HuR is a ubiquitously expressed mRNA-binding protein. Intracellular localization of HuR is predominantly nuclear, but it shuttles between the nucleus and the cytoplasm. In the cytoplasm it can stabilize certain transcripts. Because nucleocytoplasmic translocation of HuR is necessary for its activity, it was hypothesized that cytoplasmic HuR expression in cancer cells could be a prognostic marker. To test the significance of HuR in carcinogenesis of the breast, we have investigated HuR expression in a mouse mammary gland tumor model and from 133 invasive ductal breast carcinoma specimens. HuR expression was elevated in the cyclooxygenase-2 transgene–induced mouse mammary tumors, and its expression was predominantly cytoplasmic in the tumor cells. In the human carcinoma samples, high cytoplasmic immunoreactivity for HuR was found in 29% (38 of 133) of the cases. Cytoplasmic HuR expression associated with high grade (P = 0.0050) and tumor size over 2 cm (P = 0.0082). Five-year distant disease-free survival rate was 42% [95% confidence interval (95% CI), 26-58] in cytoplasm-high category and 84% (95% CI, 76-91) in cytoplasm-negative or -low category (P < 0.0001), and high cytoplasmic expression of HuR was an independent prognostic factor in a Cox multivariate model (relative risk 2.07; 95% CI, 1.05-4.07). Moreover, high cytoplasmic HuR immunopositivity was significantly associated with poor outcome in the subgroup of node-negative breast cancer in a univariate analysis (P < 0.0007). Our results show that high cytoplasmic HuR expression is associated with a poor histologic differentiation, large tumor size, and poor survival in ductal breast carcinoma. Thus, HuR is the first mRNA stability protein of which expression associates with poor outcome in breast cancer.
- Drosophila embryonic lethal abnormal vision protein
- breast cancer
- mouse models
Regulation of mRNA stability is an important element of eukaryotic gene expression (1, 2) . One of the best-characterized cis-acting elements of mRNA turnover is adenylate/uridylate-rich instability elements (ARE) located in the 3′ untranslated region of many unstable transcripts, such as those for certain proto-oncogenes, cytokines, and cytokine-response genes (2). HuR (or HuA) is an ARE-binding factor that is related to Drosophila embryonic lethal abnormal vision protein (3). Expression pattern of HuR is ubiquitous, whereas the other Hu family members (HuB/Hel-N1, HuC, and HuD) are primarily found in neuronal tissues (3, 4) . Intracellular localization of HuR is predominantly nuclear, but it shuttles between the nucleus and the cytoplasm, which suggests that HuR binds to ARE-containing mRNAs in the nucleus and that this complex is then transported to the cytoplasm (5). While HuR is bound to the transcript, it stabilizes the message and facilitates an efficient translation of the protein.
In brain tumors, HuR protein is expressed in poorly differentiated tumors, such as glioblastoma multiforme (6). In this tumor type HuR immunoreactivity is primarily nuclear with lesser cytoplasmic positivity. Similarly, elevated immunoreactivity of HuR has been detected in colorectal cancer as compared with adjacent nonneoplastic epithelium (7). Owing to nucleocytoplasmic translocation of HuR being necessary for its activity and cytoplasmic presence of HuR found in several types of carcinomas, it was hypothesized that cytoplasmic HuR expression could be a prognostic marker in cancer patients (8–10) . In our experience, nuclear HuR immunopositivity is found almost invariably in serous ovarian carcinoma specimens, whereas cytoplasmic positivity was observed only in a subset of the tumors (8). Furthermore, we and others have identified cytoplasmic expression of HuR protein to associate with poor prognosis in ovarian carcinoma patients (8, 10) . To investigate the localization and significance of HuR protein expression in breast cancer, we have analyzed HuR protein expression in a mouse mammary gland tumor model and investigated the relevance of HuR expression in a series of patients with invasive ductal breast carcinoma.
Materials and Methods
Patients. Women with invasive ductal breast carcinoma treated at the Departments of Surgery and Oncology, Helsinki University Central Hospital, in 1987 to 1990 were included into the study. Patients with in situ carcinoma, distant metastases at the time of the diagnosis, synchronous or metachronous bilateral breast cancer, malignancy other than breast cancer in history (except for basal cell carcinoma or cervical carcinoma in situ), and women who did not undergo breast surgery were excluded. This left 149 patients for the analysis, of whom sufficient clinical data and histologic specimens could be retrieved from 140 patients. All patients underwent level I and II axillary lymph node dissection. Postoperative radiotherapy was given to 63% (61 of 97) of patients treated with mastectomy and to 93% (41 of 44) of patients treated with breast sparing surgery. In the axillary node-negative group 2.4% (2 of 83) received adjuvant therapy (one woman was treated with antiestrogen therapy and one with chemotherapy) and in the node-positive group 67% (33 of 49) received adjuvant therapy (23 were treated with antiestrogen therapy, 9 with chemotherapy, and 1 received both treatments). The patients were followed up at 6-month intervals for the first 5 years after the diagnosis and then annually in an outpatient clinic. The median duration of follow-up of the patients still alive was 10.1 years (range 8.6-12.6 years) and of all patients 9.0 years (range 0.3-12.6 years). The median age at the time of the diagnosis was 59 years (range 34-89 years). The study was approved by an institutional ethical committee.
Western Blot. Mouse mammary glands were dissected from CD1 wild-type mice and from CD1 cyclooxygenase-2 (COX-2) transgenic mice, homogenized in an extraction buffer containing 1% Tween 20, following which Western blot analysis was conducted using 60 μg of protein as described previously (11). A monoclonal HuR antibody (1:10,000 dilution, 1 μg/mL, 19F12) was a kind gift of Clonegene LLC (Hartford, CT). Loading was controlled with β-actin antibody (1:10,000 dilution; Santa Cruz Biotechnology, Santa Cruz, CA).
Immunohistochemistry. Immunostaining protocol of mouse and human specimens for HuR was carried out as described previously (8). For mouse samples the protocol was identical except that the specimens were pretreated with Vector M.O.M. Basic Kit according to instructions of the supplier (Vector Laboratories, Burlingame, CA). The HuR antibody was used in a dilution of 1:20,000 (0.5 μg/mL). Four of the 140 breast cancer specimens did not contain any tumor cells, two turned out to be of nonductal histology after reevaluation of the specimens, and one specimen was from a lymph node metastasis, which left us 133 specimens for the analysis. To confirm the specificity of the staining a subset of the specimens (n = 7) were restained with and without the antigenic peptide for 19F12 (10 μg/mL; Clonegene) for 1 hour at room temperature before the staining procedure.
Evaluation of HuR Immunostaining. HuR immunoreactivity was scored independently and in a blinded manner by two investigators (A.R. and M.H.) from the 133 breast cancer specimens. HuR immunostaining score was based on the following criteria: nuclear staining only; low intensity of cytoplasmic HuR staining present (visible with 100× or a higher magnification); high intensity of cytoplasmic HuR staining present (visible with 50× or a lower magnification). Each staining set of 20 specimens contained two predetermined colon carcinoma control slides, one of which contained only nuclear staining in the tumor cells and another one with cytoplasmic immunopositivity. The percent agreement between the two independent and blinded investigators in allocation of the tumors into cytoplasm-negative or -low versus cytoplasm-high categories was 94%. All specimens with discordant scores (n = 8) were reevaluated by the two investigators using a multiheaded microscope, and the consensus score was used for further analyses.
Statistical Analysis. The association between HuR staining and clinically relevant and prognostic variables was assessed by using the χ2 test. Life tables were computed according to the Kaplan-Meier method. Distant disease-free survival was calculated from the date of the diagnosis to the occurrence of metastases outside the regional area or to death from breast cancer. Overall survival was calculated from the date of the diagnosis to death from breast cancer. Patients who died from intercurrent causes were censored at the date of death. Survival curves were compared with the log-rank test. Multivariate survival analyses were done with the Cox proportional hazards model by entering the covariates that were statistically significant in a univariate survival analysis: HuR expression (only nuclear or low cytoplasmic versus high cytoplasmic), tumor size in centimeters (continuous), the number of axillary node metastases (continuous), and histologic grade (well differentiated versus moderate to poorly differentiated). Cox regression was done using a backward stepwise selection of variables, and a P of 0.05 was adopted as the limit for inclusion of a covariate.
HuR Protein Expression in Mouse Mammary Gland Tumors. Recently, we developed transgenic mice that overexpress human COX-2 gene in the mammary glands using the murine mammary tumor virus promoter (11). The transgene is expressed in the mammary glands and its expression is particularly high during pregnancy and lactation, and multiparous females exhibit exaggerated tumor formation in the mammary glands. As shown in Fig. 1A , there is no marked difference of HuR expression in virgin mice when compared with pregnant, lactating, or weaning mice. Owing to COX-2 expression being elevated in pregnant and lactating mice ( Fig. 1A), this suggests that COX-2 expression alone does not have a profound effect on HuR protein expression in mouse mammary tissue. However, in a tumor sample derived from a transgenic mouse the level of HuR protein was substantially higher than in mammary gland tissue derived from a wild-type mouse. This tempted us to investigate localization of HuR in the mammary gland tissues. In wild-type mouse mammary glands HuR expression was mostly absent or localized in the nuclei of epithelial cells (n = 5; Fig. 1B). In contrast, in transgenic mammary gland tumors HuR was strongly expressed in the cytoplasm of the tumor cells (n = 2; Fig. 1C). These observations are in line with the hypothesis that during tumorigenesis HuR expression translocates from the nucleus to the cytoplasm.
HuR Protein Expression in Human Invasive Ductal Breast Carcinoma. Cytoplasmic HuR immunoreactivity was absent in 60% (80 of 133), low in 11% (15 of 133), and high in 29% (38 of 133) of the invasive ductal breast carcinoma specimens. Representative immunostaining results are shown in Fig. 2A and B . Only one of the 133 tumor samples had no nuclear HuR staining. Specificity of the HuR antibody was confirmed by using the antigenic peptide as a blocking reagent, and this control procedure blocked both nuclear and cytoplasmic immunoreactivity of HuR (data not shown).
Association of HuR Expression with Clinicopathologic Variables. As shown in Table 1 , high cytoplasmic HuR expression was more frequent in poorly differentiated carcinomas (P = 0.005) and in cases with tumor diameter over 2 cm (P = 0.0082). No significant association was found between high cytoplasmic HuR expression and age at diagnosis, axillary lymph node status, or estrogen receptor expression (P > 0.1 for all comparisons).
Univariate Survival Analysis. Presence of metastatic axillary lymph nodes (P < 0.0001), a large primary tumor size (P = 0.0015), high histologic grade (P = 0.03), and high cytoplasmic HuR expression (P < 0.0001, log-rank test or log-rank test for trend) were associated with poor survival in a univariate survival analysis, whereas age at diagnosis, estrogen or progesterone receptor expression, and adjuvant therapy were not significantly associated with outcome in the present series.
Five-year distant disease-free survival of the patients with no cytoplasmic HuR expression was 85% (n = 80; 95% CI, 77-93), in low cytoplasm-positive category 79% (n = 15; 95% CI, 57-100), and in high cytoplasm-positive category 42% (n = 38; 95% CI, 26-58; P < 0.0001). When the patients with low or absent cytoplasmic HuR expression were analyzed as a single group, 5-year distant disease-free survival rate was 84% (95% CI, 76-91). Patients with high cytoplasmic HuR expression turned out to have markedly reduced survival when compared with the patients with low or absent cytoplasmic HuR ( Fig. 3 ). When we used overall survival instead of distant disease-free survival, 5-year survival in the cytoplasm-negative or -low category was 93% (95% CI, 87-98) and in the high cytoplasm-positive category 58% (95% CI, 42-74; P < 0.0001). To further characterize the role of HuR in breast cancer, we analyzed certain subgroups in respect of survival in the two HuR categories. These data suggest that HuR is an important prognostic variable in subgroups of axillary lymph node-negative breast carcinomas, in carcinomas with tumors of the size below 2 cm, and in those with estrogen receptor expression ( Table 2 ).
Multivariate Analysis. To find out whether HuR expression is an independent prognostic factor in breast cancer, HuR expression was entered into a multivariate model together with axillary lymph node status, tumor size, and histologic grade (i.e., the variables that were significantly associated with outcome in the univariate survival analyses). In the multivariate survival analysis high cytoplasmic HuR immunopositivity (P = 0.035), axillary lymph node status (P = 0.0007), and tumor size (P = 0.01) were identified as independent prognostic variables, whereas histologic grade was not ( Table 3 ).
The present results show that cytoplasmic HuR expression associates with poor outcome in patients with invasive ductal breast carcinoma. We also show that HuR expression is elevated in a mouse mammary gland tumor model and that expression of HuR is localized in the cytoplasm of the tumor cells. These observations are in line with the hypothesis that during tumorigenesis HuR expression translocates from the nucleus to the cytoplasm (8–10) . Our data are also in line with another mouse tumor model in which urethane-induced neoplasias of the lung showed cytoplasmic expression of HuR (12). In addition, recent results published by Lopez et al. (9) suggest that HuR plays a role in tumorigenesis owing to overexpression of HuR promoting growth of human colorectal cancer cells in nude mice. All this indicates that identification of cytoplasmic HuR expression could provide useful information about the behavior of malignant tumors.
We found cytoplasmic HuR immunoreactivity in 40% of the 133 invasive ductal breast cancer specimens, and 29% of these samples showed high level of expression. This is in line with our previous data on serous ovarian carcinoma, in which we found cytoplasmic HuR expression in 52% of the cases (8). Similarly, Denkert et al. (10, 13) reported cytoplasmic HuR immunoreactivity in 45% of ovarian carcinomas and in 30% of breast carcinoma specimens. Association of cytoplasmic HuR immunoreactivity with poor differentiation seems to be a general phenomenon that has been previously shown in ovarian and breast carcinomas, but in these previous studies an association with reduced survival was found only among ovarian cancer patients (8, 10, 13) . This discrepancy with respect of its marked prognostic role in breast cancer may partially be due to differences in patient series and treatments or use of different antibody preparations or immunohistochemistry scoring criteria.
In respect of survival analysis, we made an important observation showing that although high cytoplasmic HuR expression is more frequent in poorly differentiated cancer and when the tumor size is over 2 cm ( Table 1), the prognostic significance of this marker remains significant in the subgroups of cancers below 2 cm and without axillary lymph node metastasis ( Table 2). In addition, whereas frequency of estrogen receptor expression did not correlate with HuR staining pattern ( Table 1), the prognostic significance of high cytoplasmic HuR expression was evident in the estrogen receptor–positive subgroup of tumors, but not in the estrogen receptor–negative ones ( Table 2). However, although these data support a prognostic role of HuR among patients with early-stage and estrogen receptor–positive breast carcinoma, the subgroup analyses should be viewed with caution due to the small patient numbers.
Cytoplasmic HuR immunoreactivity is associated with COX-2 expression in breast and ovarian cancer (8, 10, 13) , and high COX-2 expression itself is a poor prognostic variable in these malignancies (14–17) . Because of HuR regulating COX-2 expression in ovarian and breast cancer cells (8, 18) , it is possible that part of its function is mediated via induction of COX-2 that can promote carcinogenesis of the breast via various mechanisms (19). In addition to COX-2, a HuR consensus binding motif has been identified in a number of other transcripts encoding for growth factors, cytokines, and cell cycle regulators (20), which are also likely mediators of HuR action in carcinogenesis. It is not known how HuR is retained in the cytoplasm of breast cancer cells, but one mechanism was recently described in human colorectal cancer cells where nuclear import of HuR was modulated by AMP-activated protein kinase (21). In addition, induction of the p38 mitogen-activated protein kinase can promote HuR-mediated mRNA stabilization (22), and the extracellular signal-regulated kinase induced binding of HuR to the transcript of p21 (23).
In summary, the results show that high cytoplasmic HuR expression is associated with a high histologic grade, a large primary tumor size, and poor survival of patients with in invasive ductal breast cancer. Thus, HuR is the first mRNA stability protein of which expression associates with poor outcome in breast cancer. Owing to HuR expression having prognostic influence in the subgroups of small cancers and among patients with node-negative cancer, where the need for prognostication is the greatest, HuR expression may prove to be a clinically valuable prognostic variable. However, due to the relatively small patient series larger studies are warranted.
Grant support: Academy of Finland, the Finnish Cancer Foundation, Helsinki University Central Hospital Research Funds, the Sigrid Juselius Foundation, and NIH grants HL70694 and HL49094 (H. Furneaux and T. Hla).
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.
We thank Tuija Hallikainen, Elina Laitinen, and Päivi Peltokangas for excellent technical assistance.
- Received October 22, 2004.
- Revision received December 31, 2004.
- Accepted January 6, 2005.
- ©2005 American Association for Cancer Research.