Classification, Morphology, Molecular Pathogenesis, and Outcome of Premalignant Lesions of the Pancreas Open Access

Cystic lesions of the pancreas are common, and with ever-improving radiologic techniques, many are discovered incidentally during imaging for another indication.³  Although some pancreatic cysts do not pose a risk to the patient, intraductal papillary mucinous neoplasm (IPMN), mucinous cystic neoplasm (MCN), and intraductal tubulopapillary neoplasm are true neoplasms that can progress to PDAC. Excitingly, we now know that the earlier we can detect, correctly classify, and treat these cystic neoplasms, the less likely they are to progress to invasive PDAC.

Intraductal papillary mucinous neoplasms are grossly visible cystic neoplasms (>1 cm) that communicate with the pancreatic ductal system. The cysts are lined by neoplastic mucinous columnar epithelium that often forms papillary structures that protrude into the cyst.⁴  Although IPMNs may occur anywhere in the pancreas, they are more common in the pancreatic head. These neoplasms are equally prevalent in men and women, and the usual age of diagnosis is 60 to 70 years.^5–7 

Many IPMNs are discovered incidentally in asymptomatic patients, whereas other IPMNs come to clinical attention because they produce abdominal-localizing symptoms including abdominal pain, weight loss, pancreatitis, diarrhea (secondary to pancreatic exocrine insufficiency), jaundice, nausea/vomiting, and new-onset diabetes mellitus (secondary to pancreatic endocrine insufficiency).^5,7,8  Radiologic features that favor a diagnosis of IPMN over other cystic lesions include involvement of the duct system, dilation of the main pancreatic duct, clusters of cysts indicating dilated branch ducts, and surrounding parenchymal atrophy.⁹  An endoscopic ultrasound is often performed with the aim of confirming the IPMN diagnosis. The finding of mucin excretion from the ampulla of Vater (fish-mouth appearance) is practically pathognomonic for IPMN.^6,10  Endoscopic ultrasound also allows for fine-needle aspiration of the lesion and for cyst fluid sampling.

The hallmark feature in cytology for a neoplastic mucin-producing cyst (either MCN or IPMN) includes the presence of thick extracellular mucin (neoplastic mucin; Figure 1, a).^11,12  Stelow et al¹³  reviewed the fine-needle aspirations of 31 IPMNs and found extracellular mucin in a majority of these lesions. Other cytologic features for IPMN (and MCN) include mucinous glandular epithelial cells in groups, papillary clusters, single cells with varying degrees of cytologic atypia, and cyst debris (Figure 1, b). The aspirate samples can be cellular or paucicellular, with high-grade lesions demonstrating increased cellularity and decreased mucin. The lining epithelium of an IPMN can be of gastric-foveolar, intestinal, oncocytic, or pancreaticobiliary types, but definite characterization of the lining epithelium is not required in the cytologic evaluation.

Genevay et al¹²  evaluated 112 histologically confirmed mucin-producing cysts of pancreas and demonstrated that high-grade atypia on cytologic evaluation was the most sensitive predictor of malignancy in mucinous cysts. High-grade dysplasia is defined by the presence of atypical epithelial cells that are arranged in 3-dimensional aggregates, small papillary clusters, or single epithelial cells with a loss of the normal honeycomb architecture (Figure 1, c), with increased nuclear to cytoplasmic ratio, irregular nuclear membranes, abnormal chromatin pattern, crowding, and molding, with or without intracellular mucin (Figure 1, d). Cysts demonstrating high-grade dysplasia may also show background tumoral necrosis.

In addition to cytologic evaluation from a fine-needle aspiration specimen, cyst fluid analysis for carcinoembryonic antigen and amylase levels is extremely helpful in the diagnosis of neoplastic mucin-producing pancreatic cysts. An elevated cyst fluid carcinoembryonic antigen level reflects the presence of mucinous epithelium and is seen in both MCNs and IPMNs. Brugge et al¹⁴  reported that a cutoff carcinoembryonic antigen level of 192 ng/mL has a 73% sensitivity and 84% specificity in differentiating mucinous from nonmucinous pancreatic cysts. This same study¹⁴  found that among all the pancreatic cyst fluid diagnostic parameters currently in use, carcinoembryonic antigen concentration was the most accurate test for the diagnosis of a nonmalignant mucinous cyst, even more so than cytologic evaluation. Because IPMNs are connected to the pancreatic ductal system, amylase levels may also be elevated in this neoplasm and may be useful in differentiating IPMN from MCN.¹⁵ 

The risk of an IPMN having associated invasive carcinoma varies depending on the location and size of the cystic lesion. Clinical and radiologic criteria are used to assess this risk preoperatively, and approximately 30% of IPMNs that are surgically resected ultimately demonstrate invasive disease.^16–20  A subset of IPMNs are not recommended for surgical resection because they are at low risk of progression; these include small, branch-duct IPMNs.³  Clinical symptoms related to the pancreas, such as abdominal pain, jaundice, and weight loss, are worrisome for an associated invasive PDAC. Imaging features that are more often seen in IPMN with an associated PDAC include main pancreatic duct or distal common bile duct dilation (Figure 2, a), cysts larger than 3.0 cm, numerous cysts, and the presence of a solid component or mural nodule.^8,19,21,22  The single most important factor for the surgical pathologist to evaluate in resected pancreata with IPMN is the presence or absence of an associated invasive carcinoma. This starts with a careful gross evaluation for the presence or absence of mural or parenchymal masses, and extends to thorough, if not complete, sampling of the IPMN for histology.

One way to classify IPMNs, although in and of itself not prognostically significant, is by the type of mucinous lining seen on histologic examination.²³  The 4 histologic subtypes of IPMN are gastric (Figure 2, b), intestinal (Figure 2, c), pancreatobiliary (Figure 2, d), and oncocytic; these may be found in isolation or be mixed within a single lesion.^4,8,24,25  Gastric-type IPMNs tend to have an undulating (rather than papillary) lining. The cells are cuboidal to columnar with small, basally oriented nuclei and a foveolar mucin cap identical to the foveolar mucinous cells present in gastric mucosa. Intestinal-type IPMNs unsurprisingly resemble colonic mucosa, with cigar-shaped nuclei, villouslike papillae, and often colonic-type goblet cells. Pancreatobiliary IPMNs have increasing architectural complexity with papillae and sometimes bridging or cribriforming of the epithelial lining. The cells themselves are more cuboidal than columnar, with basophilic cytoplasm, minimal mucin, and enlarged nucleoli. In contrast, oncocytic IPMNs, also known as intraductal oncocytic papillary neoplasms, are characterized by cells with abundant eosinophilic cytoplasm; however, like the pancreatobiliary subtype, they also contain cells with atypical nuclei forming complex architecture.²⁴ 

Intraductal papillary mucinous neoplasms with the intestinal or pancreatobiliary subtypes are more likely to harbor an associated invasive PDAC, whereas purely gastric IPMNs are the least likely to have an associated invasive malignancy. The invasive PDACs that arise in intestinal-type IPMNs can be tubular or colloid. Additionally, intestinal-type IPMNs more often involve the main pancreatic duct, whereas branch-duct–type IPMNs usually show gastric-type differentiation.^26,27  These distinctions in epithelial subtype may be useful in conjunction with other features for preoperative planning; however, a substantial minority of IPMNs have been shown to demonstrate a mixed or unclassifiable phenotype when the completely resected cyst is present for evaluation. Because of the heterogeneity inherent in this neoplasm, the clinical and prognostic utility of epithelial subtypes is limited.²⁵ 

The degree of dysplasia, low versus high, should also be documented.²⁸  In general, as an IPMN progresses from low-grade to high-grade dysplasia and eventually to invasive PDAC, cytologic and architectural atypia increase within the lesion. Nuclei in high-grade dysplastic lesions are larger, more irregular, and more hyperchromatic than their low-grade counterparts. Additionally, high-grade nuclei show “loss of polarity,” or stratification, rather than sitting along the basement membrane. Finally, the architecture of a high-grade IPMN has increased complexity of papillary structures and cribriforming.²⁴  Gastric IPMNs are usually low-grade lesions, intestinal-type IPMN may be low or high grade, but pancreatobiliary and oncocytic IPMN are almost always high grade.²⁸  Oncocytic IPMNs are distinct from the pancreatobiliary type in that even if an associated invasive PDAC is present, patients tend to have a better outcome, but these data are limited by the rarity of this subtype.^27,29,30 

The above categorizations and attempts to calculate risk for a particular patient have grown out of our increased understanding of the natural history of IPMNs. First, patients who have invasive PDAC associated with an IPMN have an estimated 5-year survival of approximately 30% to 50%. Although this is a better prognosis than that of patients who have PDAC in the absence of a concurrent IPMN, it is a much lower survival rate than that of patients who have a noninvasive IPMN, who have an estimated 5-year overall survival of 70% to 90%.^7,18,31  For this reason, extensive sampling of an IPMN with adjacent pancreatic parenchyma is vital for accurate diagnosis and prognostication.²³  Second, IPMNs are often multifocal, and patients are therefore at risk for both synchronous and metachronous disease throughout the entire pancreas.^6,20,32  In a patient who has undergone partial pancreatic resection for a noninvasive IPMN, the risk of recurrence or new disease in the remnant pancreas is less than 20%.^7,33,34  In contrast, patients who are found to have invasive PDAC with an IPMN have a recurrence rate of almost 50%, usually as metastatic disease.³⁵  These data highlight the critical importance of following patients, even those with noninvasive disease, after surgery. Although adequate data are lacking for firm recommendations for follow-up of patients who have partial pancreatectomy, the consensus recommendations are currently for at least repeat examinations at 2 years and 5 years postoperatively.^20,36 

The genetic alterations found in IPMN parallel, to a large extent, those found in invasive PDAC, with 2 differences: GNAS and RNF43 are more commonly present in IPMNs and their associated invasive cancers than in invasive PDACs not associated with an IPMN precursor. The discovery of the GNAS oncogene mutation is exciting for 2 reasons. First, GNAS mutations appear to be specific for IPMN when compared with other cystic lesions of the pancreas. Second, patients with McCune-Albright syndrome have postzygotic GNAS mutations resulting in mosaicism, and recent data indicate that these patients may have a higher incidence of IPMN than the general population and therefore may warrant increased surveillance.^6,37–39  As with invasive PDAC, activating point mutations in KRAS occur in IPMN, most often at codon 12. This appears to be an early event in IPMN formation, and KRAS mutations are found in the majority of grossly visible and even smaller (<1 cm) “incipient” IPMNs.^37,40  A variety of other genetic alterations can be found in IPMNs: LKB1/STK11, PK3CA, RNF43, p16/CDKN2A, TP53, and SMAD4/DPC4.^30,41–44 

The importance of the above findings cannot be overstated. First, increased understanding of molecular alterations within IPMNs provides insight into the pathogenesis, spread, and progression of this neoplasm. A recent study by Pea et al³²  used targeted DNA sequencing to investigate the relationships between primary IPMN and progressive disease, including invasive PDAC, in 13 patients who underwent initial IPMN resection followed by completion pancreatectomy at a later date. The authors found that, based on the mutational profile of the neoplasms within the same patient, 7 patients had likely independent lesions, 3 patients had likely related lesions, and 3 patients had lesions of an indeterminate relationship. Only 1 of 8 patients with a negative resection margin had a likely related progressive lesion. Two of 5 patients with a positive resection margin showing high-grade dysplasia had likely related progressive lesions; the remaining 3 patients with positive margin had low-grade dysplasia and did not have genetically related progressive lesions. These results suggest 3 possibilities for IPMN progression in the remnant pancreas. First, as in the patients with high-grade dysplasia at the margin, a patient may have transection of a single neoplasm, and the portion of the transected neoplasm that remains in the remnant pancreas later recurs as a mass. Second, as in the patients with negative margins but related lesions, there may be intraductal “metastasis” of a primary lesion. In these cases, the presence of a negative margin highlights the skip nature of these neoplasms. Finally, as in the patients with genetically distinct neoplasms, there may be independent growth of multiple primary IPMNs.³²  As mentioned before, but worth reemphasizing here, these data highlight the importance of clinically following patients after the resection of an IPMN.

Increased understanding of molecular alterations within IPMNs is also valuable because the mutations can help guide preoperative diagnosis and management. Recent studies of cyst fluid obtained from IPMNs during endoscopic ultrasound have shown that molecular markers can accurately differentiate IPMNs from other cystic lesions of the pancreas.³⁹  Because of the high specificity of GNAS for IPMN, the presence of GNAS in cyst fluid is diagnostic of an IPMN.^39,45  Mutations in RNF43 are similarly specific for IPMN.³⁹  Additionally, because IPMNs connect with the ductal system of the pancreas, secretin-stimulated pancreatic juice collected from the ampulla of Vater during endoscopic ultrasound can also be analyzed for mutations associated with high-grade dysplasia and invasive PDAC.⁴⁶  Finally, a recent study of cyst fluid telomerase activity shows promise for predicting the presence of high-grade dysplasia or invasive carcinoma associated with a cystic lesion.⁴⁷  As these molecular diagnostics become incorporated into clinical practice, we will be better able to provide appropriate preoperative clinical recommendations for patients with pancreatic cysts.

Mucinous cystic neoplasms are grossly visible cystic neoplasms that, in contrast to IPMNs, do not communicate with the pancreatic ductal system. Mucinous cystic neoplasms are lined by neoplastic mucinous columnar epithelium with an underlying distinctive and pathognomonic spindle cell ovarian-type stroma (Figure 3, a and b). This type of cyst is much more common in the body or tail of the pancreas and is much more common in women than in men (∼20:1).^48–50 

Some patients are found to have an MCN incidentally during abdominal imaging, but patients with an MCN can also present with abdominal pain or the feeling of abdominal fullness. Because MCNs do not communicate with the ductal system, patients with an MCN are less likely to present with pancreatitis, jaundice, or new-onset diabetes mellitus than are patients with an IPMN.⁴⁹  Radiologic features that favor a diagnosis of MCN include a solitary cyst with thick cyst walls, internal septations, and peripheral calcifications.¹⁵  As previously described, the cytologic features of MCN are similar to those of IPMN, and it is therefore difficult to distinguish MCN from IPMN based on cytologic features alone. The ovarian-type stroma of an MCN that is seen on histology is rarely encountered on a cytology sample. Cytology can, however, confirm the presence of a mucin-producing cyst using the criteria discussed in the above section on IPMNs.

Mucinous cystic neoplasms have an associated invasive PDAC in approximately 15% to 30% of cases, but the risk of invasion can be adjusted for a particular patient based on clinical, imaging, cytology, and surgical pathology findings.^15,20,50  In general, increasing age is a risk factor for invasive PDAC arising from an MCN. This relationship of carcinoma with age is not surprising given that MCN is a known precursor of invasive PDAC.^15,49,51  On imaging studies, larger cyst size, thickening of the septations within an MCN, and presence of an intracystic or adjacent solid mass are radiologic features that may suggest an associated invasive PDAC.^15,50,52  The cytopathologic features that are worrisome for invasive carcinoma arising from an MCN are similar to those for IPMN: nucleomegaly, clumped and irregular chromatin distribution, irregular nuclear contours, and crowding.

As was true for IPMNs, once resected, the single most important factor for the pathologist to determine is the presence or absence of an associated invasive carcinoma. On sectioning of the specimen, a solid stromal component or an intracystic nodule, often with papillary features, is commonly associated with an invasive component. High-grade dysplasia is a histologic feature that raises concern for an adjacent invasive PDAC, as is the loss or effacement of the classic ovarian-type stroma beneath the mucinous epithelium. For these reasons, extensive, if not complete, sampling of the cyst wall and of any nodular areas adjacent to the cyst is necessary to fully rule out the presence of invasion.⁵⁰  Noninvasive MCNs are graded as low- or high-grade neoplasms, with those formerly considered intermediate grouped in the low-grade category based on the recommendations of an expert panel.²⁸ 

Standard of care for MCN is usually a distal pancreatectomy with a lymph node dissection.²⁰  Patients who are found to have only noninvasive MCN on the resection specimen do not need any additional follow-up, because, in contrast to IPMNs, MCNs are solitary lesions and the 5-year overall survival rate for a patient with noninvasive MCN approaches 100%. The presence of an associated invasive PDAC lowers the 5-year survival to 25% to 35%.^20,50,51  The presence of minimally invasive carcinoma, defined in these cases as microscopic foci of adenocarcinoma that do not invade the pancreatic parenchyma but are limited to the ovarian stroma of the MCN, confers an excellent prognosis. Although minimally invasive carcinoma is likely cured by resection, these lesions may rarely recur.⁵³  Again, this underscores the importance of extensive sampling of the cyst wall for accurate diagnosis and prognostication.

Activating point mutations in KRAS are an early genetic alteration in the development of MCN, whereas inactivation of TP53 and SMAD4 occur as late events. These alterations are similar to those found in other epithelial neoplasms of the pancreas, and therefore are not specific markers for MCN. In fact, presurgical diagnosis of MCN often relies on what is not present in the cyst, rather than what is: MCN does not have a specific genetic mutation, it does not communicate with the main pancreatic duct, and it does not present as multiple cysts.³⁹ 

Because a comprehensive review of intraductal tubulopapillary neoplasm was recently published in this journal, only a brief mention of this precursor will be made here.⁵⁴  Intraductal tubulopapillary neoplasms are grossly visible neoplasms that communicate with the pancreatic ductal system, grow in a tubular pattern, and often have high-grade dysplasia and comedo-like necrosis. Because this is an uncommon entity, data on molecular diagnostics and clinical prognosis are limited.

Although the above sections have focused on grossly visible precursor lesions, the remainder of the review will cover what is known about the microscopic and noncystic lesion known as pancreatic intraepithelial neoplasia (PanIN). Because PanIN is microscopic and often multifocal, it cannot be treated with surgical resection in the manner of a cyst, even though we now believe that most PDACs arise from PanIN lesions. Nevertheless, understanding the stepwise molecular changes that occur in the transformation of normal ductal epithelium through the stages of PanIN to invasive cancer provides insight not only into potential treatments for PDAC, but also into the molecular machinery of invasive malignancy. Because the research available on molecular changes in PanIN is vast, only the most common genetic mutations and deletions will be discussed below.

Pancreatic intraepithelial neoplasias have been recognized histologically for more than a century as lesions histologically distinct from the cuboidal cells of normal ductal epithelium (Figure 4, a). Although a variety of terms have been given to PanIN lesions, consensus has been reached on both terminology and grading of this histologically identifiable lesion.^55,56  The grading system originally proposed for PanIN was a 3-tiered system based on nuclear and architectural features. The first tier was stratified by architecture. Pancreatic intraepithelial neoplasia 1A consists of flat ductal lesions with columnar cells that have small, basally located nuclei. Pancreatic intraepithelial neoplasia 1B has the same cytologic features, but allows for papillary or micropapillary architecture (Figure 4, b). Pancreatic intraepithelial neoplasia 2 has nuclear atypia including loss of polarity, nuclear crowding, and some hyperchromasia with either a flat or papillary architecture (Figure 4, c), whereas PanIN-3 is almost always papillary or micropapillary, may have cribriforming, and has marked nuclear atypia that exceeds that seen in PanIN-2, and may also include macronucleoli and abnormal mitotic figures (Figure 4, d).

Recently, a simplified 2-tier classification has been proposed for both PanIN and IPMN. In this system, PanIN-1 and PanIN-2 are considered low grade and PanIN-3 is termed high grade.²⁸  This change in terminology is largely based on clinical relevance of PanIN lesions. In surgically resected specimens, the presence of Pan-IN 1 or 2 (low-grade PanIN), even when present at a surgical resection margin, has not been correlated with the presence of residual or recurrent PDAC in the remnant pancreas. Pancreatic intraepithelial neoplasia 3 (high-grade PanIN), on the other hand, is frequently found in the presence of invasive PDAC, and should therefore be reported.^28,57,58  In fact, lesions called PanIN-3 may sometimes represent “cancerization of ducts,” or intraductal spread of an invasive PDAC. Adequate sampling of the pancreas should be able to confirm the presence of the invasive component.²⁸ 

Pancreatic intraepithelial neoplasia lesions are often surrounded by lobular parenchymal atrophy.^59,60  When atrophy is multifocal, or solitary of sufficient size, it may be detected on endoscopic ultrasound. In patients at high risk for PDAC, such as those with a strong family history or known genetic risk factor, multifocal atrophy may serve as a surrogate for multifocal PanIN lesions and possible risk of PDAC. Studies suggest that the progression from initial mutation within a pancreatic cell to the invasive event of malignancy takes at least a decade, if not longer.^2,61  Screening a selected subset of the population at risk, then, may actually allow for detection and treatment of high-grade PanIN lesions prior to invasion.^1,59,62 

The genetic changes identified in PanIN correlate with the progression of histologic dysplasia. The earliest alterations found in low-grade PanIN lesions are shortening of telomeres and activating point mutations in KRAS.^63–66  Telomeres are repeated sequences at the ends of chromosomes that are responsible for maintaining the integrity of the genomic material by prevention of breakage during cell division. Telomere shortening and abnormalities of telomerase activity have been identified in a variety of cancers.^66,67 

Mutated KRAS proteins allow for dysregulation and constitutive signaling activation within the mitogen-activated protein kinase and AKT signaling cascades.⁶⁸  Mouse models have shown that the KRAS G12D mutation is sufficient for producing PanIN lesions that spontaneously progress to invasive carcinoma, and studies of human PDAC tissues have repeatedly shown that KRAS codon 12 mutations are highly prevalent in low-grade PanIN as well as high-grade lesions and invasive PDAC.^64,69–71 

Mutations in the tumor suppressor gene CDKN2A usually occur after KRAS mutations, and therefore are more prevalent in high-grade than in low-grade PanIN lesions.^68,71  The protein product of CDKN2A, p16, is a necessary component of the cell cycle that, under normal conditions, inhibits cyclin D1. In PanIN and PDAC pathogenesis, p16 function may be lost secondary to deletion or promoter methylation silencing of CDKN2A.^72,73  Immunohistochemical studies, where nuclear p16 loss serves as a surrogate for CDKN2A loss or silencing, have demonstrated that cell cycle dysregulation is present in a subset of PanIN lesions in the setting of chronic pancreatitis, and have confirmed that the prevalence of cell cycle dysregulation increases with increasing histologic dysplasia.^74,75 

Finally, the mutations that are found almost exclusively in high-grade PanINs (and invasive PDAC) occur in the tumor suppressor genes TP53 and SMAD4.^63,64,76  Under normal conditions, the p53 protein is necessary for several vital cell functions, including the cell cycle, DNA repair, and the apoptotic pathway. Without functional p53, cells may continue to proliferate, even if they are damaged. Loss of p53 is almost always due to loss of one TP53 allele and mutation of the second allele.^68,77  SMAD4/DPC4 is lost in a similar manner to TP53, although homozygous deletion of the gene also may occur. Without a functional Smad4/Dpc4 protein, the transforming growth factor β signaling pathway is interrupted, and tumor growth can continue unchecked. SMAD4 gene inactivation or deletion is associated with a worse prognosis in invasive PDAC.⁷⁸ 

As with the previously mentioned cyst fluid studies for IPMN, techniques to collect pancreatic juice fluid endoscopically from the duodenum and analyze it for mutations common to PanIN and PDAC are in development. Currently, KRAS mutations and elevated telomerase activity can be identified in pancreatic secretions of patients who are known to have PDAC.^47,79  A small proportion of patients who are “normal controls” have also been found to harbor KRAS mutations.⁷⁹  Presumably, these control patients have PanINs that may one day be able to be followed with enhanced imaging and secretion analysis with the goal of preventing progression to invasive carcinoma.

A note about acinar-ductal metaplasia and atypical flat lesions is warranted so that the reader is familiar with these terms. Both acinar-ductal metaplasia and atypical flat lesions have been identified in mouse models of PDAC and the pancreata of human patients who have a familial predisposition to PDAC.²⁸  Because these lesions sometimes harbor KRAS mutations, it is thought that they may represent a metaplastic-dysplastic sequence, but as of yet the true clinical significance of these histologic entities is unknown.^80,81 

Overall, the focus of this review has been on the importance of PDAC precursors as they relate to improved therapies for patients. The timely recognition of PDAC precursors through imaging and cytopathology may allow for intervention before these lesions progress to an invasive malignancy. The single most important issue in evaluating surgical resection specimens is determining whether or not there is an associated invasive carcinoma. Finally, increased knowledge of the molecular alterations and pathways involved in the progression from precursor to invasive malignancy will hopefully lead to new, targeted therapies for patients with PDAC.