This study included 48 surgical samples identified as ACC of the salivary glands at the Department of Oral Pathology, Dental Hospital, Yonsei University from 1990–2015. This study was approved by the Institutional Review Board of the Faculty of Dental Sciences, Yonsei University, South Korea (2-2015-0031). The clinical information of all patients was collected retrospectively through medical records (Table 1). All cases were confirmed histologically in accordance with the guidelines of the latest World Health Organization classification of salivary gland tumors, based on hematoxylin and eosin-stained tissue sections.15 There was no age, sex, ethnicity, or tumor stage restriction on patient enrollment. Follow-up was calculated as the time from the first visitation at the institution for the primary tumor to the latest follow-up or deceased date.

The staining was performed on 4-µm-thick sections using an EnVision-HRP detection system (Dako, Carpinteria, CA, USA). All procedures were performed at room temperature. Sections were deparaffinized through a series of xylene baths and rehydrated in graded concentrations of alcohol. To retrieve antigenicity, the slides were steamed with 10 mmol/L of citrate buffer (pH 6.0; DakoREAL™). Tissue sections were treated with 3% hydrogen peroxide to block endogenous peroxidase activity, followed by incubation in 5% bovine serum albumin. The sections were then incubated with primary antibodies diluted with phosphate buffered saline for 90 min in a humid chamber. All information on primary antibodies is listed on Supplementary Table 1 (only online). The slides were then incubated with secondary EnVision reagent (Dako) for 30 min, followed by incubation with diaminobenzidine chromogen. Next, the sections were counterstained with Mayer hematoxylin for visualization. The phosphate-buffered saline solution without primary antibody treatment was applied as the negative control for all antibody staining. Human lymphocytes and normal salivary glands were used as positive controls for ATM staining as recommended by Villaruz, et al.16 A paraffin-embedded breast cancer cell line was used to determine appropriate dilution ratio and as a positive control for p53 and phospho-p53 (S15).

With a cutoff of >25% positive cells, positivity of nuclear Myb expression was determined according to previously reported guidelines,17 whereas “loss of ATM expression” was determined as tumors showing >25% ATM-negative cells.18 Because mutant p53 proteins generally have longer half-lives than wild-type p53 protein, which leads to their nuclear accumulation, tumors showing >20% nuclear p53 or phospho-p53 (S15)-positive tumor cells were determined as having positive p53 or phospho-p53 (S15) expression.19

Correlation between clinical findings and patient survival was analyzed by Fisher's exact test. Cumulative survival probabilities were estimated using the Kaplan-Meier method. All statistical analyses were performed using statistical software (SPSS version 21.0, IBM Corp., Armonk, NY, USA), and the level of statistical significance was set at p<0.05.

The clinicopathological characteristics of the patients are listed in Table 1. The age range of the patients was 26–71 years; median age at diagnosis was 55 years. Among the 48 patients, 25 (52.1%) were male; 26 (54.2%) were younger than 55 years old. Thirty-six tumors (75%) were found in minor salivary glands, and most of them (47.9%) were in the palate, followed by the submandibular gland (16.7%), floor of the mouth (12.5%), and parotid glands (6.3%). Histologically, 13 tumors (27.1%) were diagnosed as ACC solid type; 35 tumors (72.9%) were diagnosed as ACC tubular and cribriform type. perineural invasion was found in 29 cases (60.4%) of ACC samples.

The median follow-up period for the patient population was 5 years. During that time, local recurrence and distant metastasis were identified in two (4.2%) and 18 (37.5%) of 48 patients, respectively. Five (10.4%) of the 48 patients died during follow-up (Table 2). Of the deceased patients, four (80%) were older than 55 years, and all tumors were found in minor salivary glands: two of them were solid type, three of them had perineural invasion, and four of them showed distant metastasis. Since the hallmarks of poor prognosis are solid histological type, perineural invasion, and distant metastasis,2 Kaplan-Meier analysis was performed to evaluate whether these factors affected prognosis. Histological type and perineural invasion were not significantly correlated with survival rate, whereas distant metastasis was found to be significantly correlated with poor survival (p=0.044) (Supplementary Fig. 1A-C, only online).

All 48 ACC patient samples were screened for Myb staining. Myb expression was not detected in nontumoral salivary gland parenchyma. Overall, 41 of 48 ACC samples (85.4%) were positively stained for Myb. Myb expression was detected in the nucleus and ranged from negative to strongly positive in intensity. In tubular foci, positive signals were predominantly found in myoepithelial cells (Fig. 1). Myb expression did not significantly differ depending on the anatomic site of tumors or clinical features of the patients. However, Myb expression differed significantly based on the histological type of ACC (Table 1). Thirty-three cases of tubular and cribriform ACC (94.3%) showed Myb positivity, whereas eight cases of solid type ACC (61.5%) showed Myb positivity (p=0.011), according to Fisher's exact test. Myb expression had no significant effect on survival rate when Kaplan-Meier analysis was performed (Supplementary Fig. 1D, only online).

Antibody specificity was validated using human lymph node tissue as a positive control. For negative control, the entire procedure was conducted with the omission of ATM antibody (Fig. 2A and B). For normal salivary glands, ATM protein was detected in the nucleus of acinar and ductal cells (Fig. 2C). ATM expression was reduced/absent in cancer cells and stromal fibroblasts in 58.3% and 60.4% of patients, respectively, compared with that in adjacent normal tissue (Fig. 2D-G). ATM levels in cancer cells were negatively correlated with age (p=0.003). Moreover, perineural invasion was significantly associated with the loss of ATM expression in stromal fibroblasts (p=0.008), according to Fisher's exact test, as shown in Table 3. Kaplan-Meier analysis revealed that low ATM expression in cancer cells was correlated with poor survival (p=0.037). Low ATM expression in stromal fibroblasts, however, was not significantly related to patient survival (Fig. 3A and B). Furthermore, low ATM expression, either in cancer cells or stromal fibroblasts, was not significantly correlated with metastasis-free survival (Fig. 3C and D).

Of the 48 patients, 38 (79.2%) had p53 expression and 30 (62.5%) had phospho-p53 (S15) expression (Fig. 4). No positive correlation was observed between the expression patterns of these proteins and clinical findings, except for histological type, according to Fisher's exact test (p=0.016) (Table 3). Although statistical significance was not attained, positive p53 expression or negative phospho-p53 (S15) expression showed tendencies of being correlated with poor survival rate (Fig. 5A and B).

Upon analyzing ATM expression stratified by positive p53 expression or negative phospho-p53 expression, loss of ATM and positive p53 expression significantly reduced overall survival (p=0.017) (Fig. 5C). Loss of ATM and negative phosphop53 expression was associated with worse prognosis, although statistical significance was not attained (Fig. 5D). In addition, loss of ATM, given positive p53 or negative phospho-p53 expression, reduced metastasis-free survival, although statistical significance was not attained (Supplementary Fig. 2, only online). As shown in Table 3, all five deceased patients exhibited a loss of ATM expression, as well as positive p53 and negative phospho-p53 expression, confirming that these markers can predict poor prognosis in ACC patients.