Gastrointestinal tract (GI), pancreas, and lung are among the most common sites to develop well-differentiated neuroendocrine tumors (NET). The incidence of GI NET is increasing, based on recent Surveillance, Epidemiology and End Results (SEER) data [1]. Patients with NET may initially present as liver metastasis in up to 19% of cases [2, 3], and the clinical management and outcome depend on multiple factors, including the primary tumor site [1, 4]. Although primary NET often shows a predominate architectural growth pattern at a particular site, tumor heterogeneity with mixed growth patterns do occur [5]. Using morphological features to diagnose the NET with an unknown primary can be challenging based on the hematoxylin and eosin staining alone, and a panel of immunohistochemistry (IHC) is often required to determine the site of primary tumor. General neuroendocrine markers including chromogranin and synaptophysin are very useful markers to determine the neuroendocrine differentiation. However, they lack specificity for the lineage or site of the tumor. Caudal type homeobox 2 (CDX-2) has been investigated previously by multiple groups and showed high sensitivity and specificity for small intestinal NET, but variable sensitivity for colorectal and pancreatic NETs. [6–9] PAX-8 and Islet-1 (ISL-1) have been investigated for their role to identify primary and metastatic pancreatic neuroendocrine tumors and variable sensitivity and specificity has been reported [10–13]. An extended panel of biomarkers including CDX-2, pancreatic and duodenal homeobox 1 (PDX-1), TTF-1, and NESP-55 has been reported in NETs from the pancreas, GI and lung, and these results also showed relatively low sensitivity for the tumors from the rectum and pancreas [14]. More recently, the expression pattern of CDX-2, PAX-6, ISL-1, ER, and PR in the GI and pancreatic NET was investigated and showed small differences between the primary and metastatic NETs [15, 16]. However, these markers showed relatively low sensitivity for pancreatic NETs. Investigation of more biomarkers to distinguish common NETs of different site remains necessary.

For this purpose, we focused on a panel of transcription factors (TFs) including CDX-2, NK2 homeobox 2 (NKX2.2), PDX-1, and pancreas specific transcription factor 1a (PTF1A) [17, 18], and simultaneously investigated their protein expression pattern in the NETs of the most common sites, including GI, pancreas and lung. PDX-1 is a transcription factor essential for both pancreatic ductal and islet cell development during early embryogenesis [18]. After development, PDX-1 is predominantly expressed in the islet cells of pancreas, but not in the ductal epithelium or acinar cells [19]. Interestingly, NKX2.2 is critical for the differentiation of beta cells in the islet cells of the pancreas [20–26], and important for the differentiation of enteroendocrine cells of the intestine [27, 28], as well as oligodendrocytes of the central nervous system [29, 30]. During early pancreatic development, NKX2.2 protein expression is initiated with PDX-1 and PTF1A in the dorsal and ventral pancreatic buds [31]. In the intestine, NKX2.2 is expressed in the endocrine cells located in the villi and deep crypts, and a rare subset of NKX2.2-positive cells is coexpressed with intestinal stem cell markers [32, 33]. We compared the results with the well-known transcriptional factor of intestinal differentiation, CDX-2.

In this study, four transcriptional factors including CDX-2, NKX2.2, PDX-1, and PTF1A were simultaneously investigated for their protein expression in the most common NETs from the GI, pancreatic, and lung by IHC. We found that these transcription factors were differentially expressed in NETs at these anatomic sites. NKX2.2 was a highly sensitive biomarker for neuroendocrine differentiation of pancreatic and intestinal neuroendocrine origin. PDX-1 was expressed in NETs of small intestine and appendix with variable but higher immunoreactivity at these sites, but not in the gastric, colonic or rectal NETs. NKX2.2 was positive in over 2/3 pancreatic NETs. Although PDX-1 immunopositivity was seen in less than 1/3 pancreatic NETs in previous reports, it was positive in 6% of pancreatic NETs in this study probably due to different source of antibodies, since it was positive in all internal controls. Although PTF1A was characterized as one of the pancreatic lineage-specific transcription factors (TFs), its expression was non-specific and was not a useful biomarker for NETs. Consistent with other’s results, CDX-2 can be a useful biomarker for NETs arising from the small intestine and appendix. However, it showed only 50% immunoreactivity in colonic NETs and no immunoreactivity in all rectal NETs. CDX-2 was patchy positive in 11% of pancreatic NETs similarly seen in a previous study [14]. Overall, NKX2.2 showed much higher sensitivity for both rectal and pancreatic NETs, compared to PDX-1 and CDX-2 for the tumors at these sites. Most importantly, these biomarkers were completely negative in all lung NETs, pancreatic ductal adenocarcinomas, or other carcinomas (data not shown).

Previous studies showed that NKX2.2 was negative in 13 bronchopulmonary carcinoid tumors, whereas 24 of 26 GI carcinoids (including 1 stomach, 1 duodenum, 2 ampullary, 14/16 pancreas, 5 ileum, and 1 colon), were immunoreactive for NKX2.2 [34]. In this study, we argue that a panel of PDX-1, CDX-2, and NKX2.2 would significantly increase the sensitivity of NETs from the GI tract and pancreas. Although TTF-1 was not tested in this cohort, it has been studied to determine NETs of pulmonary origin by several groups [8, 35–38]. Considering the fact that the majority of neuroendocrine tumors arise from the tubular GI tract, pancreas, and lung, we proposed a diagnostic algorithm using a panel of IHC stains including NKX2.2, PDX-1, CDX-2, and TTF-1 would aid the diagnosis of the primary tumor site in the setting of metastasis (Fig. 4). Although ISL-1, PAX-6, and PAX-8, were not investigated in the primary NETs in our study, they can be added into the panel when the results are indeterminate for the pancreatic tumor or to increase the confidence of prediction [12], although the specificity of PAX-8 and ISL-1 was variable in different studies [39, 40]. Interestingly, a recent study showed high sensitivity of NKX2.2 in 85 liver metastatic NETs of the digestive system [41].

Using IHC, we found differential immunoreactivity of NKX2.2, CDX-2, and PDX-1 to the NETs from a different anatomic site. NKX2.2 showed higher sensitivity than CDX-2 and PDX-1 in both GI and pancreatic NETs. PDX-1 was expressed mainly in the small intestinal and appendiceal NETs, occasionally in the pancreatic NETs, and not in the colorectal NETs. Importantly, NKX2.2, CDX-2, and PDX-1 were not expressed in 11 lung NETs, and we understand the limitation of small sample size for a final conclusion on the specificity of these markers. These findings suggest that a panel of biomarkers including NKX2.2, PDX-1, and CDX-2 may provide additional insights in clinical diagnosis for the most common NETs especially in the setting of metastasis with an unknown origin.