29]. In 2015, whole genome sequencing of 100 PDAC tumors revealed that alterations of chromosomal structure led to gene disruption, particularly in the genes commonly associated with PDAC [30]. Waddell et al. [30] classified tumors into four subtypes based on the number and location of the rearrangements: stable, locally rearranged, scattered, and unstable. Stable tumors had 50 structural rearrangements located randomly throughout the genome while unstable tumors contained $>$200 rearrangements. Locally rearranged tumors had intra-chromosomal rearrangements clustered on only a few chromosomes, whereas scattered tumors contained 50–200 structural rearrangements present throughout the genome. Additionally, these subtypes had potential clinical relevance in patients receiving platinum-based therapy.

Epidermal growth factor receptor (EGFR) is a transmembrane tyrosine kinase receptor (TKR) that is commonly overexpressed in PDAC. Constitutive activation of EGFR has been noted in multiple cancers and EGFR inhibitors have emerged as an effective therapy [32]. EGFR activation has been involved in pancreatic proliferation and progression. Erlotinib is an oral, selective tyrosine kinase inhibitor (TKI) that acts against EGFR. A phase III randomized trial assigned 569 patients with unresectable, locally advanced, or metastatic PDAC to receive either gemcitabine with erlotinib or a placebo [33]. Patients who received erlotinib had improved one year survival versus the placebo cohort (23% versus 17%, respectively) and longer progression free survival. Erlotinib was the first molecular-targeting agent that demonstrated a statistically significant effect among patients with PDAC. Erlotinib was subsequently studied in the CONKO-005 trial, which randomly assigned 463 patients with resected PDAC to receive adjuvant gemcitabine alone or gemcitabine with erlotinib [34]. There was no improvement in disease free or overall survival with the addition of erlotinib. In a follow up study, tumor samples from the CONKO-005 trial were sequenced [35]. A SMAD4 genetic aberration with low mRNA expression of MAPK9 was noted to be predictive of response to erlotinib. In turn, SMAD4 status may identify a subset of patients who would benefit from adjuvant erlotinib in early-stage PDAC.

Cetuximab is a monoclonal antibody against EGFR that is commonly used in colorectal cancer treatment [36]. Pre-clinical studies with PDAC mouse xenograft models demonstrated cetuximab efficacy [37]. However, these results did not translate to the early clinical trials. In a phase III randomized trial, 745 patients with locally advanced or metastatic PDAC were assigned to receive gemcitabine alone or gemcitabine with cetuximab [38]. There was no difference in median overall survival between the gemcitabine/cetuximab and gemcitabine alone cohorts (6.3 months versus 5.9 months, respectively).

Nimotuzumab is a monoclonal antibody against EGFR that has demonstrated better efficacy among patients with wild-type Ras PDAC. Qin et al. [39] randomized 92 patients with locally advanced or metastatic PDAC to receive gemcitabine with nimotuzumab or placebo. The nimotuzumab cohort had an overall survival of 10.9 months versus 8.5 months in the placebo cohort. While the differences were statistically significant, the absolute difference in survival was modest.

There may be ways to target the EGFR pathway indirectly in pancreatic cancer. For example, MASTL regulates EGFR protein stability and kinase signaling involved in the progression of PDAC [40]. In turn, this makes MASTL a potential target for pancreatic cancer treatment. Dosch et al. [41] reported that STAT3 is a potential biomarker of resistance to gemcitabine in PDAC. Additionally, inhibition of Src and EGFR pathways may be a mechanism to overcome gemcitabine resistance.

KRAS mutation is present in up to 98% of PDAC and often considered to be the inciting event in PDAC carcinogenesis [42]. KRAS activates several downstream oncogenic signaling pathways (e.g., PI3K/AKT/mTOR, Raf-MEK-ERK, RAL-PLD1, T1AM1-Rac). As such, targeting KRAS mediated pathways is an important area of investigation in PDAC research. Pre-clinical studies have demonstrated some promise [43]. However, these preclinical results have not consistently translate to the clinical setting, likely due to the difficulty in re-creating the dense stromal tumor microenvironment of PDAC in pre-clinical models. As such, inhibition of downstream pathways of KRAS may exert different effects in the pre-clinical studies compared with the clinical setting.

Several trials have evaluated the efficacy of inhibitors of the PI3K/AKT/mTOR pathway. Several PI3K inhibitors, including copanlisib, alpelisib, and buparlisib, have been demonstrated to be safe in phase I studies with low efficacy [44, 45, 46]. Everolimus, an mTOR inhibitor, has demonstrated some efficacy in the clinical setting. Kordes et al. [47] enrolled 31 patients with advanced PDAC who received capecitabine and everolimus as either first- or second-line therapy. These patients had an acceptable toxicity profile with a median overall survival of 8.9 months. However, other studies of mTOR inhibitors as a monotherapy or combination therapy have not demonstrated clinical efficacy, including other trials with everolimus [48, 49, 50, 51]. Perifosine is an AKT inhibitor that demonstrated significant activity in the preclinical setting, but failed to demonstrate efficacy in clinical trials [52, 53]. These therapies have also been tried in combination with other targeted therapies with little success [50, 54, 55, 56]. Unfortunately, none of the PI3K/AKT/mTOR inhibitors have demonstrated a survival benefit in patients with PDAC.

Mutations in KRAS can lead to inappropriate activation of the MAPK pathway (Ras-Raf-MEK-ERK) and act as genetic drivers in the initiation and progression of tumors. Unfortunately, it has been challenging to target this pathway effectively in PDAC. Tipifarnib, which showed promise in preclinical studies, has failed to demonstrate efficacy in clinical trial [57, 58, 59, 60]. Sorafenib, a TKI, demonstrated disease stability in early clinical trials, but did not improve survival when used as a monotherapy or in combination with gemcitabine [61, 62, 63]. Selumetinib, a MEK inhibitor, was well tolerated when tested in the second- or third-line setting, but has not demonstrated significant clinical activity [64, 65]. The ERK inhibitor ulixertinib was tested in combination with gemcitabine and nab-paclitaxel in treatment-naïve patients with metastatic PDAC [66]. While effective (median OS 12.23 months), the study was stopped for increased adverse events. There is currently an ongoing phase I trial evaluating ulixertinib and palbociclib as a second-line therapy in patients with solid tumors, including PDAC (NCT03454035).

Based on pre-clinical evidence of a synergistic effect, trametinib (MEK inhibitor) has been tested in combination with erlotinib (EGFR inhibitor). However, this combination only demonstrated mild clinical efficacy with a median OS 7.3 months [67]. Additionally, the combination of radiation therapy, pembrolizumab, and trametinib demonstrated slightly improved OS compared with radiation therapy and gemcitabine among patients with locally recurrent PDAC (median OS: 14.9 months versus 12.8 months, respectively) [68]. A recent phase II trial of patients with advanced PDAC who progressed on chemotherapy were treated with trametinib and an oral FAK inhibitor with a median OS of 3.6 months and progression free survival of 1.6 months [69]. Trametinib may prove to be effective if the right combination therapy or subset of patients is identified, but more work is still needed to demonstrate long term clinical efficacy.

Until now, targeting KRAS could only be done through an indirect therapeutic strategy that focused on inhibiting downstream effects, as targeting KRAS directly was not possible. However, recent research has potentially overcome this obstacle by creating a compound that can target the KRAS${}^{\text{G12c}}$ mutant allele [70]. This finding has led to the development of sotorasib and adagrasib [71, 72]. These inhibitors have largely had success in non-small cell lung cancer where the mutation frequency of KRAS${}^{\text{G12c}}$ is 13.8%. Unfortunately, KRAS inhibitors have demonstrated less success for PDAC in which the KRAS${}^{\text{G12c}}$ mutation frequency is 1% [43]. In the phase I/II trial of patients with KRAS${}^{\text{G12c}}$ mutated PDAC who failed first line chemotherapy, sotorasib had some antitumor activity [73]. The KRYSTAL-1 phase I/II trial is currently evaluating adagrasib as a second line therapy in patients with advanced solid tumors and a KRAS${}^{\text{G12c}}$ mutation. At this time, 21 of the 63 patients have PDAC. Adagrasib has been well tolerated and patients with PDAC had an objective response rate of 33.3%, disease control rate of 81%, median progression free survival of 5.4 months, and median overall survival of 8 months [74].

The KRAS${}^{\text{G12d}}$ mutation is the most prevalent variant in PDAC, occurring in approximately 41% of KRAS mutated patients [43]. MRTX1133 is a KRAS${}^{\text{G12d}}$ inhibitor that has demonstrated tumor regression in PDAC cell lines and patient derived xenografts [75]. Given the potential applicability to PDAC, work is currently underway to develop KRAS${}^{\text{G12d}}$ specific inhibitors [76]. In particular, there are several ongoing trials investigating various KRAS targeted drugs in combination therapy (NCT05379985, NCT04916236, NCT04975256).

Cancer vaccination has emerged as a potential immunotherapy for patients with solid tumors by inducing specific targeted immunity. This approach has mainly been studied as a means to personalize medicine by vaccinating patients with antigens specific to their tumor to stimulate the immune system, thereby inducing targeting of cancer cells that present those same antigens. Recent work has explored mRNA vaccines that are specific to a common mutation (e.g., KRAS mutated cancers) [76]. mRNA vaccines can also express multiple antigens at once, potentially allowing for a more sustained and stronger immune response [77]. A phase I trial is currently investigating a vaccine that targets KRAS-mutant non-small cell lung cancer, colorectal cancer, and PDAC (NCT03948763). There are ongoing trials evaluating dendritic cell and peptide vaccines for patients with KRAS-mutated PDAC (NCT03592888 and NCT04117087).

Claudins are a family of transmembrane tight junction proteins that typically reside in the apical region of the cell membrane and form a paracellular barrier [78]. This barrier regulates the passage of ions between cells, maintains homeostasis, and is associated with cell signaling pathways that influence proliferation and differentiation [79]. Claudins can be modified during carcinogenesis and may be a potential therapeutic target. Claudin-18 (CLDN18) has two variants. Zolbetuximab is a monoclonal antibody to the variant CLDN18.2 that has demonstrated some efficacy against gastric, gastric-esophageal junction, and esophageal cancers based on the FAST phase II randomized trial [80]. While the variant CLDN18.2 is not expressed in normal pancreatic tissue, it has been noted to be aberrantly expressed in 60–90% of PDACs [81, 82, 83]. There is an ongoing phase II trial evaluating its use in CLDN18.2-expressing metastatic PDAC (NCT03816163) in combination with gemcitabine and nab-paclitaxel as first-line therapy.