Insulinoma-Associated Protein-1 Expression in Lymphoepithelial Carcinoma of the Thymus: A Potential Pitfall for Diagnosis With Neuroendocrine Carcinomas of the Thymus Open Access

Thirty-four cases accessioned under the diagnosis of lymphoepithelial carcinoma or lymphoepithelioma-like carcinoma of the thymus were identified from the surgical pathology files and from the personal consultation files of the authors. Histologic sections stained with hematoxylin-eosin and representative paraffin blocks were available in all cases. Clinical and demographic information were obtained from the medical records or from the referring physician. None of the patients had a history or evidence of malignancy outside of the mediastinum, confirming the primary nature of these tumors. The study was approved by our institutional review board. More detailed clinical and pathologic information, including EBER in situ hybridization status, was previously published on 25 of the cases in this study.⁹ 

For immunohistochemical staining, whole slide sections from representative tissue blocks were cut at 4 µm, deparaffinized in xylene, hydrated in descending dilutions of ethanol, and exposed to heat-induced epitope retrieval. Immunohistochemical staining was performed with reagents from the Dako Envison FLEX kit and the Dako AutostainerPlus stainer. Following pretreatment with target retrieval solution, tissue was blocked with peroxidase-blocking agent for 5 minutes and incubated with the primary antibodies at room temperature. Appropriate positive and negative controls were run concurrently for all antibodies tested. The antibodies used included INSM1 (clone MRQ-70, prediluted, Cell Marque); chromogranin (1:100; clone DAK-A3, Dako-Agilent, Santa Clara, California); synaptophysin (SYNAPT, ready to use, Dako-Agilent); cytokeratin AE1/AE3 (clone AE1/AE3, ready to use, Dako-Agilent); p40 (1:100; polyclonal, Biocare Medical, Concord, California); CD5 (clone 4C7, ready to use, Dako-Agilent); CD117 (1:400; polyclonal, Dako-Agilent); and Ki-67 (clone MIB-1, ready to use, Dako-Agilent). Immunostaining of neuroendocrine-associated markers was scored as negative (0%), weak (5%–30%), moderate (30%–50%), and strong (>50%) depending on the percentage of cells with positive staining, limited to cells with moderate to strong intensity staining (visible from ×2 magnification). Ki-67 proliferation index was calculated by counting positive nuclei in 100 tumor cells manually.

The clinical demographic features of our patients are summarized in Supplemental Table 1 (see the supplemental digital content at https://meridian.allenpress.com/aplm in the February 2025 table of contents). There were 20 men and 14 women aged 20 to 85 years (mean, 59 years). All tumors appeared as anterior mediastinal masses on chest imaging studies without evidence or history of tumor elsewhere. All patients were treated by complete surgical resection; in some cases, postoperative radiation and/or chemotherapy was also given. Epstein-Barr virus status was positive in 1 of 21 patients tested by in situ hybridization for EBER.

The tumors were characterized by a variegation of growth patterns but with a very distinctive and uniform cytologic appearance. The most distinctive feature in all tumors was a uniform population of poorly differentiated tumor cells that were characterized by large round to oval nuclei with vesicular chromatin, prominent round eosinophilic nucleoli, and an indistinct rim of eosinophilic cytoplasm (Figure 1, A). The tumors all were mitotically active, averaging 8 mitoses per 10 high-power fields. Several growth patterns were observed, including islands, cords, and trabecula of tumor cells separated by a dense lymphoid stroma. A few of the tumors also contained areas characterized by desmoplastic stroma devoid of a lymphoid host response. Solid tumor islands with central comedo-like areas of necrosis were seen in most of the cases (Figure 1, B). Scattered foci of stromal calcifications could also be found in 7 cases. Broad anastomosing cords of tumor cells associated with areas of comedo-like necrosis and fibrous intervening stroma (Figure 1, C) or separated by a dense lymphoid infiltrate (Figure 1, D) were also present. Areas displaying small tumor nodules reminiscent of the organoid pattern of growth seen in carcinoid tumors were present in 6 cases (Figure 1, E). A fine serpiginous arrangement of tumor cells resembling the ribbon pattern of neuroendocrine tumors was also present in 3 cases (Figure 1, F). In 8 cases, remnants of involuting thymus were present at the periphery of the tumor.

The results of the immunohistochemical stains in our cohort are shown in Supplemental Table 1. All cases showed strong positivity of the tumor cells for cytokeratin AE1/AE3 and p40. Sixty-five percent of cases showed moderate to strong nuclear positivity of the tumor cells for INSM1 (22 of 34) (Figure 2, A); 20% of cases showed weak staining (7 of 34), and 5 cases (15%) were negative for this marker. The larger islands of tumor cells with central comedo-like areas of necrosis also showed strong staining of the tumor cells for INSM1 (Figure 2, B). The pattern of staining in the cases with moderate positivity showed a tendency for the positive tumor cells to be concentrated along the periphery of the cords and islands of tumor (Figure 2, C and D). Interestingly, the tumor cells located deeper within these cords and islands also showed scattered nuclear staining, but the intensity of the staining was much weaker. In the 3 tumors that displayed a serpiginous, ribbon-like growth pattern, the tumor cells also showed scattered positivity for INSM1 in at least 30% of the cells (Figure 2, E). An interesting finding observed in the involuting thymic remnants at the periphery of the tumors in 8 cases was focal but convincing nuclear positivity in small pockets of cells located within the medullary portion of the thymic cell rests (Figure 2, F). Immunohistochemical stains for chromogranin showed only focal weak positivity in 1 case (case 21); stains for synaptophysin showed focal weak positivity in 12 cases. Immunohistochemical staining for CD5 was positive in 30 of 34 cases and CD117 staining was positive in 32 of 34 cases. There was no particular correlation between the pattern of staining for INSM1 and these markers. A stain for Ki-67 showed nuclear positivity that ranged from 15% to 40% of tumor cells (mean, 32%).

Thymic carcinoma is a distinct type of primary malignant neoplasm of the thymus that is difficult to diagnose because of its rarity as well as its variegation in morphotypes.^9–13  Thirteen distinctive histologic variants of thymic carcinoma are recognized by the World Health Organization (WHO).¹³  One of these variants, designated by the WHO as lymphoepithelial carcinoma (LEC), represents a highly distinctive subtype of thymic carcinoma characterized by poorly differentiated morphology and a dense lymphoid stromal component. LECs share with other primary malignant thymic epithelial neoplasms strong immunostaining for p63 and p40 and positivity for CD5 and CD117.¹³  Although rare cases have been reported that express chromogranin and synaptophysin, most studies, including those describing results of ultrastructural examination, have failed to identify evidence of neuroendocrine differentiation in these tumors, and they are widely acknowledged to represent a poorly differentiated, nonkeratinizing variant of squamous cell carcinoma.^12,13 

Thymic LEC is characterized by a variety of growth patterns that include irregular anastomosing cords, large solid islands, small infiltrative islands, and trabecular arrangement of tumor cells. Foci of central, comedo-like areas of necrosis are often present within the tumor cell islands, reminiscent of that seen in atypical carcinoids of the thymus. Rare cases may display a ribbon-like, trabecular growth pattern closely reminiscent of neuroendocrine tumors, and stromal calcifications like those observed in thymic carcinoids can also be present. The cytology of the tumors is characterized by a uniform population of round cells with round to oval nuclei, vesicular chromatin, prominent round eosinophilic nucleoli, and an indistinct rim of eosinophilic cytoplasm. The stroma is most often composed of a dense population of polyclonal B- and T-cell lymphocytes, although some cases may be desmoplastic and devoid of a lymphoid stroma.^9,13  The size of the tumor cells and the prominence of nucleoli can overlap with atypical carcinoids and large cell neuroendocrine carcinoma. In small biopsy samples, the monotonous appearance of the tumor cells and their trabecular arrangement can be confused for a neuroendocrine neoplasm. Aberrant expression of neuroendocrine markers has been described previously in thymic carcinomas.^14,15  In the study by Lauriola et al,¹⁴  the authors found expression of synaptophysin and neuron-specific enolase in 2 cases of lymphoepithelial carcinoma of the thymus. The explanation offered for this finding, as for cases of poorly differentiated non-neuroendocrine carcinomas with aberrant neuroendocrine differentiation in other organs such as the lung, was the existence of normal neuroendocrine cells derived from the neural crest in the involved tissues. It was postulated that the general trend of aberrant neuroendocrine differentiation was likely associated with an increase in malignancy and loss of differentiation for the various neoplasms that exhibit this feature.

Immunohistochemical support for neuroendocrine differentiation is generally achieved by obtaining a positive result with chromogranin, synaptophysin, and CD56, the 3 most commonly used markers for this purpose. In recent years a novel marker, INSM1, has been introduced for the identification of neuroendocrine differentiation. INSM1 is a zinc-finger transcription factor that was initially isolated from normal human pancreas and mouse insulinoma cell lines¹⁶  and was subsequently found to participate in the development of β cells in the pancreas.¹  Since then, INSM1 RNA has been isolated from brain, retina, olfactory epithelium, thyroid, thymus, and the gastrointestinal tract.^17–19  This protein was first introduced as a potential immunohistochemical neuroendocrine marker by Rosenbaum et al,¹⁹  who reported positivity for INSM1 in 88% of 129 neuroendocrine tumors from various sites and only in 1 of 24 non-neuroendocrine tumors tested. Since then, numerous additional studies addressing the utility of this marker for the diagnosis of neuroendocrine tumors have been published, including studies on neuroendocrine tumors of the lung,^20–22  larynx,^23,24  sinonasal tract,²⁵  breast,^26,27  pancreas,^28,29  skin,^30,31  thyroid and parathyroid,³²  cervix,³³  gastrointestinal tract,³⁴  and genitourinary tract.³⁵  Owing to the high specificity and sensitivity observed among thoracic neuroendocrine tumors, INSM1 has been proposed by some as a stand-alone diagnostic marker or as a first-line marker for the demonstration of neuroendocrine differentiation.^7,36,37  More recent studies, however, have identified expression of this marker in non-neuroendocrine neoplasms, including solitary fibrous tumor,³  extraskeletal myxoid chondrosarcoma,⁴  chordoma,⁵  angiosarcoma,⁶  alveolar rhabdomyosarcoma,⁷  Ewing sarcoma,⁷  and in various types of non-neuroendocrine carcinomas.⁷  There are 2 recent studies that have also documented INSM1 expression in thymic carcinoma. The study by Tsai et al⁷  reported INSM1 positivity in 4 of 10 cases of thymic carcinoma; however, to the best of our knowledge, no information was provided regarding the specific histologic type of the carcinomas studied. A more recent study by Kashima et al⁸  reviewed 35 cases of thymic squamous cell carcinoma, along with 4 atypical carcinoid tumors and 112 thymomas, and found that INSM1 was expressed in 63% of thymic carcinomas, whereas it was rarely expressed in thymomas. The authors indicated in their study that their tumors corresponded to squamous cell carcinomas as defined by the WHO; however, it is unclear whether this included cases of lymphoepithelial carcinoma of the thymus, as this entity is currently classified as a separate category from squamous cell carcinoma of the thymus.⁸  Review of the provided images in the study by Kashima et al⁸  reveals some images that may correspond to examples of “lymphoepithelial carcinoma,” indicating that at least some of the cases in their study corresponded to this variant. However, regardless of their specific case mix, our findings reinforce their impression that thymic squamous cell carcinomas may show increased expression of INSM1.

This study, which focused exclusively on the lymphoepithelial carcinoma subtype of thymic carcinoma, showed a high level of expression of INSM1. A total of 29 of 34 cases (85%) showed expression of this marker overall; in 7 cases the expression was focal and restricted to scattered foci of tumor cells, whereas in 22 cases the expression was moderate to strong. It is of interest that normal thymus adjacent to the tumor in 8 cases also showed scattered minute pockets of positive cells, mostly restricted to the medullary areas, suggesting that normal thymic epithelial cells in the medulla may also express this marker. Of note, chromogranin was focally and weakly positive in 1 of our cases, and synaptophysin showed only focal weak positivity in scattered tumor cells in 12 cases. The areas of positivity for INSM1 and for either chromogranin or synaptophysin did not overlap and the latter were seen to be distributed at random, unlike the distinctive pattern of staining for INSM1, which showed a predilection for a peripheral distribution in the tumor cell cords and islands. The absence of staining for chromogranin and synaptophysin in most of our cases suggests that the INSM1 positivity is not necessarily correlated with neuroendocrine differentiation in the thymus.

The role of INSM1 in lymphoepithelial carcinoma of the thymus is still unclear. Unfortunately, owing to the limitations for obtaining clinical follow-up in referral material, we were unable to comment on the impact of this finding regarding the clinical behavior of the lesions, but this finding remains of importance for the histologic differential diagnosis. Because of the demonstration of INSM1 nuclear expression in tumors that are clearly not associated with neuroendocrine differentiation, it has been postulated that this protein may play a different role in the neoplastic process for such cases rather than act as a marker of neuroendocrine differentiation. We suspect the same principle operates in LEC of the thymus and that the strong and consistent expression of this protein may be related to some as yet poorly understood function involved in growth regulation rather than serving as a marker of true neuroendocrine differentiation. This finding is of importance for the management of patients with this tumor type given that neuroendocrine carcinomas of the thymus are in general regarded as being highly aggressive neoplasms that respond to multimodality therapy, including cisplatinum-based chemotherapy for neuroendocrine tumors,^38,39  whereas lymphoepithelial carcinomas of the thymus tend to be a surgical disease that can also potentially respond to other types of treatment, such as checkpoint inhibitor treatment with anti–PD-L1.^9,40 

In summary, the potential for confusion of LEC of the thymus with a primary neuroendocrine malignancy of this organ is significant, particularly in the setting of small biopsy samples where adequate assessment of growth patterns may not be easily accomplished and because these tumor types tend to be treated differently. Morphologically, LEC is generally defined by a population of tumor cells with syncytial growth and often prominence of the nucleoli.¹²  Trabecular and serpiginous arrangements often seen in these tumors may be confused for a large cell neuroendocrine carcinoma or an atypical carcinoid of the thymus.¹³  Use of immunohistochemical stains in this setting can greatly facilitate the diagnosis. Unlike thymic carcinoids and large cell neuroendocrine carcinomas, LEC is universally positive for p63/p40 in all cases.¹³  Strong INSM1 positivity in a tumor that also shows strong p63/p40 positivity should favor a diagnosis of LEC over a neuroendocrine carcinoma. Lack of staining for additional neuroendocrine-associated markers such as chromogranin and synaptophysin should raise doubts regarding a diagnosis of neuroendocrine malignancy in this context and allow for a correct diagnosis.