Pancreatic Neuroendocrine Neoplasms A Current Summary of ...

Pathology Patterns Reviews

Pancreatic Neuroendocrine Neoplasms A Current Summary of Diagnostic, Prognostic, and Differential Diagnostic Information Mark R. Wick, MD,1 and Fiona M. Graeme-Cook, MB, BCh2 Key Words: Pancreatic endocrine tumor; Islet cell tumor; Immunohistochemistry; Hormonal syndromes; Prognosis

Abstract Pancreatic endocrine tumors (PETs) continue to be challenging diagnostic and prognostic lesions in surgical pathology and clinical medicine. These neoplasms can be graded into 1 of 3 tiers, based on histologic characteristics in likeness to epithelial neuroendocrine tumors in other anatomic sites. However, grade 1 tumors are by far the most common and are the most difficult to prognosticate. The most helpful features by which to gauge the behavior of such lesions include size (3 cm or larger); mitotic activity (2 or more mitoses per 10 high-power [×400] microscopic fields); marked nuclear atypia, especially with atypical mitotic figures; predominant tumor synthesis of gastrin, vasoactive intestinal polypeptide, somatostatin, glucagon, calcitonin, or adrenocorticotropic hormone; complete nonfunctionality of the tumor at an immunohistochemical level; or invasion of blood vessels, nerves, or adjacent organs by the neoplasm. Differential diagnosis of PETs includes lesions such as solid-pseudopapillary neoplasms, acinar carcinomas, metastatic neuroendocrine tumors, and plasmacytomas.

Pancreatic endocrine tumors (PETs) have been a focus of fascination for both pathologists and clinicians for almost a century. Nicholls1 documented an example of a pancreatic neoplasm in 1902 that was termed an “islet cell adenoma,” and Fabozzi2 described a biologically malignant counterpart of that lesion the following year. In the intervening period, it has been realized that PETs likely arise from multipotential epithelial cells in pancreatic ductules, rather than the islets of Langerhans.3,4 Thus, there has been movement away from the diagnostic terms islet cell tumor and islet cell carcinoma in reference to these neoplasms. They are relatively uncommon, being seen in only 0.1% of all autopsies and accounting for approximately 15% of pancreatic neoplasms in surgical series; these percentages equate to an annual incidence of less than 1 pancreatic neuroendocrine tumor per 100,000 persons in the general world population.5 Increasing sophistication in laboratory methods has permitted ever more detailed characterization of several endocrinopathic disorders that potentially are associated with PETs. The latter include the multiple endocrine neoplasia syndrome, type 1 (MEN1; Wermer syndrome), and hyperhormonal states relating to the production of various neuropeptides by pancreatic neuroendocrine lesions. The clinical characteristics of those conditions have been detailed thoroughly in other publications4,6,7 and will not be recounted here.

Pathologic Techniques for Characterization of Pancreatic Neuroendocrine Lesions Morphologic Features of Pancreatic Endocrine Neoplasms Conventional gross and light microscopic examination continues to be the mainstay for pathologic recognition of S28

Am J Clin Pathol 2001;115 (Suppl 1):S28-S45

© American Society of Clinical Pathologists


PETs. They typically are well-demarcated masses on macroscopic examination, with a propensity (but not an exclusive tendency) to develop in the body and tail of the pancreas4,6 ❚Image 1❚. Potential histologic growth patterns seen in these neoplasms have been well-described in the previous literature,4,6,8-10 and only a brief account of them will be given in this discussion. The majority of pancreatic endocrine neoplasms show histologic patterns that reflect a relatively high level of cellular differentiation. As such, they assume well-defined insular, trabecular, “ribboning,” festooning, and pseudoglandular configurations with no evidence of necrosis—and with or without formation of true intercellular rosettes—generally fitting into the category of “organoid” growth ❚Image 2❚. Nuclei are relatively uniform in size and shape, with dispersed chromatin, inconspicuous nucleoli, and scant mitotic activity. Although the cytoplasm usually is amphophilic, oxyphilia or clear-cell change may be seen in these lesions as well. Moreover, the production of stromal amyloid (in insulin-producing PETs), melanin, psammomatous calcification (primarily in somatostatin-producing tumors), or metaplastic osteoid is detected rarely.4,10 Tumors in this category are subdivided into those that are encompassed by the adventitia of the pancreas and those that demonstrate metastasis or evidence of locoregional invasion into blood vessels, nerves, and adjacent organs. A second, less common group of PETs differs from the description just given in that it shows a higher degree of nuclear atypicality and pleomorphism, with multifocal necrosis, brisk mitotic activity, or both. Organoid growth tends to be less well-defined in this cluster of neoplasms, and local invasiveness is observed more commonly.11

A

The third and least frequently encountered type of PET is represented by poorly differentiated neoplasms that are virtually identical to small cell carcinomas of the lung.12-17 They show only vaguely organoid patterns, with very high nuclear/cytoplasmic ratios, nuclear “molding,” numerous apoptotic cells, geographic necrosis, and a high mitotic rate ❚Image 3❚. Occasional tumors in this class may show mixtures of the aforementioned growth patterns. Hence, thorough sampling of large lesions (4 cm or larger) is advisable. Histochemical Examination of Pancreatic Endocrine Tumors Histochemical methods for the recognition of neuroendocrine differentiation in pancreatic tumors are still worthwhile. Silver stains are the most helpful in that regard, and the majority of PETs demonstrate cytoplasmic labeling with the Grimelius, Sevier-Munger, or Churukian-Schenk procedure.10 Unfortunately, these techniques are not specific, inasmuch as exocrine cytoplasmic granules and mucin also may be decorated by them. Thus, they serve as screens for the presence of neuroendocrine features but are not sufficient unto themselves diagnostically. Stains for epithelial mucin, including Best mucicarmine method and the periodic acid–Schiff technique after diastase digestion, are still valuable in selected instances for separating exocrine or mixed exocrine-endocrine pancreatic neoplasms (which are mucinnegative) from pure neuroendocrine tumors.4,10,18 Ultrastructural Characteristics of Pancreatic Endocrine Neoplasms In keeping with the premise that PETs are truly

B

❚Image 1❚ A, Pancreatic endocrine neoplasm, shown on computed tomography as a mass in the body of the pancreas. B, Gross examination revealed the tumor is well-demarcated from the surrounding pancreatic parenchyma.



S29

Wick and Graeme-Cook / PANCREATIC NEUROENDOCRINE NEOPLASMS

B

A

❚Image 2❚ “Organoid” microscopic images of pancreatic neuroendocrine tumors, showing a “ribboning” growth pattern (A; H&E, ×100), sheet-like growth with internal rosette formation (B; H&E, ×250), and an insular configuration (C; H&E, ×250).

C

neuroendocrine in nature, they are largely defined by their ultrastructural characteristics. Such lesions exhibit the regular presence of dense-core neurosecretory granules in the cytoplasm, ranging from 80 to 300 nm in diameter, and with a tendency for clustering near Golgi complexes, which may be prominent in their own right. Endoplasmic reticulum is visualized easily in PETs, but other metabolic organelles are relatively nondescript. The tumor cells are joined to one another by well-defined attachment plaques.4 Two subtypes of neurosecretory granules are worthy of further discussion because they relate well to the type of neuropeptide product synthesized by PETs. The first displays a crystalline core and an irregular submembranous halo, corresponding to insulin granules ❚Image 4A❚. The second shows dual density of its internal core structure, with only an indistinct halo and no crystalline configurations; this is the image of a glucagon granule ❚Image 4B❚. Otherwise, the S30


neuropeptide contents of cytoplasmic granules in PETs have no remarkable morphologic attributes and are indistinguishable from one another by electron microscopy.10,18-21 Aside from its usefulness for the delineation of insulinor glucagon-synthesizing lesions, another reason for the continued use of ultrastructural analysis in PETs is to detect neoplasms with divergent pancreatic acinar or ductal differentiation. The former of those tumor types shows the presence of both neurosecretory and zymogen-type cytoplasmic granules, whereas the latter manifests focal ductal luminal differentiation with formation of plasmalemmal microvilli.22,23 Immunohistologic Features of Pancreatic Neuroendocrine Lesions All PETs are, by definition, epithelial in nature and, therefore, are expected to manifest immunoreactivity for keratin proteins. In particular, keratins 8 and 18 are seen in © American Society of Clinical Pathologists


❚Image 3❚ Poorly differentiated (grade 3) neuroendocrine carcinoma of the pancreas showing high nuclear/cytoplasmic ratios, nuclear “molding,” and abundant apoptosis. The overall image is comparable to that of small cell neuroendocrine carcinoma of the lung (H&E, ×160).

greatest density in such lesions. Coexpression of neurofilament proteins is also observed in more than 75% of cases, and roughly 50% also exhibit labeling for vimentin.24 In regard to generic markers of a neuroendocrine lineage, chromogranin A and synaptophysin are similarly efficacious for helping to define PET as such4,5 ❚Image 5❚. In our experience, one or both of those proteins is present in

A

approximately 90% of pancreatic endocrine neoplasms. Adjunctive immunohistochemical indicators of neuroendocrine differentiation include protein gene product 9.5, neuron-specific (gamma-dimer) enolase, and CD57 (HNK-1 antigen), but those markers are not as specific as chromogranin A and synaptophysin.25 It is of interest and practical importance that insulin- and somatostatin-producing PETs tend to be nonreactive for chromogranin A.26 Instead, they often label for chromogranin B or chromogranin C,27 but antibodies to the latter determinants, unfortunately, are poorly available commercially. In this era of cost consciousness, it may well be asked whether it is necessary to immunostain PETs for specific neuropeptide products such as insulin, glucagon, gastrin, somatostatin, pancreatic polypeptide, alpha neuropeptide subunit, neurotensin, and others. Paradoxically, the answer to this question may be “no” when clinical endocrinopathy is present. These tumors may feature symptomatic hyperinsulinemia; hypergastrinemia with the Zollinger-Ellison syndrome; watery diarrhea with hypokalemia (Verner-Morrison syndrome) due to oversecretion of vasoactive intestinal polypeptide; or the glucagonoma syndrome, with glucose intolerance, migratory necrolytic erythema of the skin, depression, and wasting.4 In those circumstances, serum measurement of the peptides in question is accomplished easily, and the immunostains in question are no longer necessary for pathologic diagnosis or prognosis in most cases. It is only for the PETs that are physiologically “silent” that labeling for specific neuropeptides is a useful undertaking. This is because overproduction of such clinically transparent

B

❚Image 4❚ A, Electron micrograph of insulin-producing neuroendocrine tumor of the pancreas, demonstrating neurosecretory granules (NSGs) with crystalloid cores (top of Image) that are typical of that hormonal product. B, Another micrograph shows a glucagon-synthesizing neoplasm, in which NSGs with “double-density” cores are apparent. (Uranyl acetate and lead citrate, ×11,700)



S31

Wick and Graeme-Cook / PANCREATIC NEUROENDOCRINE NEOPLASMS

A

B

❚Image 5❚ Diffuse immunoreactivity for chromogranin A (A; ABC immunostain, ×100) and synaptophysin (B; ABC immunostain, ×250) in a pancreatic neuroendocrine tumor.

moieties as pancreatic polypeptide ❚Image 6❚ or somatostatin may be found, providing prognostic data and potential markers of tumor growth that can be assayed serologically thereafter. As discussed subsequently, there are only debatably significant behavioral differences among truly nonfunctional, unifunctional, and plurihormonal PETs in overall survival, when insulin-producing tumors are excluded from analysis.28 Other determinants of potential interest relate to the differential diagnostic alternatives to “pure” pancreatic endocrine neoplasms. Several lesions exhibit partial or complete exocrine differentiation and morphologically may

❚Image 6❚ Immunolabeling for pancreatic polypeptide in a pancreatic neuroendocrine neoplasm in a case lacking clinical evidence of an endocrinopathy (ABC immunostain, ×250).

S32


strongly resemble PETs. Those tumors are expected to express such markers as “DU-PAN-2,” carcinoembryonic antigen, lysozyme, alpha 1 -antitrypsin, CD10, CA19-9, tumor-associated glycoprotein-72, and mutant K-ras proteins, whereas pure neuroendocrine pancreatic neoplasms generally lack such cell products.29 Molecular Abnormalities in Pancreatic Endocrine Neoplasms The molecular basis for the development of PETs is not as well elucidated as that for other more common tumors. However, significant steps have been taken toward that end with molecular biological technologies that include restriction fragment length polymorphism analysis, fluorescent in situ hybridization, the polymerase chain reaction, comparative genomic hybridization, nucleic acid sequencing, singlestrand conformational polymorphisms, and direct cytogenetic evaluation in the hospital pathology laboratory.30-35 Thus far it seems that PETs evolve from a tumor suppressor pathway rather than a mutator pathway (ie, they appear not to show microsatellite instability). Studying patients with a genetic predisposition to develop the tumor is a well-traveled route in discovering pertinent genes. PETs are common in patients with MEN1. These patients exhibit a germline mutation in one copy of a gene located on the centromeric portion of the long arm of chromosome 11 (11q13), known as the mu gene. The gene product menin appears to have a tumor suppressor function.34,36,37 Loss of heterozygosity (LOH) for menin was observed by Debelenko et al 38 in 85% of non–gastrinproducing MEN1-associated PETs and in 41% of MEN1associated “gastrinomas.” © American Society of Clinical Pathologists


Although this gene locus is frequently abnormal in MEN1-related PETs of all subtypes, sporadic (nonfamilial) endocrine neoplasms of the pancreas less frequently show mutations in, or LOH for, menin.37-44 Sporadic insulinproducing PETs uncommonly demonstrate LOH or mutations in the MEN1 gene,44 whereas sporadic gastrin-producing lesions have somatic menin abnormalities in 44% of cases. There is a suggestion in these data that functional and nonfunctional tumors may show a different molecular pathogenesis. This hypothesis is supported by a study using highresolution allelotyping by microsatellite DNA analysis. Rigaud et al45 assessed the fractional allelic loss (FAL) in PETs of various types. Overall, two groups emerged from their analysis: one had a high level of FAL and tended to be DNAaneuploid; the other infrequently showed FAL and tended to be DNA-diploid. LOH at 6q was associated with hormonally silent PET; LOH at 11q was seen in functional PET. Beghelli and colleagues46 have shown that another locus on the short arm of chromosome 17 (17p13) exhibits deletions in approximately 25% of sporadic endocrine neoplasms; the study also failed to find evidence of microsatellite instability. Chung et al47 showed that 33% of the PETs they studied exhibited allelic loss on chromosome 3p (3p25), a possible novel tumor suppressor gene. Mutations of K-ras codon 12, the p53 gene, and the CDKN2 gene are well recognized in over 50% of exocrine pancreatic carcinomas but are generally rare (