Frequent Immunohistochemical Expression of Transcriptional Repressor GATA Binding 1 in Salivary Gland Neoplasms: A Sensitive but Nonspecific Marker Open Access

We retrieved cases of SG neoplasms, benign and malignant, diagnosed between 2017 and 2022 from the departmental archives. Search terms included mucoepidermoid carcinoma, adenoid cystic carcinoma (AdCC), secretory carcinoma, acinic cell carcinoma, salivary duct carcinoma (SDC), epithelial-myoepithelial carcinoma (EMC), myoepithelial carcinoma, hyalinizing clear cell carcinoma, polymorphous adenocarcinoma, carcinoma ex pleomorphic adenoma, microsecretory adenocarcinoma, basal cell adenocarcinoma, intraductal carcinoma, oncocytic carcinoma, and adenocarcinoma, not otherwise specified, among malignant tumors, and pleomorphic adenoma, basal cell adenoma, myoepithelioma, intercalated duct adenoma, and striated duct adenoma among benign tumors. We also retrieved cases of head and neck squamous cell carcinoma (HNSCC), head and neck mucosal melanoma (HNMM), and carcinomas from various other sites, namely, breast, lung, stomach, colon, ovary, periampullary region, urinary bladder, prostate, thyroid, and kidney. Lastly, head and neck paragangliomas (HNPGLs) were also retrieved. Ethics approval to perform the analysis on archival specimens was waived due to the retrospective nature of the study.

Immunohistochemistry for TRPS1 was performed on whole tissue sections using a rabbit monoclonal antibody against TRPS1 (clone EPR16171; Abcam, Cambridge, United Kingdom). Briefly, 4-μm-thick formalin-fixed, paraffin-embedded tissue sections were deparaffinized and rehydrated. Antigen retrieval was performed using Tris ethylenediaminetetraacetic acid buffer solution (pH 9.0) for 30 minutes, in a microwave oven. Hydrogen peroxide blocking and protein blocking was done, followed by overnight incubation with the primary antibody in a dilution of 1:8000 at 4°C. UltraVision Quanto detection system HRP DAB was used as the secondary antibody detection system (Thermo Fisher, Fremont, California). Sections from normal breast tissue were used as positive controls with each batch of staining. For negative controls, the primary antibody was omitted. Immunostained slides were reviewed by 2 pathologists (S.S. and A.K.). Nuclear staining of any intensity was recorded, while only membranous staining and/or cytoplasmic staining were considered as nonspecific. Staining was scored based on the proportion of positively stained nuclei as follows: 0: no positively stained cells; 1: 1% to 10%: 2: 11% to 50%; and 3: 51% to 100%. Intensity of staining was scored as 0: negative; 1: weak; 2: moderate; and 3: strong. Immunoreactivity score (IRS) for TRPS1 immunoexpression was calculated by multiplying the proportion score and the intensity score for each case. IRS was interpreted as 0 to 1: negative, 2: low positive, 3 to 4: intermediate positive, and 6 to 9: high positive.¹¹  Cases with discordant IRSs were reviewed together by both pathologists to reach a consensus score. Mean, standard deviation, and median values of IRS were computed, and a Mann-Whitney U test was used for comparison of IRS values, using IBM SPSS statistics software. SRY-related HMG-box 10 (SOX10) immunohistochemistry was performed in non-SG carcinomas using a mouse monoclonal antibody (IHC010, GenomeMe Lab Inc., Richmond, British Columbia, Canada) in a dilution of 1:500 following antigen retrieval in citrate buffer at pH 6.

Two hundred and seventy tumors, including 148 cases of SG neoplasms, 110 cases of carcinomas of various sites, 6 HNMMs, and 6 HNPGLs were evaluated for TRPS1 immunoexpression. The distribution of the histologic types of tumors is depicted in Table 1.

SG neoplasms included 23 benign and 125 malignant tumors. Benign SG tumors included 12 pleomorphic adenomas, 6 basal cell adenomas, and 5 myoepitheliomas. All benign SG tumors showed TRPS1 positivity, with none having a negative IRS (Table 1). Seventeen of 23 cases (73.9%) had a high-positive IRS, while 5 of 23 cases (21.7%) were intermediate positive, and only 1 of 23 cases (4.3%) was low positive. The majority of pleomorphic adenomas (10 of 12; 83.3%) were high positive (Figure 1, A and B), with the remainder (2 of 12; 16.7%) being intermediate positive. In 3 cases, stronger staining was noted in myoepithelial cells on the abluminal aspect of the tubules as well as embedded in the myxoid stroma, as compared to luminal cells lining the tubules. Basal cell adenomas, similar to pleomorphic adenomas, were most frequently high positive (4 of 6 cases; 66.7%), followed by intermediate positive (2 of 6 cases; 33.3%); they, however, did not show stronger staining in abluminal cells. Myoepitheliomas showed high positivity in 3 of 5 cases (60%), intermediate positivity in 1 of 5 cases (20%), and low positivity in 1 of 5 cases (20%).

One hundred and twenty-five malignant SG neoplasms were assessed for TRPS1 staining, as detailed in Table 1. Two biopsy cases of adenocarcinoma, not otherwise specified, did not have sufficient tissue in the block to perform TRPS1 staining and were not included. All sections for TRPS1 immunostaining were from primary tumors. Of the 125 malignant SG neoplasms, 115 of 125 (92%) were immunopositive for TRPS1, with 95 of 125 (76%) high positive (Figure 1, C through L), 14 of 125 (11.2%) intermediate positive, and 6 of 125 (4.8%) low positive. Intermediate- to high-positive IRSs were seen in the majority of SDCs (22 of 22 cases; 100%), AdCCs (18 of 18 cases; 100%), secretory carcinomas (17 of 17 cases; 100%), myoepithelial carcinomas (6 of 6 cases; 100%), polymorphous adenocarcinomas (5 of 5 cases; 100%), EMCs (16 of 17 cases; 94.1%), hyalinizing clear cell carcinomas (5 of 6 cases; 83.3%), mucoepidermoid carcinomas (8 of 10 cases; 80%), carcinoma ex pleomorphic adenomas (5 of 7 cases; 71.4%), microsecretory adenocarcinomas (2 of 2 cases; 100%), and basal cell adenocarcinomas (2 of 2 cases; 100%). The histologic types of carcinomas in the carcinoma ex pleomorphic adenomas were 3 SDCs, 2 EMCs, and 2 myoepithelial carcinomas, as seen in Table 1. The only SG carcinomas with the majority of cases being negative or low positive were acinic cell carcinoma (9 of 12 cases; 75%) and oncocytic carcinoma (1 of 1 case; 100%); the latter was negative for mastermind-like transcriptional coactivator 2 (MAML2) rearrangement by fluorescence in situ hybridization and was immunonegative for androgen receptor, human epidermal growth factor receptor 2 (HER2), and rat sarcoma (RAS) Q61R mutation specific antibody.

Thus, 138 out of 148 (93.2%) SG neoplasms showed TRPS1 immunopositivity, with high IRS in 111 of 148 (75%), intermediate IRS in 20 of 148 (13.5%), and low IRS in 7 of 148 (4.7%) cases, and only 10 of 148 SG neoplasms (6.8%) were negative for TRPS1.

Among 122 tumors (Table 1) other than primary SG neoplasms in which TRPS1 immunostaining was performed, there were 5 breast carcinomas (Figure 2, A through C), and 72 nonbreast–non-SG carcinomas including 33 HNSCCs of different histologic subtypes (Figure 2, D through L), 6 HNPGLs, and 6 HNMMs (Figure 2, M through O). The hormone receptor profile of the 5 invasive breast carcinomas of no special type was as follows: 3 estrogen receptor, progesterone receptor–positive and HER2-negative tumors, and 2 TNBCs. The 33 HNSCCs included 9 conventional SCCs, 8 human papillomavirus (HPV)-unrelated nonkeratinizing SCCs, 10 spindle cell SCCs, and 3 cases each of HPV-associated SCC and basaloid SCC. Of these, 31 were mucosal SCCs and 2 were HPV-positive ocular adnexal SCCs. Among the 105 nonbreast–non-SG carcinomas, TRPS1 staining was performed on metastases in 7 cases (2 SCC, 4 thyroid, and 1 prostate carcinoma). In the remaining 98 cases, it was performed on primary tumors. In 1 case of thyroid follicular carcinoma metastatic to bone, TRPS1 was performed on decalcified sections.

All 5 breast carcinomas were high positive for TRPS1 (100%). Among nonbreast–non-SG carcinomas, HNSCCs were frequently high (11 of 33; 33%) to intermediate (4 of 33; 12%) positive, especially conventional keratinizing SCCs (Figure 2, D through F) and spindle cell SCCs (Figure 2, G through I). Both SCCs examined from metastatic sites showed high positive staining. Thyroid (9 of 17 high and 2 of 17 intermediate positive) and renal cell (4 of 12 high and 3 of 12 intermediate positive) carcinomas also showed frequent TRPS1 positivity (Figure 3, A, B, and D through G). Adjacent normal thyroid (Figure 3, C) and kidney (Figure 3, H), when present, also showed TRPS1 immunopositivity of variable intensity. Four of the thyroid carcinoma sections for TPRS1 staining were from metastatic tumors; 3 of these showed high positivity (including on a decalcified section), while 1 was negative. Three of 13 (23%) lung adenocarcinomas showed high-positive IRS, 2 of 13 (15%) were intermediate positive, and the rest were negative for TRPS1. The 8 cases that were TRPS1 negative were of papillary-acinar (Figure 3, I and J) morphology, while 4 of the 5 cases with intermediate to high positivity had signet ring cell features (Figure 3, K). All gastrointestinal adenocarcinomas (Figure 3, L through O), including stomach, colon, and periampullary carcinomas, were negative (5 of 5 cases each; 100%), as were all 6 of 6 prostatic acinar adenocarcinomas (100%), 1 of which was a metastasis. Two of 5 (40%) ovarian serous carcinomas showed TRPS1 positivity of intermediate- and high-positive IRS in 1 case each. Urothelial carcinomas were either negative (2 of 4; 50%) or low positive (2 of 4; 50%). HNMMs were negative (4 of 6; 67%) or low positive (2 of 6; 33%), while all HNPGLs were negative (6 of 6 cases; 100%).

The mean ± standard deviation IRS for TRPS1 in SG carcinomas was 6.7 ± 2.913, and the median IRS was 9. The mean ± standard deviation IRS in nonbreast–non-SG carcinomas was 2.47 ± 3.032, and the median IRS was 1, while HNMMs had a mean ± standard deviation IRS of 0.67 ± 0.9 and a median IRS of 0. HNPGLs had a mean ± standard deviation IRS of 0.17 ± 0.4 and median IRS of 0. TRPS1 IRS in SG carcinomas was significantly higher than the IRS of nonbreast–non-SG carcinomas (P < .001, Mann-Whitney U test).

SOX10 positivity was seen in 1 TNBC, 1 basaloid SCC, all HNMMs, and in the sustentacular cells of all HNPGLs. The TNBC had high-positive TRPS1 and faint SOX10 (Figure 2, A through C), while the basaloid SCC was TRPS1 negative and showed strong SOX10 positivity (Figure 2, J through L). None of the 46 TRPS1-positive nonbreast–non-SG carcinomas was SOX10 positive. Results of combined TRPS1 and SOX10 are shown in Table 2.

SG neoplasms are a heterogeneous group of benign and malignant tumors with considerable morphologic and immunohistochemical overlap among themselves, and also with neoplasms of other organs and tissues, especially those of epithelial origin. Unlike several organs, such as lung, colon, and kidney, which demonstrate immunohistochemical expression of fairly specific lineage-associated transcription factors (eg, TTF1, CDX2, PAX8), no such pan-SG marker has been identified to date. While immunohistochemical stains can determine the luminal, abluminal, or biphasic nature of most SG neoplasms, and immunohistochemical surrogates for molecular alterations have been identified in certain histologic types of SG neoplasms, they do not serve the purpose of confirming the origin of a neoplasm as being from SG tissue.⁷  Thus, the need for such a pan-SG neoplasm marker arises, which would have maximal utility in the setting of metastatic disease.

The search for a breast-specific tumor marker by Ai et al¹¹  resulted in the identification of TRPS1 as sensitive and specific marker for breast carcinoma, even for triple-negative breast cancer. Interestingly, a TMA analysis in tumors other than breast carcinoma revealed TRPS1 immunopositivity in a proportion of SG SDCs and breast AdCCs.¹¹  Hence, we aimed to investigate the expression of TRPS1 across various SG neoplasms, to determine if it could be a potential pan-SG marker. Subsequent studies consolidated the high specificity of TRPS1 for breast carcinoma; however, we postulated that if TRPS1 is expressed in a significant proportion of SG neoplasms, this would affect its specificity for breast carcinoma, impacting routine diagnostic pathology practice, particularly in the hands of the unwary. We therefore assessed TRPS1 expression in SG neoplasms, nonbreast–non-SG carcinomas, HNMMs, and HNPGLs to determine its utility in distinguishing between SG tumors of different types, as well as from non-SG tumors. Our findings indicate that TRPS1 is not specific to breast carcinoma, as TRPS1 immunoexpression was seen in 93% (138 of 148) of all SG neoplasms, irrespective of their benign or malignant nature. Staining was not limited to a particular cell type, being present in tumors of ductal as well as myoepithelial/basal origin. In biphasic tumors, TRPS1 expression was seen in both luminal and abluminal cells; however, few cases of pleomorphic adenoma, AdCC, and EMC showed stronger staining in abluminal myoepithelial cells. All benign SG tumors examined showed TRPS1 immunopositivity, with the majority (22 of 23; 96%) showing intermediate or high-positive IRS, making TRPS1 a reliable immunohistochemical marker across all types of benign SG neoplasms, irrespective of cell of origin.

To date, a single study by Tjendra et al¹⁴  has evaluated TRPS1 expression in various histologic types of SG neoplasms, at primary as well as metastatic sites. Similar to our results, they reported TRPS1 positivity in the majority of benign and malignant SG neoplasms tested. Importantly, malignant SG carcinomas at metastatic sites all showed intermediate to strong TRPS1 positivity.¹⁴  The only stark difference between their results and ours was in secretory carcinoma: while they found all 3 cases to be negative or low positive, we found all 17 to be intermediate or high positive. However, a study by Salem et al,¹⁵  although including a smaller number of SG carcinomas, also found both their cases to be positive. Yoon et al¹²  similarly showed 3 of 3 breast secretory carcinomas to be positive for TRPS1.

In addition to SG neoplasms, we performed TRPS1 immunohistochemistry in several nonbreast–non-SG tumors that are potential differential diagnoses for SG neoplasms, which is the novelty of our study. In the parapharyngeal space, where the site of tumor origin is not easily identified on imaging, the most common tumors encountered include benign SG tumors, HNPGLs, and benign nerve sheath tumors,^16–18  and it may be difficult to distinguish between them on core needle biopsy or fine-needle aspiration cytology. HNPGLs and SG tumors can both have cells with clear cytoplasm, and both could demonstrate positivity with GATA3. We found that all HNPGLs examined were immunonegative for TRPS1. Thus, although there are established immunohistochemical markers for these and several other differential diagnoses such as schwannoma versus myoepithelioma, with both being S100 protein and SOX10 positive,¹⁹  application of TRPS1 can, with a single marker, determine whether the tumor is of SG origin or not and exclude other possible alternative diagnoses without having to use a wide panel of immunohistochemical stains. Apart from the obvious diagnostic utility, this would also help conserve tissue for further ancillary testing, including molecular testing, reduce cost per case, and decrease turnaround time. We did not find rarer benign epithelial tumors such as intercalated duct and striated duct adenomas in our archives; therefore, these could not be assessed for TRPS1 staining, which remains a limitation of the study. However, these do not pose a diagnostic dilemma often.

Among malignant SG neoplasms, 92% (115 of 125) of cases showed TRPS1 immunoexpression, across 12 of 13 histologic types of carcinomas tested, making TRPS1 the first pan-SG carcinoma immunohistochemical marker. Notably, the only carcinomas largely negative or low positive for TRPS1 were acinic cell carcinomas and 1 oncocytic carcinoma. Oncocytic carcinoma is increasingly being recognized as a heterogeneous group including oncocytic subtypes of various well-defined SG carcinomas such as mucoepidermoid carcinoma, EMC, etc.^20,21  The oncocytic carcinoma included, despite extensive testing, could not be categorized as any other SG carcinoma, and its nature remains ambiguous, making it difficult to comment on the negativity for TRPS1. The possibility of it being a metastatic carcinoma can also not be excluded. Thus, due to its widespread expression across most histologic types, there does not appear to be any diagnostic value of TRPS1 in distinguishing between different primary SG carcinomas.

Next, we explored the utility of TRPS1 in distinguishing SG carcinomas from non-SG carcinomas. SG carcinomas metastasize to lymph nodes in approximately 25% to 40% of cases and develop distant metastasis in approximately 20% to 25% of cases, with lung being the most common site.^22–28  In limited material, the morphologic features alone may not be useful in indicating the primary as SG, due to the wide histologic spectrum.⁶  All breast carcinomas assessed were diffusely immunopositive for TRPS1 with high-positive IRS (100%) irrespective of hormonal status, in concordance with previous studies.^11,12  Apart from breast carcinoma, frequent TRPS1 positivity with intermediate to high IRS was seen in HNSCCs, lung adenocarcinomas, thyroid carcinomas, and renal cell carcinomas. Interestingly, the lung adenocarcinomas with intermediate to high TRPS1 IRS were anaplastic lymphoma kinase (ALK)-rearranged signet ring cell adenocarcinomas, while all those with papillary-acinar morphology were negative or low positive. On literature review, immunoexpression of GATA3, another member of the GATA family of nuclear transcription factors, has been described in a subset of mucinous lung adenocarcinomas.²⁹  All gastrointestinal and hepatobiliary adenocarcinomas, including poorly cohesive signet ring cell carcinomas, did not display TRPS1 immunoexpression, similar to prior analyses.^11,13  Ovarian serous carcinomas also showed intermediate to high positive IRS in a subset (2 of 5; 40%), as observed previously.¹¹  Overall, the mean IRS in SG carcinomas was significantly higher than the mean IRS in nonbreast–non-SG carcinomas. However, the frequent positivity in certain nonbreast–non-SC carcinomas indicates that TRPS1 expression is more widespread than previously believed, limiting its ability to detect SG origin, even if only high IRS is considered, and especially in the setting of lung metastasis. Thus, while TRPS1 staining is seen in more diverse SG neoplasms and appears to be a sensitive pan-SG marker, its utility is limited by its lack of specificity.

To overcome this pitfall, we examined SOX10 staining in nonbreast–non-SG carcinomas to assess whether a combination of TRPS1 and SOX10 would be more useful than TRPS1 alone. SOX10 is a transcription factor that is expressed by luminal as well as abluminal cells of the acini and intercalated ducts of normal salivary glands.³⁰  It also stains several SG neoplasms that originate from these cells, with SOX10-negative SG carcinomas being SDC and mucoepidermoid carcinoma.³¹  We found that none of the TRPS1-positive nonbreast–non-SG carcinomas was positive for SOX10. Thus, it appears that a TRPS1⁺/SOX10⁻ immunoprofile favors a non-SG origin in most situations. However, this would not help in distinguishing mucoepidermoid carcinoma and SDC from non-SG cancers. Mucoepidermoid carcinomas showed intermediate to high positive TRPS1 IRS in 80% of cases. Interestingly, HNSCC also showed intermediate to high positivity in nearly half the cases, thus impeding its use for differentiation between the two. Addition of SOX10 does not confer any benefit in this scenario as both mucoepidermoid carcinomas and SCCs are TRPS1⁺/SOX10⁻.

The initial study including primary SG neoplasms in their analysis of TRPS1 immunoexpression was performed on TMAs (Table 3).¹¹  TMA analysis is considered an inexpensive and reliable alternate to whole tissue sections while investigating a large number of cases.³²  While TMAs have distinct advantages, it is largely considered that confirmation of the clinical relevance of novel immunohistochemical stains should preferably be assessed on whole tissue sections.³³  Hence, we used the latter rather than a TMA for our study. Keeping intratumoral heterogeneity in mind, the use of whole tissue sections may be responsible for the higher frequency of TRPS1 immunopositivity seen in our study and in that of Tjendra et al.¹⁴  These differences could also be attributed to the antibody clones used (Table 2).

To conclude, this study reiterates the presence of TRPS1 expression in a wide spectrum of SG neoplasms, confirming that it is not specific for breast carcinoma. It also assessed TRPS1 staining in carcinomas of several other sites, revealing significantly higher mean TRPS1 IRS in SG carcinomas than in nonbreast–non-SG carcinomas. Certain nonbreast–non-SG carcinomas like those of lung, thyroid, kidney, and HNSCC do express TRPS1, but the combination of SOX10 with TRPS1 can enhance the discriminatory value as compared to TRPS1 alone.