Presented in part as a platform presentation at the 52nd Annual Scientific Meeting of the American Society of Cytopathology, Chicago, IL, November 12–17, 2004.
The cytologic differentiation between neoplastic and reactive/reparative processes in the endoscopic ultrasound-guided fine-needle aspirations (EUS-FNA) of the pancreas can be difficult. Malignant transformation of the pancreatic ductal epithelium changes the expression of apomucins. The goal of the current study was to determine an optimal immunohistochemical panel of mucin (MUC) antibodies that would allow the cytomorphologic distinction of pancreatic ductal adenocarcinoma and its differentiation from reactive/reparative processes and inadvertently sampled gastric and duodenal mucosa.
Pancreatic EUS-FNA specimens performed on 351 patients were reviewed. Expression profiles of MUC1, 2, 5AC, and 6 were examined on 56 cell block sections and 26 follow-up pancreatectomy specimens.
MUC1 and 6 expression was found in nonneoplastic pancreatic samples, whereas there was an absence of expression of MUC2 and 5AC. MUC2 was detected in mucosal goblets cells of the duodenum, MUC6 in Brunner glands, and MUC5AC in gastric foveolar cells. MUC5AC expression in differentiating ductal adenocarcinomas from benign conditions demonstrated better operating characteristics than either MUC1 or MUC6. The apomucin expression pattern both in cytology and follow-up surgical pathology specimens was similar. In surgical pathology specimens, the panel of 3 antibodies, MUC1+/MUC2−/MUC5AC+, was noted in 15 of 17 ductal carcinomas (88.2%). In nonneoplastic pancreatic tissue, the expression panel MUC1+/MUC2−/MUC5AC− was observed in 14 of 17 (82.4%) cases. In cytology specimens, the combination of MUC1+/MUC2−/MUC5AC+ was noted in 21 of 30 ductal carcinoma cases (70.0%), 3 of 6 atypical cases (50%), and 1 of 1 suspicious for malignancy cases (100%). The combination MUC1+/MUC2−/MUC5AC+ was not observed in any of the negative for malignancy or reactive cases (0 of 6).
Endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA) of the pancreas has proven to be a safe and effective procedure that in most cases can provide diagnostic material. Nevertheless, the differential diagnosis of neoplastic and reactive/reparative processes in EUS-FNA specimens of the pancreas can be extremely difficult due to overlapping cytomorphologic features and inadvertent sampling of the duodenal and gastric mucosa during the ultrasound-guided procedure. 1–6
Ductal adenocarcinoma is the most common malignancy of the pancreas. 7 Recently, it has been shown that the malignant transformation of pancreatic ductal epithelium changes the epithelial expression of apomucins.8–15 Apomucins are high molecular weight glycosylated proteins that are synthesized by the mucosal lining cells of most epithelia as membrane-bound or secretory products. To our knowledge to date, at least 14 distinct epithelial mucin genes have been identified. These show an organ-, tissue-, and cell-specific type expression and play an important role in mucosal surface protection and lubrication.16, 17 Aberrant expression as well as changes in the level and distribution pattern of mucin (MUC) proteins have been implicated in the pathogenesis of many nonneoplastic and neoplastic conditions, including the metaplasia-dysplasia-carcinoma sequence in the secretory epithelia of the gastrointestinal tract. In tissue specimens of pancreatic ductal adenocarcinomas, overexpression of MUC1 (membrane-bound pan-epithelial mucin) and MUC6 (gastric pyloric gland-type secretory mucin) and de novo expression of MUC5AC (gastric surface secretory-type mucin) as indications of gastric foveolar and pyloric gland-like metaplasias have been observed as early events in pancreatic carcinogenesis in all stages of pancreatic intraepithelial neoplasias (PanIN) and invasive ductal adenocarcinomas, whereas goblet cell metaplasia with associated MUC2 (intestinal-type secretory mucin) expression was an extremely rare event in most of the studies.8–13 Investigations regarding the mucin expression in pancreatic FNAs are rare.18, 19 In this study we examined the epithelial expression profiles of MUC1, 2, 5AC, and 6 on the cell block sections of EUS-FNA and follow-up surgical pathology specimens of different pancreatic lesions. The goal of the current study was to determine an optimal immunohistochemical panel of MUC antibodies for EUS-FNA specimens that would allow the cytomorphologic distinction of pancreatic ductal adenocarcinoma as well as its differentiation from reparative/regenerative changes in inadvertently sampled gastric and duodenal mucosa.
MATERIALS AND METHODS
The study was approved by the Institutional Review Board of the University of Pennsylvania. We retrospectively reviewed all EUS-FNA of the pancreas performed at the University of Pennsylvania Medical Center between January 1998 and May 2004. Our study included 351 patients with EUS-FNA, 198 of whom were males and 153 were females with an average age of 65.2 years (range, 12-95 years). The surgical pathology follow-up was available in 26 patients. Other patients had radiologic or clinical follow-up. A gastroenterologist performed the EUS-FNAs in all cases. The number of passes varied from 1 to 6, with an average of 3 passes. On-site evaluation of all EUS-FNA was provided by a cytopathologist and a preliminary diagnostic impression was rendered in each case. Both air-dried Diff-Quik®-stained (Baxter Diagnostics, McGaw Park, IL) smears for onsite evaluation and 95% alcohol-fixed smears for subsequent Papanicolaou staining were prepared. The residual specimen was rinsed in Normosol® solution for Millipore® (Bedford, MA) filter and cell block preparation. Average time spent for the FNA procedure in our center was 30 minutes. The final cytologic diagnosis was rendered by a consensus of cytopathologists at the departmental Cytology Conference. Fifty-six archival paraffin-embedded cells blocks of pancreatic EUS-FNAs with adequate cellular material and 47 tissue blocks of 26 pancreatectomy surgical pathology follow-up specimens were chosen for this study. Representative surgical pathology tissue blocks were selected such that the immunohistochemical reactivity of mucin antibodies could be assessed both in the neoplastic pancreatic tissue and in the nonneoplastic pancreas and duodenum as internal controls.
Immunohistochemistry was performed on 4-μm sections using mouse monoclonal antibodies to MUC1 (clone Ma695, dilution 1:75; Novocastra Laboratories, Newcastle-upon-Tyne, UK), MUC2 (clone Ccp58, dilution 1:50; Novocastra Laboratories), MUC5AC (clone CLH2, dilution 1:100; Novocastra Laboratories), and MUC6 (clone CLH5, dilution 1:50; Novocastra Laboratories). All MUC antibodies were tested on paraffin sections of positive control tissues with and without heat-induced antigen retrieval; recommended by the manufacturer (Novocastra Laboratories). Although all MUC antibodies worked well without antigen retrieval, the heat-induced antigen retrieval method was shown to enhance the detection sensitivity without any adverse effect or undesirable reactivity.
Before immunohistochemical staining, tissue sections were pretreated in a pressure cooker for 2 minutes (Biocare Decloaking Chamber, Biocare Medical, Walnut Creek, CA) in 1× citrate buffer, pH 6.0 (Lab Vision, Fremont, CA). After slides were allowed to cool, immunostaining was performed with Envision + HRP Mouse (Dakocytomation, Carpinteria, CA) in a Dako Autostainer. Primary antibody incubation was performed at room temperature for 1 hour. Appropriate positive and negative controls were used.
In cytology specimens the immunoreactivity was considered positive if any cells stained regardless of the intensity and the proportion. In the surgical pathology specimens a tumor was considered positive if >5% of tumor cells stained regardless of the staining intensity level.
The statistical analysis was performed using Student t test. P<.001 was considered statistically significant.
In the current study we included 56 cytology cases and 26 available surgical pathology follow-up cases. To establish normal and aberrant expression of the apomucins, we first examined 47 tissue blocks from surgical pathology follow-up specimens. Tissue blocks included representative sections of nonneoplastic gastric, duodenal, and pancreatic tissue and neoplastic pancreatic tissue. Among the neoplastic pancreatic tissue sections were ductal adenocarcinoma, mucinous cystic neoplasms and carcinoma, intraductal papillary mucinous neoplasm, and serous microcystic adenoma. The results of apomucin expression in the tissue sections of pancreatic ductal adenocarcinoma, nonneoplastic pancreas, and duodenum are shown in Table 1.
Ducts and acini of nonneoplastic pancreas showed apical membranous staining in 14 of 14 cases (100%), whereas MUC1 expression was absent in normal duodenum. Ductal adenocarcinoma (CA) in neoplastic pancreas showed strong and diffuse membranous and cytoplasmic staining (Fig. 1).
MUC2 was expressed in the goblet cells of normal duodenum and was absent in nonneoplastic pancreatic tissue (Fig. 1). Only 1 of 17 (5.9%) cases of ductal adenocarcinoma demonstrated MUC2 positivity, whereas the remainder were negative.
MUC5AC was expressed in gastric foveolar cells and was absent in duodenal tissue. Only 1 of 17 examined cases of nonneoplastic pancreatic tissue (5.9%) demonstated focal cytoplasmic expression in ducts with mucinous metaplasia (PanIN/L-1A) in a case of chronic pancreatitis. Fifteen of 17 cases of ductal adenocarcinoma (88.2%) demonstrated strong cytoplasmic positivity for MUC5AC. The proportion of tumor cells expressing MUC5AC was smaller than that of MUC1-positive tumor cells, but in most cases it was strong when present (Fig. 1). Interestingly, strong MUC5AC expression was also found in areas with pancreatic intraepithelial neoplasia (PanIN) and in atypical ducts surrounded by nonneoplastic pancreatic parenchyma (Fig. 2, arrow).
Expression of MUC6 with an apical membranous staining pattern was found in 14 of 14 cases of nonneoplastic pancreas in both ducts and acini (100%), as well as a cytoplasmic staining pattern in ducts with mucinous metaplasia in 1 case of chronic pancreatitis. The Brunner glands of normal duodenum also showed positive staining with MUC6. Six of 17 cases of ductal adenocarcinoma (35.3) demonstrated cytoplasmic staining with intensity similar to that of MUC 5AC (Fig. 1).
Additional surgical pathology cases examined included 1 case of mucinous (colloid) carcinoma that was found to only express MUC2 and 1 case of serous microcystic adenoma that was positive only for MUC1. In 2 cases of mucinous cystic neoplasms (MCNs), both cases demonstrated expression of MUC5AC and MUC6 and lack of expression for MUC1. One case of MCN was positive for MUC2. Also examined was a case of intraductal papillary mucinous neoplasm (IPMN), where apomucins MUC2, MUC5AC, and MUC6 were expressed but MUC1 was found to be absent.
In surgical pathology specimens, apomucin expression MUC1+/MUC2/MUC5AC+ was observed in 15 of 17 cases of ductal carcinoma (88.2%), whereas in nonneoplastic pancreatic tissue the expression of the above panel was MUC1+/MUC2−/MUC5AC− in 14 of 17 cases (82.4%), which included 3 cases of chronic pancreatitis. Also, in contrast to diffuse cytoplasmic MUC1 expression in malignant cases, MUC1 expression in nonneoplastic pancreas was apical or membranous.
The results of apomucin expression in cytology specimens with the diagnosis ductal carcinoma, atypical, suspicious for malignancy, reactive, and negative for malignancy cases are shown in Table 2.
Table 2. Mucin Expression Profile in EUS-FNA Specimens
No. of Positive Cases (%)
EUS-FNA: endoscopic ultrasound guided-fine-needle aspiration; CA: invasive ductal adenocarcinoma; NM: negative for malignancy, SUSP: suspicious for malignancy.
Nonneoplastic pancreatic parenchyma demonstrated MUC1 positivity in ductal and acinar structures in a pattern similar to that of the surgical pathology specimens. MUC1 expression was noted in 25 of 30 cases of ductal adenocarcinoma (83.3%). Both in cytology and surgical pathology specimens, 2 main patterns of mucin expression were observed. One with MUC1 positivity, along with both gastric type apomucins expressed (Fig. 3), and the other when in the tumor cells there was MUC1 positivity and aberrant expression of only MUC5AC (Fig. 4).
Strong MUC2 expression was found in only the goblet cells in strips of duodenal mucosa. (Fig. 5) and could be easily distinguished from tumor cells that were negative for MUC2 and showed strong positivity for MUC5AC (Fig. 5, arrow). Seven of 30 cases of ductal adenocarcinoma (23.3%) demonstrated positivity for MUC2, but again these tissue fragments were easily distinguished from the goblet cell staining pattern evident in normal duodenal mucosa.
MUC5AC expression was present in 3 of 5 cases in benign gastric foveolar cells (60%) and MUC5AC expression was seen in 29 of 30 cases of ductal adenocarcinoma (96.7%). Figures 6, 7, and 8 show different areas of 1 case of ductal carcinoma that on the cell block had abundant components of gastric foveolar-like cells (Fig. 6). These benign-looking cells with basally placed nuclei and abundant apical mucin stained positively only for MUC5AC. In the same specimen, the tumor cells (Figs. 7 and 8, red arrows) demonstrated an MUC1+/MUC2−/MUC5AC+/MUC6+ expression pattern, suggesting that the benign-appearing MUC1−/MUC2−/MUC5AC+/MUC6− cells shown in Figure 6, being completely negative for MUC1, are most likely from the adjacent PanIN/1A rather than gastric foveolar cells. The duodenal mucosa in this specimen could be distinguished from other epithelial fragments by scattered positivity to MUC2 in the goblet cells and negative staining with other mucin markers (Fig. 8, black arrows).
Nonneoplastic pancreatic parenchyma of demonstrated MUC6 positivity in ductal and acinar structures also showed a pattern similar to that seen in the surgical pathology specimens. Eleven of 30 cases of ductal adenocarcinoma (36.7%) demonstrated aberrant MUC6 expression.
Also examined were 1 case with a cytologic diagnosis of suspicious for malignancy and 6 cases with cytologic diagnoses of atypical. One atypical EUS-FNA specimen that was proven on surgical resection to be ductal adenocarcinoma is shown in Figure 9 (arrow). This case showed positive staining of the atypical glands with MUC1 in both apical and cytoplasmic staining patterns and positive cytoplasmic staining with MUC5AC. MUC2 and MUC6 were both negative. A similar pattern of expression is shown in the pancreatectomy specimen (Fig. 9).
The results of immunohistochemical staining of EUS-FNA specimens of other pancreatic lesions are additionally detailed in Table 3. These includes 2 cases with a cytologic diagnosis of mucinous neoplasm (MN) and 2 cases with equivocal cytologic diagnosis (CA vs. MN). Both EUS-FNA specimens show identical expression of the four examined apomucins. In contrast to scattered MUC2 positivity in the duodenal goblet cells, the tumor cells in mucinous neoplasms showed strong and diffuse staining for both MUC2 and MUC5AC. Figure 10 shows the apomucin expression pattern in the case with cytologic diagnosis of mucinous neoplasm and corresponding surgical pathology follow-up with a final diagnoses of IPMN. Acinar cell carcinoma showed positive immunostaining only for MUC6. Of 2 examined cell block samples of metastatic tumors, the only positive staining noted was for MUC1, in a case of renal cell carcinoma metastasis. In cases of neuroendocrine tumors (NET), there was positive staining for MUC1 in 1 of 4 cases with absent expression of the other examined apomucins (25%). Interestingly, the surgical pathology follow-up in 1 case with a cytologic diagnosis of CA was a neuroendocrine tumor (glucagonoma) that demonstrated strong expression of MUC1. No apomucin expression was found in 1 low-grade B-cell lymphoma sample.
Table 3. Mucin Expression Profiles in Other EUS-FNA Specimens
In cytology specimens, the combination of MUC1+/MUC2−/MUC5AC+ was seen in 21 of 30 ductal carcinoma cases (70.0%), 3 of 6 atypical cases (50%), and 1 of 1 cases that was suspicious for malignancy (100%). The combination MUC1+/MUC2−/MUC5AC+ was not noted in any of the negative for malignancy or reactive cases (0 of 6). Seven cases with a cytologic diagnosis of ductal carcinoma and 2 cases with a cytologic diagnosis of atypical cells with MUC1+/MUC2−/MUC5AC+ expression had surgical pathology follow-up with a diagnosis of pancreatic ductal carcinoma. Other cases had clinical and/or radiologic follow-up with a diagnosis of pancreatic carcinoma. Two cases with a cytologic diagnosis of atypical cells with MUC1−/MUC2−/MUC5AC−/MUC6− expression on follow-up showed surgical pathology specimens of lymphoma and pancreatitis. None of the cytology cases that had follow-up diagnosis other that ductal carcinoma demonstrated the expression pattern MUC1+/MUC2−/MUC5AC+. Our data demonstrate that the mucin immunoprofile shows a specific pattern of expression in pancreatic carcinoma cases, and can be especially useful in cases with atypical/suspicious for malignancy cytologic diagnoses.
The sensitivity, specificity, and predictive power for each antibody used in this study for both FNA and surgical pathology specimens are depicted in Figures 11 and 12. Statistical analysis showed that MUC1 and MUC5AC expression in pancreatic adenocarcinoma is significantly higher than in nonneoplastic pancreatic tissue both in cytology and surgical pathology specimens (P<.0001).
EUS-FNA of the pancreas is considered the first-line diagnostic modality to define pancreatic lesions. It has proven to be a valuable tool in the preoperative diagnosis and management of patients with pancreatic lesions. 1–6 Whereas most aspirations yield diagnostic material, the differential diagnosis of the reactive/reparative and neoplastic processes as well as their precise classification in the pancreatic EUS-FNAs specimens can be challenging. This can be explained by the considerable overlap in the cytomorphologic features of these processes in the pancreas.1–3, 19 In addition, ductal metaplasias associated with many nonneoplastic and neoplastic diseases in the pancreas create even more difficulties in interpreting the cytology specimens. Therefore, ancillary techniques, such as molecular studies and immunohistochemistry, could be useful in increasing accuracy in the evaluation of the EUS-FNA specimens.
The most common pancreatic neoplasm associated with the worst prognosis is pancreatic ductal adenocarcinoma. 7 As in other sites of the gastrointestinal tract, the metaplasia-dysplasia-carcinoma sequence appears to play a major role in pancreatic cancinogenesis. Recently, it was shown that altered synthesis and patterns of expression of mucin glycoproteins with associated metaplasias in the pancreatic ductal epithelium are related to the histogenesis and biological behavior of the pancreatic neoplasms.7–15
Ductal metaplasias that occur in the pancreas in different pathological conditions include mucinous, squamous, oncocytic, goblet cell, and acinar-cell types. 7 The most common types of mucinous metaplasia are gastric foveolar-like and pyloric gland-like metaplasias. Mucinous metaplasias can occur in the nonneoplastic conditions, such as chronic pancreatitis, but also have been shown to be an early event in the development of pancreatic ductal adenocarcinoma, as evidenced by de novo expression of MUC5AC and overexpression of MUC1 and MUC6 in both PanIN and in invasive tumors.8–13 In contrast, goblet cell metaplasia with associated MUC2 expression is extremely uncommon in pancreatic ductal adenocarcinomas.8–11 Intraductal papillary mucinous neoplasms (villous dark cell type) is the most common histologic type and is associated with both MUC5AC and MUC2 expressions, whereas 2 other histologic types of IPMN (papillary clear cell and compact cell type) have a mucin pattern expression similar to ductal adenocarcinoma.20–23 The different patterns in apomucin expression of the various pancreatic lesions reflect the histogenic abilities of the stem cells of dorsal and ventral pancreatic anlagen with corresponding lineage-associated metaplastic tendencies toward the gastric- or duodenal-type mucosa. This difference can be helpful in understanding the histogenesis of the pancreatic tumors7–15, 20–25 and providing an additional tool for subclassification and the potential for early diagnosis of pancreatic tumor.
To our knowledge, there are very few reports in the literature regarding the utility of ancillary techniques in fine-needle aspirates of the pancreas, and the majority of these are molecular studies on detection of K-ras mutation, which is the most common genetic alteration in pancreatic ductal adenocarcinomas. 26, 27 Although these molecular techniques definitely have a potential role in the improvement of the sensitivity and specificity of the pancreatic FNAs, they are not available to many laboratories as routine tests. Search for immunohistochemical markers that might be helpful in the cytologic differentiation of the pancreatic lesions in FNA specimens prompted us to choose as a potentially useful tool a panel of mucin markers comprising 4 apomucins-MUC1, 2, 5AC, and 6. These markers show very distinct patterns of expression in the tissue that is used as a control and also, in many cases, show quite distinct patterns of immunoreactivity in different pancreatic lesions.
Reports in the literature regarding the utility of immunohistochemistry are rare, with a recent report by Chhieng et al. 19 using a panel of only 2 apomucin markers. In our opinion, a panel of only 2 apomucin markers is not optimal, for several reasons. First, we found that MUC1 can show weak apical reactivity not only in the acinar cells, as Chhieng et al.19 noticed, but also in the normal ductal cells and in the ductal epithelium in cases of chronic pancreatitis. Similar patterns of MUC1 expression were found by other authors.9, 10 Second, we have observed MUC 1 staining in a sample of neuroendocrine tumor of pancreas (glucagonoma). This can be explained by the fact that both exocrine and endocrine pancreatic cells originate from ductular epithelium. Third, we found an MUC1+/MUC2− expression pattern in the case of metastatic renal cell carcinoma. So, in contrast to MUC1, apomucin MUC5AC is not normally expressed in the pancreatic tissue, and is aberrantly expressed in pancreatic ductal carcinoma.8–11, 13 In our study, MUC5AC showed immunoreactivity in 96.7% of cases of FNA cytology specimens and 88.2% of cases of surgical pathology follow-up specimens. The difference in MUC5AC expression between benign and malignant lesions was statistically significant. Our observations demonstrated that panel MUC1+/MUC2−/MUC5AC+ is highly specific for the diagnosis of ductal adenocarcinomas both in cytology and surgical pathology specimens. A MUC1+/MUC2−/MUC5AC+ pattern of expression was seen in 15 or 17 of ductal carcinomas in surgical pathology specimens (88.2%) and in 21 of 30 cases in cytology specimens (70.0%). In addition, many pancreatic ductal carcinomas showed a distribution and intensity of MUC6 immunoreactivity similar to that seen with MUC5AC immunostaining.
The results of the current study have shown that a panel of 4 mucin antibodies is useful in identifying gastric and duodenal mucosa incidentally sampled during the EUA-FNA procedure.
In EUS-FNA specimens, benign-appearing cells with basally placed nuclei and abundant apical mucin raise the differential of gastric foveolar cells, mucinous metaplasia in chronic pancreatitis, or nondysplastic mucinous lesions. Because normal and/or reactive gastric mucosa expresses MUC1 and MUC5AC only in gastric foveolar cells, and MUC6 only in pyloric glands, the observed pattern of immunoreactivity of these mucin markers was helpful in distinguishing metaplastic and/or neoplastic pancreatic epithelium from gastric mucosa. Similarly, because most pancreatic ductal carcinomas do not show reactivity for MUC2 immunostain, positivity of scattered goblet cells for MUC2 helped in identifying duodenal mucosa sampled during the FNA procedure. The current study has also shown that the panel of these 4 apomucin markers could be potentially useful in the precise classification of pancreatic tumors in cytology specimens (work in progress).
The current study data allowed us to conclude that an immunohistochemical panel for FNA specimens comprising MUC1+/MUC2−/MUC5AC+ is the most optimal for the diagnosis of pancreatic ductal carcinoma in EUS-FNA specimens. A mucin panel comprising all 4 mucin antibodies (MUC1, MUC2, MUC5AC, and MUC6) also may be helpful in differentiating normal/reactive duodenal and gastric epithelium from neoplastic pancreatic tissue. In addition, this panel could be potentially useful in the subclassification of pancreatic neoplasms, as well as their distinction from metastases in cytology specimens.