Synaptophysin-expressing Gastric Glomus Tumors

Synaptophysin-expressing visceral glomus tumors have been recently reported.^1–6 Because the stomach is one of the most common gastrointestinal sites for neuroendocrine tumors and glomus tumors, synaptophysin-expressing glomus tumors can be mistaken for neuroendocrine tumors in gastric endoscopic biopsy specimens. We report two patients with gastric glomus tumors that were originally misdiagnosed as gastric neuroendocrine tumors because they expressed synaptophysin. Synaptophysin-expressing visceral glomus tumors have important implications.

The first patient (the index case, Patient 1) was a 48-year-old man. He presented with epigastric abdominal pain for 1 month with black stool. Gastroscopy revealed an antral polypoid mass (6.5 × 5.5 cm) with mucosal ulceration. Radiologically, it showed a large infiltrating mass involving the muscle layer with lymph node involvement. Endoscopic biopsy samples were procured from the lesion. The specimen was composed of multiple pieces of antral-type mucosa and submucosa intermixed with pieces of solid nodular tumor tissue (Figure 1). The initial histopathologic diagnosis was a neuroendocrine tumor pending immunohistochemical confirmation. A limited panel of synaptophysin, KIT (CD117), CD45, keratin (AE1/AE3), and Ki67 was initially performed. The tumor cells expressed synaptophysin (Figure 2) and showed a 3% Ki67 index. They were negative for the other immunomarkers. Because neuroendocrine tumors are usually keratin-positive, the immunomarkers panel was extended to include chromogranin A, CD56, S100, CD34, DOG1, and smooth muscle actin. The tumor cells expressed smooth muscle actin but were negative for the other markers (Figure 2). Calponin and desmin were added and the tumor cells were positive for calponin, but negative for desmin. The histopathologic examination of the small endoscopic samples of the tumor did not reveal cellular atypia, an increased mitotic index or necrosis. The patient has developed liver metastasis after a 3-months follow-up. Considering the large size and clinical behavior, the final diagnosis of a malignant gastric glomus tumor was entertained.

OPEN IN VIEWER

Figure 1.

(A) The gastric endoscopic biopsy sample (Patient 1) reveals pieces of antral-type mucosa with a fragment composed of muscularis mucosae and submucosa infiltrated by solid sheets and trabeculae of tumor cells (Hematoxylin and Eosin [H & E] stain, original magnification × 100). (B) The gastric endoscopic biopsy sample (Patient 2) reveals pieces of antral-type mucosa and a piece of a solid tumor nodule (H & E × 100). (C & D) The submucosal and mucosal tumor nodules (patients 1 and 2) are composed of uniform tumor cells with central to plasmacytoid round nuclei with fine chromatin, arranged in solid organoid nesting and trabecular patterns with rosette-like acini and thin-walled small vessels; simulating gastric well-differentiated neuroendocrine tumors.

OPEN IN VIEWER

Figure 2.

(A) The tumor cells show variable moderate patchy to strong diffuse synaptophysin staining. The overlying gastric mucosa shows scattered synaptophysin-positive neuroendocrine cells (×200). (B) The tumor cells (patient 1) show a strong diffuse cytoplasmic staining for synaptophysin (×400). (C) The tumor cells (patient 2) show a moderate to strong patchy synaptophysin expression (×400). (D) Smooth muscle actin immunomarker highlights the round cells of the tumor nodule (patient 1) which dissects through the residual spindle smooth muscle fibers of the muscularis mucosae (×200). (E) The round cell tumor cells (Patient 2) show a diffuse strong cytoplasmic staining for calponin (×400). (F, G, & H) The tumor cells are negative for keratin, chromogranin and CD56 respectively (×200).

The second patient (the retrospectively retrieved case, patient 2) was a 55-year-old man. He presented with dysphagia for 2 months. Gastroscopy revealed an antral polypoid submucosal nodule (1.8 × 1.5 cm) with an intact mucosa. Radiologically, it showed a well-defined nodule without mural wall involvement, lymphadenopathy or organomegaly. The endoscopic biopsy samples revealed several pieces of antral-type mucosa and submucosa intermixed with pieces of solid nodular tumor (Figure 1). The initial diagnosis was gastric well-differentiated neuroendocrine carcinoma based on a limited panel of synaptophysin, KIT and CD45 that was initially performed, in which the tumor cells expressed synaptophysin (Figure 2). However, the tumor cells were subsequently found to express smooth muscle actin and calponin, but failed to stain for chromogranin, CD56, and keratin. The tumor tissue fragments in the endoscopic samples did not show cellular atypia, mitoses, or necrosis. The patient has lost follow-up, however considering the small size and the histopathologic parameters of the tumor, the final diagnosis was a benign gastric glomus tumor.

We performed a retrospective review over a 10-year period to confirm that all the gastric neuroendocrine tumors were correctly diagnosed and none of the previously reported tumors were mislabeled for glomus tumors because of a limited immunohistochemistry panel. We retrieved 43 patients with gastric neuroendocrine tumors. We screened the hematoxylin and eosin slides, and the immunohistochemistry slides. The specimens that revealed expression of synaptophysin, chromogranin, CD56, and keratin were confirmed as gastric neuroendocrine tumors. We found four lesions that were initially stained for a limited panel of synaptophysin, CD117, CD45, and Ki67 on the presumption of a histopathologic diagnosis of a gastric neuroendocrine tumor and expression of synaptophysin. The paraffin blocks of these four lesions were retrieved and the immunohistochemistry panel was extended to include chromogranin A, CD56, keratin, smooth muscle actin, and calponin. In three lesions, the tumor cells stained for chromogranin, CD56, and keratin; and were negative for smooth muscle actin and calponin, which confirmed the original diagnosis of neuroendocrine tumors. One tumor (patient 2) did not express chromogranin, CD56, and keratin, but expressed smooth muscle actin and calponin. The diagnosis was amended to gastric glomus tumor. We also retrieved all glomus tumors from our pathology archive files over a 10-year period to test for their expression of synaptophysin. We found 26 patients with peripheral cutaneous and soft-tissue glomus tumors. We retrospectively performed immunohistochemistry to investigate their expression for synaptophysin. None of the peripheral glomus tumors expressed synaptophysin.

Glomus tumors encompass peripheral soft-tissue lesions, and the less frequent visceral lesions, of which the gastrointestinal tract is the most common site.^1,2,5,7 Of the latter, the stomach is a frequent site for glomus tumors as well as for neuroendocrine tumors. ⁷ Both tumor types have overlapping clinical, endoscopic, and radiologic features.^2,6,7 A pericytoma-like perivascular growth pattern of uniform cells with central round nuclei is a clue to glomus tumor.^3,4,7 Several case series and single case reports have demonstrated synaptophysin-expressing gastric glomus tumors in biopsy and resected specimens.^1,2,5–7 This phenomenon was also observed in nongastrointestinal tract visceral glomus tumors, for example, the respiratory tract.^2,4,8 In particular, the classical solid-type glomus tumor, which is the most common gastric type, can histopathologically mimic gastric neuroendocrine tumors. ⁷ The correct diagnosis depends on the utilization of at least two specific immunomarkers for each of the two look-alike lesions.^7,8

There are certain implications of the aberrant, but consistent synaptophysin immunoexpression by visceral glomus tumors compared to peripheral counterparts. First, it imposes a diagnostic potential pitfall for the histopathologists. Two of the gastric glomus tumors in our series were misinterpreted as gastric neuroendocrine tumors based on synaptophysin expression. Likewise, several studies have reported the misdiagnosis of visceral glomus tumors for neuroendocrine tumor because of their expression of synaptophysin.^2,3,6,9 Because neuroendocrine tumors are common in the gastrointestinal tract and in the respiratory tract compared to the less frequent visceral glomus tumors, pathologists rely on synaptophysin as the sole specific neuroendocrine marker. A comprehensive panel of chromogranin, CD56 and keratin should be included with synaptophysin, because neuroendocrine tumors are consistently positive for these immunomarkers, while glomus tumors are typically negative.^1–10 Second, whether it may have a predictive value would be an interesting question for future studies. Some of the reported synaptophysin-expressing glomus tumors were generally large and revealed the histopathologic and clinical features of intermediate and malignant tumors.^1,5,9–11 In a series of resected specimens of 26 gastroesophageal glomus tumors, Papke et al have proposed the malignancy criteria (size ≥5.0 cm, or both atypia and mitoses ≥2/10 HPF) for visceral glomus tumors. ⁹ In this study, synaptophysin was positive in 73% of the tumors. Brikness-Gartman et al have indicated that resection specimens show heterogeneity of the cytologic atypia and mitotic rate, that the diagnostic areas of malignancy could be missed in small biopsy samples. ¹⁰ Third, why visceral glomus tumors express this specific neuroendocrine marker in contrast to peripheral glomus tumors may highlight the pathophysiologic mechanism of this phenomenon and its importance. One study found pinocytic vesicles and electron-dense granules in synaptophysin-expressing glomus tumors. ¹ They have proposed that glomus tumors which exhibit an endocrine differentiation may represent a novel variant of glomus tumors. ¹

In conclusion, synaptophysin-expressing solid-type gastric glomus tumors can be misdiagnosed as gastric neuroendocrine tumors. The pathologists should be aware of this potential pitfall. Synaptophysin should be used in conjunction with keratin, chromogranin, and CD56 to confirm the diagnosis of neuroendocrine tumors and exclude synaptophysin-expressing visceral glomus tumors. The future implications and the clinical significance of this phenomenon warrant validation.