Correlated Expression of CD47 and SIRPA in Bone Marrow and in Peripheral Blood Predicts Recurrence in Breast Cancer Patients

The numerous efforts in breast cancer research and care have improved early detection and treatment. However, breast cancer prevalence and mortality remain at a high level every year. The prevention and therapy of breast cancer among Japanese women is a crucial public health concern. The most recent statistics for Japan document over 55,000 cases per year (1), with a mortality surpassing 12,000 per year (2). Even after apparently successful localized treatments, there are long-term risks of recurrence and metastasis. To evaluate the postsurgical risk of recurrence of breast cancer, mammography, echogram, computer tomography, and magnetic resonance imaging are utilized for diagnostic imaging, and carcinoembryonic antigen (CEA), CA15-3, and NCC-ST439 are evaluated in peripheral blood as tumor markers. However, the long-term risk of relapse is largely due to clinically occult microrecurrences and micrometastases that are currently beyond detection by current conventional screening strategies. Therefore, it is important to exploit novel tumor markers that could predict recurrence and metastasis with greater reliability.

Kaplan et al. have shown that bone marrow–derived hematopoietic progenitor cells play an important role in the accumulation of premetastatic niches and the promotion of carcinogenesis and metastasis (3). We confirmed their findings by using clinical samples in which hematogenous metastasis occurred, and within these metastases, hematopoietic progenitor cells and isolated tumor cells (ITC) coexisted. This study showed the necessity of identifying metastasizing cancer cells as well as normal host side factors, such as bone marrow–derived cells and endothelial cells (4).

Recent studies showed that CD47 specifically inhibits phagocytosis and that there was a significant correlation between gene expression of CD47 and leukemia, hematopoietic stem cells, and tumor-initiating cells of bladder cancer (5–7). It can be inferred from the association between CD47 and cancer stem cells that ITC would elude the immune system by taking advantage of activation and initiation of the signal transduction cascade of CD47, resulting in inhibition of phagocytosis.

CD47 was originally identified in association with the integrin α_vβ₃, hence its alternative name of integrin-associated protein. It is also a member of the Ig superfamily, possessing a V-type Ig-like extracellular domain, five putative membrane-spanning segments, and a short cytoplasmic tail (8). CD47 seems to carry out several functions. For instance, CD47 functions as a marker of “self” on murine RBC. Erythrocytes lacking CD47 expression are rapidly removed from the bloodstream by splenic red pulp macrophages (9, 10). Signal regulatory protein α (SIRPA), a transmembrane glycoprotein, is a novel intracellular signal transducer when it is engaged by its ligand, CD47. CD47 on normal peripheral blood red cells circumvent elimination by binding to SIRPA (10). The interaction of CD47 with SIRPA occurs between host-derived cells, and is mostly related to cell signaling in the immune and nervous systems (11).

In the present study, we confirmed that high expression of CK19, a marker for ITC, had correlation with high CD47 expression. Therefore, we focused on the clinicopathologic significance of CD47 gene expression in the bone marrow and peripheral blood of breast cancer and its potential utility as a novel and specific biological marker for recurrence and/or overall survival (OS) in breast cancer patients. Investigating the characteristics of CD47 mRNA expression in bone marrow and peripheral blood in this study, we evaluated the correlation between CD47 and clinicopathologic factors in 738 breast cancer cases, and showed that the magnitude of CD47 expression could be used as a new prognostic marker for recurrence and metastasis. Moreover, we found that expression of CD47 and SIRPA were correlated in bone marrow and peripheral blood. This association has potentially important implications for clinicopathologic outcome. We suggest that the correlated expression of CD47 and SIRPA represents a dynamic process involved in the progression of breast cancer cells. This report is one of the first to show that a host factor in bone marrow confers prognostic importance.

A total of 738 breast cancer patients were identified based on their pathologic diagnosis before surgery at Kyushu Cancer Center from July 2000 to August 2005. Written informed consent was obtained from all patients according to the guidelines approved by the Institutional Research Board. Patients ranged in age from 24 to 89 years, with a mean age of 55.1 years. No patients received antihormonal treatment, chemotherapy, or radiotherapy before surgery. All patients were closely followed after surgery at regular 3- to 6-month intervals, and the follow-up periods ranged from 2 months to 6 years, with a median of 3.0 years. After surgery, all patients were clearly classified into the category of breast cancer based on the clinicopathologic criteria described by the Japanese Society for Breast Cancer. All data, including age, menopause, tumor stage, lymphatic invasion, lymph node metastasis, vascular invasion, distant metastasis, clinical stage, estrogen receptor (ER), progesterone receptor (PgR), human epidermal growth factor receptor 2 (Her2) score, and recurrence were obtained from the clinical and pathologic records. ER, PgR, and Her2 scores were obtained from immunohistochemistry staining conducted by two well-trained pathologists. Her2 status was scored using the Her2 expression criteria (Supplementary Table S1). These criteria changed in May 2009, but we applied the previous criteria to evaluate Her2 status. The primary tumors with Her2 score (2+) had their immunohistochemistry results additionally validated by fluorescence in situ hybridization. After surgical therapy, all patients were individually treated by antihormonal treatment, chemotherapy, and/or radiotherapy according to breast cancer treatment guidelines in Japan, which were based on American Society of Clinical Oncology and National Comprehensive Cancer Network recommendations. Regarding noncancer patients, we considered that inflammatory and neoplastic diseases, including benign tumors, might affect the result of our experiments, and we therefore excluded those patients from this group. We selected 19 patients who underwent surgery for elective cholecystolithiasis at the Medical Institute of Bioregulation Hospital, Kyushu University, between 1999 and 2003; all patients had a blood test before surgery, and they were confirmed to be without inflammatory symptoms. Written informed consent was obtained from all patients. All control patients had a whole body computer tomography examination to determine whether they had cancer, and their status was confirmed by assessing tumor markers in the peripheral blood. After surgery, they were followed at regular 6-month intervals, and the absence of cancer was confirmed over 3 years following surgery.

Bone marrow and peripheral blood samples were obtained from patients under anesthesia before surgery. Peripheral blood samples were taken from superficial veins on the opposite side of the breast cancer, and bone marrow samples were taken from the sternum with a 15G needle. As there was a potential for contamination by skin, bone marrow specimens were collected with another syringe after the first 2 to 3 mL were aspirated. ISOGEN-LS (Nippon Gene Co., Ltd., Japan) was added and mixed, stored for 5 minutes at room temperature, and was immediately frozen in liquid nitrogen and stored at −80°C until RNA extraction. Samples from noncancer patients were obtained with the same procedure.

The breast cancer cell lines CRL1500, MCF7, MRK-nu1, YMB1, YMB1E, SKBR3, and MDA-MB-231 were obtained from the Cell Resource Center for Biomedical Research Institute of Development, Aging and Cancer (Tohoku University, Japan).

Total RNA from bone marrow and peripheral blood was extracted from control and breast cancer patients using the ISOGEN-LS method followed by Isogen-chloroform extraction and isopropanol precipitation (12). As described previously, cDNA was synthesized from 8.0 μg of total RNA (13).

The sequences of CD47 primers were as follows: sense primer, 5′-GGCAATGACGAAGGAGGTTA-3′; antisense primer, 5′-ATCCGGTGGTATGGATGAGA-3′. The sequences of SIRPA primers were as follows: sense primer, 5′- GTTTAAGTCTGGAGCAGGCACT-3′; antisense primer, 5′-GCAGATGACTTGAGAGTGAACG-3′. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control and the sequences of GAPDH primers were as follows: sense primer, 5′-TTGGTATCGTGGAAGGACTCTA-3′; antisense primer, 5′-TGTCATATTTGGCAGGTT-3′. cDNA was synthesized from 8.0 μg of total RNA. Real-time monitoring of PCR reactions was done using the LightCycler system (Roche Applied Science) and SYBR-Green I dye (Roche Applied Science). Monitoring was done according to the manufacturer's instructions. Quantitative reverse transcriptase-PCR (RT-PCR) was done with the following cycling conditions: initial denaturation at 95°C for 10 minutes, followed by 40 cycles of 95°C for 10 seconds, annealing at 60°C for 10 seconds, and extension at 72°C for 10 seconds. After amplification, products were subjected to a temperature gradient from 68°C to 95°C at 0.2°C/second, under continuous fluorescence monitoring, to produce a melting curve of the products. All concentrations were calculated relative to the concentration of cDNA from Human Universal Reference total RNA (Takara Bio Inc., Japan). The concentrations of CD47 and SIRPA mRNAs were then divided by the concentration of the endogenous reference (GAPDH) to obtain normalized expression values (14–16). Each assay was done twice to verify the results, and the mean normalized value of mRNA expression was used for subsequent analyses.

For continuous variables, data were expressed as means ± SD. The relationship between CD47, SIRPA mRNA expression and clinicopathologic factors was analyzed using the χ² test and Student's t-test. In addition, the data were also analyzed using the nonparametric Wilcoxon rank-sum test. Survival curves were plotted according to the Kaplan-Meier method, and the generalized log-rank test was applied to compare the survival curves. Variables with a value of P < 0.05 in univariate analysis were used in a subsequent multivariate analysis using the Cox regression. All tests were analyzed using JMP 7 software (SAS version 7.0.1, SAS Institute, Inc.), and the findings were considered significant when P < 0.05.

In our study, we assessed CK19 expression in the bone marrow and peripheral blood of 738 breast cancer patients with quantitative real-time RT-PCR analysis. We detected 57 CK19-positive cases in bone marrow and 57 cases in peripheral blood. We divided patients into two groups: the CK19-negative group and the CK19-positive group. We then evaluated CD47 expression in the bone marrow and peripheral blood of 738 breast cancer patients with quantitative real-time RT-PCR and compared CD47 expression with CK19 status. CD47 expression in the CK19-positive group was higher than that in CK19-negative group (bone marrow, P = 0.04; peripheral blood, P = 0.02; Supplementary Fig. S1). CD47 expression (mean ± SD) in bone marrow was 1.91 ± 2.06 [confidence interval (CI), 1.54-2.28] in the CK19-positive group and 1.51 ± 1.34 (CI, 1.39-1.63) in the CK19-negative group. CD47 expression in peripheral blood was 2.65 ± 3.03 (CI, 2.12-3.18) in the CK19-positive group and 1.97 ± 1.74 (CI, 1.74-2.19) in the CK19-negative group.

In this study, we selected 19 patients with cholecystolithiasis as control cases. Quantitative real-time RT-PCR analysis showed higher expression of CD47 mRNA in breast cancer cases than in control cases (Fig. 1A-1). In bone marrow, the mean expression ratio (mean ± SD) of CD47/GAPDH mRNAs in breast cancer, 1.52 ± 1.43 (CI, 1.42-1.63), was significantly higher than that in noncancer cases, 0.80 ± 0.65 (CI, 0.49-1.11; Wilcoxon Rank-Sum test, P = 0.033). In peripheral blood, the mean expression ratio (mean ± SD) of CD47/GAPDH mRNAs in breast cancer, 1.83 ± 1.69 (CI, 1.67-1.99), was significantly higher than that in noncancer cases, 0.48 ± 0.49 (CI, 0.24-0.73; P = 0.0015). In Fig. 1A-2, the histogram shows the number of cases within each range of expression ratios of CD47/GAPDH. We also evaluated CD47 mRNA in 32 noncancer patients; the results of CD47 expression were similar to that from 19 patients as normal controls (Supplementary Fig. S2-a).

We assessed CD47 expression in breast cancer cell lines to determine whether the expression ratio of CD47 is affected by the number of ITC in bone marrow or peripheral blood. The human breast cancer cell lines CRL1500, MCF7, MRK-nu1, YMB1, YMB1E, SKBR3, and MDA-MB-231 were assessed. The mean expression ratio of CD47/GAPDH mRNA was 0.24 ± 0.18 (CI, 0.07-0.40) in breast cancer cell lines, and was significantly lower than those found in the bone marrow and peripheral blood samples of breast cancer cases. Next, we compared the CD47/GAPDH ratio of each histologic type of breast cancer in bone marrow (Fig. 1B). We divided patients into three types: ductal carcinoma in situ (DCIS; n = 53), invasive ductal carcinoma (IDC; n = 647), and other types of breast cancer (other; n = 38). The mean expression ratios of CD47/GAPDH mRNAs were as follows: DCIS, 0.68 ± 0.71 (CI, 0.47-0.88); IDC, 1.66 ± 1.47 (CI, 1.54-1.77); others, 1.09 ± 1.31 (CI, 0.69-1.49). The CD47/GAPDH expression ratio in IDC was significantly higher than in DCIS (P < 0.0001) and others (P = 0.03). Such variations in the CD47/GAPDH expression ratios were likely to be reflected in the bone marrow of IDC, so we focused on the IDC subtype for subsequent analyses.

Table 1 shows the results of tumor markers that were analyzed in peripheral blood samples taken before surgery. The mean values (mean ± SD) of tumor markers in breast cancer were as follows: CEA, 2.01 ± 3.37; CA15-3, 12.19 ± 11.19; and NCC-ST439, 7.45 ± 31.50. NCC-ST439 was only slightly higher than normal levels, whereas the values of CEA and CA15-3 were within normal levels. As a result, these tumor markers may not be useful as biomarkers in breast cancer cases before surgery.

Next, we divided breast cancer patients into two groups according to their CD47/GAPDH ratios. Thus, bone marrow and peripheral blood values were divided into those greater or less than the median CD47/GAPDH ratio: the high-expression group (n = 226) and the low-expression group (n = 226). In both bone marrow and peripheral blood, the mean expression ratios of CD47/GAPDH mRNAs in cancer cases were significantly higher than in control patients. Furthermore, the mean ratio of CD47/GAPDH mRNAs in the high-expression group of cancer cases was three to five times higher than in noncancer cases.

In this study, all the data were obtained from clinical and pathologic records. However, due to lack of data from breast cancer patients, particularly Her2 scores, we limited the analysis to the 452 clinical cases in which data were complete. Clinicopathologic significance of the CD47/GAPDH mRNA expression ratio in bone marrow is shown in Table 2. The incidence of triple negatives for ER, PgR, and Her2 status was significantly higher (P = 0.0097) in the high-expression group than in the low-expression group. The incidence of premenopausal patients was significantly higher (P = 0.01) in the high-expression group than in the low-expression group. The incidence of Her2 score was significantly higher (P = 0.03) in the high-expression group than in the low-expression group. The incidence of recurrence was significantly lower (P = 0.04) in the high-expression group than in the low-expression group. Conversely, no significant differences were observed regarding age, tumor stage, lymph node metastasis, lymphatic invasion, venous invasion, distant metastasis, clinical stage, estrogen receptor, and progesterone receptor.

The 5-year disease-free survival (DFS) and OS rates in patients with high CD47/GAPDH mRNA in bone marrow and patients with low CD47/GAPDH mRNA in bone marrow are shown in Fig. 1C. The survival difference between these two groups was statistically significant (DFS, P = 0.0035, log-rank test; OS, P = 0.015, log-rank test). The patients received at least one postoperative therapy (antihormonal treatment, chemotherapy, or radiotherapy). Univariate and multivariate analyses of clinicopathologic factors affecting DFS rate in bone marrow are shown in Table 3. Univariate analysis revealed a significant relationship between DFS and the following factors: lymphatic invasion, lymph node metastasis, venous invasion, estrogen receptor, progesterone receptor, Her2 score, and CD47 expression. Multivariate analysis indicated that the high expression ratio of CD47 was found to be an independent and significant prognostic factor for survival (P = 0.024). Univariate and multivariate analyses of clinicopathologic factors affecting OS rate in bone marrow are shown in Table 3. Univariate analysis revealed a significant relationship between OS and the following factors: menopause, lymph node metastasis, estrogen receptor, progesterone receptor, recurrence, and CD47 expression. Multivariate analysis indicated that the high expression ratio of CD47 was not an independent and significant prognostic factor for survival (P = 0.41).

The clinicopathologic factors analyzed in relation to CD47 mRNA expression in peripheral blood are shown in Table 2. The incidence of recurrence was significantly lower (P < 0.0001) in the high-expression group than in the low-expression group. The incidence of tumor stage was significantly lower (P = 0.0011) in the high-expression group than in the low-expression group. Conversely, no significant differences were observed regarding age, menopause, lymph node metastasis, lymphatic invasion, venous invasion, distant metastasis, clinical stage, estrogen receptor, progesterone receptor, Her2 score, and ER, PgR, Her2 status. The 5-year DFS and OS rates in patients with high CD47 mRNA and patients with low CD47 mRNA are shown in Fig. 1D. The survival difference between these two groups was not statistically significant for DFS (P = 0.18, log-rank test) and OS (P = 0.87, log-rank test). Univariate and multivariate analyses of clinicopathologic factors affecting DFS rate in peripheral blood are shown in Table 4. Univariate analysis revealed a significant relationship between OS and the following factors: tumor stage, lymphatic invasion, lymph node metastasis, venous invasion, estrogen receptor, progesterone receptor, and Her2 score, but CD47 expression was not included. Multivariate analysis indicated that the presence of lymph node metastasis was found to be an independent and significant prognostic factor for survival (P = 0.0055). Univariate and multivariate analyses of clinicopathologic factors affecting OS rate in peripheral blood are shown in Table 4. Univariate analysis revealed a significant relationship between OS and the following factors: menopause, lymph node metastasis, estrogen receptor, progesterone receptor, and recurrence. Univariate analysis indicated that the high expression ratio of CD47 was not an independent and significant prognostic factor for survival.

We investigated SIRPA expression in the same breast cancer patients and controls. Figure 2 shows the correlation between CD47 expression and SIRPA expression. In cancer cell lines, CD47 expression did not correlate with SIRPA expression (Spearman correlation = 0.0319; P = 0.95; data not shown). In control patients, CD47 expression was correlated with SIRPA in both bone marrow (P < 0.0001) and peripheral blood (P = 0.0044; Fig. 2A and B). In 32 noncancer cases, CD47 expression was correlated with SIRPA in both bone marrow (P = 0.004) and peripheral blood (P < 0.0001; Supplementary Fig. S2-b).

In breast cancer cases, CD47 expression is more strongly correlated with SIRPA in both bone marrow (P < 0.0001) and peripheral blood (P < 0.0001; Fig. 2C and D).

CD47 is expressed on the surface of a wide variety of cells such as hematopoietic cells, keratinocytes, and cells of the brain (17). CD47 is associated with α_vβ₃ integrin and is implicated in the modulation of integrin functions, such as cell adhesion, phagocytosis, and cellular migration (18–20). It is known that CD47 is a marker of self on RBC. CD47 could work as a marker of self on cancer cells, and breast cancer cells may express high levels of CD47 by themselves. Our results showed that the CD47/GAPDH expression ratio in breast cancer cell lines was significantly lower than those found in the bone marrow and peripheral blood samples of breast cancer cases. This may indicate that CD47 has various functions, and that the level of CD47 expression was affected by the cell environment rather than by the number of cancer cells. Therefore, the high expression of CD47 in the bone marrow and peripheral blood of breast cancer patients may represent the characteristic appearance of breast cancer and some evidence of a cancer-specific mechanism in the bone marrow and peripheral blood of breast cancer.

Recent reports have shown that CD47 plays a role in inhibiting macrophage phagocytosis of cancer stem cells and tumor-initiating cells (5–7). In the same manner as a cancer stem cell, the cancer cell itself may circumvent immune system surveillance by expressing CD47 as a marker of self, thereby evading natural killer cells (21, 22). In our study, we found that high CD47 expression had a correlation with high CK19 expression in the bone marrow and peripheral blood of breast cancer. This result strongly suggests that ITC of breast cancer patients may utilize the function of CD47 in circulating circumstances such as bone marrow and peripheral blood. Moreover, expression of CD47 in the bone marrow and peripheral blood of breast cancer patients was significantly higher than in control patients. Because expression of CD47 in circulating tumor cells increases exponentially with the progress of the cancer stage, CD47 derived from ITC may be an upregulating factor of breast cancer.

CD47 also promotes apoptosis, and the CD47 ligand thrombospondin (TSP) has been implicated as an antitumor and antimetastatic factor in breast cancer (23–33). Both TSP1 and a CD47 agonist peptide (4N1K, derived from TSP1) can induce a novel form of apoptosis in transformed and activated normal T cells (34, 35), chronic lymphocytic leukemia cells (36), erythroleukemia cells, and primary arterial smooth muscle cells (35). We supposed that CD47 may have a role not only as a marker of self but also as an inducer of apoptosis to inhibit phagocytosis.

In the present study, the mean ratio of CD47/GAPDH mRNAs in the high-expression group of cancer cases was three to five times higher than in noncancer cases. We suggest that CD47 may be specifically expressed in the bone marrow and peripheral blood of breast cancer patients and that CD47 expression may represent an important biomarker in breast cancer patients. As a result of the identification of the clinical significance of CD47 expression in bone marrow and peripheral blood, we found that overexpression of CD47 in bone marrow and peripheral blood correlated with the aggressiveness of breast cancer. This result might suggest that the more there are circulating tumor cells expressing increased CD47 in bone marrow and peripheral blood, the more active the primary immune system is in inducing apoptosis in tumor cells in the circulating systems. Therefore, it is important to clarify the level of CD47 expression in bone marrow and peripheral blood to indicate whether micrometastasis exists in the breast cancer cases. Thus, CD47 may be a novel biological marker that predicts the number of highly malignant circulating tumor cells that escape from the immune systems in breast cancer.

To further characterize the function of CD47 in bone marrow and peripheral blood, we examined SIRPA expression in the same breast cancer samples. In doing so, we obtained the novel finding that the expressions of CD47 and SIRPA are markedly associated. The correlation between CD47 and SIRPA was significantly stronger in breast cancer patients than in control cases. In control cases, the CD47-SIRPA signaling system is activated in bone marrow and in peripheral blood, reflecting homeostatic regulation in the hematopoietic system. In the breast cancer cases, carcinogenicity may promote the CD47-SIRPA cell signaling system in bone marrow and in peripheral blood, thereby possibly promoting micrometastases. We suggest that expression of the CD47/SIRPA signal system indicates the presence of cancer-specific microenvironmental areas that support micrometastasis.

In conclusion, our data indicate that CD47 is a significant prognostic indicator for DFS, and our study is one of the first to report a host factor in bone marrow with prognostic significance. With regard to patient care, many cases require postoperative adjuvant chemotherapy. Due to the associated adverse effects of such treatment, reliable prognostic markers for recurrence and metastasis would greatly improve patient management. We suggest that this biomarker may fill that need for enhanced patient care.

No potential conflicts of interest were disclosed.

We thank T. Shimooka, K. Ogata, M. Kasagi, Y. Nakagawa, and T. Kawano for their technical assistance, and Kelly K. Chong for her editorial assistance.

Grant Support: CREST, Japan Science and Technology Agency (JST); Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for Scientific Research, grant numbers 19591509, 19390336, 20390360, 20591547, 20659209, 20790960, 20790961 and 21791295.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.