Dog1 Expression in Neuroendocrine Neoplasms: Potential Applications and Diagnostic Pitfalls

Neuroendocrine neoplasms represent a heterogeneous group of rare tumors, more frequently arising from gastroenteropancreatic tract and lungs. At the time of diagnosis, 20% of cases are metastatic, and 10% of cases are considered as cancer of unknown primary origin. Several immunohistochemical markers are routinely used to confirm the neuroendocrine differentiation, first among all Synaptophysin and Chromogranin-A; on the other hand, different immunohistochemical markers are used to establish primary anatomical site, as TTF1, CDX2, Islet-1 and Calcitonin, but no marker is available in order to distinguish among different sites of the digestive tract. DOG1 (discovered on GIST-1) is a gene normally expressed in interstitial cells of Cajal and, in routine practice, DOG1 immunostaining is used in diagnosis of GIST (gastrointestinal stromal tumor). DOG1 expression has been described in several neoplasms other than GIST, both in mesenchymal and epithelial neoplasms. In the present study, DOG1 immunostaining has been performed in a large cohort of neuroendocrine neoplasms, including neuroendocrine tumors and neuroendocrine carcinomas, in order to evaluate frequency, intensity and pattern of expression in different anatomical site and in different tumor grade. DOG1 expression was detected in a large percentage of neuroendocrine tumors, with statistically significant association between DOG1 expression and gastrointestinal tract neuroendocrine tumors. As a consequence, DOG1 could be included in marker panel for the identification of primary site in neuroendocrine metastases of unknown primary origin; moreover, these results recommend careful evaluation of DOG1 expression in gastrointestinal neoplasms, in particular in differential diagnosis between epithelioid GIST and neuroendocrine tumors.


Introduction
Neuroendocrine neoplasms (NENs) are a heterogeneous group of rare tumors, accounting for 2% of all malignancies, with an increasing incidence of 2-5 cases per 100000 per year, with common primary sites represented by digestive tract, pancreas and lung [1][2][3][4]. Metastases are found at the time of diagnosis in gastrointestinal stromal tumor (GIST) [13,14], with high sensitivity (ranging between 87% and 94,4% depending on the antibody clone), and high speci city [15][16][17]. In literature, several studies evaluated DOG1 expression in both mesenchimal and epithelial neoplasms other than GIST; indeed, among mesenchimal neoplasms, DOG1 reactivity has been found in low-grade bromyxoid sarcoma, gastrointestinal leiomyoma, myxo brosarcoma, synovial sarcoma, myxoid liposarcoma, undifferentiated pleomorphic sarcoma and in a prostatic stromal sarcoma [18][19][20][21][22]; as concerns epithelial neoplasms, DOG1 expression has been proved in cases of salivary gland, colorectal and endometrial adenocarcinomas, in pancreatic adenocarcinoma and in pancreatic solid pseudopapillary neoplasm, in testicular tumors and in some histotypes of renal cell carcinoma, in poorly differentiated head and neck carcinoma and in a subset of adnexal tumors, with particular correlation with eccrine differentiation neoplasm[16, [23][24][25][26][27][28][29][30]. Furthermore, DOG1 has also been evaluated as a prognostic marker in breast cancer, with con icting results [31,32]. Furthermore, DOG1 expression has also been studied as a neuroendocrine marker in the spectrum of non-neoplastic neuroendocrine cells and in neuroendocrine neoplasms, with a few cases of neuroendocrine carcinomas which stained positive for DOG1 antibody [32][33][34][35]; however, in literature, no studies with focus on DOG1 expression in gastrointestinal NENs are currently available.
The aim of this study was to perform DOG1 immunostaining on a large cohort of neuroendocrine neoplasms in order to provide a systematic evaluation of frequency, intensity and expression pattern in different primary sites and different grades neuroendocrine neoplasms.

Samples
A total of 197 consecutive cases of neuroendocrine neoplasms (NENs) surgically resected or biopsied from 2015 to 2021 were retrieved from Pathology Department in Niguarda Hospital, Milan, Italy. All cases were reviewed in order to con rm the diagnosis; tumor grading was de ned according to the latest WHO guidelines available at the time of the study. We Immunohistochemical analyses and DOG1 assessment A representative formalin-xed para n-embedded (FFPE) tissue block for each case was obtained from our archives, and immunohistochemical analyses were performed on 3 µm thick sections on polarized glass slides using an automated system (Dako Omnis; Agilent) and the following antibodies: cytokeratin cocktail (clone AE1/AE3, ready to use dilution; Dako/Agilent), Synaptophysin (ready to use dilution; Dako/Agilent), Chromogranin-A (dilution used 1:1000; Dako/Agilent), Ki67 (clone MIB-1, ready to use dilution; Dako/Agilent) and DOG1 (clone SP31; dilution used 1:50; GENNOVA). Immunohistochemical analyses were evaluated by two surgical pathologists and a resident fellow.
For each DOG1 immunostaining, percentage and distribution of positive neoplastic cells (diffuse or focal) and staining pattern (membranous and/or cytoplasmatic) were de ned; next, staining intensity was scored with a semi-quantitative system, as faint (score 1+), moderate (score 2+) or intense (score 3+). Moderate or intense staining in more than 5% of tumor cells was considered positive.

Results
In this study, DOG1 immunohistochemical staining was evaluated in a total of 197 neuroendocrine neoplasms with different primary site and differentiation grade, and neuroendocrine tumors of digestive tract resulted to be more frequently positive than NETs of other primary sites, with statistically signi cance. Otherwise, DOG1 in NECs was less commonly expressed, and no signi cant difference was found between different anatomical sites. These results differ from those of the study carried out by Jansen et al [32], where only one cases of pancreatic neuroendocrine neoplasms shown positivity to DOG1 immunostaining, among a large cohort of NENs. Possible explanations of this difference are the tumor heterogeneity, which could have result in negative staining, since Jansen et al used 0,6 mm tissue cores microarrays.
As concerns nonneoplastic tissue, in the present study only endocrine cells of gastric and ileal mucosa expressed DOG1, while pancreatic and bronchial endocrine cells were constantly negative, in contrast to Ardeleanu's ndings [33]. A possible explanation could be the different speci city of the DOG1 clone, since Ardeleanu ed al used a polyclonal anti-DOG1 antibody as compared to our monoclonal antibody. Based on these ndings, DOG1 seems to be not an endocrine marker, as previously proposed [33], but more probably it could mark only endocrine cell originating from the digestive tract.
In conclusion, our data show a strong correlation between DOG1 expression neuroendocrine tumors arising in the digestive tract. In light of these ndings, DOG1 expression could be used as part of immunohistochemical panel for the purpose of de ning primary anatomical site in the challenging setting of NET of unknown origin, suggesting clinical and radiological investigation of digestive tract. On the other side, our results highlight the need of careful evaluation and interpretation of DOG1 expression in the differential diagnosis between NET and GIST, particularly with epithelioid morphology, where association of other speci c markers is recommended to avoid misdiagnosis, especially in small biopsies or in cytological specimens.

Declarations Ethical Approval
Collaterally, DOG1 expression in nonneoplastic cells was evaluated, with interesting ndings. Endocrine cells of gastric and ileal mucosa, identi ed with both morphology and Synaptophysin and Chromogranin-A, also expressed DOG1 (Fig. 1G), a peculiar nding even more evident in neuroendocrine cell hyperplasia in the setting of autoimmune atrophic gastritis. Conversely, we couldn't identify any DOG1 expression in pancreatic endocrine cells, in contrast with results described by Ardeleanu's [33].  Tables   Tables 2 is available