Biomarker immunoprofile in salivary duct carcinomas: clinicopathological and prognostic implications with evaluation of the revised classification

Salivary duct carcinoma (SDC) is an uncommon, aggressive malignant neoplasm histologically resembling high-grade mammary ductal carcinoma. SDC can arise de novo or ex pleomorphic adenoma. To clarify the correlation of biomarker immunoprofile with clinicopathological findings and clinical outcome of SDC, we conducted immunohistochemistry for EGFR, HER2, HER3, AR, CK5/6, p53, and Ki-67, along with HER2 fluorescence in situ hybridization in 151 SDCs. SDCs ex pleomorphic adenoma more commonly overexpressed EGFR, HER2, HER3, and Ki-67 than de novo SDCs (P = 0.015, < 0.001, 0.045, and 0.02, respectively). In multivariate analysis, AR− and CK5/6+ were associated with shorter progression-free survival (P = 0.027 and 0.004, respectively). Moreover, patients with p53-extreme negative/positive demonstrated poorer overall survival (P = 0.007). On assessing the revised classification by the combination of biomarker expression, the percentages of each subtype were as follows: ‘apocrine A’ (AR+/HER2−/Ki-67-low) (24%), ‘apocrine B’ (AR+/HER2−/Ki-67-high) (18%), ‘apocrine HER2’ (AR+/HER2+) (35%), ‘HER2-enriched’ (AR−/HER2+) (12%), and ‘double negative’ (AR−/HER2−) (11%). ‘Double negative’ was further subclassified into ‘basal-like’ (EGFR and/or CK5/6+) (7%) and ‘unclassified’ (3%). Consequently, patients with ‘apocrine A’ showed a better progression-free survival than those with any other subtypes. Our revised immunoprofiling classification was valuable for predicting the survival and might be useful in personalized therapy for patients with SDC.


INTRODUCTION
Salivary duct carcinoma (SDC) is an uncommon tumor, accounting for 10% of all salivary gland carcinomas, and histologically resembles high-grade breast ductal carcinoma [1].It occurs not only de novo but also as the malignant component of carcinoma ex pleomorphic adenoma (PA) [1].The standard treatment for SDC is surgical excision and post-operative radiotherapy; however, SDC exhibits clinically aggressive behavior with locoregional recurrence and distant metastasis [2][3][4].The development of salvage therapy for these patients is required in order to improve their prognosis.
Molecular subtypes defined by gene expression patterns of breast cancer using DNA microarrays are known to be of major prognostic value [17].Subsequently, the immunohistochemical classification based on a combination of estrogen receptor (ER), progesterone receptor (PR), HER2, EGFR, CK5/6, and Ki-67 status has been developed as a surrogate; this classification system has prognostic and therapeutic implications that correlate with molecular subtypes [18,19].For SDCs, two immunohistochemical classification systems corresponding to the breast cancer have been suggested, but their prognostic relevance is unclarified [14,15,20].
In the current study, we examined the correlations of the immunoexpression of biomarkers in a large series of SDC with clinicopathological features, including the histologic origin (i.e., de novo versus ex PA), and clinical outcome.Furthermore, we attempted to propose a revised classification of SDC based on the biomarker immunoprofile and assessed its impact on the survival.

Patient characteristics
Representative histologic features of SDC case are shown in Figure 1.The patients consisted of 127 males and 24 females, with a median age of 64 years (range, 26-87 years) (Table 1).One hundred and seventeen of 151 patients (77%) had tumors arising in the parotid gland, and 30 (20%) had tumors arising in the submandibular gland.Sixty-nine patients (46%) presented with T4 disease.Lymph node involvement was present in 80 patients (53%), and distant metastasis was observed in 9 patients (6%).Based on the histologic origin, 151 SDC cases were histologically classified as follows: 57 de novo cases (38%), 89 ex PA cases (59%), 5 unknown cases (3%).SDCs ex PA included 13 intracapsular SDC ex PA cases, 5 microinvasive SDC ex PA cases, and 71 widely invasive SDC ex PA cases.The median follow-up period of survivors was 3.7 years (range, 0.4-18.7 years).

Prognostic impact of biomarkers
The results of univariate and multivariate analyses of prognosis for patients with SDC are shown in Table 2.In univariate analysis, patients with AR-negative, p53extreme negative/positive, and Ki-67-high significantly decreased both overall survival (OS) and progressionfree survival (PFS).Additionally, CK5/6-positive was significantly associated with a shorter PFS.Furthermore, in the multivariate analysis, patients with p53-extreme negative/positive demonstrated a significantly worse OS.Also, AR-negative and CK5/6-positive patients were independently associated with worse PFS.Kaplan-Meier survival curves for AR, p53, and CK5/6 are shown in Figure 7A-7C.
The immunoexpression of biomarkers other than those mentioned above, including ERβ, EGFR, HER2 (Figure 7D), HER3, MUC1, and PLAG1 did not have a significant impact on the survival of patients with SDC.

DISCUSSION
Because SDC is an uncommon entity and frequently poses a diagnostic challenge for general pathologists,  the precise immunohistochemical biomarker expression profile and its correlation with the clinicopathological and prognostic significance are not fully explored [1,13].Thus, thorough large-scale series investigation is necessary to establish the convincing evidence-based data for this highly aggressive tumor.
With the recent introduction of HER2-targeted therapy for patients with SDC, the determination of HER2 status is crucial in order to select patients who may benefit from this treatment [21][22][23][24].However, the positive rate of HER2 overexpression in SDC reported previously is extremely broad, ranging from 15% to 100% [9,11,14,20,22], due seemingly to the ambiguous criteria defining the positivity.When the 2007 ASCO/CAP guideline recommending HER2 testing for breast cancer [25] is adopted for the evaluation, HER2 positivity of SDC ranges from 15% to 44% [12,14,15,20,26].In the current study, the value of 46% was slightly higher than that in previous reports, largely because we assessed the HER2 status based on both immunohistochemistry and FISH findings in accordance with the updated 2013 ASCO/CAP guideline [27], in which the cutoff immunohistochemical level for HER2 positivity was reduced from 30% to 10%.Additionally, our results revealed that SDCs showed extremely high concordance between cases showing HER2 3+ and HER2 amplification.Therefore, the expression of HER2 protein in SDC is highly influenced by the HER2 amplification status.Abbreviations: HR = hazard ratio; CI = confidence interval; HER2 = human epidermal growth factor receptor 2; AR = androgen receptor; EGFR = epidermal growth factor receptor; CK = cytokeratin.Regarding the correlation of HER2 status with clinical features, the HER2 positivity is considered to be a predictor of a poor prognosis in breast cancer [17,18].In SDCs, although Skálová et al. [9] and Jaehne et al. [5] reported that HER2 overexpression was linked to a poor survival in their analysis of 11 and 34 cases, respectively, we failed to detect any relationship between the HER2 status determined by immunohistochemistry and/or FISH analysis and the clinical outcome, comparable to that found in recent studies [12,15,26].SDC not only occurs de novo but also arises in PA [1].However, the difference in the molecular mechanisms underlying the carcinogenesis between these two sequences is still poorly understood.Recently, Chiosea et al. have revealed that using targeted next-generation sequencing, SDCs ex PA tend to have TP53 mutations or ERBB2 copy number gain, whereas de novo SDCs frequently harbor combined HRAS/PIK3CA mutations but no ERBB2 amplification [16].We verified that SDCs ex PA commonly showed not only HER2-positive but also overexpression of EGFR and HER3 as compared with the de novo type.Therefore, activation of HER family members is a more crucial event in the carcinogenesis of SDC when it arises in PA than with de novo occurrence.A further analysis of downstream events in the HER family signaling pathway is required to clarify the detailed mechanism of carcinogenesis in SDC ex PA.We did not find statistically significant relationship between p53 expression and histologic origin of SDC.
than AR-positive patients, although the ERβ expression did not influence the survival.The role of AR in the generation or progression of SDC has been under investigated [29]; however, AR has recently become a key target of androgen deprivation treatment for this tumor [2,11,21,24,33,34], as with prostate cancer.Further studies will be required to verify the effect and potential resistance mechanisms for this therapy.
CKs, intermediate filament proteins, reflect the epithelial cell type and state of tissue growth and differentiation in addition to the functional status of the tissue.In the breast and salivary gland, CK5/6 is regarded as a basal marker.To our knowledge, this is the first study to report that the overexpression of CK5/6 was an independent prognostic factor in patients with SDC, a finding that was equivalent to its role in a breast cancer study [18].
Whether or not the p53 expression can be a consistent independent prognostic biomarker of SDC is still under debate [5,8,16].Most recently, in breast cancer, Boyle et al. attempted to classify the p53 expression patterns into three groups: extreme negative, extreme positive, and non-extreme [35].They found that p53-extreme negative/ positive expression was significantly associated with a poorer OS than p53-non-extreme expression, and that combined p53-extreme negative/positive expression better predicted the OS than either pattern alone.Furthermore, in their analysis of the TP53 mutation status, detectable mutation types appeared to be related to the protein status, with a missense mutation corresponding to the extreme positive phenotype, and the nonsense mutation appearing to abrogate the protein expression, manifesting as the extreme negative phenotype.By using their methods, we disclosed that the p53-extreme negative/positive expression was an independent prognostic factor in patients with SDC.
Several immunohistochemical classification systems have been developed as surrogate methods for the molecular subtypes based on the gene expression patterns of breast cancer, which have proved useful in guiding decisionmaking for systemic therapies, predicting the biological behavior of the tumor, and determining the prognosis [18,19].Breast cancer has been proposed to be classified into several subtypes based on the expression profiles of ERα, PR, HER2, and Ki-67 LI, such as luminal A, luminal B, luminal B HER2, HER2-enriched, and triple negative (basal-like) [18,19].Unlike breast cancer, however, in SDCs, the expression of ERα and PR is almost exclusively negative [10,11,14], while AR is known to be frequently expressed, as noted in our results.Consequently, it is reasonable to postulate that AR expression in SDC is analogous to ERα reactivity in breast cancer, representing an apocrine phenotype.Given the morphologic similarity to mammary ductal carcinoma, two classification systems of SDC based on the biomarker immunoexpression profile were recently proposed [14,15,20].Di Palma et al. suggested that SDC can be classified into four subtypes by a combination of the expression of AR, HER2, EGFR, and CK5/6 as follows: 'luminal AR positive', 'HER2 positive', and 'basal-like' in addition to 'intermediate' [14,20].They did not take into account the Ki-67 LI in their system and failed to detect a correlation between the nuclear grade and subtype, except that 'basal-like' subtype SDCs were highgrade, though no prognostic significance was provided.In the breast cancer classification, the threshold for Ki-67 LI between low-and high-Ki-67 LI groups as 20% is currently accepted by most of experts [36], whereas no optimal cutoff point in SDC has been validated yet.The mean Ki-67 LI of 44% in SDCs was about two times higher than that of breast cancers reported in the literature [37].Furthermore, survival analysis revealed that the Ki-67 LI value of 40% was the most suitable cutoff point in terms of prognostic relevance.For these reasons, we adopted that value for our SDC classification.In the present study, we refined our previously reported classification system in order to reflect the increased feasibility of an appropriate personalized systemic therapy with anti-HER2, anti-AR, and/or cytotoxic drugs [15].Consequently, our revised classification system has great advantages in predicting the prognosis of patients with SDC.Di Palma classification is certainly simple for practical use but was only remotely related to the survival by our present case analysis.Since the therapeutic efficacy of our revised classification has yet to be evaluated, further studies are essential to determine its usefulness for devising a relevant treatment strategy for SDC in a clinical setting.
In conclusion, activation of HER family members is more frequently observed in SDC arising in PA than in its de novo occurrence.The immunohistochemical expression of AR, CK5/6, and p53 (of its extreme evaluation) was independent prognostic factors in SDC.However, further clinical trials are necessary to establish the optimum treatment referring to the expression profile of SDC.Our classification based on the biomarker immunoprofile was valuable for predicting the survival and might be useful in the future for selecting appropriate therapy for patients with SDC.

Patient selection
This study comprised 151 patients with SDC diagnosed and treated at 7 institutions between 1992 and 2014, except for patients who underwent anti-HER2 or anti-AR therapy as an initial treatment.Those patients include the cases previously reported by Otsuka et al [4].All tumors were confirmed to have been diagnosed correctly on a central review system by two expert pathologists (T.N. and Y.S.) according to the rigorous histomorphologic criteria for SDC (Figure 1) [1].Other entities, including high-grade transformation of various carcinomas, high-grade mucoepidermoid carcinoma, squamous cell carcinoma, and adenocarcinoma, not otherwise specified, were carefully eliminated from this study via ancillary analyses, if necessary [7,28].Moreover, we conducted histologic review of the multistep sections from the entire tumor in every case in order to enhance accuracy of identification whether histologic origin of the tumor was de novo or ex PA.In the case of SDC ex PA, pre-existing PA component frequently represented a hyalinized nodule surrounded by carcinoma.Even in such an instance, carcinoma nests enclosed within the hyalinized nodule were often rimmed by myoepithelial marker-positive benign neoplastic cells.The patients' charts were retrospectively appraised to obtain data on the age, gender, tumor site, tumor size, lymph node involvement, distant metastasis, treatment, and outcome.The tumor stage was classified according to the UICC TNM classification and staging system (2010, 7th edition).Most patients were treated surgically with postoperative irradiation and/or chemotherapy.
The present study was approved by the Institutional Ethics Review Board of the ethics committee of each of the seven institutions that participated in this study, and the need to obtain informed consent was waived owing to the retrospective nature of the analysis.

Immunohistochemistry and FISH
For immunohistochemistry, formalin-fixed, paraffin-embedded tumor tissue was cut into 3-μmthick sections.A polymer-based detection system with heat-mediated antigen retrieval was conducted using the primary antibodies shown in Supplementary Table 2. Diaminobenzidine was applied to detect antigen-antibody reactions.Appropriate positive and negative controls were employed for all conditions.
Since it has been mentioned that SDC with no immunoreactivity for AR is rare, and that such cases should be carefully diagnosed as SDC [28,30], in order to ensure the reliability we attempted AR immunohistochemistry repeatedly on multi-step sections for cases when the first trial completely failed immunoreactivity.
To examine the presence of HER2 amplification, a FISH analysis was carried out for all 151 SDC cases.A 4-μm-thick paraffin section from each block was placed onto a glass slide and subjected to FISH.HER2 amplification was performed in accordance with the manufacturer's instructions using FISH HER2 PharmDx (Dako, Glostrup, Denmark), which contained both fluorescently-labeled HER2 gene and chromosome enumeration probe 17 (CEP17).

Evaluation of HER2 status
HER2 positivity was defined as either immunohistochemically 3+ or HER2 amplification according to the American Society of Clinical Oncology/ College of American Pathologists (ASCO/CAP) guidelines for breast cancer [27].Immunohistochemically, HER2 3+ was defined as circumferential membrane staining that was complete, intense, and >10% of tumor cells (Figure 2A).Regarding the FISH analysis, 100 non-overlapping, intact interphase tumor nuclei identified by DAPI staining were evaluated, and the HER2 gene (red signal) and CEP17 (green signal) copy numbers in each nucleus were assessed.Samples were considered to be amplified when the average copy number ratio (HER2/CEP17) was ≥ 2.0 in all nuclei evaluated, or when the HER2 signals formed a tight gene cluster (Figure 2B).

Assessment of immunohistochemistry
A case was considered to be positive for AR when ≥ 20% of tumor cell nuclei showed strong staining (Figure 3).The cases were regarded as positive for ERβ when intensely positive staining of the cell nuclei was seen in ≥ 1% of tumor cells [38].
The percentage of Ki-67-positive cells was determined by counting at least 1000 tumor cells, and then recorded as the Ki-67 LI.Ki-67 LI, a value of < 40% was considered to indicate Ki-67-low, while ≥ 40% was considered to indicate Ki-67-high (Figure 4).

Figure 1 :
Figure 1: Representative histologic features of salivary duct carcinoma case.(A) Dilated ductal structures with a papillary, "Roman-bridge," or solid growth accompanied by comedo necrosis.(B) Tubular and cribriform structures with scirrhous pattern.Note that carcinoma cells display large pleomorphic nuclei and abundant eosinophilic cytoplasm.

Figure 7 :
Figure 7: Kaplan-Meier survival curves of patients with salivary duct carcinoma.(A) Three-year progression-free survival

Table 2 : Univariate and multivariate analyses for the correlation of biomarker immunoprofile with clinical outcomes in patients with salivary duct carcinoma Bio- markers Overall survival Progression-free survival
Adjusted by age, gender, primary tumor site, TNM classification, first-line treatment, and histologic origin.