CD44⁺/CD24^- phenotype contributes to malignant relapse following surgical resection and chemotherapy in patients with invasive ductal carcinoma

A total of 147 invasive ductal carcinoma samples were randomly selected from our tissue database. Patients had been treated at the Peking Union Medical College Hospital between April 2000 and December 2007. None of these patients had received neoadjuvant chemotherapy or radiotherapy. Clinical information was obtained by reviewing preoperative and perioperative medical records, or by telephone or written correspondence. Patients were staged based on the tumor-node-metastases (TNM) classification of the International Union Against Cancer, revised in 2002.[15] The use of these human materials in this study was approved by the Peking Union Medical College Hospital Medical Ethics Committee.

Patient clinical characteristics are shown in Table 1. Fresh-frozen tumor tissue samples were used for routine determination of estrogen receptor (ER), progestogen receptor (PR), and human epidermal growth factor receptor (Her2). Paraffin specimens of these tumors were collected and 5 mm thick tissue sections were cut and fixed onto silicified slides. Each sample was stained with hematoxylin and eosin (H&E) and histologically typed according to the World Health Organization (WHO) classification [16]. Tumor size and the number and location of metastatic lymph nodes were obtained from pathology reports. Basal-like features of tumor was defined as immunohistochemically negative for both SR and Her2.

Briefly, each tissue section was deparaffinized, rehydrated and incubated with fresh 3% hydrogen peroxide (H₂O₂) in methanol for 15 min. After rinsing with phosphate-buffered saline (PBS), the samples were immersed in 0.01 M sodium citrate buffer (pH 6.0) and heated in a microwave oven at 100 °C for 15 min for antigen retrieval. Non-specific binding was blocked by incubating the sections with normal goat serum for 15 min at room temperature. The samples were subsequently incubated at 4 °C overnight with different primary antibodies. The primary antibodies used included rabbit polyclonal antibody to CD44 (CD44v6, IgG, 1:50, Abcam, Cambridge, UK), mouse monoclonal to CD24 (IgG, 1:50, Thermo Electron Corp., Burlington, ON, CA), FITC linked mouse monoclonal antibody to SABC (1:50), and goat anti-rabbit Cy3 antibody (IgG, 1:20). CD44 was detected with permanent red and CD24 was detected with diaminobenzidine. ALDH1 was detected with a monoclonal rabbit anti-ALDH1 antibody (ALDH1A1, IgG, 1:100, Abcam) followed by EnVision™ on a Tech-Mate™ (DAKO). All slides were counterstained with hematoxylin to identify nuclei. All samples were scored twice by one person in a blinded fashion, with all unclear results discussed with a pathologist. If there were staining discrepancies among the three cores from the same patient, an average was used. CD44 staining was detected mainly in the membrane and CD24 staining was detected mainly in the cytoplasm. The proportion of CD44+/CD24- tumor cells was defined as the percentage of cells positive for permanent red staining but negative for diaminobenzidine staining. The results of CD44+/CD24- tumor cells proportion were classified into two groups, high and low, with a cut-off value based on the median value of their proportion.

All calculations were performed using SPSS V.14.0 statistical software (Chicago, IL, USA). Associations between the presence of CD44, CD24 or different CD44/CD24 phenotypes and clinical variables as well as breast cancer subgroups were assessed by Fisher's exact test, except for age where the Mann–Whitney U test was used. Multivariate analysis was performed using Cox proportional hazards regression to determine the prognostic effect on disease-free survival (DFS) and overall survival (OS), and the log-rank test to compare survival between two strata. Hazard ratios (HR) and their corresponding 95% confidence intervals (CI) were computed to provide quantitative information about the relevance of the results of statistical analysis. All tests were two-sided and P < 0.05 was considered statistically significant.

The baseline characteristics of the study population are given in Table 1. All patients were female, with a mean ± standard deviation (SD) age of 51.6 ± 12.5 years (range, 13.5 to 80.7 years) and a mean ± SD tumor size of 3.1 ± 1.8 cm (range, 0.4 to 9.5 cm). Lymph node involvement was positive in 115 patients (78.2%). According to TNM classification, 14 patients (9.5%) were stage I, 56 (38.1%) were stage II, 76 (51.7%) were stage III, and 1 (0.7%) was stage IV. Of the 147 patients, 57 (38.8%) were positive for ER expression, 64 (43.5%) were positive for PR, 70 (47.6%) were positive for Her2, and 39 (26.5%) were positive for basal-like features (defined as immunohistochemically negative for both SR and Her2). Of the 147 patients, 87 (59.2%) were received adjuvant chemotherapy and 95 (64.6%) were received agents targeted against estrogen receptor. Median follow-up time was 23.0 months (range, 2 to 91 months), during which 40 patients (27.2%) experienced tumor recurrence and 51 (34.7%) developed metastases.

The presence of CD44 and CD24 antigens on invasive ductal carcinoma tissues was analyzed using double-staining immunohistochemistry. Figure 1 displays representative staining patterns of CD44 and CD24. CD44 was visible primarily as membranous permanent red staining, with only eight tumors displaying cytoplasmic and membranous staining. CD24 was visible mainly as cytoplasmic diaminobenzidine staining, with only six tumors displaying membrane diaminobenzidine staining. To determine the proportion of tumorigenic CD44+/CD24- cells within each tumor, we scanned for the presence of permanent red staining without any diaminobenzidine interference. CD44+/CD24- tumor cells were present in 103 of the 147 (70.1%) tumors, but absent from the other 44 (29.9%), with the proportion of tumor cells expressing this phenotype ranging from a few to 70%, with a median proportion of 5.8%, and this median proportion was selected to categorize patients as CD44+/CD24- tumor cells high group and CD44+/CD24- tumor cells low group according to cutoff definition. The frequency of tumors with different proportions of CD44+/CD24- tumor cells is presented in Table 2. The proportions of CD44+/CD24- tumor cells in clinical specimens correlated significantly with lymph node involvement (P = 0.026) and PR expression (P = 0.038). Higher proportions of CD44+/CD24- tumor cells were observed in specimens from patients with (19.20%) than without (8.66%) lymph node involvement and with (21.06%) than without (13.09%) PR expression. There were no associations between percentage of CD44+/CD24- tumor cells and patient age, tumor size, TNM stages, ER or Her2 expression, and basal-like features.

Of the 121 samples with CD44/CD24 data, 56 (46.2%) were positive for PR expression. CD44+/CD24- status was significantly correlated with strong PR staining in all patients (P = 0.038) and in samples from patients with recurrence or metastasis (P = 0.046). Interestingly, although ER expression was observed in 50 of the 121 (41.3%) patients with CD44/CD24 data, the presence of CD44+/CD24- tumor cells was not significantly correlated with positive ER expression in all patients and in patients with recurrence or metastasis.

We found that the proportion of CD44+/CD24- tumor cells was similar in breast cancer samples with and without basal-like features (11.93% versus 18.44%, p = 0.143). However, in samples from patients with tumor recurrence or metastasis, the proportion of CD44+/CD24- tumor cells was significantly higher in breast cancer tissue with basal-like features than in tissue without such features (22.66% versus 17.70%, p = 0.05).

The results of univariate analyses of the associations between each individual predictor and DFS are shown in Table 3. The proportion of CD44⁺/CD24^-/low tumor cells (P = 0.002), PR status (P = 0.004), basal-like feature (P = 0.007), and TNM stage (P = 0.029) were strongly correlated with DFS. Kaplan-Meier analysis showed that the presence of CD44⁺/CD24^-/low tumor cells was significantly associated with shorter DFS compared with the absence of CD44⁺/CD24^-/low tumor cells (22.9 ± 2.2 months versus 35.9 ± 3.8 months; Pearson chi-square, 10.696, p = 0.001; Figure 2A). When all predictors were included in a Cox model (multivariate analysis, Table 3), the presence of CD44⁺/CD24^-/low tumor cells (hazard ratio, 1.931; P = 0.011), PR status, basal-like feature, and TNM stage retained their prognostic significance for DFS.

Meanwhile, the results of univariate analyses of the associations between each individual predictor and OS are shown in Table 4. Similarly with the relation with DFS, the proportion of CD44⁺/CD24^-/low tumor cells (P = 0.001), basal-like feature (P = 0.029), and TNM stage (P = 0.027) were strongly correlated with OS. Kaplan-Meier analysis showed that the presence of CD44⁺/CD24^-/low tumor cells was significantly associated with shorter OS compared with the absence of CD44⁺/CD24^-/low tumor cells (39.3 ± 2.6 months versus 54.0 ± 3.5 months; Pearson chi-square, 12.140, p = 0.001). When all predictors were included in a Cox model (multivariate analysis, Table 4), the presence of CD44⁺/CD24^-/low tumor cells (hazard ratio, 2.237; P = 0.002), basal-like feature, and TNM stage retained their prognostic significance for OS.

We separately analyzed the secondary lesions from 56 patients with invasive ductal carcinoma and metastasis or recurrence. We found that a significantly higher proportion of secondary than primary lesions were positive for CD44⁺/CD24^-/low tumor cells (26.9% versus 7.0%, P < 0.05).