From somatic mutation to early detection: Insights from molecular characterization of pancreatic cancer precursor lesions

Abstract

Pancreatic cancer arises from non-invasive precursor lesions, including pan creatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasm (IPMN) and mucinous cystic neoplasm (MCN), which are curable if detected early enough. Recently, these types of precursor lesions have been extensively characterized at the molecular level, defining the timing of critical genetic alterations in tumorigenesis pathways. The results of these studies deepen our understanding of tumorigenesis in the pancreas, providing novel insights into tumor initiation and progression. Perhaps more importantly, they also provide a rational foundation for early detection approaches that could allow clinical intervention prior to malignant transformation. In this review, we summarize the results of comprehensive molecular characterization of PanINs, IPMNs, and MCNs, and discuss the implications for cancer biology as well as early detection.

Introduction

Pancreatic cancer is expected to be the second leading cause of cancer-related deaths in the U.S. by 2030, with a current 5-year survival rate of only 8% [1,2]. This dismal prognosis is largely due to a lack of early clinical symptoms, with many patients already presenting with advanced disease at initial diagnosis. Therefore, early detection approaches will be critical to improve outcomes in this disease. A great opportunity for early detection is the treatment of premalignant pancreatic lesions, including pancreatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasm (IPMN) and mucinous cystic neoplasm (MCN). However, prevention of pancreatic cancer must be balanced with potential overtreatment of low-risk lesions. Recent advances in sequencing technologies have deepened our understanding of the genetic changes that characterize these lesions, which offer new opportunities for screening and early detection. In this review, we summarize the current research findings on the molecular genetics of PanIN, IPMN, and MCN, and how these discoveries are being incorporated into cutting-edge early detection approaches.

Pancreatic Intraepithelial Neoplasia (PanIN)

PanINs are the most common precursor to invasive PDAC. Approximately 16% of normal pancreata harbor a PanIN lesion [3]. This prevalence increases in older patients and in the setting of invasive PDAC [3]. PanINs are microscopic (<5 mm in diameter by definition) and arise in small pancreatic ducts [4]. Traditionally, PanIN lesions were grouped by a three-tier classification scheme: PanIN-1A/B, PanIN-2, or PanIN-3, depending on the degree of architectural and nuclear atypia [5,6]. To improve reproducibility and clinical relevance, this classification system has recently been revised and now categorizes these lesions into two groups: low-grade or high-grade PanIN [4] (Figure 1). The cause of these lesions remains unclear, but risk factors such as smoking and excessive alcohol consumption may contribute to their pathogenesis [7]. Due to their microscopic size, PanINs are rarely detected using current imaging modalities [8] and thus require alternative approaches for their identification and management. The molecular features that characterize these lesions are key to understanding PanIN tumorigenesis and early events of pancreatic cancer.

Figure 1. Histologic features of PanIN and IPMN.

(A) This low-grade PanIN has uniform basally-located nuclei and apical mucin. There is minimal cytologic atypia. (B) The cytologic atypia is much more pronounced in this high-grade PanIN, which shows nuclear pleomorphism and marked loss of nuclear polarity. (C) This low-grade IPMN shows gastric-type differentiation, with apical mucin and basal nuclei. Like the low-grade PanIN, there is minimal cytologic atypia. (D) This high-grade IPMN shows pancreatobiliary-type differentiation, with architectural complexity, nuclear pleomorphism and loss of nuclear polarity. (E) The epithelium of this low-grade MCN shows similar cytologic features to low-grade IPMN and PanIN, but the lesion is characterized by its unique ovarian-type stroma. (F) This high-grade MCN shows typical high-grade cytology, including nuclear pleomorphism and loss of nuclear polarity – ovarian-type stroma underlies the neoplastic epithelium.

Molecular Features of PanIN

Numerous studies have identified somatic mutations in KRAS to be one of the earliest drivers of PanIN progression [9–11] (Table 1). KRAS encodes a small GTPase that regulates signal transduction for many cellular activities such as growth, survival and proliferation [12]. Upon activation, this GTPase stimulates multiple cellular pathways including mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K), and Ras-like (Ral). Recently, it has been shown that over 90% of all PanINs, regardless of histological grade, contain at least one KRAS alteration [9]. Additionally, several studies have utilized mouse models to demonstrate that endogenous expression of mutant KRAS in murine pancreata is essential for initiating the development of PanIN [13–15]. Altogether, these data suggest that KRAS alterations play an important role early in PanIN development.

Table 1:

Driver gene mutation prevalence in pancreatic precursors

Lesion	Driver Gene	Mutation Prevalence	Common alterations	Reference
Low-grade PanIN	KRAS	>90%	Missense mutation (codons 12, 13, 61)	[9–11]
	p16/CDKN2A	0-12%	Inactivating mutation/LOH, homozygous deletion, promoter hypermethylation	[17–24]
	TP53	<5%	Missense mutation/LOH	[11,19,28–30]
	SMAD4	<5%	Inactivating mutation/LOH, homozygous deletion	[11,19,27–30]
High-grade PanIN	KRAS	>95%	Missense mutation (codons 12, 13, 61)	[9–11]
	p16/CDKN2A	18%	Inactivating mutation/LOH, homozygous deletion, promoter hypermethylation	[17–24]
	TP53	10-15%	Missense mutation/LOH	[11,19,28–30]
	SMAD4	<5%	Inactivating mutation/LOH, homozygous deletion	[11,19,27–30]
Low-grade IPMN	KRAS	43-89%	Missense mutation (codons 12, 13, 61)	[47–53]
	GNAS	41-77%	Missense mutation (codon 201)	[50,52,54,55]
	RNF43	10%	Inactivating mutation/LOH	[59]
	p16/CDKN2A	<5%	Inactivating mutation/LOH, homozygous deletion, promoter hypermethylation	[53,54,59]
	TP53	<5%	Missense mutation/LOH	[53,59]
	SMAD4	<5%	Inactivating mutation/LOH, homozygous deletion	[53,54,59]
High-grade IPMN	KRAS	34-71%	Missense mutation (codons 12, 13, 61)	[47–53]
	GNAS	42-72%	Missense mutation (codon 201)	[50,52,54,55]
	RNF43	25-75%	Inactivating mutation/LOH	[54,57,59]
	p16/CDKN2A	0-15%	Inactivating mutation/LOH, homozygous deletion, promoter hypermethylation	[53,54,59]
	TP53	18-20%	Missense mutation/LOH	[53,59]
	SMAD4	<5%	Inactivating mutation/LOH, homozygous deletion	[53,54,59]
Low-grade MCN	KRAS	3-26%	Missense mutation (codons 12, 13, 61)	[91–93]
	GNAS	0%	Missense mutation (codon 201)	[50,52,54]
	RNF43	12%	Inactivating mutation/LOH	[54]
	p16/CDKN2A	0-14%	Inactivating mutation/LOH, homozygous deletion, promoter hypermethylation	[93,95,96]
	TP53	0%	Missense mutation/LOH	[54,93,95]
	SMAD4	0%	Inactivating mutation/LOH, homozygous deletion	[95,98]
High-grade MCN	KRAS	50-100%	Missense mutation (codons 12, 13, 61)	[91–93]
	GNAS	0%	Missense mutation (codon 201)	[50,52,54]
	RNF43	25%	Inactivating mutation/LOH	[54]
	p16/CDKN2A	50-59%	Inactivating mutation/LOH, homozygous deletion, promoter hypermethylation	[93,95,96]
	TP53	25-56%	Missense mutation/LOH	[54,93,95]
	SMAD4	0%	Inactivating mutation/LOH, homozygous deletion	[95,98]

The loss of CDKN2A/p16 expression typically occurs after KRAS mutation and is more prevalent in high grade PanIN (Table 1). During cellular stress (i.e. DNA damage, hyperproliferative signals) the p16 protein acts as an important regulator of cell proliferation by blocking phosphorylation of RB, which inhibits passage through the G1/S cell cycle checkpoint [16]. Loss of p16 expression has been shown in approximately 27%, 55%, and 71% of cancer-associated PanIN-1A/B, PanIN-2, PanIN-3, respectively [17,18]. A more recent study, comprising a larger cohort of lesions, indicated that 32% of low-grade PanIN and 83% of high-grade PanIN show abrogation of p16 expression [19]. This loss can occur by several different mechanisms including homozygous deletion, intragenic mutation coupled with loss of the second allele, and epigenetic silencing by promoter hypermethylation [20–22]. A common occurrence alongside homozygous deletion of CDKN2A is deletion of the MTAP gene, thus producing concordant loss of both p16 and MTAP protein expression [23]. This can allow immunolabelling for both MTAP and p16 proteins to serve as a surrogate marker for homozygous deletion of CDKN2A, which has been found in 8% of PanIN lesions [23]. Hypermethlyation in the promoter region of CDKN2A has been found in 6% and 21% of low-grade and high-grade PanINs, respectively [24]. These findings suggest that loss of p16 increases with grade and is not an initiating event in PanIN tumorigenesis.

Loss of expression of TP53 and SMAD4 is found almost exclusively in high-grade PanIN and invasive PDAC (Table 1). The proteins encoded by TP53 and SMAD4 play critical roles in mediating cell cycle arrest, cellular senescence and apoptosis in response to cellular stresses [25,26]. Activation of p53 often occurs in response to DNA damage signals, while SMAD4 mediates TGF-β signaling. Several early studies used immunohistochemistry (IHC) to show aberrant p53 and SMAD4 expression in more than 30% of high-grade PanIN lesions, while expression is retained in nearly all low-grade PanINs [19,27,28]. More recently, targeted and whole exome sequencing studies have revealed that only 15% of high-grade PanINs had TP53 mutations and none harbored mutations in SMAD4 [11,29]. Additionally, IHC of these tumor suppressor genes showed rare abnormal expression in high-grade PanIN [29,30]. Discordance in TP53/SMAD4 mutation prevalence between these studies may be explained by a key confounding factor: so-called “cancerization of the ducts,” in which an invasive cancer invades into and spreads within the pancreatic duct system, histologically mimicking high-grade PanIN. This phenomenon has been reported in as many as 70% of resected PDACs [4]. Unlike the aforementioned sequencing studies, the earlier studies sampled PanIN lesions in the setting of invasive PDAC. It is therefore possible that these PanINs were misclassified and actually represented the intraductal spread of invasive carcinoma, which is histologically indistinguishable from high-grade PanIN [4].

In addition to specific gene mutations, there are several other genetic abnormalities that have been implicated in PanIN tumorigenesis. One of the earliest events in PanIN development is telomere shortening [31,32]. Telomeres maintain the integrity of the genome by protecting against chromosomal breakage and interchromosomal fusion [31]. Dysfunctional and shortened telomeres lead to chromosomal instabilities, such as an increased incidence of atypical mitosis and anaphase bridges [33]. Several studies have indicated that approximately 90% of low-grade PanIN lesions have telomere shortening [32,33]. Furthermore, this shortening is exacerbated as PanINs increase in histological grade, with the shortest telomeres belonging to invasive PDAC [33].

A later event in PanIN progression is chromosomal instability and subsequent copy number alterations. These have been widely characterized in PDACs; however, only recently have copy number alterations been described in PanIN. In 2011, Hong et al. analyzed low-grade PanINs from patients with a family history of pancreatic cancer and found that unlike PDAC, these lesions harbor very few chromosomal copy number alterations [34]. Interestingly, a later study of high-grade PanIN found widespread copy number losses in CDKN2A, TP53, and SMDAD4 [35]. There was also strong evidence for bi-allelic inactivation in CDKN2A and TP53, but not SMAD4. This study also identified chromothripsis events in PanIN – a phenomenon characterized by chromosomal shattering and chaotic reassembly. They found that nearly 40% of PanIN lesions were affected by chromothripsis-like events. The vast majority of these chromothripsis events were found in high-grade PanINs; however, it should be noted that these findings are confounded by possible intraductal spread of invasive cancer. Interestingly, Notta et al. identified chromothripsis in 65% of pancreatic cancers, suggesting that these events may promote progression to invasive cancer [36].

Intraductal Papillary Mucinous Neoplasm (IPMN)

IPMN is another common precursor to pancreatic cancer and the most common cyst of the pancreas. Originally thought to be uncommon, improvements and expanded use of imaging modalities have revealed that nearly 14% of the U.S. adult population harbors a pancreatic cyst [37,38]. IPMNs are grossly visible (>1cm in diameter), mucin-producing neoplastic cysts that arise within the main pancreatic duct or branch ducts [6]. Similar to PanINs, IPMNs were formerly grouped by a three-tier classification system based on dysplasia: low, intermediate, or high-grade [6]. This classification was revised to a two-tier system where the former intermediate-grade category is now considered low-grade [4] (Figure 1). IPMNs are also categorized by histologic subtype based on the direction of differentiation of the lining epithelium: gastric, intestinal, pancreatobiliary and oncocytic. Recently, oncocytic-type IPMNs, often referred to as intraductal oncocytic papillary neoplasms (IOPNs), were shown to be unique neoplasms from IPMNs and genetically distinct from the other histologic subtypes [39,40]. Gastric, intestinal and pancreatobiliary-type IPMNs can progress to conventional ductal/tubular carcinomas, while intestinal-type IPMNs can also give rise to colloid carcinomas characterized by extensive stromal mucin accumulation [41,42]. The risk of malignancy associated with an IPMN varies depending on numerous factors (i.e. size, location, grade of dysplasia), with 30–50% of surgically resected IPMNs harboring invasive carcinoma [43–45]. Similar to PanIN, molecular studies of IPMN have demonstrated that the progression from low-grade IPMN to high-grade IPMN is associated with an accumulation of genetic changes that eventually give rise invasive carcinoma [46]. Therefore, an understanding of the molecular drivers that characterize IPMNs is critical for developing effective early detection strategies for pancreatic cancer.

Molecular Features of IPMN

Numerous studies have identified genetic alterations that play a key role in IPMN tumorigenesis. The most common alterations in IPMN are somatic mutations in the oncogenes KRAS and GNAS (Table 1). Mutations in KRAS occur in 50–80% of all IPMNs and are thought to be an early event of IPMN development [47–53]. The identification of GNAS mutations in IPMN was pioneered by Wu et al, which defined a new pathway for pancreatic neoplasia [50]. Since then, GNAS alterations have been found in 40–70% of all IPMNs and remarkably, are not typically found in other pancreatic precursors, or in invasive PDAC not associated with an IPMN [50,52,54–56]. In association with IPMN, GNAS mutations are found in 23–37% of invasive carcinomas [52,55,57]. GNAS encodes the alpha-subunit of a stimulatory guanine nucleotide-binding protein, which activates the cyclic-AMP cascade, leading to cell growth and proliferation [58]. GNAS mutations are most prevalent in intestinal-type IPMNs, found in 70–100% of these neoplasms [50,52,59,60]. Overall, more than 90% of all IPMNs harbor a KRAS and/or GNAS mutation [54], making them important drivers of IPMN development. Moreover, their prevalence in low-grade IPMNs suggests that alterations in these oncogenes may be initiating events in IPMN formation.

Another commonly mutated gene in IPMN is RNF43, found in 10–75% of IPMNs [54,57,59] (Table 1). Alterations in RNF43 are typically inactivating (i.e. nonsense, frameshift) and accompanied by loss of heterozygosity, implicating RNF43 as a tumor suppressor gene in IPMN tumorigenesis. The RNF43 protein is a transmembrane E3 ubiquitin ligase that serves as a negative regulator of Wnt signaling, thereby inhibiting cell proliferation [61]. Whole-exome and targeted sequencing studies have identified mutations in RNF43 in 6–11% of all invasive PDACs, not just those associated with an IPMN [56,62,63]. Recently, loss of function RNF43 mutations have been reported in other cystic precursors and less frequently in PanIN [29,54].

Inactivation of CDKN2A/p16 has been shown to play a role in IPMN progression, particularly during late-stage development (Table 1). Many studies performing IHC have found p16 loss in IPMNs with high-grade dysplasia (50–100%), and less frequently in low-grade IPMNs (10–51%) [19,53,64,65]. However, several next-generation sequencing (NGS) studies have described CDKN2A mutations at a much lower prevalence: 0–2% of low-grade IPMNs and 0–15% of high-grade IPMNs [53,54,59]. This discrepancy may be explained by alternative mechanisms of CDKN2A gene silencing. For example, epigenetic silencing by promoter hypermethlyation has been described in 21% of high-grade IPMNs [66]. Additionally, allelic loss of chromosome 9p was found in 18–62% of IPMNs [67,68]. Altogether, these studies indicate that loss of CDKN2A/p16 is mediated by the same three mechanisms described in PanIN and is a later event in IPMN tumorigenesis.

Mutations in TP53 are extremely rare in low-grade IPMNs but appear much more frequently in high-grade IPMNs (Table 1). Prior to the widespread use of NGS technologies, the literature described variable p53 expression in IPMNs [69]. Several studies have found diffuse nuclear p53 labeling in invasive carcinomas but were unable to detect p53 in IPMN [70,71]. Others showed variable p53 staining in 40–50% of high-grade IPMNs [19,48]. Targeted, massively-parallel sequencing studies have identified TP53 mutations in 15–20% of high-grade IPMNs and 0–5% of low-grade IPMNs [53,59]. These data suggest that TP53 mutations are late-occurring alterations and may play a role in the malignant progression of IPMN.

Unlike the previously mentioned tumor suppressor genes, loss of SMAD4 is mainly confined to invasive carcinomas (Table 1). Several studies have analyzed the immunohistochemical expression of SMAD4 in IPMNs and invasive cancers [53,64,72,73]. They all found retained expression of SMAD4 in the vast majority of IPMNs, while typically half of invasive carcinomas show loss of SMAD4. In concordance with the IHC findings, targeted and whole exome sequencing studies also found SMAD4 mutations to be very rare in IPMN [53,54,59]. However, loss of heterozygosity (LOH) studies using polymerase-chain reaction (PCR)-based microsatellite analysis found allelic loss of 18q in 22–38% of IPMNs [67,68]. Overall, it seems SMAD4 inactivation is not involved in early-stages of IPMN development but is important for its transition to invasive carcinoma.

Mutations in several other cancer-related genes have been reported in IPMNs at low prevalence, such as PIK3CA, BRAF, PTEN and STK11 [49,74–77]. Mutations in the oncogenes PIK3CA and BRAF have also been reported to be drivers in invasive PDAC and many other cancer types [49,56,74]. Early studies reported loss-of-function mutations in PTEN at relatively low frequencies in invasive PDAC; however, more recent studies have found loss of at least one copy of the PTEN gene can help drive malignant progression of both human and mouse PDACs [78,79]. Peutz-Jeghers syndrome patients have an elevated risk of pancreatic malignancy, and commonly harbor germline mutations in the tumor suppressor gene STK11 [80,81]. Several studies have demonstrated LOH at the STK11 locus in sporadic PDACs and other cancers of the gastrointestinal tract, breast, and ovaries [82–85].

Mucinous Cystic Neoplasm (MCN)

MCN is the third most common precursor to PDAC and make up approximately 10% of pancreatic cysts [54,86]. MCNs are mucin-producing lesions and are distinguished from IPMNs by multiple gross and microscopic features, such as lack of involvement with the pancreatic ductal system and characteristic ovarian-type stroma [86,87]. Akin to PanINs and IPMNs, MCNs were previously categorized into three groups based on dysplasia, but this has since been revised into a two-tiered classification system [4]. Approximately 11–16% of surgically resected MCNs have been reported to harbor invasive carcinoma, demonstrating the importance of understanding the common molecular features that characterize these cysts [86,88,89].

Molecular Features of MCN

Like PanIN and IPMN, MCNs harbor genetic changes that lead to progression of these tumors and pancreatic carcinogenesis. KRAS mutations are reported most frequently, found in 3–100% of MCNs [50,52,90–93]. The frequency of KRAS alterations also seems to increase with grade of dysplasia [92–94]. Unlike IPMNs, mutations in GNAS have not been found in MCNs [50,52,54]. While alterations in RNF43 have only been interrogated in a small number of MCNs, Wu et al found RNF43 mutations in 12% of low-grade MCNs and 25% of high-grade MCNs in their cohort [54]. Loss of CDKN2A/p16 may play a role in later stages of MCN progression. One study found loss of CDKN2A in 5/9 high-grade MCNs, but not in low-grade [93]. Of note, only 2 of these high-grade cysts had biallelic deletion of CDKN2A and complete loss of p16 expression. Promoter hypermethylation of CDKN2A also seems to be more prevalent in high-grade MCNs, reported in 14% of low-grade and 50% of high-grade lesions [95,96]. Aberrant expression of TP53 also appears to occur late in MCN development. Early studies have shown a lack of p53 overexpression in low-grade MCNs, yet found strong positive staining concentrated in areas of high-grade dysplasia and invasion [92,97]. More recently, sequencing approaches have found 25–56% of high-grade/invasive MCNs harbor mutations in TP53, but not in low-grade MCNs [54,93,95]. Similar to IPMNs, loss of SMAD4 expression appears predominantly in invasive MCN and PDAC. In one study, which examined 36 MCNs, SMAD4 expression was retained in both low and high-grade MCNs, while 86% of invasive carcinomas arising from MCNs showed loss of SMAD4 expression [98]. Another study used both a sequencing-based approach and IHC to identify mutations in exons 1–11 of SMAD4 [95]. These researchers found 71% of invasive MCNs harbored somatic mutations in SMAD4. Several other less frequently altered genes have been reported in MCN, including PIK3CA and PTEN [99,100].

Genetic Heterogeneity and Multifocal Neoplasia

Genetic heterogeneity of driver gene alterations in PanIN and IPMN have recently been described due to increased sensitivity of NGS approaches. Several investigators have identified multiple KRAS mutations within a single PanIN lesion [11,29,30]. Remarkably, more than one KRAS mutation occurs in as many as 30% of PanINs, suggesting apparent genetic heterogeneity with respect to driver genes in these pancreatic precursors. In IPMN, Wu et al. used targeted sequencing to analyze cyst fluid from 19 patients [50]. They found 11% of IPMNs contained two different KRAS mutations, 2% contained three different KRAS mutations and 4% contained two different GNAS mutations. Additionally, Felsenstein et al. microdissected epithelium from two distinct regions of IPMN and subsequently performed deep targeted sequencing of pancreatic driver genes [57]. These studies reported that 23% of IPMNs had multiple KRAS and/or GNAS mutations, and remarkably one of these IPMNs contained four unique KRAS mutations. The finding of multiple KRAS and GNAS mutations within a single PanIN/IPMN suggest a more complex pattern of clonal evolution in these precursor lesions than previously appreciated.

In addition to the description of genetic heterogeneity within single precursor lesions, the results of several studies have provided evidence for multifocal neoplasia in the pancreas. First, Hosoda et al. analyzed high-grade PanIN and adjacent low-grade PanIN using targeted NGS in ten patients [29]. Surprisingly, they found only one pair of low-grade/high-grade PanIN that were ‘likely related,’ while the vast majority of pairs were genetically independent. This suggests the possibility of an alternative mechanism for PanIN progression – de novo development of high-grade PanIN adjacent to an independent, unrelated low-grade PanIN [7]. Next, Pea et al. performed targeted sequencing on IPMN and PDAC lesions from 13 patients who developed disease progression in their remnant pancreas following resection of IPMN [101]. Analysis of both the primary and recurrent IPMN revealed that more than half of the cases had genetic alterations not shared between these neoplasms and were classified as ‘likely independent.’ Finally, Felsenstein et al. described the genetic relationship between invasive carcinoma and co-occurring IPMN [57]. Interestingly, 18% of PDACs and co-occurring IPMNs did not share any driver gene mutation and were considered ‘likely independent,’ despite their close anatomic proximity. While this has not been investigated in detail with respect to MCN, clinical data suggests these lesions tend to be unifocal [89]. Altogether, these studies support the concept of a field defect in a subset of patients, which leads to an increased risk of neoplasia throughout the pancreas.

Early detection of pancreatic precursors

As mentioned previously, PanIN, IPMN, and MCN are common precursors of PDAC and therefore represent key targets for early detection approaches. Imaging modalities, such as computed tomography (CT), magnetic resonance imaging (MRI), and endoscopic ultrasound (EUS) are commonly used to detect lesions in the pancreas. In a multi-institution study conducted by Canto et al., these imaging methods were used to screen 225 asymptomatic, high-risk individuals (HRI) [102]. They found that 42% of HRIs had at least one pancreatic mass or a dilated pancreatic duct. Among these, proven or suspected lesions were later identified in 92% of patients. These imaging modalities can also be augmented to target specific structures and molecules. For example, Neesse et al. designed a fluorochrome that specifically targets claudin-4 [103], a protein known to be upregulated in pancreatic neoplasia [104,105].

While these imaging-based approaches are useful to detect pancreatic cysts, they may not reliably differentiate cyst type or important histological features (i.e. grade of dysplasia), which can better predict likelihood of progression. This distinction is clinically important because pancreatic cysts represent a diverse group of lesions, some of which are low-risk while others progress to invasive carcinoma. As a result, a more reliable determination of precursors with a higher malignant potential will be critical. Many studies have demonstrated the value of collecting cyst fluid by EUS-fine needle aspiration from patients diagnosed with pancreatic cysts. Several reports highlight the importance of cytological evaluation of cyst fluid for atypical epithelial cells, which can serve as a predictor of malignancy [106–108]. Others have reported on the diagnostic value of biochemical markers in cyst-fluid for differentiating likely benign, serous cysts from mucinous cysts which have greater risk of malignancy [109–111]. Furthermore, two independent studies used targeted NGS to analyze cyst-fluid [100,112]. The investigators used a combination of molecular markers to categorize a cyst as IPMN with 76–100% sensitivity and 84–96% specificity or as MCN with 100% sensitivity and 75% sensitivity. This approach identified IPMNs with high-grade dysplasia or invasive carcinoma with 88% sensitivity and 69–97% specificity. These results highlight the ability of cyst fluid sequencing to preoperatively determine cyst type as well as predict grade of dysplasia in premalignant cysts. Another marker that could be used to differentiate grade of dysplasia in IPMN is telomere fusion, which frequently occurs in critically short telomeres. Hata et al. developed a real-time quantitative PCR method to detect telomere fusion in cyst-fluid [113]. They detected telomere fusions in 27% of high-grade IPMNs, but not in low-grade IPMNs. Additionally, several studies have used cyst fluid to identify other molecular changes such as telomerase activity and microRNA levels [114,115]. As a result from these findings and others, several institutions are implementing NGS-based molecular tests using cyst fluid to aid in the clinical evaluation and diagnosis of pancreatic cysts.

While these studies provide important data for the early detection of IPMN, they are limited by sampling only cyst-fluid. As previously mentioned, Felsenstein et al found that a substantial portion of PDACs with co-occurring IPMNs are unrelated [57]; therefore, analysis of cyst fluid may not detect the true precursor of the cancer. Furthermore, traditional imaging tests are unable to detect PanIN lesions altogether. A promising approach is genetic analysis of secretin-stimulated pancreatic juice collected from the duodenum. Kanda et al. determined the prevalence of TP53 mutations in pancreatic juice from individuals undergoing pancreatic evaluation [116]. The prevalence of mutant TP53 in pancreatic juice samples from patients whose highest grade lesion was PanIN-3 (40%) is very similar to its prevalence in primary resected PanIN-3 lesions (48%). More recently, Yu et al. developed a digital NGS approach to detect low-abundance mutations in pancreatic juice samples [117]. In two cases of high-risk individuals, digital NGS was able to detect SMAD4 or TP53 mutations more than one year before their pancreatic cancer diagnosis. Finally, Suenaga et al. performed digital NGS using a targeted 12-gene panel to evaluate mutation concentrations in pancreatic juice samples [118]. Mutation concentrations in genes other than KRAS/GNAS were higher in patients with PDAC or high-grade precursors relative to all other subjects. Consistent with previous studies and the molecular progression of PanIN and IPMN, this pancreatic juice analysis found several predictors of pancreatic cancer or a high-grade precursor: presence of SMAD4 mutations, high SMAD4/TP53 mutation concentrations and high overall mutation concentrations. While the lack of these mutations does not indicate absence of disease, these results highlight the potential value of a pancreatic juice NGS screening test for patients undergoing pancreatic evaluation. Finally, analysis of molecules and cells in the blood also does not require sampling of a specific lesion in the pancreas and thus is an alternative approach to early detection. For example, some groups have reported analysis of circulating epithelial cells (CECs) in the bloodstream in patients with pancreatic cancer as well as IPMN [119,120]. This method has been used to detect CECs in patients with cystic lesions prior to the clinical diagnosis of invasive PDAC; however, the sensitivity of this method to detect PanIN is not known but is likely to be low considering the lack of access of these precursor lesions to the bloodstream. Numerous serum markers have also been suggested as potential biomarkers for early detection of pancreatic neoplasia, but these remain to be systematically evaluated [121].

In the future, more comprehensive analyses of the genetic heterogeneity in pancreatic cancer precursor lesions can further elucidate their clonal evolution and neoplastic progression. Additionally, the identification and validation of molecular markers that can reliably distinguish low-risk lesions from lesions with a high risk of malignant progression will be fundamental for effective early detection approaches. These promising approaches require a deep understanding of the molecular alterations that occur during pancreatic tumorigenesis via the PanIN, IPMN and MCN pathways. Pathologists will play a key role in the implementation of these approaches for the better management of these precursor lesions and ultimately the prevention of pancreatic cancer.