One hundred CMT samples were collected from 100 bitches underwent regional or unilateral mastectomy and ovariohysterectomy in the Veterinary Teaching Hospital of China Agricultural University from September 2012 to October 2014. Normal mammary gland tissues (NMGTs) were collected (at necropsy) from 8 intact healthy dogs that were used for general surgery practice courses for undergraduate students at the College of Veterinary Medicine. These procedures were approved by the Animal Care Committee of China Agricultural University (ID: 1114120800096). Survival of CMT patients was determined by telephoning the dog owners one year after the surgery.

After routine histologic processing and hematoxylin-eosin staining, CMT samples were evaluated by a board-certified pathologist and classified and graded as described by Goldschmidt et al. [5] and Peña et al. [18], respectively. The presence of necrosis and lymphatic invasion in the histological sections was also recorded.

A custom RT² Profiler PCR Array (CAFP12858; Qiagen, Germany) that could simultaneously detect 30 genes related to the Wnt signaling pathway in canines (Table 1) was used. Each array plate contained 2 sets of 48 wells for 2 tests. Genomic DNA contamination, reverse transcription, and positive PCR controls were included in each 48-well set on each plate. Glyceraldehyde-3-phosphate-dehydrogenase (GAPDH) was used as the assay reference gene.

Thirty-five CMT samples (see Table 2) and 5 NMGT samples were used in the assays. Total RNA was isolated from each frozen tissue sample using TRIzol Reagent (Life Technologies, USA) and quantified with Nanodrop 2000 (Thermo Fisher Scientific, USA). Then total RNA (1 µg) was reverse transcribed in a final volume of 20 µL with an RT² First Strand Kit (catalog No. 330401; Qiagen). The RT² Profiler PCR Array tests were performed following the instructions of the manufacturer. Briefly, all cDNA in each sample (20 µL) was first diluted with RNase-free water to a volume of 111 µL. Then 51 µL of the diluted cDNA was mixed with 675 µL 2 × RT² SYBR Green ROX qPCR Mastermix (catalog No. 330523; Qiagen) and 624 µL RNase-free water. Of that mixture, 25 µL per well was added into the 48 wells of the array plate. The qPCR was carried out using the Applied Biosystems 7500 Real-Time PCR System (software ver. 2.0.6; Applied Biosystems, USA) under the following thermal cycling conditions: 10 min at 95℃, followed by 40 cycles of 15 sec at 95℃ and 1 min at 60℃. The exported Ct values were input to a template Excel file provided by SABiosciences (Qiagen) and uploaded for the online analysis. After data review, qualified data from 35 CMT and 5 NMGT samples were analyzed by applying the 2^−ΔΔCt method.

Paraffin sections of CMT and NMGT samples were dewaxed in xylene and subsequently rehydrated through an ethanol series. Antigen retrieval was carried out by microwave treatment in 0.1 M citrate buffer (pH 6.0) for β-catenin or EDTA buffer (pH 9.0) for cyclin D1. Endogenous peroxidase activity was blocked with 3% hydrogen peroxide for 20 min. Then normal goat serum was applied on the slides to block non-specific binding for 20 min. Rabbit anti-β-catenin polyclonal antibodies (1:5,000 dilution, ab6302; Abcam, USA) or rabbit anti-cyclin D1 polyclonal antibodies (1:1,000 dilution, ab185241; Abcam) were applied on the sections overnight at 4℃. The sections were visualized by using the Polink-2 HRP plus rabbit DAB detection system (Golden Bridge International, USA) and counterstained with hematoxylin. Negative controls in which PBS replaced the primary antibody were set at the same time.

The IHS of β-catenin was evaluated by determining the percentage of positive cells and the intensity of the cytoplasmic staining in accordance with the description of Khramtsov et al. [11]. The scores for both membrane-associated and cytoplasm-associated β-catenin ranged from 0 to 3. The IHS of cyclin D1 was assessed by determining the percentage of stained nuclei. The staining was scored as 0 when the sample had no nuclear staining or if < 10% of its nuclei were stained; whereas tumor samples with > 10% of its nuclei stained were scored as 1 [17]. The percentage of positive staining was estimated by counting no less than 500 cells in 3 randomly selected high-power fields.

Eleven CMT samples (including 4 benign CMTs and 7 malignant CMTs) and 1 NMGT sample were selected for the detection of expressions of β-catenin, TCF7, LEF1, and cyclin D1. Proteins were extracted with RIPA lysis buffer (Macgene, China) and the soluble protein was resolved by performing electrophoresis through 12% SDS-PAGE gels followed by transferring to polyvinylidene difluorid membranes by semi-dry blotting. Non-specific binding sites were blocked by incubation for 1 h in 5% w/v milk powder in PBS, after which membranes were sliced and separately probed with rabbit polyclonal antibodies against β-catenin (diluted at 1:4,000; ab6302; Abcam), LEF1 (diluted at 1:600; ab83964, Abcam), TCF7 (diluted at 1:4,000; ab30961, Abcam), or cyclin D1 (diluted at 1:1,000; ab185241, Abcam). HRP conjugated goat anti-rabbit IgG or goat anti-mouse IgG were used as the secondary antibody. Blots were developed in ECL reagent (Macgene) and exposed to X-ray film.

For statistical purposes, Grade I tumors were assigned to the low malignant (LM) group, while Grades II and III tumors were clustered together and assigned to the highly malignant (HM) group. Similarly, some tumor subtypes were grouped together based on similarities in morphology and/or biological behavior, leading to the formation of 4 subtype groups (detailed in Table 2). SPSS (ver. 20; IBM, USA) was used for statistical analysis. A one-way ANOVA with Tukey's post hoc test was performed for analysis of the fold changes of genes in the RT² Profiler PCR Array test results. A chi-squared test and Fisher's exact test were used for analysis of the associations of β-catenin or cyclin D1 expressions with clinicopathological features. A significant difference was expressed as ^*p < 0.05 while a highly significant difference was expressed as ^**p < 0.01.

Histopathological examination of the 100 CMT samples revealed 37 benign tumors and 63 malignant tumors, which were distinguished in four subtypes as summarized in Table 2. Among the 100 tumors, 37 were included in subtype group 1, 17 in subtype group 2, 10 in subtype group 3, and 36 in subtype group 4. Among the 63 malignant tumors, 30 were classified as LM and 33 as HM. Necrosis and lymphatic invasion were detected in 41 and 9 malignant cases, respectively. The 8 NMGT samples were confirmed to be non-neoplastic, with only mild hyperplastic changes (lobular hyperplasia) detected in 4 samples.

Based on the grading and classification of the CMT samples (Table 2), 9 benign tumors, 26 malignant tumors, and 5 NMGT samples were subjected to RT² Profiler PCR Array assays. Among the 30 simultaneously detected genes, 9 exhibited significant changes in expression (Fig. 1). Expression of WNT5A was significantly upregulated in LM tumors (panel A in Fig. 1). Expression of DKK1 was correlated with the degree of malignancy, showing a 16-fold change in LM tumors and a 29-fold change in HM tumors (panel A in Fig. 1). In contrast, SFRP4 was significantly downregulated in both LM and HM tumors (7.1-fold and 7.0-fold changes, respectively, panel B in Fig. 1). For β-catenin and the destruction complex-related factors investigated in this study, only β-catenin showed an obviously upregulated expression in HM tumors (6.8-fold increase, panel D in Fig. 1). A significant difference was detected in the expression of transcription factors TCF7 and LEF1 between the benign tumor and the NMGT samples (panel E in Fig. 1). Expression of the Wnt target gene MMP7 was increased in the benign CMTs (20-fold change) while it decreased in LM tumors and HM tumors (9-fold and 6.5-fold changes, respectively; panel F in Fig. 1). The data were reanalyzed after grouping the tumor samples into 4 subtypes, and the results of the analysis showed that, compared with the expressions in other subtypes and NMGTs, expressions of SFRP1 and DKK1 were highest in carcinoma and malignant myoepithelioma (CMM) subtypes. The expression pattern of β-catenin was similar to that of the SFRP1 and DKK1 (panel G in Fig. 1).

Expressions of β-catenin and cyclin D1 were further evaluated by performing immunohistochemical analysis. Compared to NMGTs (panel H in Fig. 2), significantly reduced membrane expression of β-catenin (IHS as 0 and 1) was observed in all tumor samples (panel A–G in Fig. 2; p = 0.000). Most (8/10) CMMs also showed weak to moderate intensity staining of cytoplasm (panel A in Fig. 2); while moderate to strong cytoplasm staining of β-catenin (IHS as 2 and 3) was detected in the other CMT subtypes (panel B–G in Fig. 2). Expression of β-catenin in the cytoplasm was significantly higher in CMTs than in NMGTs (Table 3). There was an obvious shift of β-catenin localization from membrane to cytosol in the comparison of CMTs with NMGTs (p = 0.000; Fig. 3). Nuclear expression of cyclin D1 (scored as 1) was observed in 84% (84/100) of CMTs (25 benign and 59 malignant tumors; panel I–K in Fig. 2), while its expression was absent or less than 10% (scored as 0) in the nuclei of NMGTs (panel L in Fig. 2), a statistically significant difference (p = 0.000; Table 3). There was significantly higher nuclear expression of cyclin D1 in malignant tumors than in benign tumors (p = 0.001; Table 3). With regard to the expression of β-catenin in cytoplasm and the nuclear staining of cyclin D1, there were no significant differences detected among the different histological grades, between presence/absence of necrosis or lymphatic invasion in malignant CMTs, and between death or survival at one year after surgery (Table 3).

The expression of β-catenin increased significantly in 10 of 11 detected CMTs; the exception, a decreased expression, was detected in a case with simple tubulo-papillary carcinomas. Increased LEF1 expression was detected in complex adenomas, intraductal papillary carcinomas, and solid carcinomas, while no (or slight) increases of LEF1 expression were detected in simple adenomas, complex carcinoma, and 1 simple tubulopapillary carcinomas (Grade I). The expression levels of TCF7 were similar to those of LEF1 in CMTs. Decreased cyclin D1 expression was only detected in a case of simple tubulo-papillary carcinomas (Fig. 4).